Split (1)

train · 150k rows

file stringlengths 27 28	text stringlengths 1 5.17M
PMC009xxxxxx/PMC9004451.txt	==== Front Environ Chem Lett Environ Chem Lett Environmental Chemistry Letters 1610-3653 1610-3661 Springer International Publishing Cham 35431713 1434 10.1007/s10311-022-01434-9 Original Paper SARS-CoV-2 monitoring by automated target-driven molecular machine-based engineering Fan Zhenqiang 1 Xie Minhao 13 Pan Jianbin jbpan@nju.edu.cn 2 http://orcid.org/0000-0002-7021-5395 Zhang Kai zhangkai@jsinm.org 1 1 grid.412676.0 0000 0004 1799 0784 NHC Key Laboratory of Nuclear Medicine, Jiangsu Key Laboratory of Molecular Nuclear Medicine, Jiangsu Institute of Nuclear Medicine, Wuxi, Jiangsu, 214063 China 2 grid.41156.37 0000 0001 2314 964X State Key Laboratory of Analytical Chemistry for Life Science and Collaborative Innovation Center of Chemistry for Life Sciences, School of Chemistry and Chemical Engineering, Nanjing University, Nanjing, 210023 China 3 grid.89957.3a 0000 0000 9255 8984 Department of Radiopharmaceuticals, School of Pharmacy, Nanjing Medical University, Nanjing, 211166 China 12 4 2022 2022 20 4 22272233 10 1 2022 8 3 2022 © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2022, corrected publication 2022Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Biosensors based on nucleic acid-structured electrochemiluminescence are rapidly developing for medical diagnostics. Here, we build an automated DNA molecular machine on Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au composite, which alters the situation that a DNA molecular machine requires laying down motion tracks. We use this DNA molecular machine to transduce the target concentration information to enhance the electrochemiluminescence signal based on DNA hybridization calculations. Complex bioanalytical processes are centralized in a single nucleic acid probe unit, thus eliminating the tedious steps of laying down motion tracks required by the traditional molecular machine. We found a detection limit of 0.68 pM and a range of 1 pM to 1 nM for the analysis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) specific DNA target. Recoveries range between 96.4 and 104.8% for the analysis of SARS-CoV-2 in human saliva. Supplementary Information The online version contains supplementary material available at 10.1007/s10311-022-01434-9. Keywords SARS-CoV-2 monitoring Ti3C2-based composites Automated molecular machine Saliva environmental testing Modular reaction http://dx.doi.org/10.13039/501100001809 National Natural Science Foundation of China 21705061 Zhang Kai Jiangsu Provincial Health Care Commission Scientific Research ProjectM2021035 Zhang Kai Wuxi “Taihu Light” science and technology (medical and health technology) researchY20212049 Zhang Kai issue-copyright-statement© Springer Nature Switzerland AG 2022 ==== Body pmcIntroduction The coronavirus disease 2019 (COVID-19) has caused the collapse of health systems in many parts of the world. It is spreading rapidly worldwide like modern infectious diseases before it (Yan et al. 2021). Currently, countries around the globe are actively collaborating and working together to control the spread of the global outbreak. Developing and optimizing methods for rapid and accurate diagnosis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) become particularly important to test a country’s public health capacity of rapidly identifying infected individuals and taking appropriate measures to contain the virus in a pandemic (Ali and Alharbi 2020; Delafiori et al. 2021). Recently, nucleic acid-structured electrochemiluminescence biosensing platforms have been widely applied in bioanalysis, food safety, and environmental contaminant monitoring due to the characteristics of fast response, simple operation, compact device, and high sensitivity (Fan et al. 2021b; Qi and Zhang 2020). These nucleic acid-involved electrochemiluminescence sensings rely on programmed DNA probes to participate in molecular recognition and signaling (Feng et al. 2017). Programmed DNA tetrahedron reveals enormous potential due to their stiffness and size programmability, which significantly improves the low attachment efficiency, uneven spatial distribution and reduce the aggregation of conventional DNA such as single-stranded DNA or double-stranded DNA (Fan et al. 2021a, 2020, 2022, 2021c; Li et al. 2014). Typically, the diagnosis procedure of SARS-CoV-2 consists of the following steps. Firstly, the samples to be tested are collected by oropharyngeal or nasopharyngeal swab from the suspected COVID-19 patients. Secondly, the virus RNA is extracted through an RNA extraction step (Mattioli et al. 2020). Then, target DNA is obtained from viral RNA by enrichment steps containing reverse transcription-polymerase chain reaction or reverse transcription recombinase polymerase amplification (Byrnes et al. 2021; Mahas et al. 2021). In detecting SARS-CoV-2, both upstream RNA and downstream DNA can be used as targets to diagnose COVID-19. Herein, we have developed an interfacial DNA machine containing an inverted DNA tetrahedron using three capture probes as reaction probes to monitor the SARS-CoV-2 downstream DNA. When target DNA is added to the reaction system, the DNA machine produces a correlation effect on the electrochemiluminescence signal output using Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au as electrochemiluminescence luminous material on the glassy carbon electrode based on DNA hybridization calculations, by which the concentration of the SARS-CoV-2 could be deduced. Experimental Principle of the proposed electrochemiluminescence biosensor The electrochemiluminescence emitting material (Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au) is coated onto the glassy carbon electrode surface, and the electrochemiluminescence intensity is used as the sensor’s signal output (Fig. 1). Then, we modify the automated molecular machine on the surface of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au. The inverted tetrahedron has only one vertex modified by the thiol group, while possessing three capture probes where two of the strands (H and P) are partially complementarily paired with H’ and P’, respectively. The one remaining capture probe (S’-S) forms a hairpin structure by itself and its S’ end carries the quenching motif ferrocene of the electrochemiluminescence signal. When the target DNA binds to the exposed toehold sequence of the anticodon H’, H is released. Then, H binds to the toehold sequence of the anticodon P’, which allows P to be released. Immediately afterward, P binds to the anticodon S’-S, producing a site that is recognized and cleaved by the nuclear endonuclease (Nt.BbvCI), resulting in the release of quenched motif ferrocene and reuse of the P. In this process, the repeatedly generated P enables the cyclic cleavage of the S’-S and the cyclic release of the ferrocene, resulting in the generation of a transduction signal that allows the assessment and quantitative analysis of target concentration.Fig. 1 Synthesis of the Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au composite (Ti3C2/PEI-Ru@Au) and the electrochemiluminescence (ECL) process of automated molecular machine in the case of target (T) input on glassy carbon electrode (GCE). The box in the middle of the scheme is a concise frame diagram of the actual electrode reaction presented at the bottom of the scheme. The triangles represent the plane of the tetrahedron connecting the H’·H, P’·P, and ferrocene-S–S’ (Fc-S–S’) and the dashed arrows represent the direction in which the nucleic acid chain is to attack. Characterization of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au We have characterized the synthesis of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au. Ti3C2 transmission electron microscope image (Fig. 2A) shows monolayers or multilayers of nanosheets. In contrast, the Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au transmission electron microscope image shows that the gold nanoparticles are spread on the surface of Ti3C2 (Fig. 2B). The magnified Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au transmission electron microscope image (Fig. 2C) and its inset show that the gold nanoparticles are uniformly distributed with a particle size of approximately 4.3 nm. To further validate the synthesis of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au, we characterized the elemental distribution (Figs. S1A-H). The elemental distribution also shows that Au (Fig. S1H) and Ru elements (Fig. S1E) are distributed on the surface of Ti (Fig. S1C) and C (Fig. S1D), further indicating that Au particles and Ru(dcbpy)32+ are uniformly distributed on the Ti3C2 nanosheets. We have performed X-ray photoelectron spectroscopy analysis of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au nanocomposite. The X-ray photoelectron spectroscopy survey spectrum of the complex (Fig. 2D) exhibits Ti 2p and C 1 s peaks, which are the elemental peaks of Ti3C2. The formation process of Ti3C2 produces F and Cl elements; therefore, F 1 s and Cl 2p peaks are observed. The complex also has Ru 3p, Ru 3d, and Au 4f characteristic peaks due to the introduction of Ru(dcbpy)32+ and the reduction of Au3+ to gold nanoparticles. We also elaborate the bonding interaction between the complexes by fitting C 1 s, O 1 s, N 1 s and Au 4f. Four fitted peaks have appeared in the C 1 s peak of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au (Fig. 2E), which mainly correspond to the characteristic bonds of O-C=O, C-O/C-N, C–C and C-Ti. Similarly, the X-ray photoelectron spectroscopy spectra at O 1 s (Fig. 2F) show the typical bonds of C-Ti-(OH), C-Ti-Ox and TiO2. The X-ray photoelectron spectroscopy spectra at N 1 s (Fig. 2G) also show the characteristic bonds of R-NH2, R-NH-R, R = N-R and N-Ti. The double peaks of Au (Fig. 2H) at Au 4f7/2 (82.83 eV) and Au 4f7/2 (86.51 eV) also verify the presence of gold particles. All these X-ray photoelectron spectroscopy analysis data indicate the formation of the Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au complexes. We also confirm the synthesis of the complex by ultraviolet–visible spectral analysis (Fig. 2I). It shows that the Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au complex has both Ru(dcbpy)32+-polyethyleneimine (characteristic peaks at 300 nm and 475 nm) and Au nanoparticles (characteristic absorption peak at 525 nm) peaks, also verifying the synthesis of Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au complex.Fig. 2 Transmission electron microscope images (A–C), x-ray photoelectron spectroscopy (D–H) and ultraviolet–visible absorption spectra (I) are employed to illustrate the synthesis of the Ti3C2/polyethyleneimine-Ru(dcbpy)32+@Au nanocomposite (Ti3C2/PEI-Ru@Au). Specifically, A Ti3C2 nanosheets and B Ti3C2/PEI-Ru@Au. C The magnified image of Ti3C2/PEI-Ru@Au and the inset is the size distribution of Au nanoparticles in Fig. 2C. D X-ray photoelectron spectroscopy survey and high-resolution C 1 s (E), O 1 s (F), N 1 s (G), Au 4f (H). I The ultraviolet–visible absorption spectra of Ti3C2, Au nanoparticles (Au NPs), polyethyleneimine (PEI), polyethyleneimine and Ru(dcbpy)32+ composite (Ru + PEI) and Ti3C2/PEI-Ru@Au Characterization of the inverted DNA tetrahedral scaffold We also characterize the inverted DNA tetrahedra by atomic force microscope (Fig. 3A). The DNA tetrahedra formed by four strands (H-T1, P-T2, S’-S-T3, T4) appear granular and uniform in size. Meanwhile, we characterize the heights of the selected tetrahedral particles relative to the silicon wafer base (Fig. 3B-F). We define the silicon wafer substrate as the control, and the height of the tetrahedra relative to the substrate as the actual height of the DNA tetrahedra. The statistical height ranges between 8.43 nm and 9.82 nm, indicating the successful synthesis of inverted tetrahedra with relatively homogeneous dimensions. Fig. 3 A 2D atomic force microscope image is used to illustrate the formation of the inverted DNA tetrahedron. The characterization is carried on a clean silicon wafer. B–F are height maps plotted along the tangents at points a-e in Fig. 3A. The dotted line marks the vertical height of the inverted DNA tetrahedron relative to the silicon wafer surface which is served as the control Feasibility of DNA reactions exploited by the automated molecular machine We next verify the possibility of cascade reactions on the automated molecular machine (Fig. S2A). Lane 4 shows that when T and H’-H-T1 are present together, which can be effective against P’-P-T2, eventually producing three major new bands: T-H’, T1-H-P’ and P-T2. However, lane 2 indicates that H’-H-T1 and P’-P-T2 do not react when T is absent. Likewise, lane 3 T does not act on P’-P-T2 when the H’-H-T1 intermediate is absent. Lanes 1–3 are raw material strips. Upon adding H-T1 with P’-P-T2 to S’-S-T3 with nicking endonuclease (lane 4), we observe the disappearance of S’-S-T3 and the appearance of three new bands: P-T2, T1-H-P’ and cleaved S’-S-T3 (Fig. S2B). It indicates the successful linkage of the nicking reaction to the previous one. Owing to the essential H-T1 linkage, the modular cascade reaction on the DNA molecular machine proceeds smoothly. At the same time, the newly generated P-T2 is available for the continuous action of S’-S-T3 until its depletion. The evidences provided by the two images suggest that the reaction occurs as expected on the automated molecular machine. Results and discussion Performance, specificity and stability The automated molecular machine exhibits highly sensitive electrochemiluminescence variations for target concentrations (Fig. 4A, B). The increased electrochemiluminescence intensity is enhanced with the logarithmic value of the target concentration ranging from 1 pM to 1 nM. The linear equation of the calibration plot is y = 431.5 + 1255.9lgCtarget, where y represents the increased electrochemiluminescence intensity. The calculated limit of detection is 0.68 pM according to 3σ method. We compare our strategy in terms of detection methods, sensitivity and detection range with other viral DNA detection methods (Table S1). It notes that our detecting method performs well in these three aspects. Besides, this method as a fundamental strategy in combination with other signal amplification techniques such as hybridization chain reaction, loop-mediated isothermal amplification (LAMP), the catalytic hairpin assembly will achieve higher sensitivity. We further evaluate the specificity of DNA automated molecular machine for the target monitoring. The sensing performances of three non-specific strands, including Bat SARS-related CoV isolate bat-SL-CoVZC45 (M1), BM48-31/BGR/2008 (M2) and SARS-CoV, are investigated (Fig. 4C). We see that the electrochemiluminescence intensity of the target is higher than the DNA intensity of the various mismatched strands, which also illustrates that our automated molecular machine has superb accuracy. Besides, our automated molecular machine possesses excellent reproducibility (Fig. 4D) by performing continuous potential scans for 15 cycles, with relative standard deviations of only 2.54% and 2.13% for 50 pM and 0.5 nM, respectively. Fig. 4 Assay performance (A, B), specificity (C) and repeatability (D) are investigated to validate the superior characteristics of our sensors. A Electrochemiluminescence-Time (ECL-Time) and Electrochemiluminescence (ECL)-Potential (ECL-Potential, the inset of Fig. 4A) curves of the strategy when monitoring target with different concentrations (1 pM, 10 pM, 20 pM, 50 pM, 0.1 nM, 0.2 nM, 0.5 nM and 1 nM). B The variation of increased electrochemiluminescence intensity (ΔECL) with target concentration and the logarithm of target concentration (the inset of Fig. 4B) in the detection region from 1 pM to 1 nM. C Specificity of strategy at 0.1 nM target and 1 nM non-targets. D Sensor stability over 15 cycles of potential sweeps in the presence of 50 pM (blue curve) and 0.5 nM target (red curve). The relative standard deviations (RSD) were obtained by the peak values of the 15 cycles at the corresponding concentrations Analysis of saliva dilutions Since viral DNA is usually obtained from viral RNA by techniques such as reverse transcription-polymerase chain reaction or reverse transcription recombinase polymerase amplification during actual virus detection, some non-specific substances interfere with the diagnosis of COVID-19. Therefore, we discuss the feasibility of the assay under complex environments. The pharyngeal swabs method usually collects viral or bacterial samples in saliva. To verify the ability of the DNA automated molecular machine to monitor targets in the saliva environment, we add targets with different concentrations to 20-fold or 50-fold human saliva dilutions and verify their feasibility (Table S2). After repeated experiments (n = 3), the obtained recoveries are between 96.4 and 104.8%, which further validates the reliability of our DNA automated molecular machine platform for target monitoring. During the experiments, targets in the range of 5 pM to 5 nM are successfully monitored, demonstrating the universality of our DNA automated molecular machine in the complex saliva environments. Conclusion Herein, we have developed an inverted tetrahedron-based DNA molecular machine for SARS-CoV-2 target monitoring, which demonstrates excellent interference resistance and performs well in a saliva dilution environment. Only one thiol group is required for inverted tetrahedra, significantly reducing the experimental fee due to nucleic acid modifications. Moreover, the sophisticated biochemical reactions are amplified on a molecular machine with high programmability and controllability, which is expected to be promoted to medical diagnostics and environmental monitoring. This assay strategy has a promising application in determining other viral and protein biomarkers by designing the three probes that extend out. The molecular device enables the detection of contaminants in the environment and clinical specimen dilutions, which paves a novel way to advance public health development and assess their hazards in the environment. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 2742 kb) Acknowledgements This work was supported by the National Natural Science Foundation of China (21705061), the Jiangsu Provincial Health Care Commission Scientific Research Project (M2021035), the Wuxi “Taihu Light” science and technology (medical and health technology) research (Y20212049) and the Jiangsu Provincial Key Medical Discipline (Laboratory) (ZDXKA2016017). Author contributions ZF involved in data curation, formal analysis, and writing—review and editing. MX involved in validation. JP involved in supervision. KZ involved in conceptualization and supervision. Funding National Natural Science Foundation of China, 21705061, Kai Zhang ,Jiangsu Provincial Health Care Commission Scientific Research Project, M2021035, Kai Zhang, Wuxi “Taihu Light” science and technology (medical and health technology) research, Y20212049, Kai Zhang. Declarations Conflict of interrest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. The original publication has been revised due to inclusion of Supplementary materials Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Ali I Alharbi OML COVID-19: Disease, management, treatment, and social impact Sci Total Environ 2020 728 138861 10.1016/j.scitotenv.2020.138861 32344226 Byrnes SA Gallagher R Steadman A Multiplexed and extraction-free amplification for simplified SARS-CoV-2 RT-PCR tests Anal Chem 2021 93 9 4160 4165 10.1021/acs.analchem.0c03918 33631932 Delafiori J Navarro LC Siciliano RF Covid-19 automated diagnosis and risk assessment through metabolomics and machine learning Anal Chem 2021 93 4 2471 2479 10.1021/acs.analchem.0c04497 33471512 Fan Z Ding Y Yao B Electrochemiluminescence platform for transcription factor diagnosis by using CRISPR–Cas12a trans-cleavage activity Chem Commun 2021 57 65 8015 8018 10.1039/D1CC03071J Fan Z Lin Z Wang Z Dual-wavelength electrochemiluminescence ratiometric biosensor for NF-κB p50 detection with dimethylthiodiaminoterephthalate fluorophore and self-assembled DNA tetrahedron nanostructures probe ACS Appl Mater Interfaces 2020 12 10 11409 11418 10.1021/acsami.0c01243 32067445 Fan Z Yao B Ding Y Electrochemiluminescence aptasensor for Siglec-5 detection based on MoS2@Au nanocomposites emitter and exonuclease III-powered DNA walker Sens Actuat B 2021 334 129592 10.1016/j.snb.2021.129592 Fan Z Yao B Ding Y Rational engineering the DNA tetrahedrons of dual wavelength ratiometric electrochemiluminescence biosensor for high efficient detection of SARS-CoV-2 RdRp gene by using entropy-driven and bipedal DNA walker amplification strategy Chem Eng J 2022 427 131686 10.1016/j.cej.2021.131686 34400874 Fan Z Yao B Ding Y Entropy-driven amplified electrochemiluminescence biosensor for RdRp gene of SARS-CoV-2 detection with self-assembled DNA tetrahedron scaffolds Biosens Bioelectron 2021 178 113015 10.1016/j.bios.2021.113015 33493896 Feng Q-M Guo Y-H Xu J-J Self-assembled DNA tetrahedral scaffolds for the construction of electrochemiluminescence biosensor with programmable DNA cyclic amplification ACS Appl Mater Interfaces 2017 9 20 17637 17644 10.1021/acsami.7b04553 28471159 Li Z Zhao B Wang D DNA nanostructure-based universal microarray platform for high-efficiency multiplex bioanalysis in biofluids ACS Appl Mater Interfaces 2014 6 20 17944 17953 10.1021/am5047735 25299733 Mahas A Wang Q Marsic T A novel miniature crispr-cas13 system for SARS-CoV-2 diagnostics ACS Synth Biol 2021 10 10 2541 2551 10.1021/acssynbio.1c00181 34546709 Mattioli IA Hassan A Oliveira ON On the challenges for the diagnosis of SARS-CoV-2 based on a review of current methodologies ACS Sensors 2020 5 12 3655 3677 10.1021/acssensors.0c01382 33267587 Qi H Zhang C Electrogenerated chemiluminescence biosensing Anal Chem 2020 92 1 524 534 10.1021/acs.analchem.9b03425 31789502 Yan L Yi J Huang C Rapid detection of COVID-19 using maldi-tof-based serum peptidome profiling Anal Chem 2021 93 11 4782 4787 10.1021/acs.analchem.0c04590 33656857
PMC009xxxxxx/PMC9004452.txt	==== Front J Med Toxicol J Med Toxicol Journal of Medical Toxicology 1556-9039 1937-6995 Springer US New York 35415804 890 10.1007/s13181-022-00890-7 Original Article A Brief Educational Intervention to Increase ED Initiation of Buprenorphine for Opioid Use Disorder (OUD) Khatri Utsha G. Utsha.khatri@mountsinai.org 1234 Lee Kathleen 35 Lin Theodore 5 D’Orazio Joseph L. 6 Patel Mitesh S. 2789 Shofer Frances S. 310 Perrone Jeanmarie 311 1 grid.25879.31 0000 0004 1936 8972 National Clinician Scholars Program, University of Pennsylvania, Philadelphia, PA USA 2 grid.410355.6 0000 0004 0420 350X Corporal Michael J. Crescenz VA Medical Center, Philadelphia, PA USA 3 grid.25879.31 0000 0004 1936 8972 Department of Emergency Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA USA 4 grid.59734.3c 0000 0001 0670 2351 Icahn School of Medicine at Mount Sinai, 1 Gustave Levy Place, New York, NY 10029 USA 5 grid.25879.31 0000 0004 1936 8972 Penn Medicine Center for Digital Health, Center for Health Care Innovation, Perelman School of Medicine, Philadelphia, PA USA 6 grid.264727.2 0000 0001 2248 3398 Department of Emergency Medicine, Lewis Katz School of Medicine, Temple University, Philadelphia, PA USA 7 grid.25879.31 0000 0004 1936 8972 Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA USA 8 grid.25879.31 0000 0004 1936 8972 Health Care Management, Wharton School, University of Pennsylvania, Philadelphia, PA USA 9 grid.25879.31 0000 0004 1936 8972 Penn Medicine Nudge Unit, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA USA 10 grid.25879.31 0000 0004 1936 8972 Department of Epidemiology & Biostatistics, Center for Public Health, University of Pennsylvania, Philadelphia, PA USA 11 grid.412701.1 0000 0004 0454 0768 Penn Medicine Center for Addiction Medicine and Policy, Philadelphia, PA USA Supervising Editor: Leslie R. Dye, MD 12 4 2022 7 2022 18 3 205213 16 7 2021 7 3 2022 10 3 2022 © American College of Medical Toxicology 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Background Despite the evidence in support of the use of buprenorphine in the treatment of OUD and increasing ability of emergency medicine (EM) clinicians to prescribe it, emergency department (ED)-initiated buprenorphine is uncommon. Many EM clinicians lack training on how to manage acute opioid withdrawal or initiate treatment with buprenorphine. We developed a brief buprenorphine training program and assessed the impact of the training on subsequent buprenorphine initiation and knowledge retention. Methods We conducted a pilot randomized control trial enrolling EM clinicians to receive either a 30-min didactic intervention about buprenorphine (standard arm) or the didactic plus weekly messaging and a monetary inducement to administer and report buprenorphine use (enhanced arm). All participants were incentivized to complete baseline, immediate post-didactic, and 90-day knowledge and attitude assessment surveys. Our objective was to achieve first time ED buprenorphine prescribing events in clinicians who had not previously prescribed buprenorphine in the ED and to improve EM-clinician knowledge and perceptions about ED-initiated buprenorphine. We also assessed whether the incentives and reminder messaging in the enhanced arm led to more clinicians administering buprenorphine than those in the standard arm following the training; we measured changes in knowledge of and attitudes toward ED-initiated buprenorphine. Results Of 104 EM clinicians enrolled, 51 were randomized to the standard arm and 53 to the enhanced arm. Clinical knowledge about buprenorphine improved for all clinicians immediately after the didactic intervention (difference 19.4%, 95% CI 14.4% to 24.5%). In the 90 days following the intervention, one-third (33%) of all participants reported administering buprenorphine for the first time. Clinicians administered buprenorphine more frequently in the enhanced arm compared to the standard arm (40% vs. 26.3%, p = 0.319), but the difference was not statistically significant. The post-session knowledge improvement was not sustained at 90 days in the enhanced (difference 9.6%, 95% CI − 0.37% to 19.5%) or in the standard arm (difference 3.7%, 95% CI − 5.8% to 13.2%). All the participants reported an increased ability to recognize patients with opioid withdrawal at 90 days (enhanced arm difference .55, 95% CI .01–1.09, standard arm difference .85 95% CI .34–1.37). Conclusions A brief educational intervention targeting EM clinicians can be utilized to achieve first-time prescribing and improve knowledge around buprenorphine and opioid withdrawal. The use of weekly messaging and gain-framed incentivization conferred no additional benefit to the educational intervention alone. In order to further expand evidence-based ED treatment of OUD, focused initiatives that improve clinician competence with buprenorphine should be explored. Trial Registration ClinicalTrials.gov Identifier: NCT03821103. Supplementary Information The online version contains supplementary material available at 10.1007/s13181-022-00890-7. Keywords Buprenorphine Opioid use disorder Behavioral incentives issue-copyright-statement© American College of Medical Toxicology 2022 ==== Body pmcIntroduction Emergency departments (EDs) are uniquely positioned to expand access to treatment for patients with opioid use disorder (OUD). Overdose rates in recent years have escalated and much of this is attributed to the increase of fentanyl in the drug supply [1, 2]. A recent analysis of 25 EDs across multiple states found that opioid overdose visit rates increased by 28.5% in 2020 during the COVID-19 pandemic compared to 2018–2019 [3]. These patients are most vulnerable, experiencing a 5.5% risk of death 1 year following an ED visit for non-fatal opioid overdose, suggesting EM clinicians should prioritize treatment engagement [4]. Treatment of OUD with buprenorphine decreases both withdrawal symptoms and cravings for opioids and can be safely and effectively initiated from the ED [5–7]. Patients who receive buprenorphine are less likely to suffer a non-fatal or fatal overdose, use illicit opioids, develop Hepatitis C or HIV, suffer other infectious complications, or have contacts with the criminal justice system [8, 9]. However, access to buprenorphine is limited. At the time of our intervention, in order to prescribe buprenorphine, clinicians were required to complete an 8-h training course and receive a “DATA-waived” registration to their U.S. Drug Enforcement Agency (DEA) license (“X-waiver”) through the Substance Abuse and Mental Health Services Administration (SAMHSA). Buprenorphine prescribing has been low, with over half of waivered physicians not prescribing to capacity or not prescribing at all [10]. One study of providers in Washington demonstrated that in the 7 months following the full X waiver training, only 28% had ever prescribed buprenorphine. This gap between training and practice exacerbates the problem of insufficient treatment access. Additionally, negative attitudes about buprenorphine from providers are common and even those who report positive attitudes are unlikely to prescribe, with a lack of clinical confidence or institutional support cited as barriers [11–13]. Incentives have been shown to be effective at augmenting the implementation of ED buprenorphine programs, both in the number of X-waivered EM clinicians and the number of ED-buprenorphine prescriptions [14]. The ED patients with OUD who receive buprenorphine have twice the rate of treatment engagement at 30 days compared to those who receive discharge referral alone, suggesting a significant opportunity for EM clinicians to increase patient opportunities for treatment [15]. Studies have found that lack of training and experience treating OUD with buprenorphine is a major barrier to ED-initiated buprenorphine among EM-clinicians [16, 17]. One study reported that only 44% of EM clinicians felt they were prepared to discuss addiction treatment while another found that among clinicians who reported lack of training as a barrier, 64% reported that having training would increase their likelihood of initiating buprenorphine [13, 18]. In a multi-state survey in 2020, Zuckerman et al. found that among non-waivered ED physicians only 34% reported feeling comfortable initiating buprenorphine treatment and nearly half (47%) felt that it was not their job [18]. While buprenorphine administration was shown to have increased in EDs between 2002 and 2017, the prevalence of buprenorphine use in the ED remains low [19]. Myths and confusion about regulations create further barriers to ED buprenorphine administration. While the Biden administration removed the requirement of federally specified training in order to be permitted to prescribe buprenorphine to less than 30 patients at any one time, physicians must still apply for a waiver through the DEA. Additionally, buprenorphine may be administered in the ED by a non-waivered clinician through “the 72-h rule” (Title 21, Code of Federal Regulations, Part 1306.07(b)), which allows all clinicians “to administer narcotic drugs for the purpose of relieving acute withdrawal symptoms when necessary while arrangements are being made for referral to treatment” [20, 21]. Thus, regardless of their waiver status, the ED clinicians are permitted to administer buprenorphine to their patients to treat opioid withdrawal. To target the barriers preventing EM clinicians from administering buprenorphine to eligible patients (patients with OUD who exhibit symptoms of at least mild opioid withdrawal), we piloted a randomized control trial (RCT) testing the impact of behavioral incentives to enhance a didactic intervention to inform EM clinicians about the pharmacologic characteristics of buprenorphine, the physiological considerations of opioid withdrawal, the regulations around ED-administration of buprenorphine, and the framework of a “warm hand-off.” Our objective was to achieve first-time ED buprenorphine prescribing events in clinicians who had not previously prescribed buprenorphine in the ED as well as to improve EM-clinician knowledge and perceptions on ED-initiated buprenorphine. We assessed baseline knowledge and attitudes and evaluated the effect of the educational intervention immediately after the session and again at 90 days. We hypothesized that a brief educational intervention would increase knowledge and improve attitudes, translating into practice change (the outcome of administering buprenorphine to a patient for the first time) among all EM clinicians. Furthermore, we hypothesized that utilization of additional behavioral-economics-framed nudges and weekly texts with buprenorphine-evidence statements would contribute to retention of knowledge and favorable attitudes and enhance the outcome of change in practice (buprenorphine administration) compared to educational training alone. Notably, this study was designed prior to the recognition of the changed drug supply from heroin to fentanyl and any subsequent considerations of buprenorphine use in patients using fentanyl. Methods Participants Emergency medicine (EM) clinicians (residents, advanced practice providers (APPs), fellows, and attendings) practicing in EDs in the Greater Philadelphia area and in attendance of one of the targeted conferences or meetings were eligible to participate in this pilot study. The study was deemed exempt by the Institutional Review Board of the University of Pennsylvania and was pre-registered at ClinicalTrials.gov (Identifier: NCT03821103). Participants were recruited at educational or administrative conferences from February 2019 to October 2019. Eligibility was screened through an automated text-based survey. Clinicians were eligible to participate if they answered ‘no’ to the question “Have you ever administered (ordered and/or prescribed) buprenorphine or buprenorphine-naloxone (Suboxone) for a patient with opioid use disorder?” Individuals who were already “X-waivered” were included if they had not yet administered buprenorphine. We only included clinicians who had not yet administered buprenorphine (regardless of their waiver status) because we believed this group of clinicians would most benefit from training on the clinical and operational considerations of treating OUD in the ED, a competence that ED clinicians who administer buprenorphine routinely likely already possess. Educational Intervention The training didactic was offered during existing educational conference time or administrative department meetings for EM residents, fellows, attending physicians, and APPs. Attendees were recruited to participate and asked to complete a text-based eligibility screening survey. While we did not record how many total individuals were in attendance at the meetings, 227 individuals texted in response to our prompting to determine if they were eligible to participate. The survey was administered through Mosio (Seattle, WA), a software company specializing in mobile solutions for research. The training consisted of: 1. A baseline survey; 2. An evidence-based didactic and ED case-based review of buprenorphine, and 3. A post-intervention survey. The content was summarized from the DATA 2000 waiver course and included an overview of the opioid crisis, pharmacology of buprenorphine and naloxone, clinical signs and symptoms of opioid withdrawal, information on the “72-h rule,” introduction to the concept of a “warm hand-off”, and an interactive clinical case discussion. All the eligible participants completed a baseline text-based survey prior to the didactic training. Survey questions solicited demographic information (level of training, institution, and X-waivered status) and answers to six multiple-choice queries to assess knowledge of buprenorphine, and three Likert-scale assessments on views towards ED-initiated buprenorphine (Appendix A). All the participants underwent the 30-min didactic training described above and then were prompted to complete a second survey reassessing their knowledge and attitudes immediately after the didactic. The participants were given the second survey approximately 45 min after the first survey. The knowledge assessment questions in both surveys were the same but the order in which they were asked varied. Upon completion of the 2 surveys and didactic training, the subjects were compensated with an electronic code to a $15 gift certificate. For consistency, one author led all of the trainings [22]. Following the training, the participants were randomized into one of two arms using a random number generator: a standard arm and an enhanced arm. The participants were unaware of their randomization and of the other study arm, and had no additional contact with other participants through our study. The subjects in the standard arm received no further communication until 90 days, when they received the final survey. Those in the enhanced arm were notified via text message that they would receive an additional financial incentive ($20), if they texted to report administering their first dose of buprenorphine in the ED. All the participants had been instructed during the didactic that patients must be in at least mild opioid withdrawal (with a COWS score of at least 8) to be clinically eligible for buprenorphine. Additionally, the participants in the enhanced arm were enrolled to receive weekly text messages that included clinical pearls, as well as social norming, stigma-mitigating, and salience highlighting content about opioid use disorder and buprenorphine [23]. At the end of 90 days, both groups received SMS-based surveys on the same initial knowledge and attitude questions as well as additional questions on whether or not they administered buprenorphine during the study period, what influenced their decisions to administer or not administer buprenorphine, their overall rating of the training, and feedback comments. All the participants received an electronic code for a $25 gift certificate for completing this final 90-day survey. The purchase of gift certificates was funded through a grant from the Medical Toxicology Foundation of the American College of Medical Toxicology. Evaluation A thorough search of the relevant literature did not reveal any pre-existing instruments that specifically addressed our study question. Therefore, we based our items on expert input as well as previously published substance use attitude questions [24].Knowledge assessments were graded based on percent questions out of six that were answered correctly. Attitude-related responses were based on a five-point Likert scale (1 = strongly disagree, 3 = neutral, 5 = strongly agree). Statistical Analysis Summary descriptive statistics include frequency and percentage for categorical variables and means ± standard deviation for scores. To assess changes in knowledge and attitudes over the 90-day period between the study arms, analysis of variance in repeated measures was used. To minimize type I error, post hoc pairwise comparisons using Tukey–Kramer tests were performed to examine differences at time points and study arm. Outcomes are reported as mean difference with 95% confidence intervals. A sample size calculation was not performed, as this was a pilot study. To assess reported buprenorphine administration, Fisher’s exact test was used. A p-value of < 0.05 was considered statistically significant. The data were approximately normally distributed. All the statistical analyses were performed using SAS statistical software (Version 9.4, SAS Institute, Cary, NC). Figures were created in Harvard Graphics 3.0. Results A total of 227 EM clinicians were screened for eligibility after they texted in to our platform, and 41% were excluded for having previously administered buprenorphine in the ED, and 8.9% were excluded for not completing the initial survey (Fig. 1). In total, we enrolled 104 participants from 11 hospitals: 51 were randomized to the standard arm and 53 to the enhanced arm. All the hospitals were located in the Greater Philadelphia area and included academic and community emergency departments. The majority of the institutions represented were academic EDs (8/11). Of enrollees, 48.1% (n = 50) were resident physicians, 39.4% (n = 41) were attending physicians, and 12.5% (n = 13) were advanced practice providers. A minority of participants (7.7%) were previously X-waivered but had never ordered buprenorphine in the ED.Fig. 1 Consort diagram. Prior to the didactic intervention, baseline knowledge scores were similar in the standard arm (67.7% ± 19.0%) and the enhanced arm (68.4% ± 17.1%). At the end of 90 days, 59% (n = 61) of enrolled participants had completed all three assessments and 72.3% (n = 73) had completed the final survey. Clinical knowledge about buprenorphine improved for all clinicians immediately after the didactic intervention 68.4% vs 87.9%, (difference = 19.%, 95% CI 14.4% to 24.5%), but this was not sustained from baseline at 90 days in the enhanced arm 90-day mean 76.9% (difference 9.6%, 95% CI − 0.37% to 19.59.2%,) or in the standard arm 90 day-mean 73.2%, (difference 3.7%, 95% CI − 5.8% to 13.2%) (Fig. 2).Fig. 2 Changes in knowledge scores on buprenorphine. In evaluating perceptions, the participants in both groups expressed an increased ability to recognize patients who are physiologically in withdrawal and would meet the criteria for a dose of buprenorphine (Question 2). This increase was seen immediately after the intervention and sustained at 90 days (enhanced arm, baseline to 90-day difference: 0.55, 95% CI: 0.01–1.09, standard arm baseline to 90-day difference: 0.85 95% CI: 0.34–1.37; Fig. 3). While the participants also demonstrated an increase in favorable attitudes toward medications for patients with OUD (Question 1) and ED-initiated buprenorphine (Question 3), these changes were not statistically significant in either arm (Fig. 3). Additional analyses were performed looking at differences between residents and attending physicians. Advanced practice providers (APPS) were excluded from this analysis due to incomplete data. There was no difference between attending and residents with regard to changes in attitudes from baseline to 90-day surveys (Question 1, difference in change between residents and attendings 0.56, 95% CI − 0.34 to 1.47, Question 2, difference in change between residents and attendings 0.46 95% CI − 0.019 to 1.12, Question 3 difference in change between residents and attendings 0.07 95% CI − 0.60 to 0.73).Fig. 3 Changes in attitudes toward ED-initiated buprenorphine. In assessing buprenorphine administration, one-third (33%) of all the participants reported administering buprenorphine for the first time during the study period. More participants in the enhanced arm reported administering buprenorphine compared to the standard arm, but this difference was not statistically significant (40% vs. 26.3%, p = 0.319). Among those who reported not administering buprenorphine during the study period, the most common reason provided was not encountering an appropriate patient (75.5%). No participant reported lack of preparation as the reason for not administering buprenorphine. When the participants were asked to rate how likely they were to recommend the initial didactic intervention to a colleague from 0 to 10 (10 being most likely) the mean score overall was 8.3 ± 1.4 and were not statistically different between the enhanced arm and standard arm (8.5 ± 1.5 vs. 8.2 ± 1.8, p = 0.42). Discussion Our study demonstrated that a brief educational intervention and text-based survey enrollment and follow-up was feasible and associated with 33% of participants administering buprenorphine for the first time in the study period following the intervention. While there is no “baseline” rate of readiness for buprenorphine administration among EM clinicians overall, our rate of one-third of EM clinicians administering buprenorphine suggests an increase from the 20.9% ED clinicians who recently reported high readiness to initiate buprenorphine in another recent study [17]. Additionally, all the participants reported an increased ability to recognize opioid withdrawal, and this was sustained at 90 days. Such an improvement in self-reported clinical comfort is critical to combating the practitioner-level barriers to expanding ED initiation of buprenorphine. We also showed that utilizing text-delivered surveys to serially assess knowledge improvements and perceptions about buprenorphine were possible, though participant retention for serial text-based surveys was a challenge. Additional methods such as a financial incentive to report first-time buprenorphine administration and SMS-delivered clinical pearls and social norming demonstrated no significant additional benefit to the educational intervention alone for rates of first-time administration, though our study was not powered to detect a difference for this outcome. This educational intervention was modest, requiring only 30 min of conference time. Content covered could be delivered by any clinician with experience administering buprenorphine and could easily be converted to an online format (see Appendix). Of note, a significant portion of our study participants reported that they did not encounter an eligible patient to whom they could administer buprenorphine. This was an unexpected yet important result. Previous studies have demonstrated the efficacy of behavioral economics-based interventions on shaping clinician behavior, including for opioid and antibiotic stewardship [25, 26]. While institutions develop interventions to nudge clinician behavior to improve access to buprenorphine for ED patients, a robust understanding of how often patients with OUD present to the ED with opioid withdrawal and/or seeking treatment initiation is critical to developing realistic, feasible, and clinically meaningful outcome targets [14, 23]. Training providers to recognize patients who might have opioid use disorder, even though that might not be the primary reason for their visit, is also important as previous work has demonstrated many patients who receive buprenorphine in the ED do not present with a chief complaint of opioid withdrawal [27]. Furthermore, given the predominance of fentanyl in the opioid supply since the conception of our study, updating buprenorphine induction protocols to minimize the risk of precipitated withdrawal is critical. ED treatment of opioid use disorder is an opportunity reflective of the expertise of EM clinicians in recognizing time-sensitive conditions and initiating evidence-based treatment with minimal additional training. On April 28, 2021, the United States Department of Health and Human Services put into effect guidelines that allow some practitioners to forgo the 8-h training course when applying for an X-waiver. While the removal of this regulatory burden is a welcome policy change, experts in the field of addiction medicine have proposed the need for focused and specialty-specific training on OUD and buprenorphine [28, 29]. The comfort of providers in initiating buprenorphine varies according to their years in practice and practice setting, with practitioners with more years in practice and those in non-academic settings feeling less comfortable [18]. Our intervention represents one model by which to address this need for focused and brief training among ED clinicians to increase their familiarity with buprenorphine administration independent of the necessity for waiver training. As demonstrated by a recent analysis of calls to the California Poison Control System’s OUD hotline, once practitioners incorporate buprenorphine prescribing to their clinical practice, additional support may be needed for complicated buprenorphine starts, such as among special populations or patients with polysubstance use [30]. As demonstrated by our results, modest incentives and motivational reminders alone are unlikely to result in sustained practice pattern change regarding administration of buprenorphine. This suggests that there is a need to address larger institutional and health-system level barriers preventing adoption of this practice, which may include local ED culture around OUD treatment, availability of outpatient follow-up, and social and health system navigation support from social workers or peer recovery specialists. Furthermore, our results showing the only sustained impact of our intervention was the ability to recognize opioid withdrawal suggests that concerted efforts to improve education on recognizing and treating OUD are necessary, and should be routinely incorporated into medical school and residency training. Our study has notable limitations. First, we did not include a control group without any training to which we could compare the results of our intervention. Additionally, we relied on self-reported data on buprenorphine administration rather than actual observed changes in administration rates, risking social desirability bias. We did not ascertain when in the 90-day period, buprenorphine was administered or additional demographic or training details of the providers. Because this was a provider-focused intervention, we did not collect patient-level demographics. Given the pre-post nature of our study design, we are unable to attribute clinician behavior changes to our intervention alone, as we cannot account for secular trends. We did not study a comprehensive opioid use disorder curriculum but focused on patients presenting with opioid withdrawal and buprenorphine treatment. The 30-minute curriculum was brief and conducted by a single presenter, which may limit the reproducibility. Additional studies are needed to improve education on prevention, stigma and outpatient management of opioid use disorder, and assessments of necessary knowledge of buprenorphine should be tested and validated. While we included content on “the 72-hour rule” in our didactic intervention, we did not assess if our participants were already aware of this important policy that facilitates ED-administration of buprenorphine without an X-waiver. Additionally, despite a financial incentive, the completion rate of all three surveys was only 59%. While the literature is scarce on text-based surveys, previous studies have demonstrated that physician survey response rates are lower than the general population, can be improved with financial incentives, and that response rates decrease with additional surveys [31–34]. Because the enhanced group was offered an incentive to report buprenorphine administration, it is possible that financial inducement affected administration in this arm. However, when the participants were asked what influenced their decision to administer buprenorphine, no participant reported the financial incentive in the top three factors affecting their decision. In addition, we did not ascertain from the participants if they encountered OUD patients who were not yet in opioid withdrawal in the ED, and thus clinically ineligible for buprenorphine administration, which is a common clinical scenario. The patients who use fentanyl, in particular, are at risk for precipitated withdrawal, especially when they present with mild withdrawal symptoms [35]. At the time of our didactic, we used a COWs score of “8” to signal mild withdrawal in which buprenorphine could be used, which was consistent with other national protocols [36]. However, recent clinical experience during the rise of fentanyl suggests that a higher COWS score (“13”) is safer to prevent precipitated withdrawal. We did not include home inductions in our study, since that requires having a DATA 2000 waiver to prescribe buprenorphine, although it is a valuable practice to reach patients who are not in withdrawal while in the ED. Although the platform we used to send the messages provided confirmation that the carrier received the messages, we are unable to know if the recipient read the messages. Lastly, this pilot study was not powered to find differences as small as those seen in administration rates. Conclusion This intervention targeted to EM clinicians at all levels of clinical practice improved clinical knowledge of and attitudes towards ED administration of buprenorphine only transiently. It did, however, result in a sustained perception of ability to recognize opioid withdrawal. Our results do not suggest that the utilization of behavioral economics-grounded principles of social norming and gain-framed incentivization confers additional benefit to the educational intervention alone in promoting sustained change in clinical knowledge. In order to further expand evidence-based ED treatment of OUD, focused initiatives that improve clinician competence with buprenorphine should be explored. Additional studies to understand effective strategies beyond individual provider-level barriers that will promote ED administration of buprenorphine are warranted. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 79 kb) Supplementary file2 (DOCX 91 kb) Author Contribution UK, KL, and JP conceived the study, designed the trial, supervised the conduct of the trial, and data collection. JP obtained the research funding. JP designed and conducted the training sessions with support from KL and JD. KL managed the technical support, and KL and MP provided the innovation and design expertise. UK, JP, JD, and KL undertook the recruitment and planning of the sessions. FS provided the statistical advice on study design and analyzed the data; UK, KL, and TL managed the data acquisition. UK drafted the manuscript, and all the authors contributed substantially to its revision. UK takes responsibility for the paper as a whole. Trial Registration: clinicaltrials.gov identifier: NCT03821103. Sources of Funding The Medical Toxicology Foundation (MTF) of the American College of Medical Toxicology (ACMT) received funding from Independence Blue Cross Foundation under the Supporting Treatment and Overdose Prevention (STOP) program to conduct this study. Dr. Utsha Khatri is funded through the Department of Veterans Affairs through the National Clinician Scholars Program. These views do not represent the views of the Department of Veterans Affairs or of the US government. Declarations Conflicts of Interest None. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Ochalek TA, Cumpston KL, Wills BK, Gal TS, Moeller FG. Nonfatal Opioid Overdoses at an Urban Emergency Department During the COVID-19 Pandemic. JAMA. 2020;324(16):1673–1674. 10.1001/jama.2020.17477 2. US Centers for Disease Control and Prevention. Increase in fatal drug overdoses across the United States driven by synthetic opioids before and during the COVID-19 pandemic. https://emergency.cdc.gov/han/2020/han00438.asp. Accessed 10 January 2021. 3. Soares WE 3rd, Melnick ER, Nath B, D'Onofrio G, Paek H, Skains RM, Walter LA, Casey MF, Napoli A, Hoppe JA, Jeffery MM. Emergency Department Visits for Nonfatal Opioid Overdose During the COVID-19 Pandemic Across Six US Health Care Systems. Ann Emerg Med. 2022;79(2):158–167. 4. Weiner SG Baker O Bernson D Schuur JD One-year mortality of patients after emergency department treatment for nonfatal opioid overdose Ann Emerg Med 2020 75 1 13 17 10.1016/j.annemergmed.2019.04.020 31229387 5. Kakko J, Alho H, Baldacchino A, Molina R, Nava FA, Shaya G. Craving in Opioid Use Disorder: From Neurobiology to Clinical Practice. Front Psychiatry. 2019;10:592. 6. Mahmoud S Anderson E Vosooghi A Herring AA Treatment of opioid and alcohol withdrawal in a cohort of emergency department patients Am J Emerg Med 2021 43 17 20 10.1016/j.ajem.2020.12.074 33476917 7. Monico LB Oros M Smith S Mitchell SG Gryczynski J Schwartz R One million screened: scaling up SBIRT and buprenorphine treatment in hospital emergency departments across Maryland Am J Emerg Med 2020 38 7 1466 1469 10.1016/j.ajem.2020.03.005 32171581 8. Sordo L Barrio G Bravo MJ Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies BMJ. 2017 357 j1550 10.1136/bmj.j1550 28446428 9. Mattick RP, Breen C, Kimber J, Davoli M. Buprenorphine maintenance versus placebo or methadone maintenance for opioid dependence. Cochrane Database Syst Rev. 2014 Feb 6;(2):CD002207. 10.1002/14651858.CD002207.pub4. 10. Huhn AS Dunn KE Why aren’t physicians prescribing more buprenorphine? J Subst Abuse Treat 2017 78 1 7 10.1016/j.jsat.2017.04.005 28554597 11. Louie DL Assefa MT McGovern MP Attitudes of primary care physicians toward prescribing buprenorphine: a narrative review BMC Fam Pract 2019 20 1 157 10.1186/s12875-019-1047-z 31729957 12. Hutchinson E Catlin M Andrilla CH Baldwin LM Rosenblatt RA Barriers to primary care physicians prescribing buprenorphine Ann Fam Med 2014 12 2 128 133 10.1370/afm.1595 24615308 13. Im DD Chary A Condella AL Emergency department clinicians' attitudes toward opioid use disorder and emergency department-initiated buprenorphine treatment: a mixed-methods study West J Emerg Med 2020 21 2 261 271 10.5811/westjem.2019.11.44382 32191184 14. Foster SD Lee K Edwards C Providing incentive for emergency physician X-waiver training: an evaluation of program success and postintervention buprenorphine prescribing Ann Emerg Med 2020 76 2 206 214 10.1016/j.annemergmed.2020.02.020 32376089 15. D'Onofrio G O'Connor PG Pantalon MV Emergency department-initiated buprenorphine/naloxone treatment for opioid dependence: a randomized clinical trial JAMA 2015 313 16 1636 1644 10.1001/jama.2015.3474 25919527 16. Lowenstein M Kilaru A Perrone J Barriers and facilitators for emergency department initiation of buprenorphine: a physician survey Am J Emerg Med 2019 37 9 1787 1790 10.1016/j.ajem.2019.02.025 30803850 17. Hawk KF D’Onofrio G Chawarski MC Barriers and facilitators to clinician readiness to provide emergency department–initiated buprenorphine JAMA Netw Open 2020 3 5 e204561 e204561 10.1001/jamanetworkopen.2020.4561 32391893 18. Zuckerman M, Kelly T, Heard K, Zosel A, Marlin M, Hoppe J. Physician attitudes on buprenorphine induction in the emergency department: results from a multistate survey. Clin Toxicol. 2020;1–7. 19. Rhee TG D’Onofrio G Fiellin DA Trends in the use of buprenorphine in US emergency departments, 2002–2017 JAMA Netw Open 2020 3 10 e2021209 e2021209 10.1001/jamanetworkopen.2020.21209 33079195 20. Title 21, Code of federal regulations, part 1306.07(b). 21. Wiegand TJ The new kid on the block—incorporating buprenorphine into a medical toxicology practice J Med Toxicol 2016 12 1 64 70 10.1007/s13181-015-0518-4 26574020 22. Levels of racism a theoretic framework and a gardener's tale Am J Public Health 2000 90 8 1212 1215 10.2105/AJPH.90.8.1212 10936998 23. Martin A Kunzler N Nakagawa J Get waivered: a resident-driven campaign to address the opioid overdose crisis Ann Emerg Med 2019 74 5 691 696 10.1016/j.annemergmed.2019.04.035 31272821 24. Saitz R Friedmann PD Sullivan LM Professional satisfaction experienced when caring for substance-abusing patients: faculty and resident physician perspectives J Gen Intern Med 2002 17 5 373 376 12047735 25. Gong CL Zangwill KM Hay JW Meeker D Doctor JN Behavioral economics interventions to improve outpatient antibiotic prescribing for acute respiratory infections: a cost-effectiveness analysis J Gen Intern Med 2019 34 6 846 854 10.1007/s11606-018-4467-x 29740788 26. Delgado MK Shofer FS Patel MS Association between electronic medical record implementation of default opioid prescription quantities and prescribing behavior in two emergency departments J Gen Intern Med 2018 33 4 409 411 10.1007/s11606-017-4286-5 29340937 27. LeSaint KT Klapthor B Wang RC Geier C Buprenorphine for opioid use disorder in the emergency department: a retrospective chart review West J Emerg Med 2020 21 5 1175 1181 32970572 28. Weimer MB Tetrault JM Fiellin DA Patients with opioid use disorder deserve trained providers Ann Intern Med 2019 171 12 931 932 10.7326/M19-2303 31766053 29. Diamond D. Biden kills Trump plan on opioid-treatment prescriptions. The Washington Post. Jan. 27, 2021. https://www.washingtonpost.com. Accessed 1 February 2021. 30. LeSaint KT Ho RY Heard SE Smollin CG California poison control system implementation of a novel hotline to treat patients with opioid use disorder J Med Toxicol 2021 17 2 190 196 10.1007/s13181-020-00816-1 33078365 31. Cook DA Wittich CM Daniels WL West CP Harris AM Beebe TJ Incentive and reminder strategies to improve response rate for internet-based physician surveys: a randomized experiment J Med Internet Res. 2016 18 9 e244 10.2196/jmir.6318 27637296 32. Flanigan T, McFarlane E, Cook S. Conducting Survey Research among Physicians and Other Medical Professionals—A Review of Current Literature. In: Proceedings of the Survey Research Methods Section. American Statistical Association; 2008. p. 4136–47. 33. Cunningham CT Quan H Hemmelgarn B Exploring physician specialist response rates to web-based surveys BMC Med Res Methodol 2015 15 1 32 10.1186/s12874-015-0016-z 25888346 34 Brtnikova M Crane LA Allison MA Hurley LP Beaty BL Kempe A A method for achieving high response rates in national surveys of U.S. primary care physicians PLOS ONE. 2018 13 8 e0202755 10.1371/journal.pone.0202755 30138406 35. Varshneya NB, Thakrar AP, Hobelmann JG, Dunn KE, Huhn AS. Evidence of Buprenorphine-precipitated Withdrawal in Persons Who Use Fentanyl. J Addict Med. 2021. 10.1097/ADM.0000000000000922. 36. Guo CZ D'Onofrio G Fiellin DA Emergency department-initiated buprenorphine protocols: a national evaluation J Am Coll Emerg Physicians Open. 2021 2 6 e12606 34877567
PMC009xxxxxx/PMC9004455.txt	==== Front Indian J Otolaryngol Head Neck Surg Indian J Otolaryngol Head Neck Surg Indian Journal of Otolaryngology and Head & Neck Surgery 2231-3796 0973-7707 Springer India New Delhi 35433404 2824 10.1007/s12070-021-02824-1 Original Article ENT On-call Equipment Bag: Updated for COVID-19 http://orcid.org/0000-0002-0159-9321 Sangha Miljyot Singh miljyot_singh@msn.com Kumta Radhika radhika.kumta@nhs.net Jumani Kiran kiranjumani@nhs.net grid.439471.c 0000 0000 9151 4584 ENT Department, Whipps Cross University Hospital, Whipps Cross road, London, E11 1NR UK 12 4 2022 10 2022 74 Suppl 2 32683272 13 7 2021 14 8 2021 © Association of Otolaryngologists of India 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Otorhinolaryngology, or Ear Nose and throat (ENT), is a specialty requiring specific equipment for the management and treatment of patients. While most hospitals provide a 24/7 ENT procedure room, there is also a need for a mobile equipment bag with objects and instruments not readily available outside of the ENT department. We look to introduce a novel ENT equipment bag and checklist which has been updated to include a list of required equipment in the setting of the current COVID-19 pandemic. For the ENT equipment bag we use a high visibility multi-compartment bag. In addition to this, the bag contains four folders, labelled “Ear”, “Nose”, “Throat” and “Drugs”. The necessary ENT equipment is divided and included within these folders. The bag also contains a copy of the “ENT bag checklist”. This is a list, designed by the Whipps Cross ENT team, which specifies the required ENT equipment, including PPE in light of the current COVID-19 pandemic. The bag is stocked once a week using this list. Using this new system we developed a portable, practical and easy to use ENT on call equipment bag that included all of the relevant PPE to manage COVID-19 patients. We also implemented a new method of bag stocking to ensure that the bag is adequately stocked. In conclusion, we present a COVID-19 updated ENT bag and checklist. We hope this will help act as a reference for other ENT teams to compare and implement during the current COVID-19 pandemic. Keywords COVID-19 SARS-CoV-2 ENT Otorhinolaryngology Equipment issue-copyright-statement© Association of Otolaryngologists of India 2022 ==== Body pmcIntroduction Otorhinolaryngology, or the Ear Nose and throat discipline (ENT), is a niche specialty requiring specific equipment for management and treatment of patients. There has always been the need for a treatment or procedure room with provision for a microscope, a flexible nasal endoscope, microscopic instruments and auxiliary equipment. While most hospitals provide a 24/7 ENT procedure room, there is also a need for a mobile equipment bag with objects and instruments not readily available outside of the ENT department. It is important to ensure that junior doctors, who are the first responders to any ENT patient, feel safe, confident and well equipped when called to the wards, A&E or other hospital sites. In these unfamiliar environments ENT equipment is often unavailable. A national survey into night emergency cover in ENT for England showed that 42% of respondents did not feel comfortable managing common ENT emergencies as the first doctor on call [1], with another study showing that 25% of ENT treatment rooms in England were not adequately stocked [2]. Having an adequately stocked bag could help mitigate both of these issues. Building upon previous work [3], we look to introduce a novel ENT equipment bag and checklist which has been updated to include a list of required equipment in the setting of the current COVID-19 pandemic. Material and Methods For this equipment bag we use a high visibility multi-compartment bag, its distinct appearance helps in preventing it from being lost or misplaced easily. In addition to this, the bag has been further divided into 4 compartments (Fig. 1).Fig. 1 ENT equipment bag This includes the “main compartment” and compartments “1”, “2” and “3”. The main compartment contains a further four folders containing equipment (Ear, Nose, Nasal packs and Drugs folders) and a headlight (Fig. 2).Fig. 2 Main four folders and head light The entire equipment included in the bag is shown in Fig. 3 and includes: (a) headlight, (b), otoscope, (c) aural speculum, (d) cophenylcaine spray, (e) xylocaine spray and nozzle, (f) naseptin cream, (g) merocel nasal tampon, (g) rapid rhino nasal pack, (i) nasopore, (j) nasal clip, (k) nasal thudicum, (l) silver nitrate cautery, (m) syringe, (n) gauze, (o) white needle, (p) scissors, (q) tongue depressor, (r) alcohol wipe, (s) optilube, (t) FFP3 mask, (u) Tilly’s forcep, (v) St Barts wax hook, (w) crocodile forcep, (x) Jobson Horne probe, (y) scalpel, (z) goggles. In addition to these instruments an umbilical clamp, foleys catheter/Brighton's balloon and a 512 Hz tuning fork could be considered though not included here.Fig. 3 ENT equipment In addition to this the bag also includes several copies of the newly developed ENT bag checklist (Fig. 4). A stock check using the checklist is completed by the on call SHO every Monday before the morning handover. The outstanding equipment is then stocked by the day team. This ensures that there is adequate stocking of the bag.Fig. 4 ENT bag checklist Discussion We aimed to design a checklist and ENT bag that was equipped to deal with the main ENT procedures (ear examination, quinsy drainage, epistaxis control etc.). There were several different considerations when designing this bag and checklist. We feel that the portability and ease of carrying the bag is important, especially when moving quickly from place to place. We also tried to ensure that the bag was not overly cluttered with more equipment than necessary as this would make it difficult to use at speed and may lead to stocking the bag becoming laborious and time consuming. We have designed for the ENT bag to be restocked once a week. We have found this to be sufficient as the equipment used mainly out of hours (nasal packs and tongue depressors) can be included in large quantities as they do not take up much room or weight. Finally, in light of the current COVID-19 pandemic and uncertainty surrounding the ongoing precautions that may be needed over the next months to years, it is important to prioritise safety of ENT doctors by providing adequate PPE whilst working. This is especially important as many ENT procedures completed day-to-day are aerosol generating procedures [4]. This checklist would help ensure the safety of both the doctors and patients during a busy ENT on call. In conclusion we present the current COVID-19 updated ENT bag and checklist for use by the on call ENT team at Whipps Cross. This will aid other ENT teams by allowing them to compare and implement these changes in to their team during the current COVID-19 pandemic. Authors’ Contributions MS: Substantial contribution to the design of the manuscript, literature search, preparing the main paper and final approval. RK: Contribution to design of manuscript. Both authors read and approved the final manuscript. Both authors have approved the manuscript before submission, including the names and order of authors. Declaration Conflict of interest The authors declares that they have no conflict of interest. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Biswas D, Rafferty A, Jassar P (2009) Night emergency cover for ENT in England: a national survey. J Laryngol Otol 123(8): 899–902. Available from: http://ovidsp.ovid.com/ovidweb.cgi?T=JS&PAGE=reference&D=emed9&NEWS=N&AN=2009606148 2. Moorthy R Magarey M Joshi A Jayaraj SM Clarke PM A study of out-of-hours facilities in otolaryngology: current provision and problems J Laryngol Otol 2005 119 3 202 206 10.1258/0022215053561585 15845192 3. Duvvi S, Mudugal A, Kumar B (2012) How we do it: a portable ENT tool kit. Internet J Otorhinolaryngol 6(1) 4. Mick P, Murphy R (2020) Aerosol-generating otolaryngology procedures and the need for enhanced PPE during the COVID-19 pandemic: a literature review. J Otolaryngol Head Neck Surg 49(1). Available from: https://journalotohns.biomedcentral.com/articles/10.1186/s40463-020-00424-7%0A. http://europepmc.org/search?query=(DOI:10.1186/s40463-020-00424-7)%0A. http://search.ebscohost.com/login.aspx?direct=true&scope=site&site=ehost-live&db=mnh&AN=32393346%0A. http://gateway.com
PMC009xxxxxx/PMC9004456.txt	==== Front Rev Manag Sci Review of Managerial Science 1863-6683 1863-6691 Springer Berlin Heidelberg Berlin/Heidelberg 550 10.1007/s11846-022-00550-8 Original Paper Economic policy uncertainty and corporate donation: evidence from private firms in Korea http://orcid.org/0000-0001-5488-0603 Chun Hongmin hmchun@sungshin.ac.kr 1 http://orcid.org/0000-0001-8071-262X Harjoto Maretno maretno.harjoto@pepperdine.edu 2 http://orcid.org/0000-0002-3828-6791 Song Hakjoon hsong@csudh.edu 3 1 grid.264383.8 0000 0001 2175 669X Business Department, Sungshin Women’s University, Sungbukgu Bomunro 34 Gil 2, Sujung Campus, Seoul, 02844 Korea 2 grid.261833.d 0000 0001 0691 6376 Pepperdine Graziadio Business School, Pepperdine University, 24255 Pacific Coast Highway, Malibu, CA 90263-4100 USA 3 grid.253556.2 0000 0001 0746 4340 Accounting, Finance, Economics and Law Department, California State University Dominguez Hills, Carson, CA USA 12 4 2022 2023 17 3 909939 10 5 2021 24 3 2022 © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. This study examines the association between economic policy uncertainty (EPU) and private firms’ corporate donations. Based on resource constraints and the conservation of resources (COR) theory, we argue that private firms are constantly facing resource constraints and their resource conservation motive becomes apparent when EPU is heightened. Therefore, we expect that corporate donations are negatively related to EPU. Using audited corporate donations from 48,903 private firms in Korea during 2002–2019, we find that private firms’ donations are negatively related to EPU. We find that private firms operating in more competitive conditions increase their donations, but this positive association between market competition and donations is moderated by EPU. We find that private firms’ donations increased when the progressive party is in power, but this positive relationship is also moderated by EPU. Our results suggest that firms reduce their level of corporate giving to conserve resources as a precautionary saving motive when they face higher EPU. Our paper contributes to the strand of literature on corporate donations and EPU by providing evidence that EPU significantly affects private firms’ donations. We also find that firms’ strategic motives and political pressure to engage in corporate donations are moderated by EPU. Keywords Economic policy uncertainty Corporate donations Private firms Resource constraints Market competition Progressive party JEL Classification M14 D64 D80 G18 Mathematics Subject Classification 62H12 Estimation in multivariate analysis 62H15 Hypothesis testing in multivariate analysis Denney Professorship2019-2021 Harjoto Maretno issue-copyright-statement© Springer-Verlag GmbH Germany, part of Springer Nature 2023 ==== Body pmcIntroduction Our world is experiencing an unprecedented shock from the coronavirus (COVID-19) outbreak that has taken many lives, accompanied by astonishing economic and political uncertainty from various issues such as the debt ceiling and a contentious presidential election in the United States, the Brexit transition, the Black Lives Matter movement, a nuclear threat from North Korea, and many others. Countries around the world are constantly re-evaluating lockdowns, social distancing, quarantine, and vaccination policies that have brought unprecedented uncertainties to firms’ operations (Donthu and Gustaffson 2020; García-Carbonell et al. 2021). Worldwide, companies are currently experiencing an extraordinary period of economic policy uncertainty (EPU hereafter) (Baker et al. 2020). At the same time, society demands that companies bear greater social responsibility (Crane and Matten 2020). Our timely study investigates this issue by examining the relationship between EPU and corporate social responsibility (CSR), specifically corporate donations, or social or charitable giving. EPU is defined as the likelihood that national economic policies will change in future years, and the potential of these changes to affect economic activities at the firm level (Baker et al. 2016). Corporate donation is defined as corporate giving to social and charitable causes within the local community (Godfrey 2005; Wang and Qian 2011). Recent research examined the effects of EPU on publicly traded firms’ investment, mergers and acquisition, information asymmetry, disclosure, cash holding, earnings management, tax avoidance, innovations, and firm value (Baker et al. 2016; Borghesi and Chang 2019; Chen et al. 2019; Jin et al. 2019; Gulen and Ion 2016; Nguyen and Nguyen 2020; Wu et al. 2020; Xu 2020). However, the literature that examines the relationship between EPU and CSR is nascent and focuses only on publicly listed firms (Dai et al. 2020; Zhang et al. 2020). Particularly, literature that focuses on private firms’ resources, charitable giving, and response to EPU, which could be quite different from that of publicly traded firms, is absent. Zhang et al. (2020) show that Chinese public firms increase their CSR engagement during high EPU periods, and Dai et al. (2020) document that U.S. publicly held firms increase their CSR activities during high EPU. In publicly traded firms, CSR could be driven by managers’ self-interests (agency problems) (Brown et al. 2006; Cespa and Cestone 2007). We focus on private firms’ corporate donations where the motive for giving is less likely to be driven by agency problems. We pose and investigate three research questions. First, how does EPU affect private firms’ donations? Second, how does EPU moderate corporate donations for firms that operate in more competitive markets? Third, how does EPU moderate the association between corporate donations and a country’s ruling political party? We use private firms in Korea as our sample for two reasons. First, in Korea, there were 29,343 public and privately held firms in 2019 that disclosed audited financial statements that include corporate charitable giving; 2058 of these firms were public (780 KOSPI and 1278 KOSDAQ listed firms1). While evidence for corporate giving by private firms worldwide is sparse due to the absence of data, Korean private firms are required to have their financial statements containing corporate donations audited and disclosed to the public by the Financial Supervisory Service (equivalent to the U.S. Securities and Exchange Commission) (Kim et al. 2011). This rich and reliable (audited) data on corporate charitable giving by private firms in Korea provide is a perfect settin to examine our research questions. Second, previous studies on corporate giving tend to pay more attention to large public firms (e.g., Lev et al. 2010; Oh et al. 2018). This approach excludes small private firms that may engage in non-monetary charitable giving to the community. Small private firms’ donations also receive less press coverage and public attention and therefore may involve a different decision-making process than that of public firms. Our focus on private firms provides new insights on the effects of EPU on private firms’ donations. Based on the resource constraints literature (Baker and Nelson 2005; George 2005; Holtz-Eakin et al. 1994a, b; Mosakowski 2002) and the conservation of resources (COR) theory (Hobfoll 1989; Hobfoll et al. 2018), we argue that private firms have significantly less resources than public firms and are more averse to losses. Therefore, they have stronger legitimacy motives to conserve their financial resources for core activities (e.g., production processes) and to minimize corporate donations. The resources conservation motive becomes even stronger for private firms when EPU is heightened because EPU is generally accompanied by a steep economic decline due to cutbacks on consumer and government spending and an upward pressure on the firm’s cost of capital (Baker et al. 2016). Therefore, we expect that private firms’ donations are negatively related to EPU. Driven by the value enhancing motive of donation, we also argue that private firms which operate in a more competitive environment tend to make more donations in order to stay competitive. However, we expect that EPU moderates the positive relationship between competitive environment and corporate donations. Consistent with the literature (Campbell 2007; Gao 2011), we also expect that private firms’ donations are influenced by political pressure. Thus, we expect corporate donations to be greater when the progressive (liberal) party is in power, which is expected to promote greater CSR and corporate donations. However, since private firms have stronger motives to conserve their resources when EPU is heightened, we also argue that EPU moderates the positive relationship between progressive politics and corporate donations. Our study contributes to the literature in several ways. First, we examine the consequences of EPU on corporate charitable giving decisions based on the resource constraints literature. Further, we apply COR theory (Hobfoll 1989) from the individual (micro) level into an organizational (meso) level to explain the resources conservation strategy when private firms face EPU. While COR theory explains individuals’ coping strategy for stresses and potential losses, our study explains how privately held firms’ corporate donations are curtailed to cope with stresses and potential losses due to EPU. Second, we expand the extensive corporate giving literature and the emerging literature on corporate giving behavior by private firms where the motive for corporate donation is less likely to be driven by agency problems. Third, while we find evidence to support the competitive advantages and value enhancing motives of corporate donations for private firms, we also provide evidence that the resource constraints for privately held firms become apparent when EPU is heightened. Fourth, studies on corporate donation mostly investigate large publicly traded firms and business groups (Lev et al. 2010; Jeong and Kim 2019; Kim et al. 2019; Oh et al. 2018). Lev et al. (2010) acknowledge that a focus on public firms’ donations misses the full picture with regard to factors that affect corporate donation. Using reliable (audited) data for corporate donations by privately held firms in Korea, our study examines the impact of EPU on private firms’ corporate donations. Finally, our study also offers policy implications as private firms curtail their corporate donations when EPU is heightened. Literature review and hypotheses development EPU indicates the likelihood that policies will change in future years, and how these changes could affect economic activities. Baker et al. (2016) demonstrate that EPU significantly affects both the macro and firm-level outcomes. Prior studies find that uncertainty in economics and politics adversely affects consumer spending, government purchases, and stock markets. Recent studies show that EPU affects firms’ investment and financing decisions. Chen et al. (2019) find a negative relationship between EPU and firms’ investment. Xu (2020) finds that high EPU has a detrimental effect on corporate innovation by driving up firms’ cost of capital. Recent studies also demonstrate that banks tend to charge higher interest rates and reduce their loans to corporations when EPU increases (Barraza and Civelli 2020). Li (2019) documents a positive relation between EPU and a firm’s cash holding because of precautionary saving motive. Nguyen and Nguyen (2020) show that firms engage in aggressive tax avoidance activities when EPU is heightened as a precautionary saving motive. Moreover, there is extensive literature that examines the rationale and motivations for firms to engage in corporate donation (Brown et al. 2006; Wang and Qian 2011; Wang et al. 2015). Brown et al. (2006) argue that there are two main motivations for corporate donations: managerial agency problem and value enhancement objective. They argue that managers utilize corporate donations to enhance their own personal reputations and self-interests. Therefore, corporate donations represent the agency costs. They also argue that corporate donations can also bring value enhancement to shareholders. Private firms have a highly concentrated ownership structure in which the owners who hold a large proportion of equity are also the managers who make decisions. Thus, private firms are less likely to suffer from the agency problems between managers and shareholders, and they face less pressure and scrutiny from the capital market, regulatory agencies, and public opinion (Burgstahler et al. 2006; George 2005). Consequently, private firms’ donations are less likely to be driven by the agency motive. George (2005) indicates that privately held firms differ in their decision-making processes from their publicly held counterparts since they tend to be undercapitalized (Holtz-Eakin et al. 1994a, b). He argues that due to fewer resources, private firms are more likely to leverage their limited resources to achieve greater efficiencies. Baker and Nelson (2005) argue that since private enterprises have less resources, they are more responsive to changes, quickly re-allocating their limited resources in order to survive (Mosakowski 2002). Existing studies have found that EPU adversely affects firms’ resources (Baker et al. 2016). Firms engage in a precautionary saving motive, especially firms with difficulty in raising external financing (Li 2019; Nguyen and Nguyen 2020). Since private firms experience greater barriers to raise external financing, especially during heightened EPU, we argue that increases in EPU are more likely to bring a significant adverse impact on private firms’ resources. Therefore, private firms, which constantly face resource constraints, are more likely to face even greater resource constraints when EPU is heightened. Hence, they need to limit their resources to activities that generate greater and more certain payoffs. Typically, private firms are small, tend to rely heavily on bank loans, and have scarce financial slack. When EPU increases, financial institutions, especially banks, tend to charge higher interest rates and lend significantly less money to businesses (Barraza and Civelli 2020). Hence, private firms experience greater capital constraints during EPU. Moreover, the benefits of corporate giving and charitable donations are uncertain and difficult to evaluate, especially when EPU is heightened. COR theory is a stress model developed by Hobfoll (1989) to explain the defensive mode (resource conservation) adopted by individuals when facing losses or threat of loss. COR theory posits that people have a built-in bias to “overweight resource loss” and “underweight resource gain” as a theory to understand stress in organizations, with resource loss being more powerful than resource gain in terms of magnitude and speed over time (Hobfoll 1989; Hobfoll et al. 2018). Since Hobfoll’s (1989) seminal study, COR theory has been recently applied to explain conservation strategies at the organizational level (Hobfoll et al. 2018; Clercq and Belausteguigoitia 2019). The main tenet of COR is that firms' resource loss is expected to be disproportionately more salient than resource gain. Similar to individuals, organizations as a whole adopt conservation of resources behavior when they are facing losses or potential losses. We use the micro (individual) level of COR and apply it at the organizational (meso) level to explain corporate donation behavior of privately held firms when they are facing higher EPU. In the context of our study, COR theory demonstrates that a firm’s resource loss is likely to be greater than the benefits from corporate donations, especially when EPU increases. Hobfoll et al. (2018) state this argument as "organizations who lack resources are more vulnerable to resource loss and less capable of resource gain" (pg. 106). Moreover, due to private firms' limited resources, resource loss is quickly more detrimental to private firms than their publicly held counterparts. EPU affects privately held firms more rapidly and it increases the likelihood of being caught in a “loss spiral” (Hobfoll 1989), while resource gains from corporate donation tend to be small and take time to develop. Therefore, based on resource constraints literature and COR theory, we argue that private firms are financially constrained and are more likely to reduce corporate donation by redirecting their limited resources to core activities such as their production processes that carry a greater certainty to improve their financial goals when EPU is heightened. Thus, our first hypothesis is stated as follows: Hypothesis 1 Corporate donation is negatively related to EPU. Literature finds that peer pressure from competitors influences firms’ corporate donations (Cao et al. 2019). Firms use corporate donations to maintain public image, consumer loyalty, and community support (Zhang et al. 2010), and to increase employee engagement (Arco-Castro et al. 2020). Firms operate in competitive markets, with advertising and CSR commitments used to gain competitive advantages (Baron et al. 2011). Thus, corporate donation is driven by the value enhancement motive. We argue that private firms which operate in more competitive markets, measured by advertising expenditures and the industry median of corporate donation where firms operate (Baron et al. 2011), are more likely to increase their donations to gain competitive advantages (value enhancement motive). We also argue that EPU can significantly affect the positive relationship between market competition and corporate donation. When EPU is heightened, private firms are forced to make COR strategic decisions to allocate their limited resources in ways that enhance their competitive advantages and increase their likelihood to survive. We argue that increases in EPU put these private firms, which are already facing fierce competition, in a more difficult position with even greater resource constraints. Private firms that operate in more competitive markets face greater resource constraints and greater stresses (loss spiral) when EPU is heightened and therefore are more likely to revert to a resource conservation mode by immediately diverting their resources away from corporate donation toward their core activities that carry greater certainty. Hence, when EPU is heightened, we expect that private firms that operate in more competitive markets are more likely to decrease their donation levels. Thus, our second hypotheses are stated as follows: Hypothesis 2a Corporate donation is positively related to market competition. Hypothesis 2b The positive relationship between corporate donation and market competition is moderated by EPU. Institutional theory of corporate philanthropy argues that corporate charitable giving is influenced by institutionalized norms and government ideological values (Campbell 2007; Gao 2011; North 1991). Consistent with institutional theory, ideological values and beliefs of public officials and congressional representatives of the ruling party can influence firms’ corporate giving and social responsibility (Borghesi 2017). When the political ideology is tilted towards the liberal or progressive party, the collective social consciousness is more likely to encourage firms to attach greater importance to social responsibility issues such as community and environmental protections and human rights (Davis and Thompson 1994).2 For instance, the progressive party that assumed power in Korea in 2017 includes a “social value” metric to evaluate the performance of both the public and private sectors. The “social value” metric score represents thirty-five percent of the total evaluation score of public institutions; public institutions engage in a variety of “social value” enhancing activities to achieve their performance benchmarks.3 Furthermore, large corporations such as SK Corporation were urged to make a large commitment by establishing The Center for Social Value Enhancement Studies, and a social incentive system was established by the government to provide private firms with funding (approximately $40,000 dollars per year for each private firm) to create social value enhancements among small and medium enterprises (SMEs).4 Progressive politics exert significant pressure on both public and private firms to engage in higher corporate donations. Specifically, Korean private firms are more likely to increase their level of corporate donations to maintain a favorable relationship with the progressive central and local governments. Therefore, we expect that private firms’ corporate donations are higher when the progressive political party is in power. However, EPU can influence the relationship between the ruling progressive party and corporate donation. Under normal conditions, private firms expect that the progressive party is more likely to provide greater support to them than the conservative party. Drawing from the extant literature (Cyert and March 1963; George 2005), we argue that private firms become more complacent and inward-looking, exhibit greater inertia, and are less resilient to EPU when the progressive party rules because of the expectation of government support. EPU is unexpected during progressive rule; hence, when EPU occurs, private firms are more likely to experience greater adverse shocks to resources and a higher likelihood of experiencing a loss spiral. Private firms are more likely to engage in the COR strategy when EPU is heightened during progressive party rule as a reaction to an unanticipated adverse shock. Therefore, we expect that the positive relationship between corporate donation and the progressive party ruling is moderated by EPU. Our third hypotheses are stated as follows: Hypothesis 3a Corporate donation is higher when the progressive political party is in power. Hypothesis 3b EPU moderates the positive relationship between corporate donation and progressive party rule. Figure 1 illustrates the three proposed sets of hypotheses. The direct impact of EPU on corporate donation is indicated by H1 and the direct effects of market competition and political leaning (progressive political party rule) on corporate donations are indicated by H2a and H3a, respectively. The moderating effect of EPU on the relationship between market competition and corporate donation is indicated by H2b, while the moderating effect of EPU on the relationship between political leaning and corporate donation is represented by H3b.Fig. 1. The impact of EPU on corporate donation and the moderating effect of EPU Research methodology Sample selection Korean private firms are relatively small in size (Haw et al. 2014), and mostly have majority shareholders who are founders of the firm and the founders are the CEOs. Kim et al. (2011) report that Korean private firms have less shareholders than public firms and therefore are less likely to suffer from the managerial agency problem. Korean private firms also rely heavily on short-term or long-term debts obtained from regional banks, commercial banks, and other private lenders (Kim et al. 2011; Haw et al. 2014).5 Since banks generally are vulnerable to credit risks during EPU (Barraza and Civelli 2020), EPU significantly increases their cost of capital and limits private firms’ ability to obtain capital. Thus, EPU creates a capital constraint to these private firms. Therefore, Korean private firms provide a unique setting to examine the impact of EPU on private firms’ corporate donation under resource constraints. We compiled a sample of private Korean firms that never went public along with their financial and corporate donation data for 2002–2019 from the Total Solution 2000 (TS2000) and Korea Investors Services (KIS) Value databases. These databases have been used by existing studies (Oh et al. 2018; Song et al. 2020). We selected our sample companies based on the following criteria: (1) complete financial and corporate donation expense data available in the TS2000 and KIS Value databases, (2) fiscal year end of December 31,6 and (3) the firms operate in a non-financial industry. Based on this selection criteria, our final sample consists of 317,724 firm-year observations across 48,903 private firms for 2002–2019. EPU measure Our independent variable of interest, EPU, is adapted from Baker et al. (2016). Baker et al. (2016) construct the South Korean EPU index based on major South Korean newspapers in the same manner as the EPU index for the U.S. is constructed based on major American newspapers. For South Korea, Baker et al. (2016) use six major newspapers: Donga Ilbo, Kyunghyang, Maeil Economic, Hankyoreh, Hankook Ilbo, and Korea Economic Daily. Baker et al. (2016) count the number of newspaper articles containing the EPU terms uncertain or uncertainty; economic, economy, or commerce; and one or more of the following policy-relevant terms: government, Blue House, congress, authorities, legislation, tax, regulation, Bank of Korea, central bank, deficit, WTO, law/bill, or ministry of finance. Baker et al. (2016) conduct all searches in the native language of the newspapers. To construct the EPU rating for each newspaper, Baker et al. (2016) scale the raw EPU counts by the number of all articles in the same newspaper and during the same month that contain the EPU terms described above. To construct the overall South Korean EPU index, Baker et al. (2016) first standardize each newspaper’s EPU rate to a unit standard deviation. Using these standardized series, Baker et al. (2016) average the EPU rates across newspapers by month and then multiplicatively rescale the resulting series to a mean of 100. We obtain the Korean version of EPU monthly index from Baker et al. (2016) website at https://www.policyuncertainty.com/. Following Nguyen and Nguyen (2020), we take log transformation of the arithmetic average of the monthly Korean EPU index during the 12 months of the calendar year t. Table 1 shows yearly EPU values or twelve-month arithmetic average of Korean EPU (KOEPU1), first month Korean EPU (KOEPU2) and twelve-month arithmetic average of U.S. EPU (USEPU). In addition, we show the Korean election events, particularly, the presidential elections and national assembly elections and regime changes in the last column of Table 1.7Table 1 Economic policy uncertainty (EPU) index and presidential election information Year KOEPU1 12 month average KOEPU2 First month USEPU 12 month average Korean election events President party 2002 109.40 136.22 105.34 Presidential election Progressive 2003 165.82 224.40 110.12 2004 131.55 145.82 93.05 National assembly election 2005 68.64 107.98 71.75 2006 90.74 76.67 71.32 2007 82.57 134.79 80.51 Presidential election 2008 140.71 135.08 127.83 National assembly election Conservative 2009 147.08 201.63 143.94 2010 148.72 170.98 155.48 2011 167.03 137.05 172.24 2012 163.27 254.89 167.83 Presidential election, National assembly election 2013 130.61 192.95 120.12 2014 81.88 105.64 87.05 2015 128.25 165.82 108.66 2016 188.81 181.94 111.45 National assembly election 2017 160.77 391.80 111.44 Presidential election Progressive 2018 136.31 95.63 119.88 2019 257.36 249.44 140.67 Corporate donation measures We measure the dependent variable, the corporate donation of individual firms (CPG1), as the natural logarithm of annual corporate donation expense for each firm following prior studies (Maung et al. 2020). Consistent with recent literature (Kordsachia 2021; Luo et al. 2017; Ren et al. 2021), we also use annual corporate donation expense divided by firms’ size (CPG2) as our alternative measure of corporate donation. Control variables We include several firm-level control variables that are expected to affect corporate giving according to prior literature (Adams and Hardwick 1998; Brammer and Millington 2004). First, the natural log of total revenue as a measure of firm size (SIZE) is expected to be positively associated with corporate donation (CPG1 and CPG2) because bigger firms are facing more public scrutiny and they are more likely to have abundant resources. Therefore, they are more likely to engage in more corporate donation. A firm’s total debt to total equity or financial leverage ratio (LEV) is used to proxy for the firms’ bankruptcy risk. Highly leveraged firms spend more cash to pay for their interest expenses and therefore have less funds for corporate giving (Adams and Hardwick 1998). Firms’ profitability, liquidity, and operating cash flow are expected to affect corporate charities (Wang and Qian 2011). We expect a negative association between firms with negative net income (LOSS) and corporate donation and we expect positive associations between firms with greater current ratio (CRATIO) and operating cash flow (OCF) and corporate donation. Lastly, following Kim et al. (2019), we construct a dummy variable for the Big N auditors when private firms’ financial statements are audited by PWC, Deloitte, KPMG, or EY, to control for the quality of external audit and monitoring. Industry dummies, defined by two-digit SIC codes, are also included in the model to control for differences in industry characteristics (Du et al. 2018). Consistent with existing EPU studies (Baker et al. 2016; Gulen and Ion 2016; Nguyen and Nguyen 2020), we do not include time fixed effects in our model because including year dummies will automatically absorb the explanatory power of the EPU variable. Empirical models We tested our first hypothesis (Hypothesis 1) using Eq. (1). The dependent variable in Eq. (1) is the corporate giving or donation (CPG1 or CPG2) and our independent variable of interest is natural logarithm value of EPU (LNEPU) in a given year (Baker et al. 2016; Nguyen and Nguyen 2020; Yung and Root 2019). We perform our baseline ordinary least square (OLS) regression analyses based on the following Eq. (1):1 CPG1i,tCPG2i,t=β0+β1LNEPUi,t+β2SIZEi,t+β3LEVi,t+β4LOSSi,t+β5CRATIOi,t+β6OCFi,t+β7BIGNi,t+IndustryFixedEffects+εi,t Based on Hypothesis 1, we expect that private firms reduce their corporate donation as EPU increases and therefore we expect that the slope coefficient for LNEPU (β1) to be negative and significant. To test Hypotheses 2a and 2b, we conducted a multivariate regression based on the following Eq. (2):2 CPG1i,t=β0+β1LNEPUi,t+β2MEDIANCPG1LNADi,t+β3TLNEPUxTMEDIANCPG1TLNADi,t+β4SIZEi,t+β5LEVi,t+β6LOSSi,t+β7CRATIOi,t+β8OCFi,t+β9BIGNi,t+IndustryFixedEffects+εi,t Cao et al. (2019) indicate that firms do not operate in isolation and their corporate policies may be the outcome of interacting with other firms in the same industry. Following existing literature, we use industry median value of donation and logarithm value of advertising expenses as proxies for market competition (Brammer and Millington 2004; Baron et al. 2011; Harjoto et al. 2015) to test Hypothesis 2a. If private firms operate in a more competitive product market or experience intense peer pressure regarding corporate donation, we expect that private firms will increase their corporate donation to gain their competitive advantages (value enhancing motive). Thus, we expect that the slope coefficient (β2) for MEDIANCPG1 (or LNAD) would have a positive sign. Further, to test the moderating effect of LNEPU on the relationship between MEDIANCPG1 (or LNAD) and corporate donation for Hypothesis 2b, we constructed the interaction variable TLNEPU × TMEDIANCPG1 (or TLNEPU × TLNAD). There are high correlations between interaction variable (e.g., LNEPU × LNAD) and moderating variable (LNEPU) and independent variable (LNAD). Following the literature (Aiken and West 1991; Harjoto et al. 2017; Ruppert 2004), we transformed the components of the interaction variable (e.g., TLNEPU and TLNAD) by subtracting its mean value from each value of independent variables, then constructed the interaction variables based on these transformed variables (e.g., TLNEPU × TLNAD). We expect that private firms that operate in a competitive market or under intense peer pressure to engage in corporate donation will reduce corporate donations during a high EPU period. Thus, we expect the slope coefficient (β3) for TLNEPU × TMEDIANCPG1 (or TLNEPU × TLNAD) to be negative. To test Hypotheses 3a and 3b, we conducted a multivariate regression as follows in Eq. (3):3 CPG1i,t=β0+β1LNEPUi,t+β2PROGi,t+β3TLNEPUxTPROGi,t+β4SIZEi,t+β5LEVi,t+β6LOSSi,t+β7CRATIOi,t+β8OCFi,t+β9BIGNi,t+IndustryFixedEffects+εi,t In Korea, there are two main political parties, the progressive Minjoo Party of Korea and the conservative People Power Party. PROG is coded one if the president in Korea is from the progressive party or zero otherwise. When the progressive party is in power, its liberal-tilted political and social values create significant pressure on Korean private firms to engage in higher corporate donations. Consistent with Hypothesis 3a, we expect that private firms’ corporate donations are higher when the progressive political party is in power. Hence, we expect the slope coefficient (β2) to be positive. Consistent with Hypothesis 3b, we expect that private firms in Korea will decrease their corporate donations under a progressive political regime when EPU increases, given resource constraints. We use transformational value of each variable (TLNEPU and TPROG) to reduce multicollinearity problem. Hence, we expected the slope coefficient (β3) for the interaction variable TLNEPU × TPROG in Eq. (3) to be negative. All variables are defined in the Appendix. All regressions were estimated using robust standard errors clustered by the firm (Petersen 2009) to alleviate the cross-sectional correlations in the error terms that are inherent in the panel data, and industry fixed effects are included to control systematic differences in corporate donation across industry. We also performed a hierarchical regression analysis according to moderation analysis procedure. To alleviate potential outlier problems, we winsorized all continuous variables below 1% and above 99%. Empirical results Descriptive statistics Table 2 provides descriptive statistics for the variables used in this study. The mean value of corporate donation expenses (DON) is $39,565 per year. The mean value of natural log of donation expenses (CPG1) is 7.71 and the mean value of donation expenses to total assets (CPG2) is 4.87.8 The mean value of EPU in Korea and LNEPU is 145.35 and 4.93, respectively; while the mean value of natural log of U.S. EPU (LNUSEPU) is 4.75. Thus, Korean EPU is slightly higher than EPU in the U.S. Elected presidents are from the progressive party (PROG) 46% during our sample period. All other control variables, such as SIZE, LEV, LOSS, CRATIO, and OCF have similar statistical properties to those in prior studies (i.e., Kim et al. 2019). Finally, approximately 13% of private firms in Korea are audited by the big four (BIG N) accounting firms and the mean natural log of annual advertising expense is 11.03.Table 2 Descriptive statistics Variable N. of Obs Mean Std. Dev 25% Median 75% DON ($) 317,724 39,565 578,261 0.00 0.00 5,217 CPG1 317,724 7.71 7.94 0.00 0.00 15.61 CPG2 317,724 4.87 14.90 0.00 0.00 2.20 EPU 317,724 145.35 45.95 128.25 140.71 163.27 LNEPU 317,724 4.93 0.32 4.85 4.95 5.10 LNEPU2 317,724 5.11 0.41 4.90 5.11 5.31 LNUSEPU 317,724 4.75 0.24 4.66 4.71 4.95 IMPEACH 317,724 0.07 0.26 0.00 0.00 0.00 PROG 317,724 0.46 0.50 0.00 0.00 1.00 CASH 317,724 0.07 0.11 0.01 0.03 0.08 SIZE 317,724 16.44 1.71 15.70 16.64 17.44 LEV 317,724 3.27 16.66 0.40 1.43 3.49 CRATIO 317,724 2.55 7.06 0.63 1.09 1.91 LOSS 317,724 0.28 0.45 0.00 0.00 1.00 OCF 317,724 0.04 0.19 − 0.02 0.04 0.11 BIGN 317,724 0.13 0.34 0.00 0.00 0.00 LNAD 317,706 11.03 8.02 0.00 14.59 17.22 This table presents the yearly distributions of our full sample of 317,724 Korean private firm-year observations over the period 2002–2019. Variable definition is in Appendix Univariate analysis Table 3 reports the Pearson correlation coefficients for the variables used in the regressions. The LNEPU, the main variable of interest in this study, is significantly negatively correlated with corporate donation measure (CPG1 and CPG2) at the 1% significance level. Therefore, the correlation results suggest that corporate donation is negatively correlated with EPU. Next, we performed baseline OLS regression analyses to examine the association between corporate donation (CPG1 and CPG2) and EPU (LNEPU) while controlling for other factors that affect corporate donation.Table 3 Pearson correlations CPG1 CPG2 LNEPU LNEPU2 SIZE LOSS LEV CRATIO OCF CPG2 0.43 (0.00) LNEPU − 0.04 − 0.03 (0.00) (0.00) LNEPU2 − 0.02 − 0.02 0.63 (0.00) (0.00) (0.00) SIZE 0.35 0.12 − 0.01 − 0.01 (0.00) (0.00) (0.00) (0.00) LOSS − 0.04 − 0.02 0.01 0.01 − 0.04 (0.00) (0.00) (0.00) (0.00) (0.00) LEV − 0.03 0.01 0.02 0.01 − 0.16 − 0.04 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) CRATIO − 0.18 − 0.08 0.02 0.00 − 0.34 − 0.01 0.00 (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) (0.58) OCF 0.07 0.06 0.00 0.00 0.21 − 0.02 − 0.01 − 0.23 (0.00) (0.00) (0.05) (0.93) (0.00) (0.00) (0.00) (0.00) BIGN 0.10 0.04 − 0.05 − 0.03 0.24 − 0.03 0.01 − 0.03 0.05 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) This table presents Pearson correlations between key variables for the pooled sample. The two-tailed p-values are below the correlation coefficients. Variable definition is in Appendix Baseline OLS regression Table 4 presents the baseline OLS regression results to test our three hypotheses. Consistent with Hypothesis 1, Column (1) of Table 4 shows that that our first measure of corporate donation (CPG1) is negatively related with EPU (LNEPU) after robust standard errors are clustered at the firm level following Xu (2020) and Nguyen and Nguyen (2020).9 Furthermore, Column (2) of Table 4 shows that slope coefficient of MEDIANCPG1 is positive, which supports H2a: firms that operate in a more competitive market tend to have higher corporate donations. The slope of TLNEPU × TMEDIANCPG1 is negative and significant at the 1% level, which support Hypothesis 2b and indicates that Korean private firms that operate in higher peer pressure for corporate donation reduce their donation when EPU is heightened. Column (3) of Table 4 also shows similar results that firms operating in a more competitive market (LNAD) tend to have higher donations (H2a). The slope coefficient of TLNEPU × TLNAD is negative and significant at the 1% level, which also supports Hypothesis 2b and indicates that Korean private firms that operate in competitive markets tend to reduce donations when EPU is heightened. Column (4) of Table 4 also shows that the slope of the progressive party (PROG) is positive, which indicates that firms tend to have higher donations when the progressive party is in power (H3a). The slope of the interaction variable, TLNEPU × TPROG, is negative and is statistically significant at the 1% level. This result supports Hypothesis 3b and suggests that the positive association between corporate donation and progressive political regime is moderated by EPU. Overall, our results are consistent with the resource constraints and COR theory.Table 4 Baseline OLS regression Variables (1) (2) (3) (4) CPG1 CPG1 CPG1 CPG1 TLNEPU × TMEDIANCPG1 − 0.020* [− 3.074] MEDIANCPG1 0.085* [21.163] TLNEPU × TLNAD − 0.014* [− 2.891] LNAD 0.188* [55.316] TLNEPU × TPROG − 1.011* [− 12.586] PROG 0.230* [7.603] LNEPU − 0.843* − 0.444* − 0.525* − 0.576* [− 20.464] [− 11.471] [− 12.846] [− 15.046] SIZE 1.508* 1.508* 1.236* 1.510* [81.716] [81.829] [66.965] [81.845] LEV − 0.014* − 0.014* − 0.011* − 0.014* [− 15.909] [− 16.052] [− 13.642] [− 15.800] LOSS − 1.266* − 1.248* − 1.314* − 1.256* [− 28.644] [− 28.247] [− 30.586] [− 28.419] CRATIO 0.031* 0.032* 0.033* 0.031* [9.311] [9.603] [10.195] [9.438] OCF − 0.569* − 0.574* − 0.238* − 0.569* [− 6.806] [− 6.869] [− 2.897] [− 6.805] BIGN 0.113 0.057 − 0.087 0.090 [1.083] [0.542] [− 0.861] [0.858] Constant − 10.327* − 13.233* − 8.937* − 11.789* [− 18.545] [− 23.813] [− 16.189] [− 21.302] Industry dummy Yes Yes Yes Yes Cluster by firm Yes Yes Yes Yes Observations 317,724 317,724 317,706 317,724 R-squared 0.144 0.145 0.175 0.144 Average (mean) VIFs 1.09 1.11 1.1 1.14 Bolded coefficients indicate the variables of interest related to the hypotheses See the Appendix for variables definition. T-statistics, reported in bracket, are adjusted for firm-level clustering. *, , * indicate, respectively, the significance at the 1%, 5%, and 10% levels Hierarchical regression analysis Table 5 presents the multivariate regression results for Hypothesis 1. We conducted a hierarchical regression analysis as our main regression. Column (1) of Table 5 presents results only with control variables. Column (2) of Table 5 shows that our first measure of corporate donation (CPG1) is negatively related with EPU (LNEPU). The magnitude of the slope coefficient of LNEPU (− 0.875) in Column (2) indicates, in terms of economic significance, that when EPU increases by 1%, corporate donation decreases by 0.875%, which represents 11.35% of the mean for CPG1 (7.71). F-statistics for the R-squared Change (F-statistics Change) reported in Column (2) of Table 5 is significant at 1% level, which implies that our regression model containing EPU (LNEPU) in Column (2) provides a better fit to the data than a model that contains no EPU in Column (1). Our untabulated result also shows that our second measure of corporate donation (CPG2) is also negatively related to LNEPU. These magnitudes of the impact of EPU on corporate donation are economically significant.Table 5 Hierarchical regression for economic policy uncertainty (EPU) and corporate donation Variables (1) (2) CPG1 CPG1 SIZE 1.510* 1.511* [173.370] [173.671] LEV − 0.015* − 0.014* [ − 18.328] [ − 18.130] LOSS − 1.196* − 1.184* [ − 37.741] [ − 37.359] CRATIO 0.026* 0.027* [13.646] [14.165] OCF − 0.502* − 0.498* [ − 6.860] [ − 6.818] BIGN 0.368* 0.323* [9.216] [8.081] LNEPU − 0.875* [ − 21.313] Constant − 16.825* − 12.539* [ − 114.867] [ − 50.410] Observations 317,724 317,724 R-squared 0.126 0.127 R-squared Change 0.001 F-statistics 7622.14 6607.48 F-statistics Change 454.26 Average (mean) VIF 1.11 1.09 See the Appendix for variables definition. , , indicate, respectively, the significance at the 1%, 5%, and 10% levels The impacts of our control variables are mostly consistent with the literature. We find that larger firms (SIZE) tend to have greater corporate donation (Adams and Hardwick 1998; Brammer and Millington 2006), while firms with greater leverage (LEV) tend to have lower donation (Adams and Hardwick 1998). We also find that firms with greater operating cash flow (OCF) tend to have lower donations—these firms invest in research and development and capital expenditures. Firms with a negative net income (LOSS) tend to have lower donations (Wang and Qian 2011). We also find that firms with greater liquidity (CRATIO) and audited by the Big 4 tend to make more donations. Overall, our results show that increases in EPU are associated with decreases in private firms’ corporate donations. Thus, we found strong empirical evidence to support Hypothesis 1 that corporate donation is negatively related to EPU. Our results are consistent with prior studies that firms conserve their resources by reducing corporate spending as a precautionary saving motive when EPU is heightened (Gulen and Ion 2016; Nguyen and Phan 2017; Nguyen and Nguyen 2020). Panels A and B of Table 6 present the hierarchical regression results to examine our second hypotheses (H2a and H2b) regarding the impact of market competition on corporate donation and the moderating effect of EPU. Panels A and B show results on the moderating effect of EPU on the relation between industry median of corporate donation (MEDIANCPG1) or advertising expense (LNAD) as a proxy of industry-peer competitive pressure for corporate donation and product market competition. In Column (1), we include only the control variables. In Column (2), together with the control variables, we include main test variable (MEDIANCPG1 in Panel A and LNAD in Panel B). In Column (3), together with the control variable and main test variable (MEDIANCPG1 or LNAD), we include the moderating variable (LNEPU). In Column (4), together with the control variable, MEDIANCPG1 (LNAD), and LNEPU, we include the interaction variable between MEDIANCPG1 (LNAD) and LNEPU. The variables (MEDIANCPG1, LNAD, and LNEPU) used to construct the interaction variable are mean-centered (TMEDIANCPG1, TLNAD, and TLNEPU) to mitigate multicollinearity problem as well as to facilitate the interpretation of the main effects (Aiken and West 1991; Harjoto et al. 2017; Ruppert 2004). The slope coefficients of MEDIANCPG1 and LNAD in Column (2) in Panels A and B are positive and significant at the 1% level, supporting Hypothesis 2a that private firms in Korea increase corporate donations (CPG1) to gain competitive advantages when they operate in a more competitive market, measured by industry-peer donation and advertising expenses, which are consistent with prior findings (Zhang et al. 2010; Baron et al. 2011; Cao et al. 2019). Furthermore, Column (4) in Panel A shows that the coefficient of TLNEPU × TMEDIANCPG1 is negative and significant at the 1% level, which support Hypothesis 2b and indicate that Korean private firms that operate under higher industry-peer pressure for corporate donations reduce their donations when EPU is heightened. Similarly, Column (4) in Panel B also provides evidence to support Hypothesis 2b that coefficient of TLNEPU × TLNAD is negative and significant at the 1% level. This also indicates that Korean private firms that operate in competitive markets tend to reduce donations when EPU is heightened. In short, we find evidence to support Hypotheses 2a and 2b: private Korean firms in competitive markets or under high industry-peer pressure increase their donations in normal times to pursue their value enhancing motive, but decrease their donations when EPU is heightened to conserve their limited resources. This latter finding is consistent with the resource constraints and COR theory.Table 6 Hierarchical regression for product market competition and EPU as a moderator Variables (1) (2) (3) (4) CPG1 CPG1 CPG1 CPG1 Panel A. Moderating effect of EPU on the industry median of corporate donation SIZE 1.510* 1.442* 1.445* 1.445* [173.370] [164.233] [164.410] [164.340] LEV − 0.015* − 0.014* − 0.014* − 0.014* [− 18.328] [− 17.851] [− 17.799] [− 17.808] LOSS − 1.196* − 1.145* − 1.143* − 1.142* [− 37.741] [− 36.243] [− 36.169] [− 36.149] CRATIO 0.026* 0.032* 0.032* 0.032* [13.646] [16.645] [16.702] [16.697] OCF − 0.502* − 0.394* − 0.397* − 0.396* [− 6.860] [− 5.398] [− 5.447] [− 5.436] BIGN 0.368* 0.221* 0.211* 0.208* [9.216] [5.535] [5.283] [5.202] LNEPU − 0.319* − 0.335* [− 7.472] [− 7.806] MEDIANCPG1 0.116* 0.111* 0.111* [50.200] [46.033] [46.062] TLNEPU × TMEDIANCPG1 − 0.024* [− 3.428] Constant − 16.825* − 16.552* − 15.003* − 14.927* [− 114.867] [− 113.373] [− 59.153] [− 58.639] Observations 317,724 317,724 317,724 317,724 R-squared 0.126 0.133 0.133 0.133 R-squared Change 0.007 0.0001 0.0001 F-statistics 7622.14 6945.07 6084.96 5410.34 F-statistics Change 2520.04 55.83 11.75 Mean (average VIFs) 1.11 1.11 1.12 1.11 Panel B. Moderating effect of EPU on the firms’ advertising expense SIZE 1.510* 1.229* 1.232* 1.233* [173.370] [138.588] [138.951] [138.982] LEV − 0.015* − 0.012* − 0.012* − 0.012* [− 18.328] [− 15.420] [− 15.316] [− 15.330] LOSS − 1.196* − 1.257* − 1.248* − 1.248* [− 37.741] [− 40.472] [− 40.201] [− 40.180] CRATIO 0.026* 0.028* 0.029* 0.029* [13.646] [15.225] [15.544] [15.564] OCF − 0.502*** − 0.138* − 0.139* − 0.140* [− 6.860] [− 1.925] [− 1.936] [− 1.952] BIGN 0.368* 0.079 0.053 0.051 [9.216] [2.025] [1.359] [1.300] LNEPU − 0.552* − 0.545* [− 13.675] [− 13.490] LNAD 0.197* 0.195* 0.195* [115.978] [114.774] [114.786] TLNEPU × TLNAD − 0.015* [− 3.016] Constant − 16.825* − 14.350* − 11.669* − 11.714* [− 114.867] [− 98.932] [− 47.852] [− 47.947] Observations 317,724 317,706 317,706 317,706 R-squared 0.126 0.161 0.162 0.162 R-squared Change 0.035 0.001 0.0001 F-statistics 7622.14 8730.75 7667.25 6816.52 F-statistics Change 1.30 186.99 9.09 Average (mean) VIFs 1.11 1.12 1.11 1.10 Bolded coefficients indicate the variables of interest related to the hypotheses See the Appendix for variables definition. *, , * indicate, respectively, the significance at the 1%, 5%, and 10% levels We found that the slopes of EPU (LNEPU) and the interaction variables (TLNEPU × TMEDIANCPG1 or TLNEPU × TLNAD) in hierarchical regression models presented in Column (4) of Panels A and B of Table 6 are statistically significant. Based on studies on pure and quasi moderating factor (Sharma 2003; Sharma et al. 1981), EPU (LNEPU) is considered as a quasi-moderator since it significantly affects corporate donation by itself and interacts with the product market competition measures (MEDIANCPG1 and LNAD). The F-statistics Change reported in Columns (2), (3), and (4) of Panels A and B of Table 6 indicate that the incremental R-squared Change from adding the market competition measures (MEDIANCPG1 or LNAD) and the moderating effect of EPU on the market competition measures (TLNEPU × TMEDIANCPG1 or TLNEPU × TLNAD) are statistically significant. Therefore, we find evidence that the moderating effect of EPU (LNEPU) on the relationship between product market competition and corporate donation provides a better fit to the data than a model without the moderating effect of EPU. Hence, we find evidence of the positive impact of market competition on corporate donations and the moderating effect of EPU on the impact of market competition on corporate donations, which further support Hypotheses 2a and 2b. Table 7 presents data for the third hypotheses (H3a and H3b) based on hierarchical regression analyses with moderating effect of EPU. In Column (1), we include only the control variables. In Column (2), together with the control variables, we include main test variable (PROG). In Column (3), together with the control variable and PROG, we include the moderating variable (LNEPU). In Column (4), together with the control variable, PROG, and LNEPU, we include the interaction variable between PROG and LNEPU. LNEPU and PROG used in the interaction variables are mean-centered (TLNEPU and TPROG) in the same manner as in Table 6. The slope of coefficient of PROG in Column (2) is positive and significant at the 1% level, supporting hypothesis H3a that private firms in Korea increase corporate donation (CPG1) to maintain a favorable relationship with the progressive government, which is consistent with prior studies (North 1991; Campbell 2007; Gao 2011; Borghesi 2017) that document the effect of ideological values and beliefs of the ruling party on corporate donation and social responsibility. In Column (4), the slope of the interaction variable, TLNEPU × TPROG, is negative and is statistically significant at the 1% level. This result suggests that the positive association between corporate donation and progressive political regime is moderated by EPU. This result supports Hypothesis 3b and is consistent with the resource constraints and COR theory. When the progressive political regime rules during periods of high EPU, private firms in Korea reduce their donations to conserve resources that can be used in their core marketing and production activities.Table 7 Hierarchical regression for progressive political party regime and EPU as a moderator Variables (1) (2) (3) (4) CPG1 CPG1 CPG1 CPG1 SIZE 1.510* 1.511* 1.513* 1.514* [173.370] [173.509] [173.787] [173.942] LEV − 0.015* − 0.015* − 0.014* − 0.014* [− 18.328] [− 18.306] [− 18.113] [− 17.994] LOSS − 1.196* − 1.193* − 1.181* − 1.174* [− 37.741] [− 37.635] [− 37.271] [− 37.053] CRATIO 0.026* 0.026* 0.027* 0.027* [13.646] [13.597] [14.112] [14.379] OCF − 0.502* − 0.500* − 0.497* − 0.498* [− 6.860] [− 6.831] [− 6.793] [− 6.819] BIGN 0.368* 0.363* 0.319* 0.302* [9.216] [9.091] [7.989] [7.546] LNEPU − 0.860* − 0.604* [− 20.924] [− 12.890] PROG 0.229* 0.203* 0.209* [8.681] [7.677] [7.883] TLNEPU × TPROG − 1.027* [− 11.376] Constant − 16.825* − 16.951* − 12.724* − 14.017* [− 114.867] [− 115.176] [− 50.920] [− 51.071] Observations 317,724 317,724 317,724 317,724 R-squared 0.126 0.126 0.127 0.128 R-squared Change 0.0001 0.001 0.001 F-statistics 7622.14 6545.55 5789.96 5163.09 F-statistics Change 75.35 437.82 129.42 Average (mean) VIFs 1.11 1.09 1.08 1.14 Bolded coefficients indicate the variables of interest related to the hypotheses See the Appendix for variables definition. *, , * indicate, respectively, the significance at the 1%, 5%, and 10% levels We also find that the slopes of EPU (LNEPU) and the interaction variables (TLNEPU × TPROG) in hierarchical regression models presented in Column (4) of Table 7 are statistically significant. Based on the literature (Sharma 2003; Sharma et al. 1981), EPU (LNEPU) is considered as a quasi-moderator since it significantly affects corporate donation by itself and interacts with the progressive political leaning variable (PROG). The F-statistics Change reported in Columns (2), (3), and (4) of Table 7 also indicate that there is incremental significant contribution (R-squared Change) of the main test variable (PROG) and interaction variables (moderating effects) in predicting the value of corporate donation. Therefore, based on the F-statistics, we find that progressive political leaning and the moderating effect of EPU on the relationship between progressive political leaning provide a better fit to the data than a model without the progressive political leaning and the moderating effect of EPU on the relationship between progressive party and corporate donation. Since the progressive and the conservative parties’ ruling periods overlapped with the global financial crisis (GFC) (see Table 1), as a robustness check, we excluded the global financial crisis period (2008–2009) to address the concern that our empirical results could have been driven by the GFC. Our untabulated result shows that the slope coefficient of TLNEPU × TPROG shows consistent evidence to support Hypothesis 3b that EPU moderates the positive relationship between progressive party rule and corporate donation. Overall, our empirical results are robust regardless of whether the GFC period is included or not. Robustness tests We conducted several tests to examine whether our primary results remain robust under different control variables, variable measures, subsamples, and an alternative estimation method, and after taking into account the potential serial correlation of corporate donation.10 First, corporate donation in the current year could be influenced by donation level in the previous year. Thus, we included the previous donation (CPG1t-1 and CPG2t-1) as additional control variables in the regression when the dependent variables are CPG1 and CPG2, respectively. Second, corporate donation could be influenced by political uncertainty during the presidential election in Korea, which could potentially confound our main results. Upon investigation, we found that only one out of four presidential elections that occurred during our sample periods creates a significant political uncertainty in Korea. During 2016 (one year prior to the 2017 presidential election, President Park Geun-hye was impeached and removed from her presidential power and duties.11 To address this concern, we constructed the impeachment dummy variable (IMPEACH) which takes a value of one during 2016 and zero otherwise. Third, the U.S. heavily influences the Korean economy because of Korea’s reliance on trade with the U.S., and U.S. investment and financial markets.12 Hence, we include the U.S. EPU index (LNUSEPU) to control for the confounding effect related to U.S. EPU. As shown in Column (1) through Column (4) of Table 8, LNEPU remains negative and statistically significant at the 1% level with CPG1. Thus, the explanatory power of EPU on corporate donation is not fully absorbed by any other uncertainty proxies, which highlights the robustness of our main results. We also found that political uncertainty, measured by the 2016 presidential impeachment, and the U.S. EPU adversely affected private firms’ donations.Table 8 Robustness tests: additional control variables and alternative EPU Variables (1) (2) (3) (4) (5) CPG1 CPG2 CPG1 CPG1 CPG1 LNEPU − 0.869* − 0.518* − 0.822* − 0.633* [− 19.791] [− 7.883] [− 19.886] [− 9.821] LNEPU2 − 0.379* [− 14.408] SIZE 1.522* 0.364* 1.508* 1.508* 1.505* [72.444] [19.977] [81.717] [81.710] [81.562] LEV − 0.015* − 0.006* − 0.014* − 0.014* − 0.014* [− 14.127] [− 5.362] [− 15.901] [− 15.929] [− 15.960] LOSS − 1.388* − 0.603* − 1.267* − 1.264* − 1.276* [− 28.728] [− 11.883] [− 28.661] [− 28.588] [− 28.854] CRATIO 0.033* 0.015* 0.031* 0.031* 0.030* [8.939] [4.599] [9.307] [9.286] [9.124] OCF − 1.017* 1.916* − 0.569* − 0.571* − 0.569* [− 10.096] [10.201] [− 6.800] [− 6.831] [− 6.802] BIGN − 0.061 − 0.133 0.112 0.114 0.143 [− 0.547] [− 1.304] [1.075] [1.091] [1.368] CPG1t−1 0.000* [4.205] CPG2t−1 0.602* [82.005] IMPEACH − 0.096 [− 2.472] LNUSEPU − 0.376* [− 4.197] Intercept − 10.145* − 1.400* − 10.425* − 9.582* − 12.520* [− 16.528] [− 2.780] [− 18.713] [− 16.324] [− 23.295] Industry dummy Yes Yes Yes Yes Yes Cluster by firm Yes Yes Yes Yes Yes Observations 270,265 270,265 317,724 317,724 317,724 R-squared 0.148 0.386 0.144 0.144 0.143 Average (mean) VIFs 1.08 1.08 1.10 1.39 1.09 Bolded coefficients indicate the variables of interest related to the hypotheses See the Appendix for variables definition. T-statistics, reported in bracket, are adjusted for firm-level clustering. , , indicate, respectively, the significance at the 1%, 5%, and 10% levels Following Nguyen and Nguyen (2020), we used the first month EPU (LNEPU2) as an alternative EPU measure and reran the regressions in Table 4. Column (5) of Table 8 shows a significant negative coefficient for LNEPU2 at the 1% level with CPG1, which supports our main results. Table 9 presents the results of our additional robustness tests. First, more than half of our sample of private firms had zero corporate donations. Thus, we excluded zero-donation firms and reran regressions to check whether our main results were driven by zero-donation firms. Column (1) of Table 9 shows negative and significant coefficient for CPG2 at 1%, corroborating our main results. Second, firms that make donations are self-selecting in reporting their donations. To tackle this potential self-selection bias, we ran Heckman (1979) two-stage regression as shown in Columns (2) and (3). In the first stage model, we constructed the corporate giving dummy variable CPGDUM, which is equal to one if a firm reports a corporate donation in the current and last year (and 0 otherwise) to estimate a Probit model. In addition, we used the industry median value of corporate giving (MEDIANCPG1) as a proxy for industry peer pressure to donate as our instrumental variable. Column (2) of Table 9 shows that coefficient of MEDIANCPG1 is positive and significant at the 1% level in the first-stage regression, suggesting that the likelihood of corporate giving in current and last year increases with MEDIANCPG1. In the second-stage model, we included the calculated inverse Mills ratio (IMR) from the first stage model along with other control variables. In Column (3), after controlling for IMR calculated from the first-stage model, we found a consistent negative and significant association between LNEPU and CPG1 in the second stage model. Thus, our main results remain robust even after we addressed the potential self-selection bias with the Heckman two-stage procedures.Table 9 Robustness tests: Exclude zero donation firms and Heckman two-stage regression Variables (1) (2) (3) CPG2 CPGDUM CPG1 Exclude zero donation firms Heckman two-stage LNEPU − 1.417* − 0.120* [− 8.668] [− 7.015] MEDIANCPG1 0.018* [25.908] SIZE 0.156 0.251* 1.055* [1.979] [66.111] [45.696] LEV − 0.025* − 0.003* − 0.009* [− 5.471] [− 13.874] [− 16.831] LOSS − 1.533* − 0.210* − 0.639* [− 9.425] [− 26.851] [− 24.625] CRATIO 0.100* 0.005* 0.033* [5.836] [7.866] [15.794] OCF 7.025* − 0.099* − 0.049 [13.640] [− 6.666] [− 1.136] BIGN 0.024 − 0.056* 0.160* [0.064] [− 3.273] [4.424] Inverse mills ratio (IMR) 3.400* [24.408] Intercept 11.912* − 3.884* − 3.191* [7.146] [− 39.640] [− 7.121] Industry dummy Yes Yes Yes Cluster by firm Yes Yes Yes Observations 157,188 317,724 157,188 R-squared 0.019 0.097 0.153 Average (mean) VIFs 1.09 1.88 Bolded coefficients indicate the variables of interest related to the hypotheses See the Appendix for variables definition. T-statistics, reported in bracket, are adjusted for firm-level clustering. , , indicate, respectively, the significance at the 1%, 5%, and 10% levels Finally, we acknowledge that there is a potential firms’ specific omitted variables that affect corporate donation. We address this concern by conducting firm fixed-effects panel data regression model with standard errors are clustered at the firm-level (Firm Cluster) to control for firm specific and time invariant unobservable factors in our regression (Silviera, 2021). The results on Table 10 show that EPU negatively affects corporate donation (CPG1 or CPG2), indicating that our main results are robust even after we control firm-specific fixed effects for corporate donation. We conduct the generalized method of moment (GMM) dynamic panel data regression control the endogeneity and unobservable firm-specific fixed effects (Arellano and Bond 1991; Blundell and Bond 1998; Villarón-Peramato et al. 2018). We employ the change in EPU as our instrumental variable. We perform two specification tests, the first and second-order serial correlation tests of the residuals in the differenced equations (AR(1) and AR(2)) and the Sargan test for overidentification of our instrumental variable. Our untabulated results indicates that the change in EPU is negatively related to corporate donations, which corroborates our main finding. The p-value of the AR(1) indicates statistically significant for the first order of autoregressive, but the p-value for AR(2) test is 0.829 which indicates the absence of second order of serial correlation. The p-value for the Sargan test is 0.57, which indicates the null hypothesis indicating the overidentifying restriction for the GMM cannot be rejected. Therefore, our instrumental variable in our GMM regression is valid.Table 10 Robustness tests: firm fixed-effects regression Variables (1) (2) CPG1 CPG2 LNEPU − 0.324* − 1.042* [− 9.002] [− 13.478] SIZE 0.934* 0.482* [46.554] [12.411] LEV − 0.003* − 0.004* [− 4.046] [− 3.753] LOSS − 0.378* − 0.593* [− 11.960] [− 9.347] CRATIO 0.003 0.006 [1.055] [1.019] OCF − 0.342* 1.347* [− 5.180] [8.315] Intercept − 5.923* 2.191* [− 15.958] [2.953] Firm cluster Yes Yes Observations 317,724 317,724 R-squared 0.023 0.004 Number of firms 48,903 48,903 Bolded coefficients indicate the variables of interest related to the hypotheses See the Appendix for variables definition. *, , * indicate, respectively, the significance at the 1%, 5%, and 10% levels Conclusions Our study examines the relationship between private firms’ corporate donation behavior and EPU using unique and audited (credible) private firms’ donation data in Korea. We find that EPU is negatively associated with corporate donation, which suggests that private firms reduce their corporate donations when EPU is heightened. This implies that increases in EPU adversely affect charitable giving as private firms curtail their donations to the community. We find evidence to support the competitive advantages and value enhancing motives of corporate donation in which firms that operate in more competitive markets (peer pressure on donation and greater advertising expense) tend to have greater corporate donations. We also find that the political pressure exercised by the progressive party increases private firms’ donations. However, EPU moderates the positive relationship between corporate donation and market competition, and the positive relationship between corporate donation and the progressive party ruling. This implies that heightened EPU offsets private firms’ incentive to make charitable giving even if they are operating in competitive markets and are facing greater political pressure from the ruling party to engage in corporate donation. Our study extends the literature that mostly focuses on the social capital argument (Dai et al. 2020) and the competitive advantage argument that signals firms’ prospects to build trust with their stakeholders through CSR engagements (Zhang et al. 2020) when EPU is heightened. While we find evidence to support the competitive advantage argument, our study finds that private firms are more likely to revert to a resource conservation mode and precautionary saving motive when EPU increases by reducing their corporate donations. This precautionary saving motive when EPU is heightened creates a void in community charitable giving that has to be filled by government and publicly listed firms’ involvement as private firms withdraw their donation contributions to the community. Further, private firms can reconsider the strategic role of donation to the local community rather than just reverting to a resource conservation mode even when EPU heightens. Stakeholders in society also are encouraged to rethink the role of private firms’ donation to the local community and its implications under EPU. Our study has several limitations. First, due to the absence of required data, we were unable to examine the impact of ownership structure and corporate governance. Future studies could explore more rigorous analyses when such data become available. Second, the literature indicates that the ideological orientations of private firms’ owners play a significant role in the motives for corporate donation (Luo et al. 2017). Therefore, the role of personal and professional traits and political ideologies of private business owners on corporate donation can be further examined. Third, this study focuses solely on total aggregated donation amount. These aggregated amounts do not reveal how companies allocate their donations across different categories of corporate giving such as spending on social welfare to alleviate poverty, providing access to healthcare and education, and providing community support in response to natural disasters. Appendix Variable definitions.Variables Explanation DON Firms’ donation expense (in $) CPG1 Natural logarithm of the firms’ donation expense CPG2 Donation expense divided by total assets LNEPU Natural logarithm of arithmetic average of economic policy uncertainty in the past twelve months in Korea by Baker et al. (2016) LNEPU2 Natural logarithm of economic policy uncertainty in the first month in Korea by Baker et al. (2016) LNUSEPU Natural logarithm of arithmetic average of economic policy uncertainty in the past twelve months in U.S. by Baker et al. (2016) IMPEACH When President Park Geun-hye was impeached in year 2016, then coded as one, and zero otherwise PROG If the period of a ruling president is from progressive party (for example, Democratic in U.S.), then coded one, and zero otherwise GFC Global financial crisis which includes year 2008 and 2009 SIZE Natural logarithm value of total revenue LEV Total debt divided by total equity LOSS If net income is less than zero, then coded as one, and zero otherwise CASH Cash plus cash equivalents divided by total assets CRATIO Current assets divided by current liabilities OCF Operating cash flow divided by total assets BIG N If the firm is audited by Big N audit firms which are PWC, Deloitte, KPMG and E&Y, then coded as one, and zero otherwise MEDIANCPG1 Two-digit industry median value of CPG1 LNAD Natural logarithm of advertising expense CPGDUM If corporate giving in the current and last year is greater than zero, then coded as one, and zero otherwise Acknowldegment The authors thank the Editor, Sascha Kraus, Handling Associate Editor and two anonymous reviewers for their constructive comments and recommendations. Harjoto recognizes financial support and release time from the 2019–2021 Denney Professorship from Denney Endowment at Pepperdine Graziadio Business School for this research. The authors thank Larry Bumgardner for editing this manuscript. 1 The Korea Composite Stock Price Index (KOSPI) is the index that tracks the performance of all common stocks listed on the Korean Stock Exchange, which is similar to S&P 500. KOSDAQ (Korean Securities Dealers Automated Quotations) was established in 1996 and benchmarked from the U.S. counterpart, NASDAQ. 2 In the U.S., corporations tend to be more socially responsible when the Democratic Party (the liberal or the progressive party) is ruling the country (Baron et al. 2011). Studies also find that firms with headquarters in the Democratic states emphasize CSR performance more than those with headquarters in the Republican states (Di Giuli and Kostovetski 2014; Harjoto 2017; Kim et al. 2020; Rubin 2008). 3 For instance, in April of 2018, when the Chinese government imposed a ban on the import of waste resources, a trash crisis broke out on Korea’s Jeju Island, where 50% of the total amount of waste plastic was exported to China. KOSPO (Korean South Power Co), a public institution, held a resident briefing in collaboration with the Jeju Provincial Office, and gained consent to use waste vinyl refined oil as a power generation fuel, while easing concerns from residents about environmental pollution. In addition, management doctors (consulting agents) were dispatched to support small and medium-sized businesses to improve their facilities. This initiative was successful in converting 4200t, 56% of waste vinyl, to power generation fuel. Related firms recorded sales of 520 million won and creation of six jobs. The idea of converting wastes including waste vinyl and waste plastic into eco-friendly power generation fuel has been a great lesson for other local governments. It is also remarkable that KOSPO has smoothly promoted related tasks through communication with stakeholders such as local residents, the Jeju Provincial Office, and the Provincial Council (Shindonga-2019 Public Institutions Best Practice, Giving Korea, 2015). 4 During progressive party rule in 2018, there were 64 policies and 904 detailed tasks for SMEs and small business owners. Regarding the incentive system, there is unprecedented financial support for SME employers and young entrepreneurs by the Ministry of SMEs and Startups (see https://www.korea.kr/news/pressReleaseView.do?newsId=156282665). 5 In line with strong bonding between Korean private firms and regional banks, owner-managers of private firms tend to acquire buildings or land using maximum bank credit for capital gains or rental income to support their corporate donations. 6 Following prior literature that examines Korean private firms (Kim et al. 2011; Haw et al. 2014), we choose firms that have a fiscal year end of December 31 to increase the comparability of the firms’ financial statements and the computation of yearly EPU. 7 Usually, the Korean presidential election is held in December. The period of progressive- or conservative-ruling is classified depending on whether the presidential election is held in that year. For example, if the president is changed from the progressive party to the conservative party in December 2006, 2006 is coded as the progressive party and 2007 is coded as the conservative party. 8 The number of firm-year observations that report zero donation is 160,536 and we include these observations to reduce sample selection bias. In Table 8, we conduct robustness analyses by excluding firms with zero corporate donation, which revealed consistent results. 9 To address the omitted variables problem (Xu 2020), we re-estimated our multivariate regressions using a firm-fixed effect model and the untabulated results are qualitatively the same as our main results. 10 We conducted balanced panel data analyses for 3,709 firms during 2002–2019 as a robustness test and our untabulated results are similar to those using unbalanced panel data in Table 4. 11 During President Park Geun-hye’s tenure, her aide, Choi Soon-sil, who did not have an official position in the government, had used her position to seek monetary donations from several business conglomerates and this was a main cause of President Park’s impeachment in 2016 (https://www.bbc.com/news/world-asia-55657297). The 2016 presidential impeachment created a significant political uncertainty in Korea, especially related to corporate donation. Following Baker et al. (2016), we also investigated whether presidential elections with close votes (tight presidential elections) during 2002 between Roh Moo-hyun and Lee Hoi-chang and the 2012 election between Park Geun-hye and Moon Jae-in significantly affected corporate donation. Our untabulated results did not find a significant impact on corporate donations during this time, as there was not a spike in EPU. 12 In 2018, Korea exported $72,690,000 to the U.S. and imported $59,170,000 from the U.S. These are the largest export and import totals of all countries trading with Korea (The Bank of Korea Economic Statistics System 2018). Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Adams M Hardwick P An analysis of corporate donations: United Kingdom evidence J Manage Stud 1998 35 5 641 654 10.1111/1467-6486.00113 Aiken LS West SG Multiple regressions: testing and interpreting interactions 1991 Thousand Oaks CA, Sage Publications Inc Arco-Castro L López-Pérez MV Pérez-López MC Rodríguez-Ariza L Corporate philanthropy and employee engagement RMS 2020 14 705 725 10.1007/s11846-018-0312-1 Arellano M Bond S Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations Rev Econ Stud 1991 58 2 277 297 10.2307/2297968 Baker T Nelson R Creating something from nothing: resource construction through entrepreneurial bricolage Adm Sci Q 2005 50 3 329 366 10.2189/asqu.2005.50.3.329 Baker SR Bloom N Davis SJ Measuring economic policy uncertainty Quart J Econ 2016 131 4 1593 1636 10.1093/qje/qjw024 Baker SR, Bloom N, Davis SJ, Terry SJ (2020) COVID-Induced Economic Uncertainty. National Bureau of Economic Research (NBER) Working Paper No. 26983. Available at https://www.nber.org/papers/w26983 Baron D Harjoto MA Jo H The economics and politics of corporate social performance Bus Polit 2011 13 2 1 46 10.2202/1469-3569.1374 Barraza S Civelli A Economic policy uncertainty and the supply of business loans J Bank Finan 2020 121 105983 10.1016/j.jbankfin.2020.105983 Blundell R Bond S Initial conditions and moment restrictions in dynamic panel data models J Econom 1998 87 1 115 143 10.1016/S0304-4076(98)00009-8 Borghesi R Employee political affiliation as a driver of corporate social responsibility intensity Appl Econ 2017 50 19 2117 2132 10.1080/00036846.2017.1388911 Borghesi R Chang K Economic policy uncertainty and firm value: the mediating role of intangible assets and R&D Appl Econ Lett 2019 27 13 1087 1090 10.1080/13504851.2019.1661951 Brammer S Millington A The development of corporate charitable contributions in the UK: a stakeholder analysis J Manage Stud 2004 41 8 1411 1434 10.1111/j.1467-6486.2004.00480.x Brammer S Millington Andrew Firm size, organizational visibility and corporate philanthropy: an empirical analysis Business Ethics: A European Review 2006 15 1 6 18 10.1111/j.1467-8608.2006.00424.x Brown W Helland E Smith JK Corporate philanthropic practices J Corp Finan 2006 12 5 855 877 10.1016/j.jcorpfin.2006.02.001 Burgstahler DC Hail L Leuz C The importance of reporting incentives: earnings management in European private and public firms Account Rev 2006 81 5 983 1016 10.2308/accr.2006.81.5.983 Campbell JL Why would corporations behave in socially responsible ways? An institutional theory of Corporate Social Responsibility Acad Manag Rev 2007 32 3 946 967 10.5465/amr.2007.25275684 Cao J Liang H Zhan X Peer effects of corporate social responsibility Manage Sci 2019 65 12 5487 5503 10.1287/mnsc.2018.3100 Cespa G Cestone G Corporate social responsibility and managerial entrenchment J Econom Manag Strategy 2007 16 741 771 10.1111/j.1530-9134.2007.00156.x Chen P Lee C Zeng J Economic policy uncertainty and firm investment: evidence from the US market Applied Economics 2019 51 31 3423 3435 10.1080/00036846.2019.1581909 Clercq DD Belausteguigoitia I Political skill and organizational identification: preventing role ambiguity from hindering organizational citizenship behaviour J Manag Organ 2019 10.1017/jmo.2019.31 Crane A Matten D COVID-19 and future of CSR research J Manag Stud 2020 10.1111/joms.12642 Cyert RM March JG A behavioral theory of the firm 1963 Englewood Cliffs, N.J. Prentice Hall Dai Y, Rau PR, Tan W (2020) Do firms react to uncertainties by doing good deeds? Uncertainty and CSR investment. Working paper. Available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3524398 Davis GF Thompson TA A social movement perspective on corporate control Adm Sci Q 1994 39 141 173 10.2307/2393497 Di Giuli A Kostovetsky L Are red or blue companies more likely to go green? Politics and corporate social responsibility J Financ Econ 2014 111 158 180 10.1016/j.jfineco.2013.10.002 Donthu N Gustaffson A Effects of COVID-19 on business and research J Bus Res 2020 117 284 289 10.1016/j.jbusres.2020.06.008 32536736 Du X Zeng Q Chang Y To be philanthropic when being international: evidence from Chinese family firms J Manag Organ 2018 24 3 424 449 10.1017/jmo.2017.9 Gao Y Philanthropic disaster relief giving as a response to institutional pressure: evidence from China J Bus Res 2011 64 12 1377 1382 10.1016/j.jbusres.2010.12.003 32287523 García-Carbonell N Martín-Alcázar F Sánchez-Gardey G Facing crisis periods: a proposal for an integrative model of environmental scanning and strategic issue diagnosis Rev Managerial Sci 2021 10.1007/s11846-020-00431-y George G Slack resources and the performance of privately held firms Acad Manag J 2005 48 4 661 676 10.5465/amj.2005.17843944 Giving Korea (2015) Giving Korea 2015 Report. The 15th Symposium on Giving Culture. The Beautiful Foundation. Available at https://www.beautifulfund.org/eng/ResearchPublication.php Godfrey PC The relationship between corporate philanthropy and shareholder wealth: a risk management perspective Acad Manag Rev 2005 30 4 777 798 10.5465/amr.2005.18378878 Gulen H Ion M Policy uncertainty and corporate investment Rev Financ Stud 2016 29 3 523 564 Harjoto MA Corporate social responsibility and degrees of operating and financial leverage Rev Quant Financ Acc 2017 49 487 513 10.1007/s11156-016-0598-5 Harjoto MA Laksmana I Lee R Board diversity and corporate social responsibility J Bus Ethics 2015 132 641 660 10.1007/s10551-014-2343-0 Harjoto MA Jo H Kim Y Is institutional ownership related to CSR? The nonlinear relation and its implications for stock return volatility J Bus Ethics 2017 146 77 109 10.1007/s10551-015-2883-y Haw IM Lee JJ Lee WJ Debt financing and accounting conservatism in private firms Contemp Account Res 2014 31 4 1220 1259 10.1111/1911-3846.12064 Heckman JJ Sample selection bias as a specification error Econometrica 1979 47 1 153 161 10.2307/1912352 Hobfoll SE Conservation of resources American psychologists American Psychol 1989 44 3 513 524 10.1037/0003-066X.44.3.513 Hobfoll SE Halbesleben J Neveu J Westman M Conservation of resources in the organizational context: the reality of resources and their consequences Annu Rev Organ Psych Organ Behav 2018 44 3 513 524 Holtz-Eakin D Joulfaian D Rosen H Sticking it out: entrepreneurial survival and liquidity constraints J Polit Econ 1994 102 1 53 75 10.1086/261921 Holtz-Eakin D Joulfaian D Rosen H Entrepreneurial decisions and liquidity constraints Rand J Econom 1994 25 2 334 347 10.2307/2555834 Jeong Y Kim T Between legitimacy and efficiency: an institutional theory of corporate giving Acad Manag J 2019 62 5 1583 1608 10.5465/amj.2016.0575 Jin X Chen Z Yang X Economic policy uncertainty and stock price crash risk Account Financ 2019 58 5 1291 1318 10.1111/acfi.12455 Kim JB Simunic DA Stein MT Yi CH Voluntary audits and the cost of debt capital for privately held firms: Korean evidence Contemp Account Res 2011 28 2 585 615 10.1111/j.1911-3846.2010.01054.x Kim B Pae J Yoo C Business groups and tunneling: evidence from corporate charitable contributions by Korean companies J Bus Ethics 2019 154 643 666 10.1007/s10551-016-3415-0 Kim I Ryou JW Yang R The color of shareholders' money: institutional shareholders' political values and corporate environmental disclosure J Corp Financ 2020 64 101704 10.1016/j.jcorpfin.2020.101704 Kordsachia O A risk management perspective on CSR and the marginal cost of debt: empirical evidence from Europe RMS 2021 15 1611 1643 10.1007/s11846-020-00392-2 Lev B Petrovits C Radhakrishnan S Is doing good good for you? How corporate charitable contributions enhance revenue growth Strateg Manag J 2010 31 2 182 200 Li X Economic policy uncertainty and corporate cash policy: international evidence J Account Public Policy 2019 38 6 1 13 10.1016/j.jaccpubpol.2019.106694 Luo J Xiang Y Zhu R Military top executives and corporate philanthropy: evidence from China Asia Pacific J Manag 2017 34 725 755 10.1007/s10490-016-9499-3 Maung M Miller D Tang Z Value-enhancing social responsibility: market reaction to donations by family vs. non-family firms with religious CEOs J Bus Ethics 2020 163 745 758 10.1007/s10551-019-04381-8 Mosakowski E Hitt M Ireland D Sexton D Camp M Overcoming resource disadvantages in entrepreneurial firms: when less is more Strategic entrepreneurship: creating an integrated mindset 2002 Oxford, England Blackwell Publishing 106 126 Nguyen M Nguyen JH Economic policy uncertainty and firm tax avoidance Account Financ 2020 10.1111/acfi.12538 Nguyen NH Phan HV Policy uncertainty and mergers and acquisitions J Financ Quant Anal 2017 52 2 613 644 10.1017/S0022109017000175 North DC Institutions J Econom Perspect 1991 5 1 97 112 10.1257/jep.5.1.97 Oh W Chang Y Lee G Seo J Intragroup transactions, corporate governance, and corporate philanthropy in Korean business groups J Bus Ethics 2018 153 1031 1049 10.1007/s10551-018-3913-3 Petersen M Estimating standard errors in finance panel data sets: comparing approaches Rev Financ Stud 2009 22 1 435 480 10.1093/rfs/hhn053 Ren L Zhong X Wan L Defending the shell: differential effects of delisting pressure on R&D intensity and bribery expenditure Rev Managerial Sci 2021 10.1007/s11846-021-00489-2 Rubin A Political views and corporate decision making: the case of corporate social responsibility Financ Rev 2008 43 3 337 360 10.1111/j.1540-6288.2008.00197.x Ruppert D Statistics and finance: an introduction 2004 New York, NY Springer Sharma N The role of pure and quasi-moderators in services: an empirical investigation of ongoing customer-service-provider relationships J Retail Consum Serv 2003 10 4 253 262 10.1016/S0969-6989(02)00020-6 Sharma S Durand RM Gur-Arie O Identification and analysis of moderator variables J Mark Res 1981 18 291 300 10.1177/002224378101800303 Silviera AD Corporate governance and ethical culture: do boards matter? Rev Manag Sci 2021 10.1007/s11846-021-00476-7 Song H Chun H Brodmann J Song Y Organized labor and corporate philanthropy: Evidence from Korea Bus Eth European Rev 2020 29 4 780 795 10.1111/beer.12267 Villarón-Peramato Ó Martínez-Ferrero J García-Sánchez I CSR as entrenchment strategy and capital structure: corporate governance and investor protection as complementary and substitutive factors RMS 2018 12 27 64 10.1007/s11846-016-0212-1 Wang H Qian C Corporate philanthropy and corporate financial performance: the roles of stakeholder response and political access Acad Manag J 2011 54 6 1159 1181 10.5465/amj.2009.0548 Wang S Gao Y Hodgkinson GP Rousseau DM Flood PC Opening the black box of CSR decision making: a policy-capturing study of charitable donation decisions in China J Bus Ethics 2015 128 3 665 683 10.1007/s10551-014-2123-x Wu J Zhang J Zhang S Zhou L The economic policy uncertainty and firm investment in Australia Appl Econ 2020 52 31 3354 3378 10.1080/00036846.2019.1710454 Xu Z Economic policy uncertainty, cost of capital, and corporate innovation J Bank Financ 2020 111 1 15 10.1016/j.jbankfin.2019.105698 Yung K Root A Policy uncertainty and earnings management: International evidence J Bus Res 2019 100 255 267 10.1016/j.jbusres.2019.03.058 Zhang R Zhu J Yue H Zhu C Corporate philanthropic giving, advertising intensity, and industry competition level J Bus Ethics 2010 94 39 52 10.1007/s10551-009-0248-0 Zhang J, Kong D, Qin N, Wu J (2020) Being nice to stakeholders: The effect of economic policy uncertainty on Corporate Social Responsibility. Working paper. Available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3107756
PMC009xxxxxx/PMC9004457.txt	==== Front J Fam Econ Issues J Fam Econ Issues Journal of Family and Economic Issues 1058-0476 1573-3475 Springer US New York 35431528 9842 10.1007/s10834-022-09842-3 Original Paper Economic Resources Shaping Grandparent Responsibility Within Three-Generation Households http://orcid.org/0000-0001-5299-3268 Mutchler Jan E. Jan.Mutchler@umb.edu Velasco Roldán Nidya grid.266685.9 0000 0004 0386 3207 Department of Gerontology, Center for Social & Demographic Research On Aging, Gerontology Institute, University of Massachusetts Boston, 100 Morrissey Blvd., Boston, MA 02125-3393 USA 12 4 2022 2023 44 2 461472 23 3 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. The purpose of this descriptive study was to explore factors associated with perceptions of grandparent responsibility for grandchildren in three-generation households, focusing especially on a comparison of grandparents’ and parents’ financial contributions to the household and ethnicity of grandparent(s). The analysis used information about three-generation families in the 2011–2015 American Community Survey, retrieved through the Integrated Public Use Microdata Series. In 30% of these families, grandparents said they were “primarily responsible” for the grandchildren, even though the child’s parent was also in the household. Logistic regression models showed that grandparents who contributed a larger share of household income and grandparents who were householders were significantly more likely to report being primarily responsible for grandchildren in three-generation households, suggesting that the distribution of financial resources (or resource balance) within the household was associated with perceptions of responsibility. However, grandparents’ race and ethnicity moderated this association, indicating that cultural norms may intersect with resources in shaping these reports. The findings suggest that perceived responsibilities of grandparents in three-generation households may be shaped by the balance of financial resources among household members, but also by cultural norms of grandparenting. Keywords Intergenerational families Financial resources in families Family diversity Grandparent care issue-copyright-statement© Springer Science+Business Media, LLC, part of Springer Nature 2023 ==== Body pmcIntroduction Three-generation families, including minor children, parent(s), and grandparent(s), have been on the upswing in the United States (Casper et al., 2016; Dunifon et al., 2014). Estimates from the U.S. Census Bureau indicate that in 2014, 3.2 million children under the age of 18 lived with a grandparent and one or more parents, representing a 37% increase from two decades previously (U.S. Census Bureau, Current Population Survey, n.d.) Factors contributing to the establishment of intergenerational households include the high cost of housing, economic insecurity among young families as well as among aging family members, and needs for assistance across generations associated with childcare, health limitations, or other forms of support (Burr & Mutchler, 2007; Pilkauskas & Cross, 2018). Cultural orientations impacting intergenerational roles and relationships also have contributed to household configurations (Casper et al., 2016; Harrington Meyer & Kandic, 2017; Silverstein & Lee, 2016; Silverstein et al., 2012), resulting in higher prevalence of intergenerational households among some racial and ethnic groups than others. Along with changes in longevity, fertility, and marriage that are reshaping our understanding of “family” (Seltzer, 2019), these factors have led to increasing numbers of intergenerational co-resident families and growing heterogeneity in the relationships and responsibilities embedded within them. The purpose of this paper was to explore the factors associated with grandparents reporting responsibility for grandchildren within three-generation households in the US. This study was motivated by scientific questions about intergenerational support in families, as well as by policy and advocacy interests in grandparent caregivers. In this study we drew on responses to a question in the American Community Survey asking if adults were responsible for grandchildren living with them. In considering factors associated with grandparent responsibility, our analysis focused on the distribution of resources within the household and on the racial and ethnic characteristics of the grandparent, two sets of factors that have been heavily featured in the intergenerational family literature. This topic has taken on new significance during the COVID-19 pandemic. Exposure of older adults to COVID-19 infection may be higher in multigenerational homes due to household density, economic circumstances, and work patterns of adults (Stokes & Patterson, 2020). Indeed, the rate of intergenerational co-residence is positively associated with COVID-19 fatalities in the US as well as the EU (Aparicio Fenoll & Grossbard, 2020). Even absent a pandemic, multigenerational living may have mixed consequences for families. For example, single mothers and their children may benefit financially from living with grandparents (Mutchler & Baker, 2009), while adults (and especially women) who assume caregiving responsibilities for older relatives (Smith et al., 2020) or grandchildren (Harrington Meyer & Herd, 2007) experience elevated risks of employment disruptions. Background Factors Shaping the Prevalence of Three-Generation Households Three-generation co-residential families have been more common in the US than in many European countries (Glaser et al., 2018; Pilkauskas & Martinson, 2014), but less common than in Africa, Asia, Latin America and the Caribbean (United Nations, 2019). In the US, socioeconomic disadvantage has been a key factor promoting intergenerational living (Cross, 2018; Dunifon et al., 2014; Pilkauskas, 2012; Wiemers, 2014). Recent reports have documented an increase in the number of multigenerational households in the US (Casper et al., 2016; Glaser et al., 2018), which has been linked to amplified economic pressures on families that may be eased through shared resources. Ethnicity and immigration status have also been linked to intergenerational households, and some cultural groups have been more receptive than others to intergenerational living as a means of providing mutual support (Luo et al., 2012; Pilkauskas, 2012). Factors shaping rates of multigenerational living in the US compared to other countries have been thought to include the relatively higher rate of teenage births in the US (Sedgh et al., 2015), child welfare policies that promote grandparent care and multigenerational living (Baker et al., 2008; Harrington-Meyer & Kandic, 2017), immigration policies giving preference to family reunification, and immigration flows that have been dominated by arrivals from relatively “familistic” societies (i.e., Asia and Latin America). In light of these demographic and policy factors, the US is an important focus for this work. Three-generation families may reflect strategic responses to needs for mutual assistance within the intergenerational family system (Silverstein et al., 2012). For example, a three-generation household may serve as a vehicle for providing assistance to a member of the grandparent generation who is frail or in need of assistance; in many of these cases, the grandparent may contribute to the household financially or by providing childcare. Three-generation households may also be established in response to members of the parent generation who are too young, economically insecure, or otherwise unable to provide adequate support for their own children (Casper et al., 2016; Goodman & Silverstein, 2002; Pilkauskas, 2012; Pilkauskas & Cross, 2018). Mutual assistance may occur in these settings as well, through the pooling of economic resources, time, and instrumental support across generations (Harrington Meyer & Kandic, 2017). Culturally distinctive norms and values have also contributed to differences in the prevalence of three-generation households across demographic groups (Choi et al., 2016; Fuller-Thomson et al., 1997; Pilkauskas, 2012; Pilkauskas & Cross, 2018; Silverstein et al., 2012), resulting in patterns that differ by race, ethnicity, and immigrant status. Grandparents’ Contributions within the Three-Generation Household Contents of the grandparent role are diverse across families and settings (Hayslip et al., 2017). While many grandparents provide substantial support to their grandchildren without living together, the literature points to the shared household as a setting in which the contributions of grandparents may be especially substantial, including financial support, hands-on caregiving, help with homework, and other forms of care. Themes of reciprocity and obligation both across and within households are prevalent in the literature on intergenerational families (Drake et al., 2018; Johar et al., 2015). For example, many young adult children continue to live with and receive financial support from their parents (Padgett & Remle, 2016). For some families, continued financial support for adult children may extend to financially supporting or physically caring for grandchildren in the multigenerational home. Multiple “currencies” of exchange may be operational in a mutual assistance model—for example, financial assistance on the part of the grandparent generation may be balanced by instrumental assistance on the part of the parent generation, or vice versa. Considerable policy and advocacy efforts are directed toward strengthening support for families in which grandparents take on responsibility for minor grandchildren. Yet the conceptualization and measurement of caregiver status among grandparents is often unclear. To facilitate conceptualization and measurement of grandparents who are responsible for their grandchildren, questions about grandparents as caregivers have been included in the Census Bureau’s American Community Survey (ACS) for the last two decades in an effort to help understand provisions needed for federal programs designed to assist families.1 The questions, asked of all respondents age 30 or older, are as follows:26a. Does this person have any of his/her own grandchildren under the age of 18 living in this house or apartment? (Yes/no) 26b. [If yes] Is this grandparent currently responsible for most of the basic needs of any grandchildren under the age of 18 who live in this house or apartment? (Yes/no) An additional question is asked about the length of time for which caregiving grandparents have been responsible for coresiding grandchildren. Although policy and advocacy around grandparents who are responsible for grandchildren typically refer to “skipped-generation” families in which the grandchild lives with the grandparent in the absence of their parents, statistical reports make clear that many grandparents in three-generation households describe themselves as primary caregivers to a grandchild. Indeed, between 61 and 68% of grandchildren who are in the primary care of a grandparent live in three-generation households.2 In thinking about their responsibility for coresident grandchildren, grandparents may consider a mix of factors, including the extent to which they provide hands-on childcare, financial support, cultural roles, and potentially other factors that shape ways in which adults in three-generation families work together to provide support for grandchildren. The question of what factors are associated with grandparents in these three-generation families being responsible for their grandchildren is thus an important question. Our analysis focused on two sets of factors that have been featured in the intergenerational family literature. First, we explored the association between the distribution of financial resources across generations in the household and reports of grandparent responsibility. We hypothesized that grandparents in three-generation families would be more likely to be responsible for their grandchildren when their financial contributions were more sizable relative to that of the grandchild’ parent(s), measured by the share of the three-generation family income contributed by the grandparent (Hypothesis 1a). Using the same rationale, we expected grandparents would be more likely to be responsible for the grandchild when they were the householder (Hypothesis 1b). The second factor we considered in this paper is the racial and ethnic characteristics of grandparents, which may shape how they think about responsibilities toward other family members. “Cultural templates” defining family roles, intergenerational relationships, and caregiving norms (Arber & Timonen, 2012; Choi et al., 2016; Goodman & Silverstein, 2002; Herlofson & Hagestad, 2012; Luo et al., 2012; Silverstein & Lee, 2016) yield diverse household settings along with embedded relationships and responsibilities. High rates of intergenerational co-residence among African Americans, Hispanics, and Asians, for example, are frequently attributed to a combination of resource issues and cultural values and expectations. We expected that grandparents from ethnic groups having stronger cultural norms relating to significant grandparent roles (specifically, African American, Latino, and Asian grandparents) also may be more likely to perceive responsibility for the grandchildren in three-generation households (Hypothesis 2). Cultural templates may also interconnect with unevenness in financial contributions. Groups with stronger cultural patterns of multigenerational living and stronger norms surrounding grandparent contributions to care and support of grandchildren would be expected to draw on factors that go beyond financial support when describing responsibility for grandchildren. Therefore, we hypothesized that among groups in which substantial grandparent caregiving is more common, the associations between grandparents’ providing more of the household resources and their being responsible for grandchildren were dampened (Hypothesis 3). Gaps and Contributions This paper fills a gap in the literature on grandparents’ experiences within multigenerational families. Prior literature has not addressed the question of what factors account for grandparents’ perceiving responsibility for grandchildren when the child’s parent also lives in the household. Although the current study was correlational and we could not infer causation, this paper offered unique findings about how grandparents’ holding more financial resources in the household relative to the parents, in terms of income or household headship, may be factored in when responsibility is considered. It also identified intriguing differences in propensities for taking responsibility across racial and ethnic groups. This analysis on grandparent caregiving contributed to a more comprehensive scientific understanding of the matrix of family relationships that is growing increasingly complex. As well, it added to a knowledge base essential for the development and evaluation of many programs and services meant to assist grandparent caregivers (U.S. Census Bureau, n.d.). This study made a unique contribution to both of these goals. Data, Measures, and Methods Data We analyzed factors shaping reports of grandparents’ responsibility for grandchildren in three-generation households using information from the American Community Survey (ACS). The ACS has been conducted on an annual basis by the U.S. Census Bureau, providing demographic, social, economic, and housing information. Our analysis used the 2011–2015 ACS microdata file obtained through the Integrated Public Use Microdata Series (IPUMS) (Ruggles et al., 2017), which corresponded to a 5% nationally representative sample including about 3.1 million total observations. For convenience, and because all financial data in this five-year file was expressed in 2015 dollars, we referred to the data as ACS 2015. The IPUMS data system included flags linking the co-resident spouse, mother, and father of each respondent, allowing relationships to be more clearly specified. For our study, three-generation family households were the unit of observation, and each case contained information about all three generations. Our final sample included 177,858 three-generation family households including at least one minor child (younger than 18), one or both of their grandparents, and one or both of their parents. Variables As noted above, in the ACS, every adult age 30 or older was asked if they had any grandchildren under the age of 18 living in the same home. Those who responded in the affirmative were asked if they were “currently responsible for most of the basic needs” of those grandchildren. These questions do not offer respondents a clear definition of “responsibility” in this context, and respondents therefore draw on their own interpretations and perceptions. In this study, the dependent variable was dichotomous, where 1 = grandparent was reported as being responsible and 0 = grandparent was not reported as being responsible. In households where both grandparents were present, and one or both of them reported being responsible for grandchild(ren), the couple was classified as being responsible. We considered two sets of variables associated with reports of responsibility for grandchildren among grandparents in three-generation households to test our hypotheses. The first set was captured by share of household income contributed by the grandparent, and by household headship. We calculated the percentage of household income contributed by the grandparent(s), which ranged from 0% (representing three-generation co-resident household in which all of the income was generated by the parent generation) to 100% (reflecting households in which all the income was generated by the grandparent generation). A second indicator used was householder status, reflecting the person who owns or rents the home in which the three-generation family lives (1 = the grandparent(s) was householder, 0 = the grandparent(s) was not the householder). Our analysis also took into account grandparent’s income (logged) as a resource indicator. Additional variables were created to capture cultural factors shaping grandparent support and contributions to their grandchildren. Dummy variables were established based on race and ethnicity reported by the grandparent to capture potential cultural factors shaping grandparent responsibility. Categories included non-Hispanic Whites, Hispanic (any race), non-Hispanic African American, non-Hispanic Asian, and non-Hispanic “other” race. The “other” race category included those who report American Indian or Alaska Native origin, some other race, and more than one race; it also included grandparent couples in which the spouses reported different races. Non-Hispanic Whites served as the reference group. Additionally, control variables included family characteristics that may affect perceptions of grandparent responsibility in three-generation households, including characteristics of the grandparent, parent, and grandchild generations. With respect to grandparent characteristics, we included English proficiency and number of years since immigrating to the US, which have been shown to be important correlates of intergenerational living arrangements; they also reflected aspects of acculturation relevant to family roles and relationships (Abdul-Malak, 2016; Silverstein & Lee, 2016). We included variables indicating how many grandparents were in the household, and if only one grandparent was present, we indicated whether the grandparent was male or female. We also included age of grandparent(s) (mean age of grandparents if both are present) and grandparent(s) disability status (0 = no disability, 1 = one or both grandparents had a disability). Covariates capturing parent characteristics included attributes that may further contribute to the extent to which grandparent care and support was required. These included number and gender of parents (father only, mother only, or both parents), parents’ age in years (mean age of parents if both were present), parent disability status (0 = no disability, 1 = one or both parents had a disability), and parents’ school attendance (0 = not in school, 1 = one or both parents was in school). Finally, we controlled for size and age composition of the grandchild cohort living in the household by including number of grandchild(ren) under age 18 (dummy variables for 1, 2, 3, and 4 or more grandchildren present) and age of grandchild(ren) (captured by dummy variables indicating that one or more grandchildren was age 5 or under, one or more grandchildren age 6 to 12, and one or more grandchildren age 13 to 17; note that these indicators are not mutually exclusive). As shown in Table 1, grandparents were reported as being responsible for the grandchildren in three out of ten families among the three-generation family households identified for this study. Grandparents were the householder in a majority of the three-generation families (66%). The median income of grandparents in these three-generation family households was low, at roughly $17,500 in 2015 dollars (in comparison, median income for all U.S. households in 2015 was $55,775; U.S. Census Bureau, n.d.). Yet grandparents contributed a majority of the household income, with a median of 57% contributed by the grandparent generation. Just under half of the grandparents in the population defined for this study were non-Hispanic and White, while 16% were non-Hispanic Black, 9% were non-Hispanic Asian or Pacific Islander, and 25% were Hispanic. The remaining 5% were not Hispanic, but the grandparent reported some other race, reported being multi-racial, or, in the case of two grandparents being present, the two grandparents reported different races.Table 1 Characteristics of three-generation family households in the United States Characteristics Percentage or median Grandparent responsibility status Grandparent is responsible for grandchild(ren) 29.55% Resources of Grandparent(s) Median share of the combined parent & grandparent income that is contributed by the grandparent(s) 57.03% Mean share contributed by the grandparent(s) 54.10% Median income of the grandparent(s) (in 2015 dollars) $17,494.50 Mean income of the grandparent(s) (in 2015 dollars) $25,065.86 Householder status Grandparent is the householder 65.97% Grandparent(s) race and ethnic group membership Non-Hispanic White 44.49% Non-Hispanic Black 16.43% Non-Hispanic Asian/Pacific Islander 9.12% Non-Hispanic Other race, multiple races reported, or grandparents of different races 4.87% Hispanic 25.09% Characteristics of grandparent(s) English proficiency Grandparent(s) speaks English well or very well at home 19.99% Grandparent(s) speak poor English or no English at home 17.95% Grandparent(s) speak English only at home 62.06% Migration Grandparent(s) immigrated to the U.S. more than ten years ago 26.60% Grandparent(s) immigrated to the U.S. within previous ten years 6.14% Grandparent(s) are U.S. citizens by birth 67.26% Number and gender of grandparent(s) Grandfather only 10.88% Grandmother only 49.08% Two grandparents 40.04% Median age of the grandparent(s) (in years) 60 Grandparent(s) has a disability 32.65% Characteristics of parent(s) Number and gender of parent(s) Father only 16.43% Both parents 32.95% Mother only 50.62% Median age of parent(s) (in years) 33 Parent(s) has a disability 9.04% Parent(s) is in school 17.10% Characteristics of grandchildren Number of grandchildren in the household One 53.69% Two 30.13% Three 11.19% Four or more 4.99% One or more grandchild age 5 and under 56.56% One or more grandchild age 6 to 12 46.25% One or more grandchild age 13 to 17 27.07% Number of cases (unweighted) 177,858 Calculations based on weighted data from IPUMS 2015 ACS 5-year estimates Nearly two-thirds of the grandparents spoke English only at home, and nearly seven out of ten were U.S. citizens by birth. While 40% of the grandparent generation was made up of a couple, 49% of the households included just the grandmother, and 11% included only a grandfather. In one-third of the households, one or both grandparents had a disability. Median age of the grandparent generation was 60. Characteristics of the parent generation, also shown in Table 1, indicated that in half of the three-generation co-resident families considered here, only the mother was present. However, in one-third of the families, both parents were present, and 17% included fathers only. In 9% of the families, one or both parents had a disability, and in 17% one or both parents were in school. The median age of the parent generation in these households was 33 years. Most of the three-generation families included just one grandchild (54%), while 30% included two grandchildren, 11% included three, and 5% included four or more grandchildren. More than half of the families included at least one preschool-age grandchild (age 5 or under), and 47% included at least one grandchild age 6–12. Twenty-seven percent included one or more grandchildren who were teenagers age 13–17. The bivariate association between grandparent reports of responsibility and race or ethnicity showed that the share of grandparents reporting responsibility for a grandchild within a three-generation household was highest among those who reported other or mixed-race backgrounds, at 42% (see Table 2), and among African American grandparents, at 35%. Asian/Pacific Islander grandparents and Hispanic grandparents were less likely to report responsibility for grandchildren, at 14% and 25%, respectively. Nearly one-third of non-Hispanic White grandparents in three-generation family households reported responsibility for grandchildren.Table 2 Bivariate association of race and ethnicity with reports of grandparent responsibility in three-generation households Grandparent(s) race and ethnicity Grandparent(s) reported as having responsibility for grandchildren Grandparent(s) not reported as having responsibility for grandchildren N of cases (unweighted) Non-Hispanic White 31.78% 68.22% 84,544 Non-Hispanic Black/African American 35.38% 64.62% 27,378 Non-Hispanic Asian/Pacific Islander 14.01% 85.99% 15,963 Hispanic 24.99% 75.01% 39,427 Other 42.03% 57.97% 10,546 Total (all races) 29.55% 70.45% 177,858 Calculations based on weighted data from IPUMS 2015 ACS 5-year estimates All differences are significant at p < .001, two-tailed χ2 test Empirical Strategy Logistic regression models were estimated using a hierarchical inclusion of variables strategy. The first model evaluated the direct associations between the characteristics of grandparents, parents, and grandchildren, and grandparent’s reports of responsibility. Two additional models considered whether the associations between resource balance on grandparent reports of responsibility were different across race and ethnic groups. As a means of illustrating results associated with these interactive models, we estimated the predicted probabilities of grandparents reporting responsibility for grandchildren in three-generation households. All results were based on weighted data, using centered household weights. We focused on the share of income contributed by the grandparent and household headship to assess the financial contributions to the household made by the grandparent in relation to the parent. We used logistic regressions in SPSS to perform our estimations (see Eq. 1) based on weighted data. 1 lnP(Y)1-P(Y)=b0+b1x1+b2x2+⋯+bnxn Here, PY was defined as the probability of reporting primary responsibility for grandchildren among grandparents in three-generation family households. PY ranges from 0 to 1. The factors associated with the probability of grandparents’ reporting responsibility for grandchildren (independent variables) were represented in the right side of the equation by x1 to xn. In logistic regression, P(Y)1-P(Y) is known as the odds ratio, and the coefficients b1,⋯,bn indicate the change in the expected log odds with respect to a one unit change in x1,⋯,xn, respectively. We expected that grandparents would more commonly report primary responsibility for grandchildren when resources of the grandparent(s) were large relative to those of the parent(s), within the three-generation family household. We focused on the share of income contributed by the grandparent and household headship in making this assessment. We discussed the results using odds ratios (Eq. 2) rather than the log odds because we considered it a more intuitive way of describing the association between the variables of interest.2 P(Y)1-P(Y)=e(b0+b1x1+b2x2+⋯+bnxn) We anticipated that grandparents from race and ethnic groups having stronger cultural norms relating to significant grandparent roles would be more likely to report responsibility for their grandchildren in three-generation households. As well, we hypothesized that among race and ethnic groups in which the grandparent caregiving role was more familiar, the association between financial resource balance and grandparents’ reporting responsibility for grandchildren was dampened. We estimated interactions between the race and ethnicity of the grandparent and their financial resources in two independent models, one for share of income contributed by the grandparent and one for household headship to test that hypothesis. We estimated predicted probabilities (Eq. 3) for the figures describing the results of the interacted models. 3 PY=e(b0+b1x1+b2x2+⋯+bnxn)1+e(b0+b1x1+b2x2+⋯+bnxn) Results Consistent with our first hypothesis, financial resources were associated with the likelihood that grandparents report responsibility for grandchildren in three-generation families (see Table 3). Net of grandparents’ income, grandparents who contributed a larger share of household income were significantly more likely to report being responsible for grandchildren. In addition, grandparents who were householders were more likely to report responsibility for their grandchildren. Together, these results suggest that grandparents’ reports of responsibility were more likely in households where they were contributing more of the resources relative to the parent generation.Table 3 Logistic regression results for grandparents’ reports of responsibility for grandchildren in three-generation households on race and ethnicity, income share, household headship, and other selected characteristics Model 1 Model 2 Model 3 OR CI OR CI OR CI Grandparent(s) resources Grandparent’s share of household income 1.018* [1.018,1.019] 1.021* [1.020,1.022] 1.018* [1.017,1.019] Grandparent(s) income (logged) 0.905* [0.898,0.913] 0.914* [0.906,0.922] 0.910* [0.903,0.918] Grandparent(s) is householder 3.284* [3.116,3.461] 3.366* [3.191,3.550] 4.198* [3.862,4.563] Grandparent(s) race and ethnicity Hispanic 0.810* [0.761,0.862] 1.230* [1.100,1.376] 1.222 [1.083,1.379] Non-Hispanic Black/African American 1.028 [0.987,1.071] 1.367* [1.227,1.524] 1.046 [0.911,1.201] Non-Hispanic Asian/Pacific Islander 1.206* [1.107,1.314] 2.222* [1.954,2.526] 2.043* [1.796,2.324] Non-Hispanic Other 1.122* [1.054,1.194] 1.352 [1.129,1.619] 1.279* [1.007,1.624] Interactions Share of incomeHispanic 0.994** [0.993,0.995] Share of incomeBlack 0.996** [0.995,0.997] Share of incomeAsian 0.988** [0.986,0.990] Share of incomeOther 0.997 [0.995,1.000] HouseholderHispanic 0.616** [0.547,0.693] HouseholderBlack 0.984 [0.852,1.137] HouseholderAsian 0.445*** [0.383,0.517] HouseholderOther 0.859 [0.671,1.099] Unweighted N = 177,858. Calculations based on weighted data from IPUMS 2015 ACS 5-year estimates Other variables in the model include characteristics of the grandparent(s) (English proficiency, migration status, number and gender, age, disability status); characteristics of the parent(s) (number and gender, age, disability status, in-school status); and characteristics of the grandchildren (number, age). See Table 1 for description of these variables OR (odds ratios) are exponentiated coefficients; 95% confidence intervals in brackets p < .05, p < .01, *p < .001 Significant differences were also noted for race and ethnicity of grandparents, with Asian/Pacific Islander grandparents and “other” race grandparents being more likely to report being responsible for grandchildren, and Hispanic grandparents being less likely to have done so.3 No significant difference between Black/African American and non-Hispanic White grandparents was detected once other factors were controlled. Thus partial support for Hypothesis 2 was found. Note that the coefficients shown in Table 3 were estimated net of the complete set of covariates described in Table 1 (see footnote to Table 3). Two additional models tested interactions of race/ethnicity with the share of income contributed by the grandparent (Model 2) and with householder status (Model 3). Results suggested that the association between income share contributed by grandparents and grandparent reports of responsibility was smaller for grandparents who were Black, Asian, or Hispanic. Model 3 suggested further that the association between the grandparent being the householder and grandparent reports of responsibility was smaller for grandparents who were Hispanic or Asian. The implications of these results were illustrated in Figs. 1 and 2. Figure 1 was generated from coefficients in Model 2, and reflected the probability of a grandparent being reported as responsible for grandchildren based on a “prototypical” three-generation household, defined here as being composed of a single grandmother, age 60, who only spoke English at home, was U.S. born, did not have a disability, was the householder, and had income of $17,495, living with her single daughter who was age 33, not disabled, and not in school, along with one preschool-age grandchild. The estimates conveyed that all else equal, the probability of a grandparent reporting responsibility for grandchildren was substantially higher in households where grandparents contributed more of the financial resources, but this differed across race and ethnic groups. For example, non-Hispanic White grandparents, in households where they contribute 20% of the income (representing one standard deviation below the mean), had a 14% probability of assuming responsibility, but this probability was almost 40% among White grandparents contributing 88% of the income (representing one standard deviation above the mean), nearly a three-fold increase in predicted probability for this group. The curves were flatter for the other groups, suggesting that resource balance mattered less for these groups than for non-Hispanic Whites. For example, non-Hispanic Asian grandparents contributing a small share of household income had a 22% predicted probability of reporting responsibility—considerably higher than among non-Hispanic White grandparents contributing a similar share toward household income—but among Asian grandparents contributing a large share of the resources, the probability rose just to 34%, only 1.5 times the predicted probability for Asians contributing a small share. These results made clear that resources matter among all these groups, but to different degrees.Fig. 1 Predicted probabilities of grandparents reporting responsibility for grandchildren in three-generation family households, by income share and race/ethnicity of grandparent Fig. 2 Predicted probabilities of grandparents reporting responsibility for grandchildren in three-generation family households, by homeownership and race/ethnicity Figure 2 illustrated that being the householder in a three-generation family is also associated with a grandparent’s likelihood of reporting responsibility for grandchildren in different ways across race and ethnic groups. In particular, the positive coefficient for householder status was significantly reduced among Hispanic and Asian grandparents, consistent with findings based on the balance of financial resources reported here. We conclude that as expected, families described the grandparent contributions as they played out in multigenerational families in culturally unique ways, likely drawing on distinctive norms, experiences, and understandings of responsibility. Discussion The grandparent role has been discussed extensively in recent decades with respect to its implications for theory, policy, and practice. Many grandparents routinely provide emotional support and occasional babysitting for their grandchildren, and a sizable share offer financial assistance to their children and grandchildren, sometimes sharing a home or providing custodial care when parents are unable to do so. The three-generation household, including grandparents, parents, and grandchildren living together, represents a setting in which grandparent contributions may be especially substantial. This paper contributed to filling an important gap in the literature relating to multigenerational families and grandparent contributions within them. Our findings illustrated that in three out of ten three-generation households including grandparents and minor children, grandparents reported primary responsibility for the grandchildren with whom they live, even though the child’s parent(s) were also present. Our study focused on two primary mechanisms that may contribute to this outcome and illustrated the importance of both economic resources and cultural factors. Theoretical concepts of exchange and reciprocity suggest that adults reflect on the resource balance in the household in describing grandparent contributions. This expectation was supported by our finding that net of the demographic characteristics of grandparents, parents, and children, grandparents were more likely to report being responsible for grandchildren when they contributed more resources to the household, as reflected by share of income and household headship. Yet race and ethnicity of the grandparent clearly had an association that transcended financial exchanges, potentially reflecting cultural differences in how grandparent contributions were described, responses to immigrant experiences, needs of parents for support and assistance, or other factors. We found that all else equal, grandparents who were Asian/Pacific Islander were more likely to report being responsible for grandchildren, while those who were Hispanic were significantly less likely. Moreover, we found that grandparents from different racial and ethnic backgrounds responded differently to economic resources. These findings highlighted the heterogeneity of grandparent contributions within three-generation families. Advances in research and policy are needed to adequately understand and support intergenerational families and the children embedded within them. Limitations and Future Research Some limitations of this research require mention. First, the questions in the American Community Survey serving as the basis for this study do not allow elaboration on what respondents really mean when they say that a grandparent is “responsible” for grandchildren in the home. Our findings suggest that resource distribution and cultural norms are associated with responses to this question; however, additional factors, such as time spent with the grandchild or the amount of hands-on care provided, may also factor into these responses. The data used for this study do not permit determining whether grandparents who report being primarily responsible for a grandchild in the household has actually assumed financial responsibility for the child. Second, the reported assessment of grandparent responsibility in these data may or may not reflect a consensus across household members. The American Community Survey is a household survey meant to collect data on everyone living in the household, but in many households, it is likely that one person responds on behalf of everyone living in the home. Assessment of responsibility may differ depending on who responds to the questionnaire, a factor inviting further consideration using an alternative data source that allows the respondent to be identified, which the ACS does not. A third limitation of the study is that no information is provided on non-economic exchanges that occur in the three-generation households under study. Thus we are not able to determine whether three-generation grandparents who report being responsible for grandchildren are also providing, or receiving, other types of resources. As a fourth limitation, we note that households could also include additional adults who contribute financially to the household. Although we do not believe this would impact our results, it is a potentially interesting question for future research. In addition, the data used in this paper reflect a single five-year time period (2011–2015) and data from earlier or subsequent years could yield slightly different findings. Finally, this paper is based on correlational evidence that may not reflect causal processes underlying behavior and reports of grandparental responsibility. Despite these limitations, these intriguing findings raise many questions for future research. Further study could focus on the extent to which three-generation grandparents who perceive themselves to be primarily responsible for co-resident grandchildren have a different type or quality of relationship with the grandchildren in their care. Also requiring investigation is the durability of this status. The data used here reflect a cross-sectional assessment of the composition of households and the embedded relationships and responsibilities. Yet these features may be fluid; for example, some parents in three-generation households may be relatively unstable household members, moving in and out as circumstances change, with the grandparent assuming responsibility on a more permanent basis. Understanding the extent to which grandparents reporting responsibility for grandchildren in the three-generation setting reflects this fluidity would be informative. Also informative could be an in-depth analysis of how ages of the grandchildren are associated with grandparents’ reporting responsibility for them. Although this question is beyond the scope of our current study, our analysis indicates that in families with a grandchild age 5 or under, grandparents were less likely to report responsibility for grandchildren (results not shown). Very young children require more monitoring and more hands-on care, yet respondents in this study were less likely to identify being “responsible” for grandchildren when preschool-age grandchildren were present. Deepening our understanding of the stability of three-generation households, and how the movement of members in and out of the household is associated with responsibilities and relationships, is an important step in advancing the literature. Declarations Conflict of interest The authors declare that they have no conflicts of interest. Research Involving Human Participants and/or Animals This research does not contain any studies with human participants or animals performed by either of the authors. Informed Consent This research does not contain any individually identified information. 1 See https://www.census.gov/acs/www/about/why-we-ask-each-question/grandparents/. 2 Calculated by the authors from Table B10002 of the American Community Survey for 2011–2019. 3 In light of the low percentage of Asian American grandparents who reported responsibility for grandchildren within three-generation households (see Table 2), it is perhaps surprising that the multivariate results suggest that Asian American grandparents were significantly more likely to have reported responsibility than are their non-Hispanic White counterparts, all else equal. Close investigation of this result showed that a key factor shaping this result was the high presence of both parents within Asian three-generation households. Three out of four 3-generation households among Asian Americans included both parents in the middle generation, compared to one-third or fewer of the families in the other groups considered. This scenario suppressed the prevalence of grandparent responsibility among Asian Americans in bivariate comparisons. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Abdul-Malak, Y. (2016). Health and grandparenting among 13 Caribbean (and one Latin American) immigrant women in the United States. In M. Harrington Meyer & Y. Abdul-Malak (Eds.), Grandparenting in the United States (1st ed., pp. 61–80). Routledge. Aparicio Fenoll A Grossbard S Intergenerational residence patterns and Covid-19 fatalities in the EU and the US Economics and Human Biology 2020 10.1016/j.ehb.2020.100934 Arber, S., & Timonen, V. (2012). Grandparenting in the 21st century: New directions. In S. Arber & V. Timonen (Eds.), Contemporary grandparenting: Changing family relationships in global contexts (1st ed., pp. 247–64). The Policy Press. Baker LA Silverstein M Putney NM Grandparents raising grandchildren in the United States: Changing family forms, stagnant social policies Journal of Sociology and Social Policy 2008 7 53 69 Burr JA Mutchler JE Residential independence among older persons: Community and individual factors Population Research and Policy Review 2007 26 85 101 10.1007/s11113-007-9022-0 Casper, L. M., Florian, S. M., Potts, C. B., & Brandon, P. D. (2016). Portrait of American grandparent families. In M. Harrington Meyer & Y. Abdul-Malak (Eds.), Grandparenting in the United States (1st ed., pp. 109–132). Routledge Choi M Sprang G Eslinger JG Grandparents raising grandchildren: A synthetic review and theoretical model for interventions Family and Community Health 2016 39 2 120 128 10.1097/FCH.0000000000000097 26882415 Cross CJ Extended family households among children in the United States: Differences by race/ethnicity and socio-economic status Population Studies 2018 72 2 235 251 10.1080/00324728.2018.1468476 29770726 Drake D Dandy J Loh JMI Preece D Should parents financially support their adult children? Normative views in Australia Journal of Family and Economic Issues 2018 39 2 348 359 10.1007/s10834-017-9558-z Dunifon RE Ziol-Guest KM Kopko K Grandparent coresidence and family well-being: Implications for research and policy The ANNALS of the American Academy of Political and Social Science 2014 654 1 110 126 10.1177/0002716214526530 Fuller-Thomson E Minkler M Driver D A profile of grandparents raising grandchildren in the United States The Gerontologist 1997 37 3 406 411 10.1093/geront/37.3.406 9203764 Glaser K Stuchbury R Price D DiGessa G Ribe E Tinker A Trends in the prevalence of grandparents living with grandchild(ren) in selected European countries and the United States European Journal of Ageing 2018 15 237 250 10.1007/s10433-018-0474-3 30310371 Goodman C Silverstein M Grandmothers raising grandchildren: Family structure and well-being in culturally diverse families The Gerontologist 2002 42 5 676 689 10.1093/geront/42.5.676 12351803 Harrington Meyer, M., & Herd, P. (2007). Market friendly or family friendly? The state and gender inequality in older age. Russell Sage Foundation. 10.7758/9781610443937 Harrington Meyer M Kandic A Grandparenting in the United States Innovation in Aging 2017 1 2 1 10 10.1093/geroni/igx023 30480124 Hayslip B Fruhauf CA Dolbin-MacNab ML Grandparents raising grandchildren: What have we learned over the past decade? The Gerontologist 2017 57 6 1196 10.1093/geront/gnx106 28958034 Herlofson, K., & Hagestad, G. O. (2012). Transformations in the role of grandparents across welfare states. In S. Arber & V. Timonen (Eds.), Contemporary grandparenting: Changing family relationships in global contexts (1st ed., pp. 27–50). The Policy Press. Johar M Maruyama S Nakamura S Reciprocity in the formation of intergenerational coresidence Journal of Family and Economic Issues 2015 36 2 192 209 10.1007/s10834-013-9387-7 Luo Y LaPierre TA Hughes ME Waite LJ Grandparents providing care to grandchildren: A population-based study of continuity and change Journal of Family Issues 2012 33 9 1143 1167 10.1177/0192513X12438685 Mutchler JE Baker LA The implications of grandparent coresidence for economic hardship among children in mother-only families Journal of Family Issues 2009 30 11 1576 1597 10.1177/0192513X09340527 25750469 Padgett CS Remle RC Financial assistance patterns from midlife parents to adult children: A test of the cumulative advantage hypothesis Journal of Family and Economic Issues 2016 37 3 435 449 10.1007/s10834-015-9461-4 Pilkauskas NV Three generation family households: Differences by family structure at birth Journal of Marriage and the Family 2012 74 5 931 943 10.1111/j.1741-3737.2012.01008.x 24014117 Pilkauskas NV Cross C Beyond the nuclear family: Trends in children living in shared households Demography 2018 55 6 2283 2297 10.1007/s13524-018-0719-y 30298464 Pilkauskas NV Martinson ML Three-generation family households in early childhood: Comparisons between the United States, the United Kingdom, and Australia Demographic Research 2014 30 1639 1652 10.4054/DemRes.2014.30.60 25429249 Ruggles, S., Flood, S., Goeken, R., Grover, J., Meyer, E., Pacas, J., & Sobek, M. (2017). Integrated Public Use Microdata Series (IPUMS USA: Version 7.0) . University of Minnesota. 10.18128/D010.V11.0 Sedgh G Finer LB Bankole A Eilers MA Singh S Adolescent pregnancy, birth, and abortion rates across countries: Levels and recent trends Journal of Adolescent Health 2015 56 2 223 230 10.1016/j.jadohealth.2014.09.007 Seltzer JA Family change and changing family demography Demography 2019 56 2 405 426 10.1007/s13524-019-00766-6 30838537 Silverstein, M., & Lee, Y. (2016). Race and ethnic differences in grandchild care and financial transfers with grandfamilies: An intersectional resource approach. In M. Harrington Meyer & Y. Abdul-Malak (Eds.), Grandparenting in the United States (1st ed., pp. 19–40). Routledge. Silverstein, M., Lendon, J., & Giarrusso, R. (2012). Ethnic and cultural diversity in aging families: Implications for resource allocation and well-being across generations. In R. Blieszner & V. Hilkevitch Bedford (Eds.), Handbook of families and aging (2nd ed., pp. 212–307). ABC-CLIO, LLC. Smith, P. M., Cawley, C., Williams, A., & Mustard, C. (2020). Male/female differences in the impact of caring for elderly relatives on labor market attachment and hours of work: 1997–2015. Journals of Gerontology: Social Sciences, 75(3), 694–704. 10.1093/geronb/gbz026 Stokes JE Patterson SE Intergenerational relationships, family caregiving policy, and COVID-19 in the United States Journal of Aging & Social Policy 2020 32 4–5 416 424 10.1080/08959420.2020.1770031 32489144 United Nations. (2019). Living arrangements of older persons around the world (Population Facts No. 2019/2). Population Division, Department of Economic and Social Affairs. https://www.un.org/en/development/desa/population/publications/pdf/popfacts/PopFacts_2019-2.pdf U.S. Census Bureau. (n.d.). American Community Survey (ACS). Why we ask: Grandparents as caregivers [Fact Sheet]. https://www.census.gov/acs/www/about/why-we-ask-each-question/grandparents/ U.S. Census Bureau. (n.d.). 2015: ACS 5-year estimates detail tables. Table B19049. CEDSCI. Retrieved April 19, 2021, from https://data.census.gov/cedsci/table?q=median%20income%20age%202015&tid=ACSDT1Y2015.B19049 U.S. Census Bureau. (n.d.). Current population survey, March and annual social and economic supplements, 2014 and earlier. Retrieved December 18, 2021, from https://www.census.gov/data/tables/time-series/demo/families/children.html Wiemers EE The effect of unemployment on household composition and doubling up Demography 2014 51 6 2155 2178 10.1007/s13524-014-0347-0 25421522
PMC009xxxxxx/PMC9004485.txt	==== Front JPGN Rep JPGN Rep PG9 JPGN Reports 2691-171X Lippincott Williams & Wilkins, Inc. Philadelphia, PA 35425944 00019 10.1097/PG9.0000000000000131 3 Brief Report Gastroenterology Mycophenolate-Induced Colitis in Autoimmune Polyendocrinopathy-Candidiasis-Ectodermal Dystrophy Patients Schmitt Monica M. CRNP, MSN * Ferré Elise M. N. PA-C, MPH * Sampaio De Melo Michelly MD † Cooper Megan A. MD, PhD ‡ Quezado Martha M. MD † Heller Theo MD § Lionakis Michail S. MD, ScD * From the * Fungal Pathogenesis Section, Laboratory of Clinical Immunology and Microbiology (LCIM), National Institute of Allergy & Infectious Diseases (NIAID), NIH, Bethesda, Maryland † Laboratory of Pathology, Center for Cancer Research, National Cancer Institute (NCI), NIH, Bethesda, Maryland ‡ Department of Pediatrics, Division of Rheumatology/Immunology, Washington University School of Medicine, St. Louis, Missouri § Translational Hepatology Section, Liver Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), NIH, Bethesda, Maryland. Correspondence: Michail S. Lionakis, MD, ScD, Translational Hepatology Section, Liver Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), NIH, 9000 Rockville Pike, Building 10, Room 12C103A, Bethesda, MD 20892. E-mail: lionakism@mail.nih.gov. 11 2021 08 11 2021 2 4 e131e131 2 7 2021 10 9 2021 Copyright © 2021 The Author(s). Published by Wolters Kluwer on behalf of European Society for Pediatric Gastroenterology, Hepatology, and Nutrition and North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition. 2021 Written work prepared by employees of the Federal Government as part of their official duties is, under the U.S. Copyright Act, a “work of the United States Government” for which copyright protection under Title 17 of the United States Code is not available. As such, copyright does not extend to the contributions of employees of the Federal Government. Introduction: Autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED) is a prototypic monogenic autoimmune disorder caused by AIRE deficiency-mediated impaired central immune tolerance. Although multiple endocrine and nonendocrine tissues are affected in APECED, the colon is an uncommon target of autoimmune attack. Mycophenolate is a potent immunomodulatory medication that is used to treat autoimmune manifestations in patients with APECED and other autoimmune diseases. Methods: We reviewed the clinical, laboratory, genetic, histological, and treatment data of mycophenolate-induced colitis in our cohort of 104 APECED patients. Results: Among 10 mycophenolate-treated APECED patients, 4 (40%) developed reversible biopsy-proven mycophenolate-induced colitis characterized by an inflammatory bowel disease-like and/or graft-versus-host disease-like histological pattern. Conclusion: Mycophenolate-induced colitis appears to be a common complication in patients with APECED for which clinicians should maintain a high index of suspicion. autoimmune regulator autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy autoimmune polyglandular syndrome type 1 mycophenolate colitis OPEN-ACCESSTRUE ==== Body pmcWhat Is Known Mycophenolate is known to cause colitis in up to 9% of patients treated for rheumatologic diseases or solid organ transplantation. Discontinuing mycophenolate resolves symptoms in most patients. What Is New Mycophenolate-induced colitis may be more common in APECED patients than in solid organ transplant recipients. APECED patients with gastrointestinal symptoms warrant a complete gastrointestinal work up including esophagogastroduodenoscopy and colonoscopy. Discontinuing mycophenolate results in return of normal colonic histopathologic architecture in APECED and improvement or resolution of gastrointestinal symptoms. Autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED), also known as autoimmune polyglandular syndrome type 1 (APS-1), is a monogenic autoimmune disorder caused—most often—by biallelic mutations in the autoimmune regulator (AIRE) gene (1–3). Classic features consist of chronic mucocutaneous candidiasis, hypoparathyroidism, and primary adrenal insufficiency; however, a plethora of other nonendocrine autoimmune manifestations also occur (1,4–6); notably, although the small intestine is frequently affected, the colon is an uncommon target of autoimmune attack in APECED (4,5,7). Herein, we describe 4 APECED patients with autoimmune end-organ damage requiring unlabeled use of immunomodulation with mycophenolic acid (MPA), the active metabolite of mycophenolate mofetil (MMF; Cellcept), and mycophenolate sodium (MPS; Myfortic), who developed reversible, biopsy-proven mycophenolate-induced colitis. METHODS AND PATIENTS Between 2013 and 2019, we have evaluated 104 APECED patients at the NIH Clinical Center in an IRB-approved prospective natural history study (Clinicaltrials.gov NCT01386437). Within this cohort, we evaluated the clinical course of 4 patients who were diagnosed with biopsy-proven mycophenolate-induced colitis. Each patient/family provided informed consent at enrollment into our protocol, which allowed collection, analysis, and publication of patient data. Patients underwent clinical evaluation for gastrointestinal complaints that included colonoscopy with biopsies which led to the diagnosis of MMF-induced colitis. Retrospective histopathologic analysis of MMF-induced colitis was classified according to the spectrum of histopathologic features defined by Selbst et al (8). Patient 1 A 19-year-old woman with tubulointerstitial nephritis, which affects ~5% of APECED patients (4), was treated with MPS (540 mg twice daily). Seven months after initiation of MPS, she developed watery diarrhea (4–5/day) and periumbilical cramping with nocturnal symptoms that awakened her. Infection, fat malabsorption, exocrine pancreatic insufficiency, and small intestinal bacterial overgrowth (SIBO) were ruled out (Table 1). Colonoscopy revealed atrophic, friable, and edematous colonic mucosa with several stellate superficial ulcers. Histologic examination revealed chronic colitis with crypt architectural distortion, lymphoplasmacytic and eosinophilic infiltration within the lamina propria, and increased enterocyte apoptosis (Fig. 1A). MPS was discontinued with complete resolution in symptoms in 3 weeks. Due to worsening renal function upon MPS discontinuation, MPS was restarted after 5 months at half-dose (270 mg twice daily). Abdominal symptoms recurred 2 weeks later. MPS was discontinued with complete resolution of symptoms shortly thereafter. The antimetabolite azathioprine was initiated for the management of tubulointerstitial nephritis. TABLE 1. Demographic, clinical, genetic, laboratory, radiographic, gross, and histologic data of mycophenolate-induced colitis in APECED patients Demographics Patient 1 Patient 2 Patient 3 Patient 4 Ethnicity Caucasian Caucasian Caucasian Caucasian Gender Female Female Female Male AIRE mutations Allele 1 c.967_979del13 c.967_979del13 c.967_979del13 c.967_979del13 Allele 2 c.1249_1250insC c.967_979del13 c.769C>T c.967_979del13 Development of a classic dyad of clinical manifestations (age at onset in years) 11 N/A 3 6 Chronic mucocutaneous candidiasis No No 1 3 Hypoparathyroidism 6 5 3 6 Adrenal insufficiency 11 No 3 14 Non-Triad clnical minifestations (age in hears at appearance) APECED rash 1 0.75 1.5 No Enamel hypoplasia 4 10 1 No Intestinal dysfunction No No 2 5 Additional manifesstations (age in years at appearance) Ovarian failure (14), pneumonitis (15), hypertension (19) Enamel hypoplasia (10) Gastritis (6), pneumonitis (8), alopecia (9), hypothyroidism (10), Sjogren’s-like syndrome (11) Sjogren’s-like syndrome (3), gastritis (5), enamel hypoplasia (7), hypertension (30) End organ damage requiring mpa immunomodulation (age at diagnosis in years) Tubular interstitial nephritis (17) Pneumonitis (1.5), autoimmune hepatitis (6) Tubular interstitial nephritis (6) Pneumonitis (5) MPA agent MPS MMF MMF MMF Start date (dose escalation) November 2013 June 2017 August 2015 September 2014 (December 2014) Time frame of symptom onset after initian or excalation of MPA 7 months 2 weeks 8 months 4 months Age at development of MMF-induced colitis (years) 19 14 10 52 Description of symptoms 4–5 watery stools a day occurring mostly at night with periumbilical cramping and profound foul-smelling flatulence 5–6 soft/loose stools per day without abdominal cramping, pain, excessive flatulence, greasy or floating stool Avoidance of solids, nausea, burping, regurgitation, and weight loss 10 watery stools per day with tenesmus Drug dose at which symptoms began 540 mg twice a day 1000 mg twice a day 500 mg twice a day 1500 mg twice a day Weight (kg) 51.7 70.7 28.4 60.4 Microbiological stool studies (negative) Ova and parasite, GI pathogen PCR panel Ova and parasite, GI pathogen PCR panel NP Ova and parasite, GI pathogen PCR panel Fecal fat (24-hour collection) 3 g/24 hour 1 g/24 hours NP 5 g/24 hours Fecal calprotectin, fecal lactoferrin 71 µg/g, positive NP NP 863 µg/g, positive Date of endoscopies August 2014 July 2018 June 2016 May 2015 Macroscopic findings Colon with multiple patchy areas of inflammation and small ulcers ranging from 1 to 2 mm to the largest being approximately 4 mm Patchy areas of moderately erythematous mucosa found throughout the entire colon The colonic mucosa appeared visually normal without erosions, ulcers, or polyps Colon with multiple patchy areas of inflammation and small ulcers ranging from 1 to 2 mm to the largest being approximately 4 mm Histological findings* IBD-like pattern (moderate crypt architectural distortion, lymphoplasmacytic lamina propria inflammation, crypt injury and focal cryptitis) overlapping with GvHD-like pattern (increased enterocyte apoptosis—grade 2; 11–20/100 crypts).There is increased eosinophils within the lamina propria and intraepithelial space IBD-like pattern (crypt architectural distortion, lymphoplasmacytic lamina propria inflammation, crypt injury) overlapping with GvHD-like pattern (increased enterocyte apoptosis—grade 3; >21/100 crypts). Frequent tingible body macrophages within the lamina propria are seen. There is involvement of the rectum IBD-like pattern (mild architectural distortion, crypt loss and lymphoplasmacytic and eosinophilic infiltrate within the lamina propria. Rare to absent enterocytes apoptosis is seen IBD-like pattern (moderate crypt architectural distortion, lymphoplasmacytic lamina propria inflammation and crypt injury) mixed with GvHD-like pattern (increased enterocyte apoptosis—grade 3; >21/100 crypts). Frequent eosinophils and tingible body macrophages within the lamina propria Immunohistochemical staining (negative) CMV, EBV, adenovirus CMV, adenovirus Adenovirus CMV, EBV, adenovirus CMV = cytomegalovirus; EBV = Epstein-Barr virus; GvHD = graft-versus-host disease; IBD = inflammatory bowel disease; MMF = mycophenolate mofetil; MPS = mycophenolate sodium; N/A = not applicable; NP = not performed. *Characterization of histological findings was performed based on Selbst et al (8). FIGURE 1. Histological features of mycophenolate-induced colitis in autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED) patients. A) A mixed inflammatory bowel disease (IBD)-like and graft-versus-host disease (GvHD)-like pattern associated with increased number of eosinophils within the lamina propria is seen in patient 1. A) (left panel) IBD-like predominant pattern of colitis characterized by architectural distortion, crypt drop-out, crypt injury (black arrow), and chronic inflammation including eosinophils. A) (right panel) GvHD-like predominant pattern area of colitis, characterized by frequent apoptotic bodies (black arrows). B) A mixed IBD and GvHD-like pattern of colitis associated with frequent tingible body macrophages is seen in patient 2. B) (upper panel) IBD-like predominant pattern area of colitis with architectural distortion, crypt drop-out, edematous lamina propria, and lymphoplasmacytic inflammation. B) (left lower panel) GvHD-like pattern area of colitis with numerous apoptotic bodies (black arrows) and minimal lymphoplasmacytic inflammation within the lamina propria, and frequent tingible body macrophages (red arrow). B) (right lower panel) CD68 immunostain highlights the tingible body macrophages within the lamina propria. C) IBD-like pattern of colitis in patient 3 with architectural distortion, crypt drop-out, and lymphoplasmacytic and eosinophilic infiltrates within the lamina propria. Enterocyte apoptosis is also seen. D) In patient 4, IBD-like pattern of colitis was noted with moderate crypt architectural distortion, lymphoplasmacytic lamina propria inflammation, and crypt injury mixed with GvHD-like pattern as shown by increased enterocyte apoptosis -grade 3; >21/100 crypts (black arrows) and (E) frequent eosinophils and tangible body macrophages (red arrow) within the lamina propria. Patient 2 A 14-year-old girl with autoimmune hepatitis and autoimmune pneumonitis, which both affect ~40% of APECED patients (5,6), was treated with MMF and rituximab. Two weeks upon increasing the MMF dose from 500 to 1000 mg twice daily, she experienced diarrhea (5–6/day). Infection, fat malabsorption, exocrine pancreatic insufficiency, and SIBO were ruled out (Table 1). Colonoscopy showed areas of patchy erythematous mucosa throughout the colon. Histological examination demonstrated crypt architectural distortion, apoptotic bodies, and lymphoplasmacytic and tingible body macrophage infiltration within the lamina propria (Fig. 1B). MMF was discontinued with resolution of symptoms within 2 weeks. The mTOR inhibitor everolimus was initiated for the management of autoimmune hepatitis and autoimmune pneumonitis. Patient 3 A 10-year-old girl with tubulointerstitial nephritis was treated with MMF (500 mg twice daily). Eight months after initiation of MMF, she experienced vomiting, decreased food intake, and weight loss. EGD revealed esophageal ulceration with eosinophilic infiltration on histology, which prompted a colonoscopy to rule out inflammatory bowel disease. Histological examination revealed crypt architectural distortion, apoptotic bodies, and increased eosinophils in the terminal ileum and focally in the colon (Fig. 1C; Table 1). MMF was discontinued with resolution of symptoms within 1 week. Repeat colonoscopy 4 months later demonstrated colitis resolution. Azathioprine was initiated for the management of tubulointerstitial nephritis. Patient 4 A 52-year-old chronically ill man was started on MMF (1000 mg twice daily) for autoimmune pneumonitis with gradual titration of his dose over 4 months to 1500 mg twice daily for symptom relief. Approximately 4 months after titration, he began having 10 loose stools per day with tenesmus, which represented a clinical worsening of his baseline diarrhea (~4 stools/day) that was due to SIBO. Infection, fat malabsorption, and exocrine pancreatic insufficiency were ruled out with stool studies. A flexible sigmoidoscopy revealed multiple shallow 1–2 mm ulcers in the sigmoid colon and descending colon. Histological examination revealed active colitis with apoptotic bodies, eosinophils, and tingible body macrophages within the lamina propria; immunohistochemical staining for cytomegalovirus, Epstein-Barr virus, and adenovirus was negative (Table 1; Fig. 1D, E). MMF was discontinued with return of diarrheal symptoms to pre-MMF baseline levels and resolution of tenesmus. Azathioprine was initiated for the management of autoimmune pneumonitis. DISCUSSION To our knowledge, these are the first cases of biopsy-confirmed MMF-induced colitis reported in APECED. MMF is a commonly prescribed immunosuppressive drug for solid organ transplantation and, more recently, as a steroid sparing agent for autoimmune diseases such as systemic lupus erythematosus and systemic sclerosis (9–11). In diseases characterized by inflammation and tissue damage caused by dysregulated T and B lymphocytes, MMF treatment can be beneficial due to its ability to inhibit the proliferation and activation of these lymphoid cell populations (11). MMF may cause colitis associated with alteration of the gut microbiota by increasing B-glucuronidase expressing bacteria that can lead to increased concentrations of MPA in the colon leading to inflammation (12). Intestinal dysfunction is a common manifestation in APECED, with 66% of patients describing chronic constipation, diarrhea, or an alternating pattern of both with or without abdominal pain, cramping, bloating, distention, foul smelling flatulence, and/or greasing floating pale stools (4). In patients with APECED on immunosuppressants presenting with diarrhea, there are many differential diagnoses to consider including infection, SIBO, exocrine pancreatic insufficiency, autoimmune enteritis, and drug-induced colitis. Work up should include stool studies for culture, ova and parasite, fecal fat, calprotectin, and lactoferrin, as well as esophagogastroduodenoscopy with duodenal aspirate cultures and colonoscopy with biopsy. Among 104 APECED patients who we have evaluated at the NIH Clinical Center, MMF was given for various autoimmune end-organ disease manifestations in 10 (9.6%) patients. Among these patients, 4 (40%) developed reversible biopsy-proven MMF-induced colitis. This represents a frequency greater than that reported in other patient groups such as solid organ transplantation (9%) (8,13). Specifically, in that cohort of solid organ transplant recipients, 36 of 397 patients developed biopsy-confirmed MMF-induced colitis, which represents a frequency significantly lower than that observed in our APECED cohort (P value, 0.0012 by two-sided chi-square test; P value, 0.0109 by Fisher exact test). Histologic features of MMF-induced colitis can resemble those observed in graft-versus-host disease, inflammatory bowel disease, self-limiting colitis, or a combination of the 3 (8). Of note, of the 6 APECED patients in our cohort who received MMF but did not develop MMF-induced colitis, we found no distinguishing differences in AIRE gene mutations, APECED clinical manifestations, nor the duration or cumulative dosage of treatment with MMF compared to the 4 patients reported herein who did develop biopsy-proven colitis. Whether our patients with biopsy-proven MMF-induced colitis exhibited gut dysbiosis featuring B-glucuronidase expressing bacteria is unknown. Given that gut dysbiosis has been reported in APECED patients (6,14), examination of gut microbial communities in patients with versus without MMF-induced colitis represents an important direction of future research. In summary, MMF-induced colitis appears to be a frequent complication affecting patients who suffer from the monogenic autoimmune syndrome APECED, in whom AIRE deficiency-mediated autoimmune colitis is rare (1,3,4,7). Therefore, clinicians should maintain a high index of suspicion in recognizing this reversible drug complication in MMF-treated APECED patients who develop colitis and perform the appropriate diagnostic testing. More studies will be needed to confirm these results and to investigate the potential mechanisms—whether dysbiosis-related, immunological, or other—by which AIRE deficiency may heighten the risk for MMF-induced colitis. ACKNOWLEDGMENT Informed consent was obtained from all individual participants or their legal guardians included in this study to use their samples for research and to publish de-identified genetic sequencing data, in accordance with Helsinki principles for enrollment in research protocols that were approved by the Institutional Review Board of the National Institutes of Allergy and Infectious Diseases (NIAID). M.M.S. and M.S.L. conceived the study and wrote the paper. M.M.S., E.M.N.F., M.A.C., T.H., and M.S.L. provided clinical care to the patients. M.S.D.L. and M.Q. performed pathological analyses. All coauthors made edits in the article. This research was supported by the Division of Intramural Research of the NIAID, NCI and NIDDK, NIH. The authors report no conflicts of interest. The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials. Informed consent was obtained from all individual participants or their legal guardians included in this study to use their samples for research and to publish de-identified genetic sequencing data, in accordance with Helsinki principles for enrollment in research protocols that were approved by the Institutional Review Board of the National Institutes of Allergy and Infectious Disease. ==== Refs REFERENCES 1. Constantine GM Lionakis MS . Lessons from primary immunodeficiencies: Autoimmune regulator and autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy. Immunol Rev. 2019;287 :103–120.30565240 2. Oftedal BE Hellesen A Erichsen MM . Dominant mutations in the autoimmune regulator AIRE are associated with common organ-specific autoimmune diseases. Immunity. 2015;42 :1185–1196.26084028 3. Proekt I Miller CN Lionakis MS . Insights into immune tolerance from AIRE deficiency. Curr Opin Immunol. 2017;49 :71–78.29065385 4. Ferre EM Rose SR Rosenzweig SD . Redefined clinical features and diagnostic criteria in autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy. JCI Insight. 2016;1 :e88782.27588307 5. Ferré EMN Break TJ Burbelo PD . Lymphocyte-driven regional immunopathology in pneumonitis caused by impaired central immune tolerance. Sci Transl Med. 2019;11 :eaav5597.31167928 6. Chascsa DM Ferré EMN Hadjiyannis Y . APECED-associated hepatitis: clinical, biochemical, histological and treatment data from a large, predominantly American cohort. Hepatology. 2021;73 :1088–1104.32557834 7. Orlova EM Sozaeva LS Kareva MA . Expanding the phenotypic and genotypic landscape of autoimmune polyendocrine syndrome type 1. J Clin Endocrinol Metab. 2017;102 :3546–3556.28911151 8. Selbst MK Ahrens WA Robert ME . Spectrum of histologic changes in colonic biopsies in patients treated with mycophenolate mofetil. Mod Pathol. 2009;22 :737–743.19329937 9. Goyal A Salahuddin M Govil Y . A unique case of mycophenolate induced colitis after 10 years of use. Case Rep Gastrointest Med. 2016;2016 :3058407.27668102 10. Moroncini G Benfaremo D Mandolesi A . Mycophenolate mofetil-induced colitis in a patient with systemic sclerosis. BMJ Case Rep. 2018;2018 :bcr–2018. 11. Abd Rahman AN Tett SE Staatz CE . Clinical pharmacokinetics and pharmacodynamics of mycophenolate in patients with autoimmune disease. Clin Pharmacokinet. 2013;52 :303–331.23475567 12. Taylor MR Flannigan KL Rahim H . Vancomycin relieves mycophenolate mofetil-induced gastrointestinal toxicity by eliminating gut bacterial β-glucuronidase activity. Sci Adv. 2019;5 :eaax2358.31457102 13. Farooqi R Kamal A Burke C . Mycophenolate-induced colitis: a case report with focused review of literature. Cureus. 2020;12 :e6774.32117661 14. Petersen AØ Jokinen M Plichta DR . Cytokine-specific autoantibodies shape the gut microbiome in autoimmune polyendocrine syndrome type 1. J Allergy Clin Immunol. 2021;148 :876–888.33819509
PMC009xxxxxx/PMC9004539.txt	==== Front Curr Opin Environ Sci Health Curr Opin Environ Sci Health Current Opinion in Environmental Science & Health 2468-5844 Published by Elsevier B.V. S2468-5844(22)00009-5 10.1016/j.coesh.2022.100334 100334 Article Wastewater surveillance of SARS-CoV-2 in Bangladesh: Opportunities and challenges Haque Rehnuma 12∗ Moe Christine L. 3 Raj Suraja J. 3 Ong Li 4 Charles Katrina 4 Ross Allen G. 1 Shirin Tahmina 6 Raqib Rubhana 1 Sarker Protim 1 Rahman Mahbubur 1 Rahman Mohammed Ziaur 1 Amin Nuhu 1 Mahmud Zahid Hayat 1 Rahman Mahbubur 6 Johnston Dara 7 Akter Nargis 7 Khan Taqsem A. 8 Hossain Md. Alamgir 8 Hasan Rezaul 1 Islam M. Tahmidul 5 Bhattacharya Prosun 5 1 International Centre for Diarrhoeal Disease Research Bangladesh (ICDDRB), Dhaka, 1212, Bangladesh 2 School of Medicine, Stanford University, USA 3 Center for Global Safe Water, Sanitation, and Hygiene, Rollins School of Public Health, Emory University, USA 4 School of Geography and the Environment, University of Oxford, UK 5 COVID-19 Research @KTH, Department of Sustainable Development, Environmental Science and Engineering, KTH Royal Institute of Technology, Stockholm, Sweden 6 Institute of Epidemiology, Disease Control and Research (IEDCR), Bangladesh 7 Water, Sanitation & Hygiene (WASH) Section, UNICEF, Bangladesh 8 Dhaka Water Supply & Sewerage Authority (DWASA), Bangladesh ∗ Corresponding author: 18 2 2022 6 2022 18 2 2022 27 100334100334 © 2022 Published by Elsevier B.V. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. The ongoing pandemic of the coronavirus disease 2019 (COVID-19) is a public health crisis of global concern. The progression of the COVID-19 pandemic has been monitored in the first place by testing symptomatic individuals for SARS-CoV-2 virus in the respiratory samples. Concurrently, wastewater carries feces, urine, and sputum that potentially contains SARS-CoV-2 intact virus or partially damaged viral genetic materials excreted by infected individuals. This brings significant opportunities for understanding the infection dynamics by environmental surveillance. It has advantages for the country, especially in densely populated areas where individual clinical testing is difficult. However, there are several challenges including: 1) establishing a sampling plan and schedule that is representative of the various catchment populations 2) development and validation of standardized protocols for the laboratory analysis 3) understanding hydraulic flows and virus transport in complex wastewater drainage systems and 4) collaborative efforts from government agencies, NGOs, public health units and academia. Graphical abstract Image 1 Keywords Sanitation WASH Wastewater Sewage SARS-CoV-2 Low-income countries This review comes from a themed issue on Occupational Safety and Health 2022: COVID-19 in environment: Treatment, Infectivity, Monitoring, Estimation Edited by Manish Kumar, Ryo Honda, Prosun Bhattacharya, Dan Snow and Payal Mazumder ==== Body pmcIntroduction Wastewater surveillance As a result of the COVID-19 outbreak, wastewater surveillance has garnered considerable attention worldwide. Wastewater surveillance can be used to monitor wastewater released from individual household septic tanks to community drains, surface water, and sewage treatment plants, capturing the occurrence of SARS-CoV-2 and other enteric pathogens trends [1]. There is evidence that COVID-19 infected person shed SARS-CoV-2 viral RNA through the stool and urine before the symptoms appear and persist longer [2]. The presence of SARS-CoV-2 in the feces was first reported in early February 2020 by a number of investigators and demonstrated that about 48% of patients with COVID-19, irrespective of clinical severity or symptoms, have detectable SARS-CoV-2 virus RNA in their feces [3]. Notably, viral RNA was detected in fecal samples from patients even after their respiratory specimens tested negative. Other investigators have also reported longer persistence of SARS-CoV-2 RNA in the feces than in the respiratory and serum samples [2, 3, 4]. The viral load in the feces of persons testing positive for SARS-CoV-2 was estimated to be between 5 × 103-107.6 genome copies/mL, depending on the infection course [5]. Through statistical modeling, viral RNA concentrations in wastewater samples can predict symptomatic and asymptomatic COVID infections within a community [6]. Therefore, one of the motivations to use a wastewater surveillance system is to detect viral RNA and other enteric pathogens to understand the presence of infections in the community ahead of any outbreaks. Sanitation challenges in Bangladesh Approximately 72% of people in Asia do not have access to appropriate sanitation and in South Asia alone over 610 million people still practice open defecation [7]. Since the independence of Bangladesh in 1971, the country has made significant progress in providing access to sanitation. Between 2000 and 2015, Bangladesh took concerted actions to become an open defecation free (ODF) country, and as of 2018, open defecation in Bangladesh has been virtually non-existent [8]. However, Bangladesh faces considerable socio-ecological challenges for safe disposal of fecal waste generated in urban areas. Dhaka, the capital of Bangladesh, has the highest population density in the country with 47,400 people/square kilometer. Approximately 6 million people reside in urban slums in Dhaka, and an estimated population of 4.3 million people use shared sanitation facilities [9,10]. With limited sewage treatment, the management of fecal sludge has become a major challenge. There is only one wastewater treatment plant (WWTP) in Dhaka that is serves 20 percent of the population (∼5 million people) of Dhaka city [9]. The WWTP is connected to 935 km of sewer networks with 26 sewage lift stations and 4 storm sewage stations [9]. According to the WHO/UNICEF report [8], an estimated 80% of the neighborhoods in Dhaka city are located outside the coverage of this sewerage network. Furthermore, 97% of the fecal waste in Dhaka reaches the environment untreated, with 71% of households using pit and septic tank that discharges untreated fecal waste directly or indirectly into an open drains or ditches [11, 12, 13]. Our recent findings showed that the performance of on-site sanitation (OSS) was not adequate to remove pathogens of public health significance and failure to ensure strong links throughout the fecal sludge management (FSM) service chain resulted in untreated sewage contaminating the environment and surface water [14]. Frequent flooding in areas with poor drainage systems has far-reaching deleterious impacts on human health [15]. Similarly, groundwater sources in the city are impacted by fecal contamination and there is potential for viral contamination during groundwater recharge, which increases the risk of waterborne infections. This situation mirrors Bangladesh's rural areas, where OSS is the primary sanitation option [12]. Wastewater surveillance with low sanitation coverage Most studies of SARS-CoV-2 in wastewater are in high-income countries where sanitation systems are well-structured and population catchment areas are specific (Table 1 ) [16∗, 17∗∗, 18, 19, 20, 21, 22, 23]. Bangladesh, and most low- and middle-income countries, have mixed sanitation systems, with subsequent varied characteristics of sewage-sludge. Therefore, the approaches used in high-income countries will need to be adapted for shared and private latrines, on-site sanitation, and direct dumping of fecal waste into the environment. Nevertheless, two studies from Bangladesh can support the successful identification and quantification of SARS-CoV-2 viral RNA from the wastewater collected from drains, surface water, and effluent of the septic tank samples [18,24]. Data are still limited from Bangladesh to implement wastewater surveillance at the national level. Wastewater monitoring output should present integrated rather individual research outcomes. Also, the impact of climatic variability on wastewater-based surveillance should be explored in sub-tropical environments because rainfall, humidity, and temperature may affect the viability, degradation, mutation and detection of the virus in the environment. There is evidence of SARS-CoV-2 mobility in the subsurface water and possible leaching into the groundwater [25]. Developing a common platform from the different environmental data banks would be helpful to compare, sharing each other's expertise, wastewater monitoring tools, and standard method to replicate the wastewater surveillance even in a country with inadequate sanitation facilities [26].Table 1 CoV-2 viral RNA in wastewater surveillance. Table 1Country Sources No of samples Method Kit used for RT-PCR Target gene References India WWTP1,2,3 aeration pond <20 qPCR TaqPathTM Covid-19 RT-PCR Kit N, S, ORF1 ab [16] Australia Pumping station3, WWTP3 4–5 qPCR iTaq™ Universal Probes One–Step Reaction Mix N protein [17] France WWTP4,6 31 qPCR/RNA – RdRP, E protein [19] Spain WWTP1,2 72 qPCR One Step PrimeScript™ RT-PCR Kit N protein [21] USA WWTP1 10 RT-PCR qPCR Genome sequencing – S and N protein [37] Italy River, WWTP4,6 – qPCR, Genome sequencing, cell culture 2019-nCoV Real-Time RT-PCR Diagnostic Panel ORF1 ab, N and E protein [20] Netherlands WWTP1 29 qPCR EvoScript RNA Probes Master E and N protein [23] Italy WWTP1 12 RT-PCR, qPCR Kit Platinum™ SuperFi™ Green PCR Master Mix, Thermo S protein, ORF1 aba RdRP [38] Japan River, WWTP1 13 Nested PCR qPCR, (IDEXX) Premix Ex Taq Hot Start Version ORF1ab, S and E protein, E. colia [22] Israel Raw sewage WWTP 26 qPCR StepOnePlus™ Real-Time PCR System – [39] Bangladesh Drain, Canal, Sewer 16 RT-PCR (CFX96, BioRad) Sansure RT-PCR kit ORF1ab, N genes, and RNase [18] a 1 = influent; 2 = effluent; 3 = sludge; 4 = raw sewage; 5 = Pre-treated; 6 = treated ∗ “-” not available strong evidence. A summary of potential opportunities and challenges of implementing wastewater surveillance The purpose of this present paper is to encourage researchers, governments, local and international development organizations, and practitioners to use wastewater surveillance as a low-cost approach for better understanding COVID-19 and other enteric pathogens (Figure 1 ).• Sampling points: In countries like Bangladesh, it is challenging to develop a city-wide wastewater surveillance strategy and identify sample collection sites that represent population-level feces in communities where the COVID-19 burden is largely unknown due to lack of diagnostic testing centers. Designing city-wide wastewater surveillance will need to include sampling from the wastewater treatment plant (WWTP) and pumping stations for the parts of the city that have a sewerage network and sample sites at major drains and canals that capture the wastewater from septic tanks and pit latrines [27]. Understanding hydraulic flows and virus transport in complex wastewater drainage systems in Dhaka city including a mix of sewered and non-sewered areas, is challenging [28]. The present paper posits that wastewaters both from sewer network and on-site sanitation facilities receive and harbor SARS-CoV-2 from various sources in a community or catchment. The sewersheds and catchment areas can be estimated using GIS tools and sewerage network maps. Examining drainage and flow patterns such as manholes or up and downstream drains will lead to the identification of representative sampling points for periodic sampling in the next step. • Sampling technique: Wastewater surveillance requires significant logistical and financial support in dispersed communities. Traditional sampling methods, such as autosamplers, were a decent choice in affluent countries. However, not suitable for large rapid monitoring at these low-resources settings. Therefore, Passive samplers can aid with low-cost sample collecting. Passive sampling presents a cheap, safe, and easy alternative to traditional wastewater sampling within the sewage catchment for wastewater monitoring. The installation of the passive sampler in waterbodies is simple (i.e., no specific skills are necessary), quick, and usually does not necessitate space limitations. Also, it is continuously exposed to the water column, therefore the sampling error rate is low even when single water samples are taken [29]. • Laboratory methods optimization: Sample preparation and analysis methods need to be developed and validated for treated and untreated wastewater. Accredited laboratories with skilled personnel and high biosafety protocols are needed to implement this method. Because of the funding constraint, the laboratory method also needs to be cost-effective, at the same time sensitive and reliable. Loop-Mediated Isothermal Amplification (LAMP) method can be suitable in this context. LAMP reactions are capable of detecting even a few copies of target nucleic acid sequences under isothermal conditions (usually 60–65 °C) with the help of specially designed primer sets [30,31]. Due to its simplicity, it is gaining more popularity in the diagnosis of various viral diseases. Unlike PCR, this technique can be performed in a low-resource setting by merely heating the samples and reagents in a single reaction tube. Currently, Bangladesh has developed about 118 PCR laboratories since the pandemic hit in the country [32]. Once a valid method is optimized, there are opportunities to roll over the wastewater surveillance all over the country [33]. As a result, many researchers are concentrating on improving and modifying the RT-LAMP methodology to meet the needs of COVID-19 diagnosis in laboratories with limited resources in underdeveloped nations, where uncontrolled and undetected COVID-19 spread can have unforeseeable consequences. Furthermore, mobile testing units and rapid, low-cost sensors need to be developed to support the surveillance system for any future outbreak. • End-user: The necessity for low-cost, quick monitoring of COVID-19 prevalence and trends has long been acknowledged. One example is environmental surveillance, which includes the proposed wastewater monitoring for determining COVID-19 frequency in the community. The Directorate General of Health Services (DGHS) is the key player and leads the national technical committee that manages, implements, guides, and supports COVID-19 management in Bangladesh. The technical committee at DGHS endorses all research related to COVID-19 and uses the research findings for implementation and incorporation in the national guidelines. The DGHS plays a key role in the implementation of public health interventions [32]. Sharing the results of the environmental surveillance in near real-time with health authorities, including outbreak investigators, epidemiologists, public health engineers, and sanitary inspectors, to enable a more targeted public health response to current outbreaks and endemic disease. An online platform for sharing the weekly/biweekly/monthly wastewater surveillance findings, including SARS-CoV-2 temporal trends, spatial trends, and detection of variants of concern, with designated public health entities, will support decision making prior to outbreak take place. Currently, the COVID-19 prevalence rate is determined by the number of clinical tests that are positive every day and the COVID-19 case fatality rate. By establishing wastewater surveillance, it will be able to predict trends in COVID-19 prevalence 2–6 weeks ahead of changes in healthcare utilization and mortality associated with COVID-19 [34,35]. This dashboard will help key partners to make critical public health response decisions based on viral infection dynamics reflected by trends in the environmental surveillance data. • Implementation with collaborative effort: Departments from the Ministry of Health and Family Welfare (MoH&FW) and Ministry of Local Government Rural Development and Cooperatives (MoLGRD&C) are the key partners from the government; Dhaka Water Supply and Sewerage Authority (DWASA) is responsible for water supply, sanitation and stormwater disposal services in the capital, and serves around 8.6 million of the 11 million people in the Dhaka Metropolitan Area (DMA). Support from DWASA is essential to explore the diversity of the sanitation system in Dhaka City and establish wastewater surveillance; Dhaka North City Corporation (DNCC) and Dhaka South City Corporation (DSCC) are responsible for the management of drainage canal and system, and maintenance of the Sewerage Lifting Station (SLS) and The Institute of Epidemiology, Disease Control, and Research (IEDCR) is responsible for conducting communicable disease outbreak and surveillance in Bangladesh [36]. The Institute also contributes research on communicable diseases as well as the functioning of disease control programs mainly in the form of parasitic and entomological containment of vector-borne diseases, emerging and re-emerging diseases, and their response through the application of epidemiological studies. Their engagement and collaboration are crucial to implementing and scaling up Bangladesh's environmental surveillance system. Figure 1 Wastewater surveillance proposed implementation steps in Bangladesh. Figure 1 Concluding remarks Wastewater monitoring is a noninvasive, sensitive, and more cost-effective approach to adopt in low-to middle-income countries (LMICs) to examine changes in total COVID-19 burden and other waterborne diseases in community and can inform public health decision making for responding to the current COVID-19 pandemic as well as future health threats. The COVID-19 pandemic highlighted that monitoring wastewater and other relevant environmental samples for SARS-CoV-2 and other pathogens provides a sensitive signal of the presence of the pathogens in entire communities and can also indicate whether infection rates are increasing or declining. Data from the surveillance can be used as a valuable complement to epidemiological case data. Where no such epidemiological data are available, surveillance for the presence and concentration of pathogens in wastewater can be a very effective noninvasive public health tool in low-resource countries like Bangladesh. Editorial disclosure statements Given their role as Guest Editor, Prosun Bhattacharya had no involvement in the peer-review of this article and has no access to information regarding its peer-review. Full responsibility for the editorial process for this article was delegated to Payal Mazumder. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements Dr. Rehnuma Haque is supported by the Global Health Equity Scholars Program NIH FIC and NIEHS D43 TW010540. The authors thank to UNICEF/SDC and the Life science technology platform, Science for Life Laboratory for the funding to initiate the SARS-CoV-2 WBE project at icddr, b, Bangladesh. icddr, b gratefully acknowledge our core donors including the Governments of Bangladesh, Canada, Sweden, and the UK, for their support and commitment to icddr, b research efforts. ==== Refs References Meng X. Wang X. Meng S. Wang Y. Liu H. Liang D. Fan W. Min H. Huang W. Chen A. A global overview of SARS-CoV-2 in wastewater: detection, treatment, and prevention ACS EST Water 1 2021 2174 2185 This study demonstrated the significant global overview of implementing SARS-CoV-2 wastewater surveillance. 2 Wu Y. Guo C. Tang L. Hong Z. Zhou J. Dong X. Yin H. Xiao Q. Tang Y. Qu X. Prolonged presence of SARS-CoV-2 viral RNA in faecal samples Lancet Gastroenterol Hepatol 5 2020 434 435 32199469 3 Cheung K.S. Hung I.F.N. Chan P.P.Y. Lung K.C. Tso E. Liu R. Ng Y.Y. Chu M.Y. Chung T.W.H. Tam A.R. Gastrointestinal manifestations of SARS-CoV-2 infection and virus load in fecal samples from a Hong Kong cohort: systematic review and meta-analysis Gastroenterology 159 2020 81 95 32251668 This meta-analysis detected SARS-CoV-2 genetic materials in stool from 48.1% of patients, even after respiratory samples were found negative thus, concluded fecal shedding is longer than the respiratory samples. 4 Zheng S. Fan J. Yu F. Feng B. Lou B. Zou Q. Xie G. Lin S. Wang R. Yang X. Viral load dynamics and disease severity in patients infected with SARS-CoV-2 in Zhejiang province, China, January-March 2020: retrospective cohort study BMJ 369 2020 m1443 32317267 5 Foladori P. Cutrupi F. Segata N. Manara S. Pinto F. Malpei F. Bruni L. La Rosa G. SARS-CoV-2 from faeces to wastewater treatment: what do we know? A review Sci Total Environ 743 2020 140444 32649988 This article estimates the fecal shedding from a COVID-19 infected patient is 5•103–107.6 copies/mL and reduced at least 2 copies before entering the WWTP. 6 Vallejo J.A. Trigo N. Rumbo-Feal S. Conde-Pérez K. Lopez-Oriona Á. Barbeito I. Vaamonde M. Tarrío-Saavedra J. Reif R. Ladra S. Modeling the number of people infected with SARS-COV-2 from wastewater viral load in Northwest Spain Sci Total Environ 2021 10.1016/j.scitotenv.2021.152334 7 A snapshot of sanitation, hygiene and drinking water safety in South Asia 2019 8 Progress on drinking water, sanitation and hygiene 2000–2017 2019 UNICEF 9 Bhaban W. Avenue K.N.I. Bazar K. Dhaka water supply and sewerage authority. Annual report 2018-19 2020 10 Yeasmin F. Rahman M. Luby S.P. Das J.B. Begum F. Saxton R.E. Nizame F.A. Hwang S.T. Alam M.-U. Hossain M.K. Landlords' and compound managers' role in improving and sustaining shared latrines in three Dhaka city slums Water 12 2020 2073 11 Ross A.G. Rahman M. Alam M. Zaman K. Qadri F. Can we “WaSH” infectious diseases out of slums? Int J Infect Dis IJID Off Publ Int Soc Infect Dis 92 2020 130 132 This opinion paper highlighted current sanitation situation of Dhaka City slum areas. 12 WHO/UNICEF joint monitoring programme for water supply and sanitation: Progress on sanitation and drinking-water: 2010 update 2010 World Health Organization 13 Bangladesh – water & sanitation for the urban poor 2017 14 Amin N. Liu P. Foster T. Rahman M. Miah M.R. Ahmed G.B. Kabir M. Raj S. Moe C.L. Willetts J. Pathogen flows from on-site sanitation systems in low-income urban neighborhoods, Dhaka: a quantitative environmental assessment Int J Hyg Environ Health 230 2020 113619 32942223 15 Amin N. Rahman M. Raj S. Ali S. Green J. Das S. Doza S. Mondol M.H. Wang Y. Islam M.A. Quantitative assessment of fecal contamination in multiple environmental sample types in urban communities in Dhaka, Bangladesh using SaniPath microbial approach PLoS One 14 2019 e0221193 This study highlighted wastewater fromdrain water, sediment, canal water, floodwater, sludge, supernatant, and effluent samples from septic tanks, and ABRs identified high concentrations of enteric pathogens, particularly NoV-GII, V.cholerae, and Shigella in Dhaka, Bangladesh. Kumar M. Patel A.K. Shah A.V. Raval J. Rajpara N. Joshi M. Joshi C.G. First proof of the capability of wastewater surveillance for COVID-19 in India through detection of genetic material of SARS-CoV-2 Sci Total Environ 746 2020 141326 32768790 This article from India that shares a similar sanitation system like 274 Bangladesh confirmed the presenceof wastewater surveillance for the genetic materials of SARS-CoV-2 in municipal wastewater. Ahmed W. Angel N. Edson J. Bibby K. Bivins A. O'Brien J.W. Choi P.M. Kitajima M. Simpson S.L. Li J. First confirmed detection of SARS-CoV-2 in untreated wastewater in Australia: a proof of concept for the wastewater surveillance of COVID-19 in the community Sci Total Environ 728 2020 138764 32387778 This paper is published earlier in 2020 as the first proof of concept and successful detection of SARS-CoV-2 genetic materials in the wastewater through wastewater-based epidemiology (WBE) surveillance in the community at Brisbane, Australia. 18 Ahmed F. Islam M.A. Kumar M. Hossain M. Bhattacharya P. Islam M.T. Hossen F. Hossain M.S. Islam M.S. Uddin M.M. First detection of SARS-CoV-2 genetic material in the vicinity of COVID-19 isolation Centre in Bangladesh: variation along the sewer network Sci Total Environ 776 2021 145724 33652314 19 Wurtzer S. Marechal V. Mouchel J.M. Maday Y. Teyssou R. Richard E. Almayrac J.L. Moulin L. Evaluation of lockdown impact on SARS-CoV-2 dynamics through viral genome quantification in Paris wastewaters 2020 20 Rimoldi S.G. Stefani F. Gigantiello A. Polesello S. Comandatore F. Mileto D. Maresca M. Longobardi C. Mancon A. Romeri F. Presence and infectivity of SARS-CoV-2 virus in wastewaters and rivers Sci Total Environ 744 2020 140911 32693284 21 Randazzo W. Truchado P. Cuevas-Ferrando E. Simón P. Allende A. Sánchez G. SARS-CoV-2 RNA in wastewater anticipated COVID-19 occurrence in a low prevalence area Water Res 181 2020 115942 32425251 22 Haramoto E. Malla B. Thakali O. Kitajima M. First environmental surveillance for the presence of SARS-CoV-2 RNA in wastewater and river water in Japan Sci Total Environ 737 2020 140405 32783878 23 Medema G. Heijnen L. Elsinga G. Italiaander R. Brouwer A. Presence of SARS-coronavirus-2 RNA in sewage and correlation with reported COVID-19 prevalence in the early stage of the epidemic in The Netherlands Environ Sci Technol Lett 2020 10.1021/acs.estlett.0c00357 24 Haque Rehnuma Rahman Md Mahbubur Amin Nuhu Rahman Mohammed Ziaur Rahman Mahmud Zahid Hayat Sarker Protim Raqib Rubhana Inference and forecasting of SARS-CoV-2 wastewater surveillance in Bangladesh CUGH 2021 virtual conference 2021 CUGH 25 Kumar M. Thakur A.K. Mazumder P. Kuroda K. Mohapatra S. Rinklebe J. Ramanathan Al Cetecioglu Z. Jain S. Tyagi V.K. Frontier review on the propensity and repercussion of SARS-CoV-2 migration to aquatic environment J Hazard Mater Lett 1 2020 100001 34977840 26 Bhattacharya P. Kumar M. Islam MdT. Haque R. Chakraborty S. Ahmad A. Niazi N.K. Cetecioglu Z. Nilsson D. Ijumulana J. Prevalence of SARS-CoV-2 in communities through wastewater surveillance—a potential approach for estimation of disease burden Curr Pollut Rep 2021 10.1007/s40726-021-00178-4 This opinion paper emphasized the importance of wastewater-based epidemiology surveillance for SARS-CoV-2 in the LMIC and prioritized the low sanitation areas, proposing standard international tools for surveillance methods and data interpretation. 27 Wang Y. Mairinger W. Raj S.J. Yakubu H. Siesel C. Green J. Durry S. Joseph G. Rahman M. Amin N. Quantitative assessment of exposure to fecal contamination in urban environment across nine cities in low-income and lower-middle-income countries and a city in the United States Sci Total Environ 806 2022 151273 34718001 28 Foster T. Falletta J. Amin N. Rahman M. Liu P. Raj S. Mills F. Petterson S. Norman G. Moe C. Modelling faecal pathogen flows and health risks in urban Bangladesh: implications for sanitation decision making Int J Hyg Environ Health 233 2021 113669 33578186 This paper described about city-wide wastewater sampling modelling technique in the low sanitation areas among Africa, Asia and Atlanta. 29 Schang C. Crosbie N.D. Nolan M. Poon R. Wang M. Jex A. John N. Baker L. Scales P. Schmidt J. Passive sampling of SARS-CoV-2 for wastewater surveillance Environ Sci Technol 55 2021 10432 10441 34264643 30 Augustine R. Hasan A. Das S. Ahmed R. Mori Y. Notomi T. Kevadiya B.D. S. Thakor A. Loop-mediated isothermal amplification (LAMP): a rapid, sensitive, specific, and cost-effective point-of-care test for coronaviruses in the context of COVID-19 pandemic Biology 9 2020 182 32707972 This article demonstrated losw cost point of care test for COVID-19 by using LAMP method. 31 Kaya D. Niemeier D. Ahmed W. Kjellerup B.V. Evaluation of multiple analytical methods for SARS-CoV-2 surveillance in wastewater samples Sci Total Environ 808 2022 152033 34883175 32 Anwar S. Nasrullah M. Hosen M.J. COVID-19 and Bangladesh: challenges and how to address them Front Public Health 8 2020 154 32426318 33 Ahmed W. Simpson S.L. Bertsch P.M. Bibby K. Bivins A. Blackall L.L. Bofill-Mas S. Bosch A. Brandão J. Choi P.M. Minimizing errors in RT-PCR detection and quantification of SARS-CoV-2 RNA for wastewater surveillance Sci Total Environ 805 2022 149877 34818780 This review paper drew the technical factors of minimizing laboratory methodological errors and ensuring quality control while conducting SARS-CoV-2 wastewater surveillance. 34 Kirby A.E. Using wastewater surveillance data to support the COVID-19 response — United States, 2020–2021 MMWR Morb Mortal Wkly Rep 70 2021 35 Larsen DA. Wigginton KR. Tracking COVID-19 with wastewater Nat Biotechnol 38 2020 1151 1153 32958959 36 IEDCR https://iedcr.gov.bd/ 37 Holshue M.L. DeBolt C. Lindquist S. Lofy K.H. Wiesman J. Bruce H. Spitters C. Ericson K. Wilkerson S. Tural A. First case of 2019 novel coronavirus in the United States N Engl J Med 382 2020 929 936 32004427 38 La Rosa G. Iaconelli M. Mancini P. Bonanno Ferraro G. Veneri C. Bonadonna L. Lucentini L. Suffredini E. First detection of SARS-CoV-2 in untreated wastewaters in Italy Sci Total Environ 736 2020 139652 32464333 39 Bar-Or I. Weil M. Indenbaum V. Bucris E. Bar-Ilan D. Elul M. Levi N. Aguvaev I. Cohen Z. Shirazi R. Detection of SARS-CoV-2 variants by genomic analysis of wastewater samples in Israel Sci Total Environ 789 2021 148002 34323811
PMC009xxxxxx/PMC9004614.txt	==== Front J Therm Spray Tech Journal of Thermal Spray Technology 1059-9630 1544-1016 Springer US New York 1396 10.1007/s11666-022-01396-y Editorial Special Issues of the Journal of Thermal Spray Technology 12 4 2022 2022 31 4 667671 © ASM International 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. issue-copyright-statement© ASM International 2022 ==== Body pmcThe Journal of Thermal Spray Technology has been honored to work with many guest editors during the past three decades. These professionals have volunteered their time and expertise to provide special issue content to the journal. These special issues, containing either topical articles or papers developed from conference presentations, are a valuable asset. The journal thanks the guest editors for their important work in bringing this relevant and significant content to JTST. Formation of Plasma Sprayed Coatings Pierre Fauchais, Guest Editor March 1995, Vol. 4, No. 1 Advanced Evolution of Thermally Sprayed Ceramic Coatings Akira Ohmori, Guest Editor June 1996, Vol. 5, No. 2 September 1996, Vol. 5, No. 3 Thermal Barrier Coatings of the Future William J. Brindley, Guest Editor March 1997, Vol. 6, No. 1 Thermal Spray in the Czech Republic Pavel Chráska, Guest Editor September 1997, Vol. 6, No. 3 December 1997, Vol. 6, No. 4 Thin-Film Technology versus Thermal Spraying: Toward Peaceful Coexistence Lech Pawlowski, Guest Editor June 1999, Vol. 8 No. 2 Selected Papers from the 15th International Thermal Spray Conference (ITSC 1998) Christian Coddet, Guest Editor September 1999, Vol. 8, No. 3 Materials Challenges in Developing Advanced Thermal Barrier Coatings J. Allen Haynes, Guest Editor March 2000, Vol. 9, No. 1 June 2000, Vol. 9, No. 2 Solid Oxide Fuel Cells R. Henne, T. Yoshida, and J. Heberlein, Guest Editors September 2000, Vol. 9, No. 3 Progress on Diagnostics and Instrumentation in the Thermal Spray Processes James R. Fincke, Guest Editor March 2001, Vol. 10, No. 1 June 2001, Vol. 10, No. 2 Selected Papers in Thermal Spray Christian Moreau, Guest Editor September 2001, Vol. 10, No. 3 December 2001, Vol. 10, No. 4 Ceramic Thermal and Environmental Barrier Coatings Katherine T. Faber, Robert Vassen, and Dongming Zhu, Guest Editors March 2004, Vol. 13, No. 1 September 2004, Vol. 13, No. 3 Selected Papers from the International Thermal Spray Conference (ITSC 2003) Christian Moreau, Editor March 2005, Vol. 14, No. 1 Selected Papers from the International Thermal Spray Conference (ITSC 2004) Erich Lugscheider and Yoshiki Tsunekawa, Guest Editors December 2005, Vol. 14, No. 4 Selected Papers from the International Thermal Spray Conference (ITSC 2005) Erich Lugscheider, Guest Editor June 2006, Vol. 15, No. 2 Special Issue: Proceedings of the 2006 International Thermal Spray Conference Basil Marple, Margaret Hyland, Yuk-Chiu Lau, Rogerio Lima, and Joël Voyer, Guest Editors December 2006, Vol. 15, No. 4 Special Issue on Coatings Used under Severe Conditions Armelle Vardelle and Seiji Kuroda, Guest Editors March 2007, Vol. 16, No. 1 The Laser—A Junior Member in the Family of Surface Technology Tools Steffen Nowotny and Thomas Schuelke, Guest Editors September 2007, Vol. 16, No. 3 Thermal Spraying in Europe’s Nordic Region Nicolaie Markocsan, Per Nylen, Erja Turunen, and Petri Vuoisto, Guest Editors December 2007, Vol. 16, No. 4 Special Issue: Selected and Expanded Papers from the 2007 International Thermal Spray Conference Basil Marple, Margaret Hyland, Yuk-Chiu Lau, Chang-Jui Li, Rogerio Lima, and Ghislain Montavon, Guest Editors Mid-December 2007, Vol. 16, No. 5–6 Special Issue on Solution/Suspension Thermal Spray Pierre Fauchais, Guest Editor March 2008, Vol. 17, No. 1 Selected Papers from Euromat 2007 Christian Coddet, Guest Editor September 2008, Vol. 17, No. 3 Special Issue: Selected and Expanded Papers from the 2008 International Thermal Spray Conference Basil R. Marple, Margaret Hyland, Yuk-Chiu Lau, Chang-Jui Li, Rogerio Lima, and Ghislain Montavon, Guest Editors Mid-December 2008, Vol. 17, No. 5–6 Special Issue: Recent Advances in Thermal Barrier Coatings (TBCs) Robert Vassen, Per Nylen, and Detlev Stöver, Guest Editors June 2009, Vol. 18, No. 2 Special Issue: Recent Advances in Modeling and Numerical Simulation of Thermal Spray Processes Joachim Heberlein and Armelle Vardelle, Guest Editors Mid-December 2009, Vol. 18, No. 5–6 Special Issue: 3rd Asian Thermal Spray Conference (ATSC-3) K.A. Khor, Seiji Kuroda, Changhee Lee, and You Wang, Guest Editors December 2009, Vol. 18, No. 4 Special Issue: Selected and Expanded Papers from the 2009 International Thermal Spray Conference Basil R. Marple, Margaret M. Hyland, Yuk-Chiu Lau. Chang-Jui Li, Rogerio S. Lima, and Ghislain Montavon, Guest Editors January 2010, Vol. 19, No. 1–2 Special Issue: Reliability and Consistency in Thermal Spray Klaus Landes and Basil R. Marple, Guest Editors June 2010, Vol. 19, No. 4 Special Issue: 4th Asian Thermal Spray Conference (ATSC-4) Chang-Liu Li, Seiji Kuroda, Masahiro Fukumoto, Khiam Aik Khor, Changhee Lee, and You Wang, Guest Editors December 2010, Vol. 19, No. 6 Special Issue: Selected and Expanded Papers from the 2010 International Thermal Spray Conference Basil R. Marple, Arvind Agarwal, Margaret M. Hyland, Yuk-Chui Lau, Chang-Jui Li, Rogerio S. Lima, and Ghislain Montavon, Guest Editors January 2011, Vol. 20 No. 1–2 Special Issue: Emerging and Innovative Processes in Thermal Spraying Armelle Vardelle and Robert Vassen, Guest Editors June 2011, Vol. 20, No. 4 Special Issue: Selected and Expanded Papers from the 2011 International Thermal Spray Conference Basil R. Marple, Arvind Agarwal, Margaret M. Hyland, Yuk-Chiu Lau, Chang-Jiu Li, Rogerio S. Lima, and André McDonald, Guest Editors June 2012, Vol. 21 No. 3–4 Special Issue: 5th International Workshop on Suspension and Solution Thermal Spraying (S2TS) 2011 Erick Meillot, Guest Editor December 2012, Vol. 21, No. 6 Special Issue: Selected and Expanded Papers from the 2012 International Thermal Spray Conference Basil R. Marple, Arvind Agarwal, Margaret M. Hyland, Yuk-Chiu Lau, Chang-Jiu Li, Rogerio S. Lima, André McDonald, and Filofteia-Laura Toma, Guest Editors March 2013, Vol. 22 No. 2–3 Special Issue: Coatings for Energy Applications Armelle Vardelle and Robert Vassen, Guest Editors June 2013, Vol. 22, No. 5 Special Issue: 5th Asian Thermal Spray Conference (ATSC-5) Kazuhiro Ogawa, Hiroshi Katanoda, Chang-Jiu Li, Changhee Lee, Khiam Aik Khor, Margaret Hyland, G. Sundararajan, and Seiji Kuroda, Guest Editors December 2013, Vol. 22, No. 8 Special Issue: Selected and Expanded Papers from the 2013 International Thermal Spray Conference Arvind Agarwal, Margaret M. Hyland, Yuk-Chiu Lau, Georg Mauer, Rogerio S. Lima, André McDonald, and Filofteia-Laura Toma, Guest Editors January 2014, Vol. 23, No. 1–2 Special Issue: Development and Applications of Nanocomposite Coatings Rehan Ahmed and Christopher C. Berndt, Guest Editors October 2014, Vol. 23, No. 7 Special Issue: Selected and Expanded Papers from the 2014 International Thermal Spray Conference Arvind Agarwal, Giovanni Bolelli, Yuk-Chiu Lau, Rogerio S. Lima, André McDonald, Filofteia-Laura Toma, and Erja Turunen, Guest Editors January 2015, Vol. 24, No. 1–2 Special Issue: Suspension and Solution Thermal Spraying Pierre Fauchais, Guest Editor October 15, Vol. 24, No. 7 Special Issue: 6th Asian Thermal Spray Conference (ATSC-6) Shrikant Joshi, Masahiro Fukumoto, Michael Khor, Changhee Lee, Hua Li, and Dheepa Srinivasan, Guest Editors December 2015, Vol. 24, No. 8 Special Issue: Selected and Expanded Papers from the 2015 International Conference (ITSC 2015) Kantesh Balani, Giovanni Bolelli, Yuk-Chiu Lau, André McDonald, Filofteia-Laura Toma, Erja Turunen, and Christian Widener, Guest Editors January 2016, Vol. 25, No. 1–2 Special Issue: 7th Asian Thermal Spray Conference (ATSC-7) Seiji Kuroda and Chang-Jiu Li, Guest Editors December 2016, Vol. 15 No. 8 Special Issue: Selected and Expanded Papers from the 2016 International Thermal Spray Conference (ITSC 2016) Fardad Azarmi, Giovanni Bolelli, Tanvir Hussain, Yuk-Chiu Lau, Hua Li, Jon Longtin, and Filofteia-Laura Toma, Guest Editors January 2017, Vol. 26, No. 1–2 Special Issue: Metal Additive Manufacturing Mathieu Brochu, Jean-Yves Hascoet, Bertrand Jodoin, and Todd Palmer, Guest Editors April 2017, Vol. 26, No. 4 Special Issue: Next Generation Coatings for Gas Turbines Y.C. Lau, Li Li, Robert Vassen, and Mitch Dorfman, Guest Editors August 2017, Vol. 26, No. 6 Special Issue: Cold Spray Jean-Gabriel Legoux, Amadeu Concustell, Michel Jeandin, Thomas Klassen, Heli Koivuluoto, and Julio Villafuerte, Guest Editors October 2017, Vol. 26, No. 7 Special Issue: Selected and Expanded Papers from the 2017 International Thermal Spray Conference (ITSC 2017) Fardad Arzami, Kantesh Balani, Tanvir Hussain, Yuk-Chiu Lau, Hua Li, Kentaro Shinoda, and Filofteia-Laura Toma, Guest Editors January 2018, Vol. 27, No. 1–2 Special Issue: Biomaterials: Thermal Spray Processes and Applications Andrew Ang, Rehan Ahmed, and Christopher C. Berndt, Guest Editors December 2018, Vol. 27, No. 8 Special Issue: Selected and Expanded Papers from the 2018 International Thermal Spray Conference (ITSC 2018) Fardad Azarmi, Jan Cizek, Heli Koivuluoto, Yuk-Chiu Lau, Hua Li, and Filofteia-Laura Toma, Guest Editors January 2019, Vol. 28, No. 1–2 Special Issue: Thermal Spray Applications for Extreme Environments Andrew Ang, Heidi deVilliers-Lovelock, Steven Matthews, Buta Singh Sidhu, and Harpreet Singh, Guest Editors October 2019, Vol. 28, No. 7 Special Issue: Selected and Expanded Papers from the 2019 International Thermal Spray Conference (ITSC 2019) Fardad Azarmi, Kantesh Balani, Jan Cizek, Heli Koivuluoto, Hua Li, and Filofteia-Laura Toma, Guest Editors January 2020, Vol. 29, No. 1–2 Special Issue: Advanced Residual Stress Analysis in Thermal Spray and Cold Spray Processes Andrew Ang, Seiji Kuroda, Vladimir Luzin, and Shuo Yin, Guest Editors August 2020, Vol. 29, No. 6 Special Focus: Selected and Expanded Papers Based on Abstracts Submitted for the 2020 International Thermal Spray Conference (ITSC 2020) (Event Canceled Due to the COVID-19 Pandemic) Fardad Azarmi, Kantesh Balani, Jan Cizek, Heli Koivuluoto, Yuk-Chiu Lau, Hua Li, and Filofteia-Laura Toma, Guest Editors January 2021, Vol. 30, No. 1–2 Special Issue: Aerosol Deposition and Kinetic Spray Processes Kentaro Shinoda, Frank Gaertner, Changhee Lee, Ali Dolatabadi, and Scooter Johnson, Guest Editors February 2021, Vol. 30, No. 3 Special Issue: 10th Asian Thermal Spray Conference (ATSC 2020) Lua Li, Lead Guest Editor; Kazuhiro Ogawa, Huckwan Park, Harpreet Singh, and Xiuyong Chen, Guest Editors April 2021, Vol. 30, No. 4 Special Issue Featuring Papers from the International Thermal Spray Conference (ITSC) 2021 Fardad Azarmi, Kantesh Balani, Jan Cizek, Heli Koivuluoto, Yuk-Chiu Lau, Hua Li, and Filofteia-Laura Toma, Guest Editors January 2022, Vol. 31, No. 1–2 Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
PMC009xxxxxx/PMC9004616.txt	==== Front Environmental Sustainability Environmental Sustainability 2523-8922 Springer Nature Singapore Singapore 224 10.1007/s42398-022-00224-x Review Systematic mapping on the importance of vultures in the Indian public health discourse http://orcid.org/0000-0002-7268-737X Jalihal Smriti smriti.jalihal@flame.edu.in 1 http://orcid.org/0000-0003-2396-2598 Rana Shweta shweta@flame.edu.in 1 http://orcid.org/0000-0003-4223-3972 Sharma Shailja sharmas@aiimsjodhpur.edu.in 2 1 grid.459524.b 0000 0004 1769 7131 Department of Physical and Natural Sciences, FLAME University, Pune, 412115 India 2 grid.463267.2 0000 0004 4681 1140 Department of Biochemistry, All India Institute of Medical Sciences Jodhpur, Jodhpur, 342005 India 12 4 2022 2022 5 2 135143 11 11 2020 4 3 2022 6 3 2022 © The Author(s) under exclusive licence to Society for Environmental Sustainability 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Vultures are of immense ecological significance to forest and urban ecosystems. These birds play a major role in curbing environmental contamination through scavenging on carcasses. Prevention of spread of diseases is pivotal for public health and is an inexorable economic burden for any country. We present the crucial role vultures can play in disease mitigation and public health by regulating or decreasing the spread of zoonotic diseases. We elaborate examples from three zoonotic diseases; rabies, brucellosis and tuberculosis, which spread among dogs and cattle as well as human population. We establish the viable links in the transmission of these diseases from the infected dead and alive animals to humans and their possible exacerbation in the absence of vultures. These indirect links help formulate the case for increased interventions for disease spread and control along with conservation of these scavengers. Their role as natural and effective cleaners of the environment in the Indian health discourse is of importance because they can reduce the expenses of the government in waste management and maintenance of public health. Keywords Conservation Disease Ecological importance Vultures Zoonosis issue-copyright-statement© The Author(s) under exclusive licence to Society for Environmental Sustainability 2022 ==== Body pmcIntroduction India has faced zoonotic diseases such as Swine flu (Kshatriya et al. 2018), Nipah virus encephalitis (Chadha et al. 2006; Rana and Singh 2015) in the recent past and is grappling with ongoing COVID-19 (WHO 2020). The role of scavengers in limiting the diseases at host level holds value ecologically as well as economically. Vultures contribute to ecosystem services and are beneficial to public health, primarily because they can get rid of tonnage of carcasses (Balmford 2013) and maintain adequate sanitation levels in the ecosystem (Kanaujia and Kushwaha 2013). The carrions and rotting bodies of all animals are decomposed by the action of microbes, flies, beetles, raccoons and vultures primarily. Vultures feed upon soft tissues as well as the hard parts as bones, teeth and hair of the carcasses. The ingested dead remains are released as nutrients through their faeces (Mondor et al. 2012) and this cycling of nutrients is favourable for maintaining a healthy ecosystem. Vulture’s abilities to aerial search confers advantage over terrestrial scavengers to spot carcasses from a faraway range. They are efficient consumers owing to the large size of their bodies and usually benefit from the heavy mortality in the migratory ungulate populations while following them. The steep decline in vulture population is increasing the potential of disease spread and transmission to other species (Ogada et al. 2011). Exceedingly high population of managed livestock, various wild animals along with several unorganised carcass disposal sites contribute in abundance of dead meat in India (Singh et al. 2013). The 20th Livestock Census 2019 reported an increase of 4.8% over the 2012 census, with total livestock population of 536.76 million. An increase of 4.56% and 11.19% livestock population in the rural and urban areas respectively, has been reported by Ministry of Fisheries, Animal Husbandry & Dairying, Government of India (MoFAHD 2019). Dead organisms are disease reservoirs and the disease can spill over through intraspecies or interspecies transmission. The vultures curb the disease spread in a comprehensive way as opposed to the ones they spread themselves, sometimes through their contaminated feathers and feet (Ogada et al. 2012). Vultures have a higher immunity and tolerance and their low stomach pH (1–2) act as the biological filter for pathogens. Scavengers and carnivores often have shorter guts than omnivores or herbivores, which reduces the chance of bacterial pathogen multiplication as the chances of food remaining in the gut reduces (Ogada et al. 2011; Beasley et al. 2015). Ability of vultures to scavenge prevents spread of a variety of diseases among the same species (through cannibalism if there is a dearth of resource) and to other species of higher trophic levels (Vicente and VerCauteren 2019). Some carnivores can feed on the carrions but all of them are not capable of killing pathogens efficiently as vultures and so they can become reservoirs of these microbes and can further infect humans or other animals, making it a public health hazard (Soulsbury and White 2016). Due to the decrease in the vulture population in Kenya, the number of animal carcasses increased three times, so did the amount of time to clear them out. Correlating this with the probability of disease spread, the risk posed on human health (Ogada et al. 2012) and the economic burden surged threefold. Vultures are known to dispose more than 22% of the organic waste in urban areas. Such biological controls are better alternatives to chemical treatments. They also serve to be effective in agriculture by consuming problem species, thereby reducing crop loss (O’Bryan et al. 2018). Reasons for declining vulture population The vulture population started decreasing severely during 1990s, alarming several conservationists and public health experts who understood their importance. The survey conducted by the Ministry of Environment, Forest and Climate Change (MoEFCC), India, with the coordination of State Forest Departments and Bombay Natural History Society (BNHS) reported a sharp decline in the vulture population in India. The numbers decreased from 40 million to 19,000 in a span of three decades. Between 1992 and 2007, 99% of the three species of critically endangered resident Gyps vultures; white-backed vulture, long-billed vulture and slender-billed vulture were almost wiped out with their population reducing to 6000, 12,000 and 1,000, respectively (MoEFCC 2020). The use of diclofenac on cattle and other ungulates (to increase milk production and provide immediate pain relief) has led to a massive decrease in vulture numbers globally from the early 1990s (Pain et al. 2003; Prakash et al. 2003; Swan et al. 2006; Buechley and Sekercioglu 2016; Bindra 2018). Kidney failure was reported in the vultures fed on cattle treated with these anti-inflammatory drugs (Green et al. 2004). Various reasons for threats and decline in different species of vultures in India have been summarized in Table 1. Extensive use of pesticides and insecticides in the agricultural practices are reported to have shown deleterious effects to many avian species.Table 1 Various reasons for threats and decline of vultures in India Scientific name Common name Threats and Reasons for decline IUCN Red list Status in India References Gyps bengalensis Oriental White-backed Vulture Changes in livestock husbandry, kidney failure due to diclofenac exposure, overhunting, poisoning, habitat loss, decline in food availability Critically endangered Prakash et al. (2003), Botha et al. (2017) Gyps tenuirostris Slender- billed vulture Changes in livestock husbandry, overhunting, poisoning, persecution, diclofenac exposure Critically endangered Prakash et al. (2003, 2012), Pain et al. (2003) Sarcogyps calvus Red-headed vulture Changes in livestock husbandry, overhunting, poisoning, persecution, diclofenac exposure, decline in food availability, habitat loss Critically endangered Pain et al. (2003), Cuthbert et al. (2006), Botha et al. (2017), Bindra (2018) Gyps indicus Long billed vulture Exposure to diclofenac, electrocution, poisoning, human disturbances in nesting sites Critically endangered Prakash et al. (2003), Bindra (2018) Neophron percnopterus Egyptian vulture Hunting and poisoning, pesticides (organochlorine, polychlorinated biphenyls, carbamates and organophosphorus), diclofenac exposure, electrocution Endangered Pain et al. (2003), Cuthbert et al. (2006), Ogada et al. (2011), Bindra (2018), Plaza et al. (2020) Gypaetus barbatus Bearded vulture Unintentional poisoning by feeding on carcasses, persecution and electrocution, lead poisoning, decline of food availability, habitat loss, human disturbance in nesting sites (aviation, paragliding), climate change Near threatened Buechley and Sekercioglu (2016) Gyps himalayensis Himalayan griffon vulture Exposure to diclofenac, pesticides (organochlorine, polychlorinated biphenyls, carbamates and organophosphorus) Near threatened Das et al. (2011), Prakash et al. (2012), Plaza et al. (2020) Gyps fulvus Eurasian Griffon Diclofenac, lead poisoning from spent ammunition, human interference, habitat degradation Of least concern Green et al. (2004), Buechley and Sekercioglu (2016) Dhananjayan et al. (2011) confirmed the presence of cyclodiene insecticides and organochlorine pesticides in the plasma samples of three species viz. white-backed vulture (Gyps bengalensis), Egyptian vulture (Neophron percnopterus), and Himalayan griffon vulture (Gyps himalayensis) from India. In a study carried out in southern India (Tamil Nadu), Malik et al. (2018) found p,p´-DDE (Dichlorodiphenyldichloroethylene) in concentrations associated with the eggshell thinning and premature hatchings of 106 species of birds. Poharkar et al. (2009) concluded malaria as major reason for the decreasing vulture numbers as they isolated an intracellular malarial parasite from the alive as well as dead white-backed vultures in the Indian subcontinent. Mobile towers have also resulted in decreasing avifaunal diversity (Kale et al. 2012) and almost 50% reduction in vulture numbers (Verma et al. 2018). The elctromagnetic radiaitons from the towers can interfere with the sense of direction and altitude of birds, and therefore disable the avian compass disorienting them (Balmori 2015). Recreational hunting and lead poisoning from spent ammunition have been identified as one of the main contributors in decline of vultures in Europe, Africa and America (Margalida et al. 2013; Garbett et al. 2018). Pain et al. (2003) reviewed eight Gyps species and evaluated accidental poisoning, hunting, human interferences and electrocution as likely causes of decline in south east Asia, western Europe and Africa. Data collection and methods The present study extracted data from extensive research available on search engines such as Google Scholar, PubMed, research databases as EBSCO, Jstor and websites of World Health Organisation (WHO), Centers for Disease Control and Prevention (CDC), United Nations Environment Programme (UNEP), Government of India (GOI) directory and ScienceDirect. The websites of Ministry of Health and Family Welfare, Centre for Disease Prevention and Control (CDC), Ministry of Fisheries, Animal Husbandry and Dairying and Ministry of Environment and Forest and Climate Change; Government of India, were referred for the disease transmission, livestock numbers and management data, and vulture conservation plans respectively. The inclusion criteria are based on the articles on ecosystem services by vultures, causes for their decline, spread of zoonotic diseases, conservation plans and actions, English language/translated articles only and articles post 1990. The exclusion criteria was articles on decline, conservation and action plans for birds other than vultures and articles before 1990. The key words vulture, scavenger, decline, human, environment, diclofenac, zoonotic diseases, brucellosis, rabies, tuberculosis, conservation, public health, ecological value, economic load, waste management, contamination, carcass, forest, rural and urban ecosystem, prevention were used to check the title, abstracts and the full texts of the articles and reports. The articles meeting the inclusion criteria were selected for the assessment of full text and abstracts. We referred to 351 articles from the mentioned sources and out of which 234 articles meeting the inclusion criteria were selected. Substantial literature was checked to understand the extent and the harm zoonotic diseases can have on the humans and other animals with context to vulture population decline. All articles looked at were after 1990, when the vulture population started to steeply decline in India. A total of 78 articles (journals-69; reports-United Nations (CMS-UNEP-Convention on the Conservation of Migratory Species of Wild Animals-United Nations Environment Programme)-1, WHO-1; CDC-1; Government of India-3; e-books-3) were included in the final study (Fig. 1).Fig. 1 Prisma Diagram showing the identification, screening, eligibility and inclusion of the articles for the review Disease spread due to carcass decay Zoonotic diseases are transmitted from animals to humans, so controlling the zoonoses, pathogen and their vector becomes critical. They have resulted in high morbidity and mortality rates across the world among human and animal populations. Such emerging healthcare threats present high prevention and curing costs for governments around the world. It is thus important to curb these diseases at the level of animals, before they spread to human populations (Belay et al. 2017). Some of the zoonotic diseases that have posed a large social and economic costs in the recent years are zika virus disease (Qureshi 2017), COVID-19 (Chaudhary et al. 2020), Nipah virus encephalitis and swine flu (Chadha et al. 2006). Transmitted from animals to humans first, they have resulted in high rates of human-to-human transmission thereafter. Scavengers play role in disease regulation by reducing host and vector densities, through local competitive exclusion, minimising the pathogen numbers or feeding on the hosts directly. For example, leopards in Mumbai (India) have played an important role in controlling of dog population and thus, reduction in the cases of dog bites and transmission of rabies to humans (O’Bryan et al. 2018). On the basis of existing evidences, many studies reinforce the fact that absence of vultures will increase the transmission of infectious diseases to humans as well as their livestock (Markandya et al. 2008; Ogada et al. 2011, 2012; Moleón et al. 2014). Decaying carcasses are undoubtedly a huge lingering threat to human health (Shearer et al. 2018). They act as breeding grounds for a large number of pathogens that can lead to transmission of infections. Waterborne diseases easily spread through the fouling of watercourses by rotting carcasses, however, this link is not that easy to establish (Markandya et al. 2008). The reduction in the number of vultures leave quite a few carcasses un-scavenged, invariably increasing other organisms and abiotic resources exposed to a large number of diseases. The absence of vultures can cause the “piling up of corpses in the land of the living” (Van Dooren 2010). Anaerobic pathogens are the largest threat to human health as they spread through the soil from carcasses. Clostridium perfringens an anaerobic human pathogen, has been isolated from water, soil, air, dust, fresh meat, milk, and vegetables. It has been found that humans are most susceptible to get infected by this bacterium in poor hygienic conditions, similar to what is fostered by uncleared carcasses (Haagsma 1991). Additionally, carcasses are an easy feast for rodents, which has led to the increase in the rat population, thus escalating the diseases spread by them among both, the wildlife and human populations (Speer 2015). A reduction in prime scavenger species like vultures disturbs the food chain and leads to an increase in mesoscavenger species, many of them being disease-spreading pests (O’Bryan et al. 2019). Carcasses can play a role in the introduction and reintroduction of some viruses in the air, water, and soil. An excessive number of decaying carcasses cause environmental contamination, which can further interfere with the functioning of the endocrine system among humans and animals and is strongly linked to the spread of emerging diseases (Movalli et al. 2018). Exposed carcasses can easily spread zoonotic diseases to humans, animals in urban and rural areas and wild animals in the forests. Farmers and people engaged in animal husbandry, who live in close proximity to the cattle are most susceptible to these diseases (Singh et al. 2013). Diseases like rabies, brucellosis, tuberculosis, and others are spread by pathogens (Sokolow et al. 2019) that travel from carcasses into soil and water and further enter the human body through different means; such as by the consumption of water, food or through direct contact (Swan et al. 2006; Makandya et al. 2008; Plaza et al. 2020). Mudur (2001) has linked the possibility of human anthrax outbreaks with the decline in vulture population. The contact with the carcass was ascribed as one of the reasons for the disease spread. The rapid increase in stray dogs and rabies has coincided with the rapid decline in vulture population (Pain et al. 2003; Prakash et al. 2003). Most street dogs in India are not adequately vaccinated and the incidents of dog bites are common. This spreads rabies, from dogs to humans, which increases both mortality and morbidity. Preying on or contact with dead rabid dogs can transmit the pathogen to other animals and humans. (Markandya et al. 2008; Lembo et al. 2008). Theimer et al. (2017) reported that laboratory studies have demonstrated the potential of rabies transmission via ingestion of rabid animals. Bindra (2018) reported incidences of stray dogs homing and breeding in the garbage dumps and decaying carcasses led to increased disease spread in dogs as well as transmission of infections to humans. India has approximately 25 million dogs, with an estimated dog: human ratio of 1:36 (Menezes 2008) as compared to mean ratio of 1:9.5 and 1:12.3 for Asia and Africa regions respectively (Knobel et al. 2005). Since 1985, India has reported 25,000–30,000 deaths by rabies annually (Sudarshan 2004). Approximately INR1046 billion is reported to be the economic burden associated with dog bites, rabies spread in humans and simultaneous vulture decline between 1992 and 2006 (Markandya et al. 2008; Brookes et al.2019). Though rabies is also caused by the scratches and bites of cats, bats, rabbits etc., the high death rate is mainly attributed to the large stray dog population in the country (NHP India 2018). Brookes et al. (2019) reported outbreak of rabies in dairy cattle and buffaloes and death of farmers in a village in Punjab, India. The role of vulture as a scavenger becomes more critical and economical in the urban areas co-inhabited by stray and unvaccinated dogs. Heever et al. (2021) concluded that vultures can regulate the spread of rabies and brucellosis and cut on costs required for their treatment and prevention. Brucella melitensis is the main cause of brucellosis in humans (Franco et al. 2007). Humans get the disease by direct contact with infected live or dead cattle, eating or drinking contaminated milk or meat (Assenga et al. 2015) or by inhaling the airborne pathogen (Zhang et al. 2020). Hence it becomes more critical to prevent it’s spread in humans by controlling it in the dead or alive host animal. Brucellosis is a massive public health problem mainly because of the fact that there is a lack of a cohesive plan to respond to such cases on the part of hospitals and the government (Rossetti et al. 2017). This disease has been a major problem in parts of the Mediterranean, Western Asia, Latin America and Africa and is emerging as one of the most severe health problems in India. Like any other lesser-known disease, there is a strong link between brucellosis and economic loss, as the government needs to spend more money to diagnose it and find cures (Singh et al. 2018). Measures as controlling cattle herding, movement and milk production due to its massive economic implications are unviable. Brucellosis can also spread from dogs to humans through direct contact with infected individuals, or through their blood, vaginal or seminal fluids. Brucella canis can continue to be an unrecognized pathogen, in both dogs and humans. Vultures can clear carcass of infected dogs and cattle, and effectively regulate the spread of disease (Hensel et al. 2018; Brookes et al. 2019). Gumi et al. 2012 studied the zoonotic transmission between pastoralists and their livestock in Ethiopia. The study that was conducted over a span of 2 years, reported 32 suspected tuberculosis (TB) cases for transmission of the pathogen through their cattle. Though Mycobacterium tuberculosis is the most commonly found cause of TB, people in direct contact with wound of dead or alive cattle may also acquire another form of the disease known as bovine tuberculosis, caused by M. bovis. Factsheet released by CDC has also mentioned it’s spread to humans by inhaling the bacteria directly from the infected animals (https://www.cdc.gov/tb/publications/factsheets/general/mbovis.pdf). The spread of this disease is a large public health concern, especially in rural areas (Amanfu 2006; Gumi et al. 2012; Cowie et al. 2015). Across countries, this disease has also been the cause of economic losses. The proportion of bovine TB in humans was less than 5% of the total number of cases till 2012. These cases were predominantly reported in areas of Chad, Niger, Ghana, Uganda, and Ethiopia (Ukwaja et al. 2012). Once a human is infected with the disease, it can rapidly spread among the population through coughing and sneezing (Hassan et al. 2014). Srinivasan et al. (2018) in their systematic review and meta-analysis reported that there may be an estimated 21.8 million cattle infected with bovine tuberculosis in India-a population greater than the total number of dairy cows in the United States. The decrease in vulture population can cause the cattle carcass numbers to increase, which can have indirect linkage to increased spread of bovine tuberculosis in human population (Markandya et al. 2008; Vicente and VerCauteren 2019). Conservation programs and future strategies Owing to high ecological and economic value of vultures, their conservation is crucial. Majorly the attention and resources in India are diverted to the big species of tourism importance like elephant, asiatic lion, royal bengal tiger, rhinoceros, Olive Ridley Sea turtle etc, but not vultures, which carry least aesthetic appeal. Appreciating their significance, the Government has recently initiated action plans and policies for conservation. In 2006, the governments of India, Pakistan, Nepal and Bangladesh banned use of diclofenac on cattle (Cuthbert et al. 2014). In 2010 these countries negotiated further coordination and improved action plans for vulture conservation (Balmford 2013). After the ban on diclofenac and implementation of captive breeding programs, population of the critically endangered vulture species in India have started increasing (Prakash et al. 2012; Bindra 2018). The MoEFCC has devised the Action Plan for Vulture Conservation which advocates for the prevention of cattle poisoning, increased monitoring and vulture breeding programs and describing the role of the state and central governments in curbing the leak of diclofenac by the veterinary industry. Vulture conservation breeding centres have been established in eight states in India (Haryana, West Bengal, Assam, Madhya Pradesh, Gujarat, Odisha, Telangana and Jharkhand) that are being managed by State Forest Departments with support from BNHS and MoEFCC. Financial support of INR125.3 million has been given to five states (Punjab, Haryana, Kerala, Uttarakhand and West Bengal) and a mega project has been sanctioned to the Indian Veterinary Research Institute, Uttar Pradesh on assessing the safety and conservation of vultures. The status upgradation of white-backed, long-billed and slender-billed vultures from Schedule IV to Schedule I of the Wild Life (Protection) Act, 1972 along with various initiatives in mass educating and spreading awareness have been implemented. The action plan credits the vulture as the most efficient soldier of the ‘Swachh Bharat Abhiyaan’ (MoEFCC 2020). Successful vulture conservation programs across the world have reinforced the awareness towards the dwindling vulture population. About a third of all the vulture species globally are successful outcomes of captive breeding and reintroduction projects (Houston 2005). Margalida et al. (2010) modelled the effects of sanitary policies on European vulture conservation and stressed upon the creation for vulture programs with special feeding centres and policy level strategies, such as regulating the use of animal by products. Vulture management in Southern Europe has recovered more than 200 percent of Eurasian Griffon Vultures (Margalida et al. 2014). As per the Memorandum of Understanding on the Conservation of Migratory Birds of Prey in Africa and Eurasia, and Convention on Migratory Species have adopted the Vulture Multi-species Action plan (MsAP), which is a strategic conservation plan covering 128 states of all 15 species of migratory African-Eurasian vultures in a coordinated international effort to save them from further decline and extinction (Botha et al. 2017). Safford et al. (2019) proposed flagship projects to support vulture MsAP. Badia-Boher et al. (2019) have evaluated the success rate of conservation projects coupled with awareness campaigns and long-term monitoring, highlighting the improvement in vulture population in Europe. According to Becker et al. (2020) conservation practices can be refined by studying the microbiome of the vultures. Yee et al. (2021) have suggested examining ethical perspectives of the stakeholders responsible for conservation for effective decision making in conservation. Complementation of national and global efforts is the key for effective conservation (Santangeli et al. 2020). Conclusion Effective utilisation of the vultures as biological agents to control the diseases at the host level contributes towards disease regulation and may reduce the economic burden in maintaining the public health. Conserving vultures becomes more important in an agrarian society like India where people are often in close proximity with livestock and thus at a higher risk of exposure to zoonotic diseases, which can spread from live as well as dead animals. The international and national conservation efforts seem promising but consistency is required to maintain the sustainability of these programs. For future research, quantification of disease control by vultures is recommended. We need more studies establishing direct linkages between dwindling vulture population, escalation of zoonotic infections and consequent health threats. Economic valuation of the role of vultures in disease regulation is required to reassure further strengthening of vulture conservation policies and action plans. The consequences of decline in vulture population must be assessed with reference to economics linked with public health and animal care. At present both are lacking or largely not accessible to citizens in developing countries like India. Acknowledgements The authors thank the editor and two anonymous reviewers for their valuable feedback. Their inputs immensely helped improve the quality of our article. Author contributions SJ: data collection, manuscript writing; SR: data collection, manuscript writing, critical revision; SS: data collection, critical revision. Funding Not applicable. Availability of data and material Data sharing not applicable to this article as no datasets were generated or analysed during the current study. Declarations Conflicts of interest The authors declare that they have no conflict of interest in the publication. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Amanfu W The situation of tuberculosis and tuberculosis control in animals of economic interest Tuberculosis 2006 86 3–4 330 335 10.1016/j.tube.2006.01.007 16644282 Assenga JA Matemba LE Muller SK Malakalinga JJ Kazwala RR Epidemiology of Brucella infection in the human, livestock and wildlife interface in the Katavi-Rukwa ecosystem, Tanzania BMC Vet Res 2015 11 189 10.1186/s12917-015-0504-8 26253151 Badia-Boher JA Sanz-Aguilar A de la Riva M Gangoso L van Overveld T García-Alfonso M Luzardo OP Suarez-Perez A Donázar JA Evaluating European LIFE conservation projects: Improvements in survival of an endangered vulture J Appl Ecol 2019 56 5 1210 1219 10.1111/1365-2664.13350 Balmford A Pollution, politics, and vultures Science 2013 339 6120 653 654 10.1126/science.1234193 23393251 Balmori A Anthropogenic radiofrequency electromagnetic fields as an emerging threat to wildlife orientation Sci Total Environ 2015 518–519 58 60 10.1016/j.scitotenv.2015.02.077 Beasley DE Koltz AM Lambert JE Fierer N Dunn RR The evolution of stomach acidity and its relevance to the human microbiome PLoS ONE 2015 10 7 e0134116 10.1371/journal.pone.0134116 26222383 Becker AAMJ Harrison SWR Whitehouse-Tedd G Budd JA Whitehouse-Tedd KM Integrating gut bacterial diversity and captive husbandry to optimize vulture conservation Front Microbiol 2020 10.3389/fmicb.2020.01025 Belay ED Kile JC Hall AJ Barton-Behravesh C Parsons MB Salyer S Walke H Zoonotic disease programs for enhancing global health security Emerg Infect Dis 2017 23 1 S65 10.3201/eid2313.170544 Bindra P (2018) Declining vulture population can cause a health crisis. Mongabay-India. https://india.mongabay.com/2018/02/declining-vulture-population-can-cause-a-health-crisis/. Accessed 9 June 2020 Botha AJ, Andevski J, Bowden CGR, Gudka M, Safford RJ, Tavares J, Williams NP (2017) Multi-species action plan to conserve African-Eurasian vultures. CMS Raptors MOU Technical Publication no. 5. CMS Technical Series no. 35. Abu Dhabi: Coordinating Unit of the CMS Raptors Brookes VJ Gill GS Singh BB Sandhu BS Dhand NK Aulakh RS Ward MP Challenges to human rabies elimination highlighted following a rabies outbreak in bovines and a human in Punjab, India Zoonoses Public Health 2019 66 3 325 336 10.1111/zph.12568 30779303 Buechley E Sekercioglu C The avian scavenger crisis: Looming extinctions, trophic cascades, and loss of critical ecosystem functions Biol Conserv 2016 198 220 228 10.1016/j.biocon.2016.04.001 CDC-Mycobacterium bovis (Bovine Tuberculosis) in Humans, pp 1–2. https://www.cdc.gov/tb/publications/factsheets/general/mbovis.pdf. Accessed 15 July 2021 Chadha MS Comer JA Lowe L Rota PA Rollin PE Bellini WJ Ksiazek TG Mishra AC Nipah virus-associated encephalitis outbreak, Siliguri, India Emerg Infect Dis 2006 12 2 235 10.3201/eid1202.051247 16494748 Chaudhary M Sodani PR Das S Effect of COVID-19 on economy in India: some reflections for policy and programme Jo Health Manag 2020 22 2 169 180 10.1177/0972063420935541 Cowie CE Gortázar C White PC Hutchings MR Vicente J Stakeholder opinions on the practicality of management interventions to control bovine tuberculosis Vet J 2015 204 2 179 185 10.1016/j.tvjl.2015.02.022 25910515 Cuthbert RJ Green RE Ranade S Saravanan S Pain DJ Prakash V Cunningham AA Rapid population declines of Egyptian vulture (Neophron percnopterus) and red-headed vulture (Sarcogyps calvus) in India Anim Conserv 2006 9 3 349 354 10.1111/j.1469-1795.2006.00041.x Cuthbert RJ Taggart MA Prakash VB Chakraborty SS Deori P Galligan T Kulkarni M Ranade S Saini M Sharma AK Shringarpure R Green RE Avian Scavengers and the threat from veterinary pharmaceuticals Philos Trans R Soc Lond Biol Sci 2014 10.1098/rstb.2013.0574 Das D Cuthbert R Jakati R Prakash V Diclofenac is toxic to the Himalayan Vulture Gyps himalayensis Bird Conserv Int 2011 21 1 72 75 10.1017/S0959270910000171 Dhananjayan V Muralidharan S Jayanthi P Distribution of persistent organochlorine chemical residues in blood plasma of three species of vultures from India Environ Monit Assess 2011 173 803 811 10.1007/s10661-010-1424-5 20221793 Franco MP Mulder M Gilman RH Smits HL Human brucellosis Lancet Infect Dis 2007 7 12 775 786 10.1016/S1473-3099(07)70286-4 18045560 Garbett R Maude G Hancock P Kenny D Reading R Amar A Association between hunting and elevated blood lead levels in the critically endangered African white-backed vulture Gyps africanus Sci Total Environ 2018 630 1654 1665 10.1016/j.scitotenv.2018.02.220 29550066 Green RE Newton IAN Shultz S Cunningham AA Gilbert M Pain DJ Prakash V Diclofenac poisoning as a cause of vulture population declines across the Indian subcontinent J Appl Ecol 2004 41 5 793 800 10.1111/j.0021-8901.2004.00954.x Gumi B Schelling E Berg S Firdessa R Erenso G Mekonnen W Zoonotic transmission of tuberculosis between pastoralists and their livestock in South-East Ethiopia Eco Health. 2012 9 2 139 149 10.1007/s10393-012-0754-x 22526748 Haagsma J Pathogenic anaerobic bacteria and the environment Rev Sci Tech 1991 10 3 749 764 10.20506/rst.10.3.569 1782427 Hassan AS Garba SM Gumel AB Lubuma JS Dynamics of Mycobacterium and bovine tuberculosis in a human-buffalo population Comput Math Methods Med 2014 2014 1 20 10.1155/2014/912306 Heever LVD Thompson LJ Bowerman WW Smit-Robinson H Shaffer LJ Harrell RM Ottinger MA Reviewing the role of vultures at the human–wildlife–livestock disease interface: an African perspective J Raptor Res 2021 55 3 1 17 Hensel ME Negron M Arenas-Gamboa A Brucellosis in dogs and public health risk Emerg Infect Dis 2018 24 8 1401 10.3201/eid2408.171171 30014831 Houston DC (2005) Reintroduction programs for vulture species. In: Proceedings of the International Conference on Conservation and Management of Vulture Populations (eds D.C. Houston & S.E. Piper). Natural Hi story Museum of Crete, Thessaloniki, pp 87–97 Kale M, Dudhe N, Kasambe R, Chakane S, Bhattacharya P (2012) Impact of urbanization on avian population and its status in Maharashtra state, India. Int J Appl Environ Sci 7(1): 69–86. http://www.diva-portal.org/smash/get/diva2:530657/FULLTEXT01.pdf. Accessed 12 June 2020 Kanaujia A, Kushwaha S (2013) Vulnerable vultures of India: population, ecology and conservation. In: Rare Animals of India, Bentham Science Publishers, pp 113–144. 10.2174/9781608054855113010009; Knobel DL, Cleaveland S, Coleman PG, Fèvre EM, Meltzer MI, Miranda ME, Shaw A, Zinsstag J, Meslin FX (2005) Re-evaluating the burden of rabies in Africa and Asia. Bull World Health Organ 83: 360–368. https://pubmed.ncbi.nlm.nih.gov/15976877/. Accessed 12 June 2020 Kshatriya RM, Khara NV, Ganjiwale J, Lote SD, Patel SN, Paliwal RP (2018) Lessons learnt from the Indian H1N1 (swine flu) epidemic: Predictors of outcome based on epidemiological and clinical profile. J Family Med Primary Care. 7(6): 1506. http://www.jfmpc.com/text.asp?2018/7/6/1506/246514. Accessed 12 July 2020 Lembo T Hampson K Haydon DT Craft M Dobson A Dushoff J Ernest HR Kaare M Mlengeya T Mentzel C Cleaveland S Exploring reservoir dynamics: a case study of rabies in the Serengeti ecosystem J Appl Ecol 2008 45 4 1246 1257 10.1111/j.1365-2664.2008.01468.x 22427710 Malik A Dharaiya N Espín S Is current information on organochlorine exposure sufficient to conserve birds in India? Ecotoxicology 2018 27 1137 1149 10.1007/s10646-018-1969-6 30083996 Margalida A Donázar JA Carrete M Sánchez-Zapata JA Sanitary versus environmental policies: fitting together two pieces of the puzzle of European vulture conservation J Appl Ecol 2010 47 4 931 935 10.1111/j.1365-2664.2010.01835.x Margalida A Arlettaz R Donázar AJ Lead ammunition and illegal poisoning: further international agreements are needed to preserve vultures and the crucial sanitary service they provide Environ Sci Technol 2013 47 11 5522 5523 10.1021/es401544j 23672642 Margalida A Campión D Donázar J Vultures vs livestock: conservation relationships in an emerging conflict between humans and wildlife Oryx 2014 48 2 172 176 10.1017/S0030605312000889 Markandya A, Taylor T, Longo A, Murty MN, Murty S, Dhavala K (2008) Counting the cost of vulture decline—an appraisal of the human health and other benefits of vultures in India. Ecol Econ 67(2):194–204. http://hdl.handle.net/10036/4350. Accessed 21 Nov 2019 Menezes R Rabies in India CMAJ 2008 178 5 564 566 10.1503/cmaj.071488 18299543 MoEFCC (2020) http://moef.gov.in/wp-content/uploads/2020/11/Action-Plan-for-Vutlure-Conservation-In-India-2020-2025-soft-copy-for-MoEFCC-2.pdf. Accessed 9 June 2021 MoFAHD (2019) 20th Livestock Census: Animal husbandry statistics. Department of Animal Husbandry and Dairying, Ministry of Fisheries, Animal Husbandry & Dairying, Govt. of India http://dahd.nic.in/animal-husbandry-statistics. Accessed 9 June 2021 Moleón M Sánchez-Zapata JA Margalida A Carrete M Owen-Smith N Donázar JA Humans and scavengers: the evolution of interactions and ecosystem services Bioscience 2014 64 5 394 403 10.1093/biosci/biu034 Mondor EB, Tremblay MN, Tomberlin JK, Benbow EM, Tarone AM, CrippenT L (2012) The ecology of carrion decomposition. Nat Educ Knowl 3(10): 21. https://www.nature.com/scitable/knowledge/library/the-ecology-of-carrion-decomposition-84118259/. Accessed 9 June 2021 Movalli P Krone O Osborn D Pain D Monitoring contaminants, emerging infectious diseases and environmental change with raptors, and links to human health Bird Study 2018 65 1 96 109 10.1080/00063657.2018.1506735 Mudur G (2001) Human anthrax in India may be linked to vulture decline. Br Med J. 322(7282): 320. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1173210/. Accessed 9 June 2021 NHP India (2018) World Rabies Day 2018 \| National Health Portal of India. Nhp.gov.in. https://www.nhp.gov.in/world-rabies-day-2018_pg. Accessed 22 Sept 2020 O’Bryan CJ Braczkowski AR Beyer HL Carter NH Watson JE McDonald-Madden E The contribution of predators and scavengers to human well-being Nat Ecol Evoln 2018 2 2 229 236 10.1038/s41559-017-0421-2 O'Bryan CJ Holden MH Watson JE The mesoscavenger release hypothesis and implications for ecosystem and human well-being Ecol Lett 2019 10.1111/ele.13288 Ogada DL Keesing F Virani MZ Dropping dead: causes and consequences of vulture population declines worldwide Ann N Y Acad Sci 2011 10.1111/j.1749-6632.2011.06293.x Ogada DL Torchin ME Kinnaird MF Ezenwa VO Effects of vulture declines on facultative scavengers and potential implications for mammalian disease transmission Conserv Biol 2012 26 3 453 460 10.1111/j.1523-1739.2012.01827.x 22443166 Pain DJ Cunningham AA Donald PF Duckworth JW Houston DC Katzner TL Causes and effects of temporospatial declines of gyps vultures in Asia Conserv Biol 2003 17 3 661 671 10.1046/j.1523-1739.2003.01740.x Plaza PI Blanco G Lambertucci SA Implications of bacterial, viral and mycotic microorganisms in vultures for wildlife conservation, ecosystem services and public health Int J Avian Sci 2020 162 4 1109 1124 10.1111/ibi.12865 Poharkar A, Reddy PA, Gadge VA, Kolte S, Kurkure N, Shivaji S (2009) Is malaria the cause for decline in the wild population of the Indian White-backed Vulture (Gyps bengalensis) Curr Sci 96(4): 553–558. http://www.jstor.org/stable/24105469. Accessed 22 Sept 2020 Prakash V Pain DJ Cunningham AA Donald PF Prakash N Verma A Si G Sivakumar S Rahmani AR Catastrophic collapse of Indian white-backed Gyps bengalensis and long-billed Gyps indicus vulture populations Biol Cons 2003 109 3 381 390 10.1016/S0006-3207(02)00164-7 Prakash V Bishwakarma MC Chaudhary A Cuthbert R Dave R Kulkarni M Kumar S Khadananda P Ranade S Shringarpure R Green RE The population decline of gyps vultures in India and Nepal has slowed since veterinary use of diclofenac was banned PLoS ONE 2012 7 11 e49118 10.1371/journal.pone.0049118 23145090 Qureshi AI (2017) Zika virus disease: From origin to outbreak. Academic Press. https://www.elsevier.com/books/zika-virus-disease/qureshi/978-0-12-812365-2. Accessed 21 Sept 2020 Rana S, Singh S (2015). Nipah virus: effects of urbanization and climate change. In: 3rd International Conference on biological, chemical & environmental sciences (BCES-2015); 2015, pp 64–68. http://iicbe.org/upload/7575C0915051. Accessed 22 Sept 2020 Rossetti CA Arenas-Gamboa AM Maurizio E Caprine brucellosis: A historically neglected disease with significant impact on public health PLoS Negl Trop Dis 2017 11 8 e0005692 10.1371/journal.pntd.0005692 28817647 Safford R Andevski J Botha A Bowden C Crockford N Garbett R Maragalida A Ramirez I Shobrak M Tavares J Williams N Vulture conservation: the case for urgent action Bird Conserv Int 2019 29 1 1 9 10.1017/S0959270919000042 Santangeli A Girardello M Buechley ER Botha A Minin ED Moilanen A Importance of complementary approaches for efficient vulture conservation: reply to Efrat et al Conserv Biol 2020 34 5 1308 1310 10.1111/cobi.13579 32588448 Shearer JK Griffin D Cotton SE Humane Euthanasia and Carcass Disposal Vet Clin Food Anim Pract 2018 34 2 355 374 10.1016/j.cvfa.2018.03.004 Singh BB Ghatak S Banga HS Gill JP Singh B Veterinary urban hygiene: a challenge for India Rev Sci Tech 2013 32 3 645 656 10.20506/rst.32.3.2251 24761721 Singh BB Khatkar MS Aulakh RS Gill JPS Dhand NK Estimation of the health and economic burden of human brucellosis in India Prev Vet Med 2018 154 148 155 10.1016/j.prevetmed.2018.03.023 29685439 Sokolow SH Nova N Pepin KM Peel AJ Pulliam JR Manlove K Ecological interventions to prevent and manage zoonotic pathogen spillover Philos Trans R Soc B 2019 374 1782 20180342 10.1098/rstb.2018.0342 Soulsbury CD, White PC (2016) Human–wildlife interactions in urban areas: a review of conflicts, benefits and opportunities. Wildl Res 42 (7), 541–553. http://eprints.lincoln.ac.uk/id/eprint/17462/. Accessed 9 Aug 2020 Speer B Current therapy in avian medicine and surgery—E-Book 2015 1 Saint Louis Elsevier-Health Sciences Division Srinivasan S Easterling L Rimal B Niu XM Conlan AKJ Dudas P Kapur V Prevalence of bovine tuberculosis in India: A systematic review and meta-analysis Transbound Emerg Dis 2018 65 6 1627 1640 10.1111/tbed.12915 29885021 Sudarshan MK Assessing burden of rabies in India WHO sponsored national multi-centric rabies survey2003 Indian J Community Med 2004 30 100 101 10.4103/0970-0218.42864 Swan G Naidoo V Cuthbert R Green RE Pain DJ Swarup D Removing the threat of diclofenac to critically endangered Asian vultures PLoS Biol 2006 4 3 66 10.1371/journal.pbio.0040066 Theimer TC Dyer AC Keeley BW Gilbert AT Bergman DL Ecological potential for rabies virus transmission via scavenging of dead bats by mesocarnivores J Wildl Dis 2017 53 2 382 385 10.7589/2016-09-203 28094609 Ukwaja KN Modebe O Igwenyi C Alobu I The economic burden of tuberculosis care for patients and households in Africa: a systematic review Int J Tuberc Lung Dis 2012 16 6 733 739 10.5588/ijtld.11.0193 22410546 Van Dooren T (2010) Vultures and their people in India: equity and entanglement in a time of extinctions. Austral Hum Rev 22(2):130–145. http://australianhumanitiesreview.org/2011/05/01/vultures-and-their-people-in-india-equity-and-entanglement-in-a-time-of-extinctions/. Accessed 9 Aug 2020 Verma S Sao S Singh R Impact of mobile tower radiation on birds in District Rajnandgaon and Dongargarh area of Chhattisgarh World J Pharm Res 2018 10.20959/wjpr201811-12523 Vicente J VerCauteren K Olea P Mateo-Tomás P Sánchez-Zapata J The role of scavenging in disease dynamics Carrion ecology and management 2019 Wildlife Research Monographs, Springer 161 182 WHO (2020) (COVID-19) virus—Coronavirus disease 2019 (COVID-19) https://www.who.int/docs/default-source/coronaviruse/situation-reports/20200423-sitrep-94-covid-19.pd. Accessed 15 June 2021 Yee N Shaffer LJ Gore ML Harrell RM Expert perceptions of conflicts in african vulture conservation: implications for overcoming ethical decision-making dilemmas J Raptor Res 2021 55 3 359 373 10.3356/JRR-20-39 Zhang T Liang X Zhu X Sun H Zhang S An outbreak of Brucellosis via air-born transmission in a kitchen waste disposing company in Lianyungang, China Int J Infect Dis 2020 96 39 41 10.1016/j.ijid.2020.03.008 32171949
PMC009xxxxxx/PMC9004713.txt	==== Front Radiologia (Engl Ed) Radiologia (Engl Ed) Radiologia 2173-5107 SERAM. Published by Elsevier España, S.L.U. S2173-5107(22)00047-7 10.1016/j.rxeng.2022.03.001 Introduction Breast radiology: New horizons in times of pandemics☆ Radiología mamaria: Nuevos horizontes en tiempos de pandemiaPina L. Editor de Área de Imagen Mamaria 12 4 2022 3 2022 12 4 2022 64 23 © 2022 SERAM. Published by Elsevier España, S.L.U. All rights reserved. 2022 SERAM Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcIn these times of pandemic, with the focus of news headlines and a significant proportion of the scientific literature on COVID-19, we are launching this Special Issue on Breast Imaging. I hope this move will underline the fact that all the other disease processes that already affected us prior to the start of the pandemic continue to exist. Unfortunately, the saturation of health services has led to a significant increase in waiting lists for primary care and all the different specialist areas. As far as Breast Imaging is concerned, there have been undeniable delays in screening campaigns and in breast disorder consultations, resulting in harm to patient health. We therefore face a great challenge to try to recover or even improve pre-pandemic indicators. We also cannot forget our essential ongoing training; despite COVID-19, knowledge in all the other specialist areas continues to advance. This Special Issue includes recent articles accepted for publication in our Journal over the past few months. I would like to highlight the interesting study by Dr García Mur’s team, entitled La visión del residente de Radiodiagnóstico en España sobre la Radiología Mamaria [The Diagnostic Radiology resident’s view of Breast Radiology in Spain]. It concludes that the residents’ evaluation is clearly positive, and that they highly rate the dynamic nature of the rotation. However, it is surprising that most of them would not want to work in Breast Imaging. Their reasons are: their preference for CT; the need for sub-specialisation; medical-legal issues; and contact with patients. Medicine in general has a vocational component, but perhaps dedication to Breast Imaging requires it to a greater extent. There is no doubt that breast radiology has very rewarding aspects, such as multimodal work, establishing an adequate correlation between the different imaging techniques, and the feedback when we find out the results of the biopsies we indicate and perform. Additionally, working as a team with other specialists (oncologists, surgeons, gynaecologists, etc) as part of multidisciplinary committees is a significant motivating factor. Nonetheless, as residents are aware, there is a great deal of responsibility involved in interpreting Breast Imaging, and it is not exempt from medical-legal issues. A sentiment I cannot share, however, is that contact with patients is an inconvenience. On the contrary, I consider it to be one of the main motivations for our work, as it makes us doctors in the broadest sense of the word, and not just specialists in very specific imaging techniques. Carrying out diagnostic tests, especially ultrasound and biopsies, and reporting the results, makes us highly visible to patients and establishes a very interesting and human relationship between doctor and patient, which unfortunately is quite absent in other fields of Radiology. In another of the articles, Dr Vázquez-Caruncho's team report their experience with an interesting percutaneous treatment technique: cryoablation. This is just one example of the fact that our speciality never stops growing, incorporating new techniques and progressing from diagnosis to treatment. The advance of percutaneous interventional techniques for the breast has been unstoppable; from fine needle to core needle, the incorporation of vacuum-assisted biopsy and therapeutic techniques guided by stereotaxy, tomosynthesis, ultrasound or MRI, drainage of fluid collections, preoperative location of non-palpable lesions, radioguided occult lesion localisation (ROLL), etc. Ultrasound is an exceptionally useful tool in breast disease that has seen great technological advances in recent years. Gone are the old ultrasound machines that were no use for much more than clarifying the solid or cystic nature of a nodule. Today, ultrasound has multiple indications, and it is emerging as the technique of choice for the study of the axilla. In the article Ecografía axilar prequirúrgica en pacientes con carcinoma de mama. Estudio prospectivo [Preoperative axillary ultrasound in patients with breast carcinoma. A prospective study], the authors demonstrate the ability of ultrasound to distinguish between axillas with high or low tumour burden, with everything that means for therapeutic management. Also included in this Special Issue are articles discussing breast implants. As radiologists, we are definitely coming across more and more patients with implants. In the article ¿Es la ecografía mamaria una buena alternativa a la RM en la valoración de la integridad protésica? [Is breast ultrasound a good alternative to MRI in assessing prosthesis integrity?], the authors conclude that ultrasound does indeed fulfil that role. When there are pressing waiting lists, it is good to be able to make space in the MRI schedules for breast cancer patients and assess implants with a more accessible technique like ultrasound. The article Dosis glandular promedio en glándula mamaria y dosis de radiación en tiroides y cristalino en mujeres con y sin implantes mamarios [Average glandular dose in the mammary gland and radiation dose in the thyroid and lens in women with and without breast implants] presents the authors' results, which reveal a higher dose due to the presence of the implants. When deciding to have breast implants, perhaps patients should be warned of these consequences, and of the added difficulty of detecting breast cancer. Finally, this Special Issue also reports on very rare benign and malignant tumours. Among the benign, adenomyoepithelioma, and among the malignant, anaplastic large cell lymphoma associated with breast implants, which really is extremely rare. Without further ado, I hope you enjoy this Special Issue. ☆ Please cite this article as: Pina L. Radiología mamaria: Nuevos horizontes en tiempos de pandemia. Radiología. 2022;64:2–3.
PMC009xxxxxx/PMC9005018.txt	==== Front Acta Neurol Belg Acta Neurol Belg Acta Neurologica Belgica 0300-9009 2240-2993 Springer International Publishing Cham 35414154 1949 10.1007/s13760-022-01949-6 Letter to the Editor Anti-CASPR2 myeloencephalitis with active replication of hepatitis B virus in the central nervous system: a case report Liu Xu 1 Wang Lu 2 Zhong Di 1 http://orcid.org/0000-0003-2562-826X Yan Bo yanbo96@163.com 13 http://orcid.org/0000-0002-0372-1634 Hao Xiaoting sherryhao@wchscu.cn 1 1 grid.13291.38 0000 0001 0807 1581 Department of Neurology, West China Hospital, Sichuan University, Chengdu, 610000 Sichuan China 2 grid.460068.c 0000 0004 1757 9645 Department of Neurology, The Affiliated Hospital of Southwest Jiaotong University and The Third People’s Hospital of Chengdu, Chengdu, 610031 Sichuan China 3 Department of Neurology, Chengdu Shangjin Nanfu Hospital, Chengdu, 611730 Sichuan China 12 4 2022 2022 122 5 13851387 25 2 2022 27 3 2022 © The Author(s) under exclusive licence to Belgian Neurological Society 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. http://dx.doi.org/10.13039/501100009029 Adaptable and Seamless Technology Transfer Program through Target-Driven R and D 2018YFC1311400 Yan Bo issue-copyright-statement© The Author(s) under exclusive licence to Belgian Neurological Society 2022 ==== Body pmcIntroduction Autoimmune encephalitis (AE) involves inflammation of the central nervous system, and its most common forms involve auto-antibodies against the N-methyl-d-aspartate receptor, leucine-rich glioma-inactivated protein 1, or contactin-associated protein-like 2 (CASPR2) [1]. CASPR2 forms part of the voltage-gated potassium channel complex in both the central and peripheral nervous systems [1], so anti-caspr2 disease usually involves extensive damage to both systems, leading to limbic encephalitis, Morvan syndrome, cerebellar syndrome, peripheral nerve hyperexcitability, and autonomic dysfunction [2]. AE can be triggered by neurotropic viruses, and this has been well documented for encephalitis involving auto-antibodies against the N-methyl-d-aspartate receptor [3] Chronic infection with hepatitis B virus (HBV) has been linked to neurological complications such as myelitis and Guillain-Barré syndrome [4, 5], but it has never been clearly linked to AE. Here we report what appears to be the first case of anti-CASPR2 encephalitis linked to chronic HBV infection. Case presentation A 20 year-old Han Chinese woman presented at our hospital because of headache and fever, followed one week later by bilateral lower limb weakness, urinary retention, abnormal delirium, and rapid onset of somnolence. She reported that her body weight had decreased by 7 kg over the preceding three weeks. Physical examination at admission showed gross vertical nystagmus in both eyes, weakened bilateral pharyngeal reflex, poor muscle strength in the upper limbs (grade 2 on the left side, 3 on the right side). Bilateral muscle tone was reduced and tendon reflexes were reduced in both upper limbs and absent in both lower limbs. The patient was positive for meningeal irritation and negative for Babinski’s sign. She was normal in other neurological and physical examinations. 21 days after onset, patient presented with positive Babinski sign, increased tendon reflex and muscle tone. Magnetic resonance imaging revealed multiple lesions in the hippocampus, bilateral basal ganglia, brainstem, cerebellar hemisphere, as well as cervical and thoracic spinal cord (Fig. 1). Lumbar puncture showed normal cerebrospinal fluid pressure of 250 mmH2O, normal cell count (134 × 106/L) and elevated protein (1.22 g/L; normal range, 0.15–0.45 g/L). A cell-based assay revealed an anti-CASPR2 antibody titer of 1:32 in serum, but no such antibody was detected in cerebrospinal fluid. The patient was negative for auto-antibodies against the N-methyl-d-aspartate receptor, glutamate receptors 1 or 2-aminobutyric acid B receptor, a-amino-3-hydroxy-5-methyl-4-isoxazol-propionic acid (AMPA) receptors, leucine-rich glioma-inactivated protein 1, dipeptidyl-peptidase-like protein 6, Iglon5, myelin oligodendrocyte glycoprotein, glial fibrillary acidic protein, aquaporin 4, myelin basic protein, and flotillins 1 or 2. The patient was also negative for oligoclonal bands. Metagenomic next-generation sequencing (mNGS) of cerebrospinal fluid detected 28 copies of HBV-DNA, and HBV load in serum was 2.91 × 107 IU/ml. Chest computed tomography did not reveal HBV-associated occupancy.Fig. 1 a–b Magnetic resonance imaging of the patient, showing multiple lesions in the hippocampus, bilateral basal ganglia, brainstem, and cerebellar hemisphere. c T2-weighted magnetic resonance imaging revealing lesions in the thoracic spinal cord. d Cell-based assay of anti-CASPR2 antibodies in serum Based on these findings, the patient was diagnosed with anti-CASPR2 myeloencephalitis and chronic HBV infection. She was given multiple courses of immunoglobulin therapy and entecavir, and her condition gradually improved by the time of discharge. The patient complained only of mild sleep disorder, which had resolved by follow-up at 4 months after discharge. However, at this time, we tested autoimmune antibody again using cell-based assay, which revealed an anti-CASPR2 antibody titer of 1:100 in serum and 1:10 in cerebrospinal fluid. Discussion To our knowledge, this appears to be the first case of anti-caspr2 myeloencephalitis apparently linked to chronic HBV infection. Interestingly, we found that the titer of anti-CASPR2 antibody in serum and cerebrospinal fluid increased despite improvement in the patient’s condition. Previous studies suggested that there was no definite linear relationship between severity of disease and antibody titer [1]. How AE begins remains unclear. Some AE patients present tumors, such as ovarian teratoma in a patient with encephalitis involving antibodies against the N-methyl-d-aspartate receptor [6], and a thymoma in a patient with anti-caspr2 encephalitis [2]. Viral infection also appears capable of triggering AE: cases of encephalitis involving antibodies against the N-methyl-d-aspartate receptor have been linked to infection by herpex simplex virus, H1N1 influenza virus, influenza B virus, and severe acute respiratory syndrome coronavirus 2 [3, 7]. Some patients have been reported to have chronic HBV infection as well as neurological disorders including acute disseminated encephalomyelitis, neuromyelitis optica spectrum disorders, and Guillain-Barré syndrome [4]. When the body begins to generate antibodies against the invading virus, some of those antibodies may recognize auto-antigens on the surface of cells. For example, epitopes on HBV appear to resemble areas on myelin basic protein, such that the body’s anti-HBV antibodies may bind to the myelin sheath on axons, leading to demyelination and inflammation [8]. This model may explain how viral infection can lead to AE. Consistent with this idea, anti-CASPR2 antibodies in patients with AE recognize extracellular epitopes of CASPR2 [1]. Indeed, chronic HBV infection in patients with neurotic myotonia (Isaacs syndrome) can elicit production of auto-antibodies against the voltage-gated potassium channel complex in which CASPR2 functions [9]. Therefore we speculate that HBV infection of the nervous system may trigger production of auto-antibodies against CASPR2, inducing AE. Conclusion HBV may trigger anti-CASPR2 myeloencephalitis. Our case suggests that patients with chronic infections such as HBV who are suspected of having encephalitis should be screened by mNGS and assay of AE-related antibodies in the cerebrospinal fluid. Author contributions XL (Department of Neurology, West China Hospital, Sichuan University): acquired clinical data, reviewed the literature, created the figure, and drafted the manuscript. LW (The Affiliated Hospital of Southwest Jiaotong University & The Third People’s Hospital of Chengdu): participated in drafting and reviewing the manuscript. DZ (Department of Neurology, West China Hospital, Sichuan University: acquired clinical data. BY (Department of Neurology, West China Hospital, Sichuan University): performed neuroradiological assessment, and revised the manuscript. XH (Department of Neurology, West China Hospital, Sichuan University): collected and analyzed immunohistology data, conceived the figure, and revised the manuscript. Funding This work was supported by the National Key R&D Program of China (2018YFC1311400). Declarations Conflict of interest The authors declare no competing interests. Ethical approval The study was approved by the institutional review board. Consent to participate Informed consent was obtained from all participants included in the study. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xu Liu and Lu Wang contributed equally to this manuscript. ==== Refs References 1. van Sonderen A Petit-Pedrol M Dalmau J The value of LGI1, Caspr2 and voltage-gated potassium channel antibodies in encephalitis Nat Rev Neurol 2017 13 290 301 10.1038/nrneurol.2017.43 28418022 2. van Sonderen A Ariño H Petit-Pedrol M The clinical spectrum of Caspr2 antibody-associated disease Neurology 2016 87 521 528 10.1212/WNL.0000000000002917 27371488 3. Schwenkenbecher P Skripuletz T Lange P Intrathecal antibody production against epstein-barr, herpes simplex, and other neurotropic viruses in autoimmune encephalitis Neurol Neuroimmunol Neuroinflamm 2021 10.1212/NXI.0000000000001062 34697224 4. Inoue J Ueno Y Kogure T Analysis of the full-length genome of hepatitis B virus in the serum and cerebrospinal fluid of a patient with acute hepatitis B and transverse myelitis J Clin Virol 2008 41 301 304 10.1016/j.jcv.2008.01.002 18291715 5. Pettersson J Piorkowski G Mayxay M Meta-transcriptomic identification of hepatitis B virus in cerebrospinal fluid in patients with central nervous system disease Diagn Microbiol Infect Dis 2019 95 114878 10.1016/j.diagmicrobio.2019.114878 31451314 6. Dalmau J Tüzün E Wu H Paraneoplastic anti-N-methyl-d-aspartate receptor encephalitis associated with ovarian teratoma Ann Neurol 2007 61 25 36 10.1002/ana.21050 17262855 7. Phillips O Tubre T Lorenco H Limbic encephalitis in a child with ovarian teratoma and influenza B. Case report and critical review of the history of autoimmune anti-N-methyl-d-aspartate receptor encephalitis J Neuroimmunol 2021 360 577716 10.1016/j.jneuroim.2021.577716 34517152 8. Fujinami R Oldstone M Amino acid homology between the encephalitogenic site of myelin basic protein and virus: mechanism for autoimmunity Science 1985 230 1043 1045 10.1126/science.2414848 2414848 9. Basiri K Fatehi F Isaacs syndrome associated with chronic hepatitis B infection: a case report Neurol Neurochir Pol 2009 43 388 390 19742398
PMC009xxxxxx/PMC9005020.txt	==== Front J Behav Med J Behav Med Journal of Behavioral Medicine 0160-7715 1573-3521 Springer US New York 35415775 310 10.1007/s10865-022-00310-9 Article College students’ influenza vaccine hesitation: a reasoned action investigation with quantitative and qualitative data http://orcid.org/0000-0003-3981-1559 Mongeau Paul A. Paul.Mongeau@ASU.edu 1 https://orcid.org/0000-0002-9436-8652 Liu Yanqin 12 Hashi Emi C. 1 Roberto Anthony J. 1 1 grid.215654.1 0000 0001 2151 2636 Hugh Downs School of Human Communication, Arizona State University, Tempe, AZ 85287-1205 USA 2 grid.418204.b 0000 0004 0406 4925 Present Address: Banner MD Anderson Cancer Center, Gilbert, AZ 85234 USA 12 4 2022 2023 46 1-2 6575 4 9 2021 21 2 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. This two-wave longitudinal study (performed pre-COVID), using both quantitative and qualitative data, investigated college students’ influenza vaccine hesitancy and confidence using the theory of planned behavior (TPB). At Time 1, college students (n = 277) completed TPB measures and reported past influenza vaccine behavior. At Time 2 (30 days later), participants indicated whether they received the influenza vaccine since Time 1. At Time 2, participants who indicated that they had not received the influenza vaccine since Time 1 also described their most important reasons for not doing so. The TPB model fit the quantitative data well; direct paths from attitude and norms to intention, and from intention to future behavior, were strong and significant. The TPB model explained 71% of the variance in intention and 28% of the variance in future behavior. Neither perceived behavioral control nor past behavior improved the model’s ability to predict intentions or future behavior. From the qualitative data, participants’ reasons for not getting vaccinated focused on perceived behavioral control (e.g., time cost) and attitudes (e.g., unimportance and low susceptibility). Theoretical implications for message development are discussed. Keywords College students Flu shot Influenza vaccine Theory of planned behavior Vaccine hesitancy issue-copyright-statement© Springer Science+Business Media, LLC, part of Springer Nature 2023 ==== Body pmcIntroduction Vaccine hesitancy, or “the reluctance to receive recommended vaccination because of concerns and doubts about vaccines” (Dubé et al., 2021, p. 176), has existed since vaccines were first developed and represents an important public health issue (World Health Organization [WHO], 2014). Moreover, specific concerns vary across vaccines, locations, populations, and times (Dubé et al., 2021). Although COVID-19 is the most recent and pressing source of vaccine hesitancy there are other examples, such as human papillomavirus; measles, mumps, and rubella; and our specific focus, influenza vaccinations (i.e., the flu shot) among U.S. college students (National Foundation for Infectious Diseases [NFID], 2016, 2017). Motivating college students to get the influenza vaccine is challenging (Cornally et al., 2013; NFID, 2017). Although healthy college-age students typically perceive themselves as being at low risk (Czyz et al., 2019), influenza (or simply, the flu) is serious and highly contagious. The influenza virus spreads quickly in the close quarters common on college campuses. Flu symptoms can last up to eight days, cause students to miss class and/or work, and may lead to serious health complications (Nichol et al., 2010). Flu vaccination rates among college students are typically low and, therefore, the NFID called for “research to better understand and quantify…student motivators and influencers” (2017, p. 5). Answering this call, our primary goal is to investigate university students’ flu vaccination hesitancy using quantitative and qualitative data. Specifically, we use the theory of planned behavior (i.e., TPB) as a lens as it focuses on both individual and social determinants (Ajzen, 1985, 1991; Fishbein & Ajzen, 2010). Influenza and the influenza vaccine The flu is a serious, contagious, illness caused by the influenza virus (Centers for Disease Control and Prevention [CDC], 2019). The CDC estimated that between October 1, 2018 and March 9, 2019, nearly 30 million people in the U.S. became ill due to, and up to 35,500 people died from, the flu. Despite the flu vaccine’s efficacy, college-aged students typically do not comply with health recommendations (NFID, 2017). The flu is an important case from which to study vaccine hesitancy using the reasoned action approach (i.e., RAA). First, although college students underestimate the threat of the flu (Czyz et al., 2019; NFID, 2016), they are likely in close proximity with hundreds, if not thousands, of people daily in shared housing, classrooms, and face-to-face social interactions (e.g., study groups to sporting events). Second, seemingly healthy people can infect others one day before, and up to 5 to 7 days after, demonstrating flu symptoms (CDC, 2019). Third, the flu virus mutates constantly, necessitating annual vaccinations. Fourth, young-adult college students may be unfamiliar with independently making health decisions and, as a consequence, may not have taken control of health decisions (James et al., 2020). Finally, college students may perceive a lack of resources (e.g., time, money, and health insurance) to get the flu vaccine (NFID, 2016; Schmid et al., 2017). Reasoned action approach Considerable research focused on vaccinations in general (e.g., Brewer et al., 2017; Xiao & Wong, 2020) and flu vaccination in particular (e.g., Schmid et al., 2017). Although research utilized a variety of theories (Brewer et al., 2017), one popular framework is the RAA, which focuses on behavior, performed in a particular place and time, based on the information a person has at the time (Ajzen & Fishbein, 1980, 2005; Fishbein & Ajzen, 2010). The RAA encompasses both the theory of reasoned action and the TPB. Our investigation focuses on the latter, however, to fully understand it, a discussion of the former is necessary. Theory of reasoned action In its day, the theory of reasoned action (TRA: Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) was groundbreaking because it claimed that the sole determinant of behavior (i.e., actions in a particular situation) was not attitude, but behavioral intention (i.e., a person’s readiness to perform a behavior in the future). Considerable research across behavioral domains and audiences indicated that intentions are a strong, though imperfect, predictor of behavior (Fishbein & Ajzen, 2010). Behavioral intention, in turn, is determined by a person’s attitude toward the behavior (i.e., favorable or unfavorable behavioral evaluations) and subjective (or injunctive) norms (i.e., how important others as a whole evaluate behavioral performance; Fishbein & Ajzen, 1975; see the unshaded boxes in Fig. 1).Fig. 1 Path Model representation of the theory of reasoned action and the theory of planned behavior variables on flu shot intention and behavior. Note Unshaded boxes represent TRA variables (Ajzen, 1991; Ajzen & Fishbein, 1980). The lightly-shaded box, perceived behavioral control, was added in the TPB. Past behavior (the dark gray box) was suggested by Albarracin et al. (2001). For simplicity’s sake, paths among attitude, subjective norms, and perceived behavioral control are omitted from this diagram (Ajzen & Fishbein, 2005; Fishbein & Ajzen, 2010) In the TRA, both attitude and subjective norms have two predictors. Attitude is determined by behavioral beliefs (i.e., perceived consequences of behavior performance) and outcome evaluation (i.e., the positive or negative evaluation of each consequence). Subjective norms are based on normative beliefs (i.e., perceptions of what important individuals or groups expects him or her to do) and motivation to comply with each important individual or group. Therefore, beliefs—right or wrong—impact attitudes and norms (and ultimately intention and behavior) that are ripe targets for health communication interventions. Theory of planned behavior The TPB (Ajzen, 1985, 1991) extended TRA by adding a third predictor of intentions, i.e., perceived behavioral control (i.e., the extent to which a person believes that he or she is capable of, or has control over, behavioral performance; Fishbein & Ajzen, 2010; the lightly-shaded box in Fig. 1). Perceived behavioral control is “virtually identical” to self-efficacy (Fishbein & Ajzen, 2010, p. 161). In the present context, high perceived behavioral control likely represents vaccine confidence. Perceived behavioral control is a proxy for, and is easier to measure than, actual control and is influenced by two factors: control beliefs, (i.e., factors thought to influence behavioral performance or control) and the power of each control belief (i.e., how strongly each control belief might facilitate or impede behavioral performance or control). In sum, the TPB adds perceived behavioral control as a predictor of both behavior and intentions and adds two determinants for perceived control. Meta-analytic results and past behavior Several meta-analyses assessed TRA/TPB variables’ ability to predict an array of health-related behaviors and intentions among a plethora of audiences (e.g., Albarracin et al., 2001; Armitage & Conner, 2001; Downs & Hausenblas, 2005; McEachan, et al., 2011; Rich et al., 2015). Moreover, Xiao and Wong (2020) meta-analyzed TRA/TPB studies that focused on vaccination intentions. Together, results from these six TRA/TPB meta-analyses (five general and one on vaccinations) provide three important insights that can help frame our results. First, the six meta-analyses provide effect size benchmarks that provide useful comparisons for our data. Table 1 presents the range of effect size (i.e., mean weighted correlations) for all TPB relationships for the five general TRA/TPB meta-analyses and, separately, from Xiao and Wong’s (2020) vaccination meta-analysis. For the five general TRA/TPB meta-analyses, ranges for all effects fell into Cohen’s (1988) medium-to-large range. In the Xiao and Wong (2020) vaccination meta-analysis, mean weighted correlations between both attitudes and subjective norms and intentions were larger than those reported in the general TRA/TPB meta-analyses. The mean weighted correlation between perceived behavioral control and intentions in the vaccination meta-analysis, however, was at the lower end of the range in the general TRA/TPB meta-analyses.Table 1 Meta-analytic results on the influence of TPB variables on intentions and behavior Effect Genera meta-analyses Vaccination meta-analysis Attitude → intention .41–.57 .64 Subjective norms → intention .28–.40 .61 Perceived behavioral control → intention .41–.54 .42 Perceived behavioral control → behavior .24–.37 – Intention → behavior .41–.54 – Variance explained in intention 33%–44% 51.9% Variance explained in behavior 9%–27% – Notes Meta-analytic range = the range of average weighted correlations from past TRA/TPB meta-analyses (i.e., Albarracin et al., 2001; Armitage & Conner, 2001; Downs & Hausenblas, 2005; McEachan, et al., 2011; Rich et al., 2015). Vaccination meta-analysis: average weighted correlation coefficients from the Xiao and Wong (2020) meta-analysis of TRA/TPB studies specifically on vaccinations Second, meta-analyses provide benchmarks for variance explained in intentions and future behaviors (see Table 1). Specifically, in the Xiao and Wong (2020) vaccination meta-analysis, TPB predictors (attitude, subjective norms, and perceived behavioral control) explained more variation (52%) in intention than did the five general TRA/TPB meta-analyses (33–44%). With the addition of intentions, the general TRA/TPB meta-analyses were able to explain between 9 and 27% of variation in future behavior. The Xiao and Wong (2020) vaccination meta-analysis could not consider behavior because so few studies measured future behavior. Third, McEachan et al. (2011) note that past behavior (the dark gray box in Fig. 1) may inform attitude, subjective norms, perceived behavioral control, intentions, and future behavior (see also Albarracin et al., 2001). The weighted-mean correlations between past behavior and attitudes (r¯ = 0.32), subjective norms (r¯ = 0.22), perceived behavioral control (r¯ = 0.33), intention (r¯ = 0.47), and future behavior (r¯= 0.50) were, for the most part, in the moderate-to-strong range (Cohen, 1988). McEachan et al. added past behavior as an exogenous variable in a TPB model (see Fig. 1). Including past behavior in the meta-analytic TPB model increased the proportion of variance explained in both intentions (5%) and future behavior (11%), but attenuated relationships between all predictors and both intention and future behavior. Reasons for non-vaccination According to WHO (2014), vaccine hesitancy stems from a complex interplay of factors that influence the decision to accept none, a few, or all vaccines. The RAA asserts that the most important factor in (in)action is an individual's beliefs about the behavior in question (i.e., getting the flu vaccine) as they compose attitudes toward the behavior, subjective norms, and perceived behavioral control. These beliefs are often identified as barriers to flu vaccination (Schmid et al., 2017). This investigation’s second goal, then, is to understand college students’ beliefs related to vaccine hesitancy by investigating their reasons for not getting the flu vaccine. The 3C’s model (confidence, complacency, and convenience; WHO, 2014) represents a simple typology for describing people’s reasons for not getting vaccinated. Confidence focuses on trust (or lack of the same) in the vaccine, the health-care system, and policy makers. Complacency focuses on risks associated with the disease in question and the necessity of vaccination. Finally, convenience centers upon accessibility, affordability, and understandability of vaccination services. Schmid et al. (2017) report that each of these broad categories are identified as barriers to vaccination uptake among at-risk groups. Considering reasons for not being vaccinated from the TPB perspective and the 3C’s model should generate a clearer understanding of college students’ thinking that can help focus persuasive messaging. Research questions and hypotheses The primary goal of this longitudinal study is to investigate vaccination hesitancy by assessing TPB variables’ ability to predict and explain college students’ flu vaccine intentions and future behavior. A second goal is to assess college students’ most important reasons for not getting the flu vaccine. Thus, we advance the following hypothesis and research questions:H1: TPB variables (unshaded or lightly-shaded boxes in Fig. 1) will predict college students’ flu vaccine intentions and future behavior. RQ1: Does past behavior increase the predictive power of TPB variables on college students’ flu vaccine intentions and future behavior? RQ2: What are college students’ reasons for not getting the flu vaccine? Method Participants Time 1 Undergraduate students (n = 565), enrolled in three lower-division classes at a large southwestern US university,1 volunteered to participate for a small amount of course credit during late September 2018 (i.e., pre-COVID-19). Participants (n = 83) who had already received the flu vaccine were directed to an unrelated study. Thus, 482 participants completed the Time 1 survey (52.0% male, 47.6% female, and 0.4% other/non-binary; mean age = 19.04, SD = 1.59). Participants were 69% white, 13.5% Asian, 5.5% African American, 1.0% native Hawaiian or other Pacific Islander, 0.8% American Indian, and 9.9% other. Additionally, 20.7% identified as Hispanic or Latino/a. Time 2 Thirty days later2 (i.e., early November, 2018), the 482 students in the same three classes were invited to complete the Time 2 survey for additional course credit. Over half of eligible participants (n = 277; 57.5%) participated at Time 2. Non-response analysis of continuous measures revealed no significant differences: attitude (t = -0.21, df = 481, p > 0.05), subjective norms (t = -0.55, df = 481, p > 0.05), perceived behavioral control (t = 0.37, df = 481, p > 0.05), intention (t = 0.70, df = 481, p > 0.05), or age (t = -0.30, df = 481, p > 0.05). Similarly, chi-square tests revealed no differences for categorical variables: gender (χ2 = 3.49, df = 2, p > 0.05), race (χ2 = 8.05, df = 6, p > 0.05), or ethnicity (χ2 = 0.01, df = 1, p > 0.05). Participants who completed both Time 1 and Time 2 surveys were slightly less likely to have ever received a flu shot than those who completed only the Time 1 survey (χ2 = 3.92, df = 1, p = 0.048). Instrumentation Time 1 survey All TPB measures were adapted from Ajzen and Fishbein (1980) and Fishbein and Ajzen (2010). Attitude was assessed using five, five-point, semantic differential items (e.g., “To me, getting the flu shot is: harmful-beneficial”). All other continuous variables were measured with Likert-type scales (i.e., 1 = “strongly disagree” to 5 = “strongly agree”): subjective norms (four items; e.g., “My close friends think that I should get the flu shot”); perceived behavioral control (five items; e.g., “Getting the flu shot is completely up to me”), and intention (four items; e.g., “I intend to get the flu shot in the next 30 days”). Past behavior was assessed with one item (i.e., “Have you ever received a flu shot?”) with response options of “yes,” no,” and “I do not remember.” Finally, both Time 1 and Time 2 surveys included demographic items and two linking items (day of the month of their birth and the last two digits of their phone number) that facilitated pairing participants’ Time 1 and Time 2 responses. Time 2 Survey The Time 2 survey was administered 30 days after the Time 1 survey. Future behavior (i.e., using Time 1 as a reference) was measured using a single item: “Have you received a flu shot in the past 30 days (i.e., since completing Part 1 of this study)?” Response options were yes and no. Participants who answered no also provided responses to an open-ended item that read, in part: “In the space below, please indicate the most important reasons why you did not get a flu shot in this time frame.” The Time 2 survey also included demographic and linking items. Table 2 contains descriptive statistics, reliabilities, and correlations among all RAA variables.Table 2 Reliability, descriptive statistics, and zero-order correlations for model variables α M SD 1 2 3 4 5 1. Past behavior (1 item) NA 0.77 .42 – 2. Attitude (5 items) .95 3.60 1.08 .50* – 3. Subjective norms (4 items) .86 2.92 1.00 .44* .63* – 4. Perceived behavioral control (5 items) .75 4.22 .69 .20* .41* .31* – 5. Intention (4 items) .97 2.74 1.33 .39* .74* .70* .36* – 6. Future Behavior (1 item) NA 0.19 .39 .18* .32* .24* .17* .35* Notes Past behavior, attitude, subjective norms, perceived behavioral control, and intention were measured at Time 1; future behavior at Time 2. All variables, except past and future behavior (coded as 0 = No, 1 = Yes), were measured with five-point scales. Sample sizes for correlations containing past behavior are n = 246; correlations between all other variables are n = 277 p < .001 Procedures These data were part of a larger investigation of college students’ flu vaccination behavior (Roberto et al., 2019). Procedures were approved by the relevant institutional review board. Participants completed two surveys: The Time 1 survey included theoretical variables and the Time 2 (30 days later) survey included flu vaccination behavior since Time 1. Participants who had not received a flu vaccine at Time 2 also described their reasons for not having done so. Data analytic plan Hypothesis 1 (i.e., testing the TPB causal model; see Fig. 1) was tested with the path analyses option in Mplus 8.3, a structural equation modeling package (Muthén & Muthén, 2019). First, analyses tested the extent to which the model (without past behavior) fit the data. Second, path analysis tested the extent to which past behavior improved model fit. Participants (n = 31) who responded “don’t remember” to the Time 1 past behavior item were removed from TPB model tests of past behavior. Using a mean- and variance-adjusted weighted least squares estimator (WLSMV, the default option in Mplus for a dichotomous dependent variable; Kline, 2016; Muthén & Muthén, 2017), chi-square, comparative fit index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR) assessed model fit. A nonsignificant chi-square test, both RMSEA and SRMR ≤ 0.08, and CFI > 0.95 suggest strong fit (Hu & Bentler, 1999; Kline, 2016). R2 assessed the models’ predictive power for intentions and future behavior. To answer Research Question 2, one author generated TPB-based categories after reviewing participants’ reasons for not receiving a flu shot during the previous 30 days. To pilot-test initial codes, two authors independently analyzed 10% of the 213 qualitative responses, compared coding decisions, and resolved discrepancies through discussion to consensus. The same two authors then refined the codebook by modifying existing, and adding new, categories. Both authors independently analyzed all remaining responses and discussed discrepancies to consensus. This iterative process generated 15 categories based on TPB and past behavior (see Table 4). When participants provided multiple reasons, all responses were placed into the categories. Results On average, participants reported mildly positive attitudes toward the flu vaccine, neutral subjective norms, positive perceived behavioral control, and neutral behavioral intentions (see Table 2). Moreover, 19% of the Time 2 sample indicated that they received a flu vaccination during the previous 30 days. Consistent with the TPB, attitudes, subjective norms, and perceived behavioral control were positively and significantly related to behavioral intentions. Both behavioral intentions and perceived behavioral control were significantly and positively correlated with future behavior. Testing the TPB model Hypothesis 1 predicted that TPB variables (measured at Time 1) will predict college students’ flu vaccine behavior (i.e., future behavior, measured at Time 2). Fit between the model and the data was strong, χ2 (2) = 6.09, p = 0.048, RMSEA = 0.09, CFI = 0.98, SRMR = 0.03; for regression coefficients, see Table 3. Attitude and subjective norms significantly predicted behavioral intention which, in turn, strongly predicted future behavior. Perceived behavioral control did not significantly predict either behavioral intention or future behavior. Predictors accounted for considerably more variance in intentions (R2 = 0.71) than behavior (R2 = 0.28). Overall, H1 was supported.Table 3 Correlations and path coefficients with and without past behavior included in the model Effect Correlation Standardized path coefficients Without past behavior With past behavior Attitude → intention .74 .60 .53* Subjective norms → intention .70 .31* .37* Perceived behavioral control → intention .36 .02 .05 Perceived behavioral control→ future behavior .17 .12 .09 Intention → future behavior .35 .47* .46* Past behavior → attitude .50* Past behavior → subjective norms .44* Past behavior → perc. beh. control .20* Past behavior → intention − .04 Past behavior → future behavior .11 R2 for intention .71 .68 R2 for future behavior .28 .31 *p < .001 Past behavior and the TPB Research Question 1, whether adding past behavior to the TPB model would increase predictive power, was investigated with a second path analysis. Paths between past behavior and all TPB variables were added (see Fig. 1) and results are presented in Table 3. Fit between the model and the data was also strong, χ2 (2) = 5.94, p = 0.051, RMSEA = 0.09, CFI = 0.99, SRMR = 0.02. Past behavior significantly predicted attitudes toward the behavior, subjective norms, and perceived behavioral control, but neither intention nor future behavior. Again, predictors accounted for more variance in intentions (R2 = 0.68%) than future behavior (R2 = 0.31%). In short, past behavior did not improve the predictive power of the TPB model. College students’ reasons for not receiving the influenza vaccine Research Question 2 focused on participants’ reasons for not getting a flu shot and how those reasons related to TPB constructs. Participants (n = 224) provided 323 reason(s) for not getting a flu vaccine (M = 1.44 reasons per participant). All provided reasons were placed into one of 15 main categories (see Table 4; excluding other and no response categories). Over half of all responses fell into nine categories that were relevant to TPB’s attitudes toward the behavior. Four categories fell within perceived behavioral control (approximately one-third of all responses), while a single category (3% of all responses) fell within subjective norms. Finally, past behavior was used as a rationale for not getting a flu vaccine by 6% of participants.Table 4 Reasons participants did not get a flu shot Reason Definition Example Frequency (%) Attitude-based reasons Unimportance The flu shot wasn’t of value, necessary, or beneficial to participants I don’t believe I need it 45 (13.9) Low susceptibility Participants believe they are healthy, or they won’t get sick/get the flu without a flu shot I have not had the flu in years and am pretty healthy 34 (10.5) Lack of priority Participants mentioned that they prioritized other commitments, or considered getting a flu shot but forgot It’s the last thing to do on my list 25 (7.7) Side effects Individual side effects stop participants from getting a flu shot or they believe the flu shot will make them sick It hurts my arm for a few days 23 (7.1) Low salience Participants do not think about getting the flu shot or were not motivated to get the flu shot The thought was not on my mind 20 (6.2) Ineffectiveness Participants do not believe or are hesitant that a flu shot is effective at protecting them from the flu The flu shot doesn’t stop me from getting the flu 11 (3.4) Fear of needles Participants have a fear of needles or shots Truthfully, I just have an irrational fear of needles. I get scared just thinking about needles 8 (2.5) Alternative recommendations Participants think they can avoid the flu through other strategies (e.g., washing hands) I do not believe in getting shots for sicknesses that can be avoided simply by washing your hands 6 (1.9) Global vaccine concerns Participants are against vaccines in general, or have never had any vaccinations I am against all vaccinations in general, I think they are just putting in a bunch of harmful chemicals into your body and injecting the sickness 3 (0.9) Subjective norms-based reasons Family influence Parents or family members influenced participants’ decision not to get a flu shot My parents don’t believe in flu shots 11 (3.4) Perceived behavioral control-based reasons Time cost Participants “did not have time” or it would be inconvenient/take too much time to get a flu shot I didn’t have time to go out and get it 69 (21.4) Monetary cost Participants did not have money for a flu shot or did not want to pay for a flu shot I’ve been super broke over the past month. I can’t afford the $40 it normally takes at the moment 10 (3.1) Lack of knowledge Participants did not have enough knowledge about the flu shot I’m not sure if my insurance will cover it 9 (2.8) Contraindications Participants believe that an allergy prevents them from getting a flu shot I am allergic to eggs, which makes it not possible for me to get the flu shot as I would have a severe allergic reaction 2 (0.6) Past behavior Previous experiences or past behaviors are used as justification as to why participants have not gotten the flu shot this year I never have and never really got the flu, so why should I? 20 (6.2) Five categories encompassed over 60% of responses. The most common reason was time cost (not having the time to get a flu vaccine; 21.4%). Second, unimportance indicated that participants did not consider the flu vaccine as urgent, important, or necessary (13.9%). Third, in low susceptibility responses, participants described themselves as healthy and having a strong immune system (10.5%). Fourth, participants expressed concerns about the flu vaccine’s side effects (7.1%). Finally, participants referred to their past behavior (e.g., never or rarely having received a flu vaccine) as a reason for not having received a vaccination within the past 30 days. Discussion Vaccine hesitancy represents an important public health issue that varies across vaccines and populations (WHO, 2014). Consequently, considerable research, from a number of theoretical frames (for a review, see Brewer et al., 2017), focused on bolstering intention and vaccination uptake (e.g., Capasso et al., 2021; Conner et al., 2017). The present longitudinal study, focused on flu vaccine confidence and hesitancy among U.S. college students, was driven by two goals. Our first goal was to quantitatively assess vaccine hesitancy and confidence using the TPB (Ajzen, 1985, 1991; Fishbein & Ajzen, 2010). In the TPB, attitudes toward the behavior, subjective norms, and perceived behavioral control are predicted to influence behavioral intentions. Intentions and perceived behavior control, in turn, are predicted to influence future behavior. We also investigated past behavior as a predictor of all TPB variables (Albarracin et al., 2001; McEachan et al., 2011; see Fig. 1). This study’s second goal was to identify college students’ reasons for not receiving a flu vaccine. These qualitative data speak to specific elements underlying flu vaccine hesitancy. We discuss results as they relate to our goals before considering implications for health messaging directed toward college students as well as the study’s strengths and limitations. Theory of planned behavior From the RAA perspective (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 2010), our results are curiously mixed. On the positive side, consistent with Hypothesis 1, the TPB model closely fit the data and explained considerable variation in intentions and future behaviors (see Table 3). Both attitudes toward the behavior and subjective norms strongly and positively predicted behavioral intentions, which, in turn, strongly predicted future behaviors. Consistent with both the TRA and TPB, participants with positive attitudes and close friends and family who support flu vaccination had strong intentions to get a flu shot. Although the TPB model fit our data well, neither perceived behavioral control nor past behavior significantly predicted behavioral intentions or future behavior. The perceived behavioral control result is curious not only because it was inconsistent with TPB, but mean values of perceived behavioral control were highest of all measured variables (i.e., 4.22 on a five-point scale). Furthermore, past behaviors significantly predicted participants’ attitudes, subjective norms, and perceived behavioral control, but neither intentions nor future behaviors. Although past and future behavior are significantly correlated in the present data, their relationship is weak when compared with the McEachan et al. (2011) meta-analysis. College students’ betwixt and between status in dealing with their own health may explain, in part, why neither perceived behavioral control nor past behavior predicted behavioral intention or future behavior. Before coming to college, students’ health care interactions (e.g., appointments, tests, and vaccinations) were likely orchestrated by parents or guardians (Curtis, 2015). Once arriving at college, students act more independently, however, they remain tethered to family patterns and influences (Curtis, 2015; James et al., 2020). As a consequence, students’ independent experiences with health care are likely limited. So, even though a college student might have received the flu vaccine in the past, it may not have been their decision. This explanation likely speaks to the difference between perceived and actual behavioral control. New students might feel that getting a flu vaccine is up to them, however, they may lack the experience interacting with the health-care system to be able to do so effectively. Reasons for not getting a flu shot Coding identified 15 categories that described participants’ most important reasons for not getting the flu vaccine within the past 30 days (see Table 4), which have multiple implications for vaccine hesitancy and health-intervention messaging. The four largest response categories (i.e., time cost, unimportance, low susceptibility, and lack of priority) clearly indicate that college students’ flu-related risk assessments do not generate a sense of urgency (NFID, 2016, 2017; Schmid et al., 2017). Given that time is a precious commodity, there is little space for what are seen as superfluous activities. Given other responsibilities (e.g., school, work, and relationships), taking the time to get the flu vaccine is not seen as worth the effort, especially for healthy participants who consider their immune system to be robust. Second, from the TPB perspective, a vast majority of responses reflected individual (i.e., attitudes toward the behavior and perceived behavioral control), rather than social, considerations. Put another way, participants rarely mentioned other people in their reasons for not getting the flu vaccine. Family members (especially parents) were the only specific individuals mentioned. Thus, subjective norms were represented solely by family influence. The few cases that mentioned peers typically described a friend’s past experience with the flu, rather than beliefs about the flu vaccine. The absence of others’ influence beyond parents seems inconsistent with the quantitative results indicating that subjective norms influenced intentions. Third, reasons for not getting the flu vaccine are clearly consistent with the 3C’s Model (WHO, 2014; see also Schmid et al., 2017). Confidence concerns are evident in the side effects, ineffectiveness, and global vaccine concern categories. There is little evidence of a lack of confidence in health infrastructures. Complacency concerns are reflected in unimportance, low salience, and low susceptibility reasons. Finally, convenience is represented in the time cost, monetary cost, and lack of knowledge categories. The lack of knowledge category is particularly interesting and sheds further light on college students betwixt and between status in terms of their own health. Participants’ open-ended responses shed light why neither perceived behavioral control and past behavior were unrelated to both intentions and behaviors in the SEM results. For example, some participants (particularly out-of-state students) were unsure whether their parents’ health insurance would cover vaccination cost or how to complete the requisite paperwork (James et al., 2020; Schmid et al., 2017). Other categories indicated that some participants lacked resources (e.g., money3 or knowledge) necessary to get the flu vaccine, or that it was inconvenient to do so. That all of the major components of the 3C’s model appear in our qualitative findings reflects the complexity of college students’ vaccine hesitancy (Schmid et al., 2017; WHO, 2014). Specifically, given the scarcity of students’ time and resources, getting a flu vaccine appears to be an unnecessary luxury, or even a health hazard. Finally, responses reflecting the confidence component indicate that flu vaccine hesitancy applies specifically to the flu vaccine (e.g., side effects, ineffectiveness, and contraindication), rather than global vaccination concerns (e.g., anti-vaccination beliefs; Dubé et al., 2021). Interestingly, some participants’ reasons related to side effects reflected misinformation (e.g., the flu shot, or the chemicals in it, can give you the flu). Implications for health communication campaigns Our results provide multiple suggestions for messages designed to persuade college students to get their annual flu vaccine. Most importantly, messages should attempt to increase students' flu vaccination intentions by creating a sense of urgency. Messages should emphasize the importance of getting the flu vaccine by highlighting students’ susceptibility (e.g., environmental factors). Brewer et al. (2017) note that few studies have utilized fear appeals to motivate vaccination uptake. Such messages could also emphasize both response efficacy (e.g., vaccines reduce the threat of getting the flu) and self-efficacy (e.g., the individual’s ability to get the flu vaccine) by including information on the availability of vaccinations and how students can use their own (or their parents’) health insurance (Curtis, 2015; James et al., 2020). Second, given very limited global anti-vaccination sentiment, messages can attempt to increase flu vaccine confidence by directly addressing reasons for college students’ hesitancy. For example, messages should highlight the extent to which, and how, the flu vaccine provides protection. What is more, messages should attempt to combat misinformation (e.g., flu shots, or chemicals in them, do not cause the flu; flu shots take two weeks to provide protection). Finally, many participants provided multiple reasons for not getting the flu vaccine. This indicates that college students’ hesitancy suggests a complicated decision-making process (Dubé et al., 2021; WHO, 2014). Health campaign designers should consider utilizing arguments that combat multiple reasons (rather than a single determinant) to enhance message effectiveness. Strengths and limitations This investigation adds to existing literatures given four key strengths. First, this study is based on, and extends, the RAA to an important but understudied topic and audience (i.e. college students and the flu vaccine). Xiao and Wong’s (2020) meta-analysis of TRA/TPB vaccination studies included only three studies on the flu vaccine. Second, this study’s longitudinal design assessed actual behavior 30 days after other TPB variables. This design allows for the prediction of future behavior, something of a rarity, as most TRA/TPB vaccination studies either measured all variables at the same time or do not measure behavior at all (Xiao & Wong, 2021). This is important because multiple meta-analyses indicate that intentions correlate much less strongly with future behavior than past behavior. In the present study, however, past behavior predicted neither intentions nor future behavior. Confidence in our data and analyses also stems from three sources. First, we used established instruments (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 2010) with high reliabilities (i.e., alphas = 0.75–0.95). Second, our final sample size (n = 277) provides ample power for moderate-to-large effects reported in TRA/TPB meta-analyses (e.g., Xiao & Wong, 2020). Third, both qualitative and qualitative data adds depth to our interpretations. For example, inconsistency between quantitative (i.e., reflecting high perceived behavior control) and qualitative data (i.e., reflecting less control) may highlight the difference between actual and perceived control. This study’s final strength is generalizability to other vaccine-preventable diseases faced by college students (e.g., human papillomavirus and meningitis B; Xiao & Wong, 2020). Our results suggest that college students’ lack of experience in independently managing their health care might also interfere with vaccinations beyond influenza.4 One potential limitation to this study is Time 1 to Time 2 attrition. Only 277 of the original 482 participants (57.5%) completed the Time 2 survey 30 days later. While not ideal, dropouts are nearly inevitable and nonresponse analyses indicated that the only significant difference was that individuals completing the Time 2 survey were slightly less likely to have ever received the flu vaccine.5 Participants may have thought that they could not receive extra credit twice for the same study (a typical policy) or might not have needed, or benefited from, the small amount of extra credit promised one month later in the semester. Directions for future research Future research on persuading college students to get the flu vaccine should utilize the most recent iteration of the RAA the integrative model of behavioral prediction (IM; Fishbein, 2000, 2008; Fishbein & Ajzen, 2010; Institute of Medicine, 2002). The integrative model adds two factors that influence actual control over a behavior: (1) knowledge and skills and (2) environmental constraints. Knowledge and skills are individual factors that affect an ability to perform the behavior (e.g., understanding how, or having the expertise, to perform the behavior). For example, messages for new students should explain how health insurance can be used. Environmental factors, on the other hand, are external factors that facilitate or inhibit behavior. Examples include relationships (e.g., social ties), community (e.g., the physical environment), and societal factors (e.g., public policy) that influence behavior. Flu-shot campaigns should also make it as easy and inexpensive as possible to perform the behavior. Further, like perceived behavior control, both knowledge and skills and environment factors influence the strength of the relationship between intention and behavior. Open-ended responses indicate that some students see the flu shot as something they should do, but just do not have the resources to make it happen. Making it easy for students to get the flu vaccine will help reduce slippage between intentions and behaviors. Finally, future research should investigate college students’ reasons for not getting, and for getting, the flu shot. Open-ended responses highlighted barriers college students face when considering the flu vaccine (Schmid et al., 2017). Asking why students did get a flu vaccine would likely be equally insightful (e.g., underscoring parental influence in health decisions). Such data would broaden our understanding of factors that influence this important behavior. Conclusion The goals of this study were (1) to quantitatively assess the TPB’s ability to predict college students’ annual flu vaccination behavior and (2) to qualitatively identify their primary reasons for not getting the flu vaccine. In tandem, results suggest that TPB provides important insights regarding this topic and audience, and also identifies over a dozen specific concerns that need to be addressed when developing messages and interventions. Finally, many of our findings should generalize to other vaccine-preventable diseases for this population. Author contributions PM was involved in study design, and was primarily responsible for writing the final drafts of the manuscript as well as facilitating revisions. YL was involved in study design, data analysis, assisted in writing the original manuscript and participated in revisions. EH was involved in study design, qualitative analyses, and commented on the completed, and revised, manuscript. AJR was involved in study design; writing early drafts of the theory, method, and discussion sections; and edited and commented on revisions. Funding The authors have not disclosed any funding. Data availability Data and materials are available upon request. Code availability Not applicable. Declarations Conflict of interest The authors declare that they have no conflict of interest. Human and animal rights and Informed consent This study was ethically approved by the Institutional Review Board at Arizona State University (USA) and performed in accordance with the criteria defined by the rules of the committee. 1 Vaccinations were available from all of the university’s Student Health Services locations (and multiple vaccination events around campus) during the flu season (including the data-collection period). Student Health Services repeatedly communicated flu vaccination availability, using multiple media, during that time. 2 A 30-day data-collection period was chosen for several reasons. First, multiple meta-analyses report significantly larger intention-behavior associations when studies utilize time frames of four weeks or less (Downs & Hausenblas, 2005; McEachan et al., 2011; Topa & Moriano, 2010). Second, the period provided participants a chance to decide whether (or not) to get a flu shot at the beginning of the flu season (as recommended by Student Health Services). Third, it allowed us to complete the study in time to return extra credit information to instructors before the end of the semester. 3 The university provided free flu shots for students with proof of health insurance and charged $20 for students without such proof. During the data-collection period, off-campus businesses (e.g., drug and grocery stores) charged $20-$50 for the same service. 4 Our study, performed during Fall Semester 2018, preceded COVID-19 emergence, which facilitates understanding of college students’ vaccine decisions without concerns about COVID-19 as a confounding factor. 5 Given the small difference, a probability level of .048, and the number of statistical tests performed in nonresponse analyses, this effect could represent a Type 1 error. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Ajzen I Kuhl J Beckman J From intentions to actions: A theory of planned behavior Action control: From cognition to behavior 1985 Springer 11 39 Ajzen I The theory of planned behavior Organizational Behavior and Human Decision Processes 1991 50 179 211 10.1016/0749-5978(91)90020-T Ajzen I Fishbein M Understanding attitudes and predicting social behavior 1980 Prentice-Hall Ajzen I Fishbein M Albarracin D Johnson BT Zanna MP The influence of attitudes on behavior The handbook of attitudes 2005 Lawrence Erlbaum Associates Publishers 173 221 Albarracin D Johnson BT Fishbein M Muellerleile PA Theories of reasoned action and planned behavior as models of condom use: A meta-analysis Psychological Bulletin 2001 127 142 161 10.1037/0033-2909.127.1.142 11271752 Armitage CJ Conner M Efficacy of the theory of planned behaviour: A meta-analytic review British Journal of Social Psychology 2001 40 471 499 10.1348/014466601164939 11795063 Brewer NT Chapman GB Rothman AJ Leask J Kempe A Increasing vaccination: Putting psychological science into action Psychological Science in the Public Interest 2017 18 149 207 10.1177/1529100618760521 29611455 Capasso M Caso D Conner M Anticipating pride or regret? Effects of anticipated affect focused persuasive messages on intention to get vaccinated against COVID-19 Social Science and Medicine 2021 289 114416 10.1016/j.socscimed.2021.114416 34562773 Centers for Disease Control and Prevention (2019). Influenza (flu). Retrieved from https://www.cdc.gov/flu/index.htm Cohen J Statistical power analysis for the behavioral sciences 1988 2 Lawrence Erlbaum Associates Conner M Sandberg T Nekitsing C Hutter R Wood C Jackson C Godin G Sheeran P Varying cognitive targets and response rates to enhance the question-behaviour effect: An 8-arm randomized controlled trial on influenza vaccination uptake Social Science and Medicine 2017 180 135 142 10.1016/j.socscimed.2017.03.037 28347938 Cornally N Deasy EA McCarthey G McAuley C Moran J Weathers E Student nurses’ intention to get the influenza vaccine British Journal of Nursing 2013 22 1207 1211 10.12968/bjon.2013.22.21.1207 24280920 Curtis AC Defining adolescence Journal of Adolescent and Family Health 2015 7 1 39 Czyz SE Miller JY Muniz HM Abraham SP Gillum DR College students’ perceptions of influenza vaccination and childhood immunizations International Journal of Studies in Nursing 2019 4 66 75 10.20849/ijsn.v4i2.582 Downs DS Hausenblas HA The theories of reasoned action and planned behavior applied to exercise: A meta-analytic update Journal of Physical Activity and Health 2005 2 76 97 10.1123/jpah.2.1.76 Dubé È Ward JK Verger P MacDonald NE Vaccine hesitancy, acceptance, and anti-vaccination: Trends and future prospects for public health Annual Review of Public Health 2021 42 175 191 10.1146/annurev-publhealth-090419-102240 33798403 Fishbein M The role of theory in HIV prevention AIDS Care 2000 12 273 278 10.1080/09540120050042918 10928203 Fishbein M A reasoned action approach to health promotion Medical Decision Making 2008 28 834 844 10.1177/0272989X08326092 19015289 Fishbein M Ajzen I Belief, attitude, intention, and behavior 1975 Addison-Wesley Fishbein M Ajzen I Predicting and changing behavior: The reasoned action approach 2010 Psychology Press Hu L Bentler PM Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives Structural Equation Modeling: A Multidisciplinary Journal 1999 6 1 55 10.1080/10705519909540118 Institute of Medicine Speaking of health: Assessing health communication strategies for diverse populations 2002 National Academies Press James TG Sullivan MK Dumeny L Lindsey K Cheong J Nicolette G Health insurance literacy and health service utilization among college students Journal of American College Health 2020 68 200 206 10.1080/07448481.2018.1538151 30526397 Kline RB Principles and practice of structural equation modeling 2016 4 The Guilford Press McEachan RRC Conner M Taylor NJ Lawton RJ Prospective prediction of health-related behaviours with the theory of planned behaviour: A meta-analysis Health Psychology Review 2011 5 97 144 10.1080/17437199.2010.521684 Muthén, L. K.., & Muthén, B. O. (2019). Mplus (Version 8.3). Available from http://www.statmodel.com/ Muthén LK Muthén BO Mplus statistical analysis with latent variables: User's guide 2017 8 Muthén & Muthén National Foundation for Infectious Diseases. (2016). Addressing the challenges of influenza vaccination on U.S. college campuses. https://www.nfid.org/wp-content/uploads/2019/08/college-flu-summit-report-2.pdf National Foundation for Infectious Diseases (2017). National survey on college students and flu. http://www.nfid.org/idinfo/influenza/college-students-flu-survey.html Nichol KL Tummers K Hoyer-Leitzel A Marsh J Moynihan M McKelvey S Modeling seasonal influenza outbreak in a closed college campus: Impact of pre-season vaccination, in-season vaccination and holidays/breaks PLoS ONE 2010 5 e9548 10.1371/journal.pone.0009548 20209058 Rich A Brandes K Mullan B Hagger MS Theory of planned behavior and adherence in chronic illness: A meta-analysis Journal of Behavioral Medicine 2015 38 673 688 10.1007/s10865-015-9644-3 25994095 Roberto AJ Mongeau PA Liu Y Hashi E Fear the flu, not the flu shot: A test of the extended parallel process model Journal of Health Communication 2019 24 829 836 10.1080/10810730.2019.1673520 31646953 Schmid P Rauber D Betsch C Lidolt G Denker M-L Barriers of influenza vaccination intention and behaviour—A systematic review of influenza vaccine hesitancy, 2005–2016 PLoS ONE 2017 12 e0170550 10.1371/journal.pone.0170550 28125629 Topa G Moriano JA Theory of planned behavior and smoking: Meta-analysis and SEM model Substance Abuse and Rehabilitation 2010 1 23 33 10.2147/SAR.S15168 24474850 World Health Organization. (2014). Report of the SAGE working group on vaccine hesitancy. https://www.who.int/immunization/sage/meetings/2014/october/1_Report_WORKING_GROUP_vaccine_hesitancy_final.pdf Xiao X Wong RM Vaccine hesitancy and perceived behavioral control: A meta-analysis Vaccine 2020 38 5131 5138 10.1016/j.vaccine.2020.04.076 32409135
PMC009xxxxxx/PMC9005021.txt	==== Front J Gen Intern Med J Gen Intern Med Journal of General Internal Medicine 0884-8734 1525-1497 Springer International Publishing Cham 35415791 7560 10.1007/s11606-022-07560-y Perspective Closer to or Farther away from an Ideal Model of Care? Lessons Learned from Geographic Cohorting http://orcid.org/0000-0001-5107-3857 Kara Areeba MD, MS akara@iuhealth.org 1 Kashiwagi Deanne MD, MS 23 Burden Marisha MD 4 1 grid.257413.6 0000 0001 2287 3919 Division of General Internal Medicine and Geriatrics, Indiana University School of Medicine, Indianapolis, IN USA 2 grid.66875.3a 0000 0004 0459 167X Mayo Clinic, Rochester, MN USA 3 grid.66875.3a 0000 0004 0459 167X Sheikh Shakhbout Medical City Hospital (Mayo Clinic Abu Dhabi), Rochester, MN USA 4 grid.241116.1 0000000107903411 Division of Hospital Medicine at Denver Health, University of Colorado, Denver, CO USA 12 4 2022 9 2022 37 12 31623165 30 11 2021 30 3 2022 © The Author(s), under exclusive licence to Society of General Internal Medicine 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Geographic “cohorting,” “co-location,” “regionalization,” or “localization” refers to the assignation of a hospitalist team to a specific inpatient unit. Its benefits may be related to the formation of a team and the additional interventions like interdisciplinary rounding that the enhanced proximity facilitates. However, cohorting is often adopted in isolation of the bundled approach within which it has proven beneficial. Cohorting may also be associated with unintended consequences such as increased interruptions and increased indirect care time. Institutions may increase patient loads in anticipation of the efficiency gained by cohorting—leading to further increases in interruptions and time away from the bedside. Fragmented attention and increases in indirect care may lead to a perception of increased workload, errors, and burnout. As hospital medicine evolves, there are lessons to be learned by studying cohorting. Institutions and inpatient units should work in synergy to shape the day-to-day work which directly affects patient and clinician outcomes—and ultimately culminates in the success or failure of the parent organization. Such synergy can manifest in workflow design and metric selection. Attention to workloads and adopting the principles of continuous quality improvement are also crucial to developing models of care that deliver excellent care. issue-copyright-statement© The Author(s), under exclusive licence to Society of General Internal Medicine 2022 ==== Body pmcGeographic “cohorting,” “co-location,” “regionalization,” or “localization” refers to the practice of assigning a hospitalist team to a specific inpatient unit with the expectation that the majority of the team’s patients will be on their assigned unit. The benefits are thought to be rooted in the enhanced physical proximity between clinicians, bedside nurses, patients, and the interprofessional team—with gains expected in efficiency, communication, collaboration, and patient centeredness.1,2 Pre-pandemic, cohorting was adopted by nearly a third of the non-teaching services of US hospital medicine groups surveyed.3 Cohorting is complex and like therapeutic decisions is associated with benefits, risks, and unintended consequences. Examining this complexity provides insights that may allow us to design better models of care. Each inpatient unit can be viewed as a clinical microsystem—the functional unit of the entire organization—the place where the work happens and where the outcomes that coalesce into the success or failure of the organization originate.4 Models of care utilizing bundled unit-based interventions to improve the care of hospitalized patients have demonstrated improvements in lengths of stay, costs of care, and mortality.5,6 In these models, cohorting was deployed alongside other mutually reinforcing interventions such as interdisciplinary rounding and leadership dyads, which become practical only when the proximity facilitated by cohorting and the creation of a team is assured. Yet, the adoption of unit-based interventions to improve care appears to be piece-meal across institutions with few deploying a bundled approach and many instituting cohorting alone.3 A survey of hospitalists in the USA revealed that the strong positive perceptions of cohorting cluster around the benefits of collaboration with bedside nursing colleagues, improved nursing satisfaction, increased patient centeredness, and improved efficiency and team building. Strong negative perceptions were reported around increases in interruptions, erosion of group camaraderie, discontinuity in patient care, and issues related to implementation. Academic practices and longer durations of cohorting were associated with positive perceptions while higher patient loads were associated with negative perceptions.2 Studies investigating the impact of cohorting as a stand-alone intervention have shown some results supporting and others refuting these perceptions. The proportion of bedside nursing colleagues agreeing with the statement “I experience good collaboration with house staff” increased from 10 to 40% following the implementation of cohorting.7 More patients perceived that their physicians spent more than four minutes with them and discussed their anxiety and emotions following cohorting.7 Cohorting has also been associated with increases in the likelihood of repeated visits to a patient in a day and increased time spent on the unit.8 Cohorting, however, is not a panacea—with the gains accompanied by downsides. Despite intending to foster patient-centered care, cohorting has not been associated with improvements in Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) scores and in some settings may be associated with increases in length of stay. 9–11 In a single-center time-motion study, cohorted hospitalists were interrupted as often as once every eight minutes—rates similar to those seen in Emergency Department settings—and were also noted to spend more time in computer interactions than their non-cohorted counterparts.8 These findings are consequential—interruptions erode attention, increase perceived workload, increase the risk of errors, and increase the time it takes to complete tasks.12 Tasks that detract from direct patient care contribute to burnout—rates of which have increased among hospitalists since the onset of the pandemic. 13 Fragmented attention can lead to bias and failure to recognize the declining trajectory of a patient.14 Interruptions, inattention, and their consequences are difficult to measure—with few studies in hospital medicine quantifying their burden and impact. With careful attention to design and implementation, cohorting may be successful in improving communication without increasing unnecessary interruptions—but such refinement requires close monitoring and continuous improvement which are often lacking in strained hospital medicine environments. Workload, communication, and outcomes are inexorably linked in hospital medicine. While cohorting may be associated with modest increases in the duration of each patient care encounter, these gains are fragile—and may be easily lost or reversed by increases in patient loads.8 The evidence also suggests that while cohorting increases shallow availability or “reachability” and the quantity of communication, it may not alone ensure deeper interpersonal communication or improve the quality of communication.14,15 Perversely, this increased reachability and decreased travel time may be used to rationalize increases in daily patient loads for cohorted teams. A focus on increasing productivity in turn may further increase interruptions, decreasing attention and impacting downstream outcomes that are not routinely monitored—such as the quality of communication, cognitive load, cognitive bias, diagnostic errors, and satisfaction with a job well done. “Every system is perfectly designed to get the results it gets”—and it is time to scrutinize the systems in which hospitalists work every day. The complexities of geographic cohorting we have examined provide insights that may allow us to design better models of care. We propose attention to the following principles (Fig. 1): Strengthening synergies between the clinical microsystem and the institution Fig. 1 The connectedness between the inpatient unit, institution, patient, and outcomes. In many instances, the COVID-19 pandemic clearly demonstrated what effective synergies can achieve. Driven by the crisis of the pandemic and potential personal protective equipment shortages, many institutions successfully and rapidly deployed hospitalist cohorting, a feat many previously struggled to achieve. However, in many cases, cohorting was quickly dismantled—highlighting the barriers that institutions and hospitalists face to prioritize and sustain geography—and which neither can overcome alone.16 While each inpatient unit represents a microcosm of the parent organization and drives its outcomes, it in turn relies on its parent organization—and the links between the work done within the clinical microsystem every day with that of the organization need to be strengthened. Workspace design, staffing targets, electronic medical record performance, and non-clinical administrative tasks all impact cognitive load and outcomes but are beyond the control of individuals. These complex issues require monitoring, feedback between the frontline and administrators, and a commitment to drive change at every level of the institution. 2. Defining and standardizing measures of success to reflect shared priorities Effective collaboration between the clinical microsystem and the institution is also crucially conveyed by what is measured and organizations signal their priorities by the metrics audited. To date, hospitalist literature has focused heavily on length of stay, and cohorting has been associated with increases, decreases, or no changes in length of stay. Such findings raise the question of whether the intervention was well-designed to impact the outcome measured and/or whether different metrics would better reflect the benefits of the intervention. Selected metrics should represent the shared mission of the frontline clinicians and the organization. Hospital medicine groups should carefully evaluate how they (or others) measure their quality and value, and what the measures drive. There are pitfalls in metric selection that may frustrate hospitalists, and metrics should reflect what is valued, impactful, within the locus of control of hospitalists and not based on what is expedient to measure.17 As hospitalists evolve into problem solvers, communicators, educators, researchers, advocates, and boundary spanners, our metrics should mature in tandem to prevent stagnation and drive progress. This evolution will require a thoughtful investment in the infrastructure of each hospital medicine group. 3. Re-imagine and re-define optimal workload Few studies have evaluated optimal daily patient loads for hospitalists—with fifteen patients per day often cited as the threshold past which outcomes suffer.18 However, the landscape in hospital medicine has changed seismically—nursing shortages and turnover impede team building and team communication, acuity of illness continues to increase, text-based messaging may have further increased the quantity of communication, and the COVID-19 pandemic has amplified the focus on length of stay and hospital capacity while eroding the optimism and resilience of the workforce. These factors necessitate an urgent reevaluation of optimum hospitalist workloads. In trying to maximize short-term productivity measured by the numbers of patients seen and relative value units generated, we may jeopardize the very gains we are trying to achieve. For example, increasing patient loads are associated with negative hospitalist perceptions about cohorting’s impact on patient safety, collaboration with nursing colleagues, and hospitalist satisfaction2 whereas reducing patient loads for hospitalists may actually yield cost savings for institutions.19 Initiatives to increase productivity must be accompanied by an assessment of the impact on the hospitalist, and on patient and institutional outcomes. As we reimagine workloads, we must account for the cognitive intensity of the hospitalist workday. In addition to patient volume, the cognitive burden is also influenced by patient acuity, hospitalist experience, the work environment and processes, interruptions, tasks, and the performance of the electronic medical record—factors that on some days may outstrip the impact of patient numbers alone. 4. Adopting a continuous quality improvement approach to drive improvements Certain other principles emerge as we create frameworks for the way forward. Before deploying practice models, the purpose should be clearly defined—is it a way to improve patient experience? to improve the quality of communication? Studies on cohorting have measured and reported outcomes as diverse as the number of steps walked in a day, the number of pages received, agreement on the plan of care between physicians and nursing colleagues, and length of stay. Each institution may have its own unique priorities that need to be addressed, and the problem that is being solved for should be explicitly identified and the solution optimized specifically to address the issue. Without such forethought, plans may be subverted by the expectation of creating a “silver bullet” intervention—a solution viewed as the answer to multiple problems—and thus fall short by the resulting dilution of the original intent by the tacking on of adjacent issues. Interventions need to be specific not only to the issues, but to each setting. The environment of each hospital and each hospital unit is unique, and interventions should be tailored accordingly. For example, when nursing or physician turnover is high, how do you form relationships and foster psychological safety within the team? Cohorting alone may not overcome the barriers to team building in such a setting. Continuous improvement also requires attention to the current and emerging data around models of care. Adopting cohorting alone, without the associated interventions that have been linked with improved outcomes, may invoke all the downsides without achieving potential gains. Different combinations of elements of care, some of which may not include cohorting at all, could influence specific outcomes more than others.20 When interpreting literature, we should be mindful that many investigations report favorable short-term pre-post outcomes but do not reflect the downstream emergence of unintended consequences. An infrastructure that supports the continuous monitoring of outcomes, surveillance for unintended consequences, and agile course correction when needed should be developed and deployed alongside models of care. Lessons learned from examining the strengths and weaknesses of cohorting provide a roadmap for building better systems. The stressors that undermine the gains from unit-based interventions may be beyond the locus of control of any inpatient unit and require synergy between the unit and the organization. This synergy is reflected in patient loads, workspaces, and metric selections that impact the models we deploy at the level of the unit. What we do every single day—and how we do it—has implications for our patients, our communities, our wellbeing, and the future of hospital medicine. Declarations Conflict of Interest The authors declare that they do not have a conflict of interest. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Singh S Fletcher KE A qualitative evaluation of geographical localization of hospitalists: how unintended consequences may impact quality J Gen Intern Med 2014 29 7 1009 16 10.1007/s11606-014-2780-6 24549518 2. Kara A Johnson CS Hui SL Kashiwagi D Hospital-Based Clinicians’ Perceptions of Geographic Cohorting: Identifying Opportunities for Improvement Am J Med Qual 2018 33 3 303 312 10.1177/1062860617745123 29241347 3. O’Leary KJ Johnson JK Manojlovich M Astik GJ Williams MV Use of Unit-Based Interventions to Improve the Quality of Care for Hospitalized Medical Patients: A National Survey Jt Comm J Qual Patient Saf. 2017 43 11 573 579 10.1016/j.jcjq.2017.05.008 29056177 4. Nelson EC Batalden PB Huber TP Mohr JJ Godfrey MM Headrick LA Wasson JH Microsystems in health care: Part 1. Learning from high-performing front-line clinical units Jt Comm J Qual Improv. 2002 28 9 472 93 10.1016/s1070-3241(02)28051-7 12216343 5. Kara A Johnson CS Nicley A Niemeier MR Hui SL Redesigning inpatient care: Testing the effectiveness of an accountable care team model J Hosp Med. 2015 10 12 773 9 10.1002/jhm.2432 26286828 6. Stein J Payne C Methvin A Bonsall JM Chadwick L Clark D Castle BW Tong D Dressler DD Reorganizing a hospital ward as an accountable care unit J Hosp Med. 2015 10 1 36 40 10.1002/jhm.2284 25399928 7. Olson DP Fields BG Windish DM Geographic Localization of Housestaff Inpatients Improves Patient-Provider Communication, Satisfaction, and Culture of Safety J Healthc Qual. 2015 37 6 363 73 10.1111/jhq.12054 26042748 8. Kara A Flanagan ME Gruber R Lane KA Bo N Kroenke K Weiner M A Time Motion Study Evaluating the Impact of Geographic Cohorting of Hospitalists J Hosp Med 2020 15 6 338 344 10.12788/jhm.3339 31891555 9. Bryson C Boynton G Stepczynski A Garb J Kleppel R Irani F Natanasabapathy S Stefan MS Geographical assignment of hospitalists in an urban teaching hospital: feasibility and impact on efficiency and provider satisfaction Hosp Pract (1995) 2017 45 4 135 142 10.1080/21548331.2017.1353884 28707548 10. Siddiqui Z Bertram A Berry S Niessen T Allen L Durkin N Feldman L Herzke C Qayyum R Pronovost P Brotman DJ Geographically Localized Medicine House-Staff Teams and Patient Satisfaction J Patient Exp 2019 6 1 46 52 10.1177/2374373518771361 31236451 11. Singh S Tarima S Rana V Marks DS Conti M Idstein K Biblo LA Fletcher KE Impact of localizing general medical teams to a single nursing unit J Hosp Med. 2012 7 7 551 6 10.1002/jhm.1948 22791661 12. Coiera E The science of interruption BMJ quality & safety 2012 21 5 357 360 10.1136/bmjqs-2012-000783 13. Dugani SB Geyer HL Maniaci MJ Fischer KM Croghan IT Burton C Psychological wellness of internal medicine hospitalists during the COVID-19 pandemic Hospital practice (1995) 2021 49 1 47 55 10.1080/21548331.2020.1832792 33012183 14. Kissler MJ Kissler K Burden M Toward a Medical “Ecology of Attention” N Engl J Med 2021 384 4 299 301 10.1056/NEJMp2027190 33503689 15. Mueller SK Schnipper JL Giannelli K Roy CL Boxer R Impact of regionalized care on concordance of plan and preventable adverse events on general medicine services J Hosp Med. 2016 11 9 620 7 10.1002/jhm.2566 26917417 16. Linker AS, Kulkarni SA, Astik GJ, et al. Bracing for the Wave: a Multi-Institutional Survey Analysis of Inpatient Workforce Adaptations in the First Phase of COVID-19 [published online ahead of print, 2021 May 28]. J Gen Intern Med. 2021;1-6. 10.1007/s11606-021-06697-6 17. Kara AY, Rohde JM. Professional Performance Audit and Feedback for Quality Improvement: Necessary but Insufficient. Joint Commission Journal on Quality and Patient Safety. 2021. 10.1016/j.jcjq.2021.12.004. 18. Elliott DJ Young RS Brice J Aguiar R Kolm P Effect of hospitalist workload on the quality and efficiency of care JAMA Intern Med. 2014 174 5 786 93 10.1001/jamainternmed.2014.300 24686924 19. Kamalahmadi, M., Bretthauer, K., Helm, J., Mills, A., Coe, E., Judy-Malcolm, A., & Pan, J. (2019). Mixing it up: Operational impact of hospitalist workload. Baruch College Zicklin School of Business Research Paper, (2019-10), 02. 20. Bhasin A Smith G Gupta A Effects of Changes to Hospitalist Admitter Staffing on Hospitalist Perception of Workload J Gen Intern Med 2021 36 2488 2489 10.1007/s11606-020-05961-5 32583339
PMC009xxxxxx/PMC9005024.txt	==== Front Cancer Causes Control Cancer Causes Control Cancer Causes & Control 0957-5243 1573-7225 Springer International Publishing Cham 1578 10.1007/s10552-022-01578-7 Original Paper Barriers to colorectal cancer screening in Ghana: a qualitative study of patients and physicians http://orcid.org/0000-0002-9126-1518 Lussiez A. alussiez@med.umich.edu 17 Dally C. K. dallykc@yahoo.com 23 Boateng E. A. eaboateng@hotmail.co.uk 2 Bosompem K. kingsleybosompemmd@gmail.com 2 Peprah E. ebpeprah@gmail.com 2 Hayward L. haywardl@med.umich.edu 4 Janes L. linnyj@med.umich.edu 4 Byrnes M. mabyrnes@med.umich.edu 1 Vitous A. vitousc@med.umich.edu 1 Duby A. agay@med.umich.edu 1 Varlamos C. cvarlamo@med.umich.edu 1 Ma L. lindsma@med.umich.edu 1 Darkwa D. stakigh@msn.com 2 Aitpillah F. fraitp@yahoo.com 2 Gyasi-Sarpong K. C. gaysek@yahoo.com 3 Opoku B. K. baafuoropoku@yahoo.com 35 Raghavendran K. kraghave@med.umich.edu 16 Kwakye G. 16 1 grid.214458.e 0000000086837370 Department of Surgery, University of Michigan, 1500 East Medical Center Dr, Ann Arbor, MI 48109 USA 2 grid.415450.1 0000 0004 0466 0719 Department of Surgery, Komfo Anokye Teaching Hospital (KATH), Okomfo Anokye Road, Kumasi, Ghana 3 grid.9829.a 0000000109466120 Kwame Nkrumah University of Science and Technology (KNUST), Kumasi, Ghana 4 grid.214458.e 0000000086837370 University of Michigan Medical School, Ann Arbor, MI USA 5 grid.415450.1 0000 0004 0466 0719 Department of Obstetrics and Gynecology, Komfo Anokye Teaching Hospital (KATH), Kumasi, Ghana 6 grid.214458.e 0000000086837370 Michigan Center for Global Surgery, University of Michigan, Ann Arbor, MI USA 7 grid.214458.e 0000000086837370 University of Michigan, Taubman Center, Floor 2 Reception C, 1500 E Medical Center Dr SPC 5331, Ann Arbor, MI 48109-5331 USA 12 4 2022 18 23 7 2021 25 3 2022 © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Purpose The incidence of colorectal cancer (CRC) in Ghana has increased eightfold since the 1960s. In 2011, national guidelines were set forth recommending all patients aged 50–70 years old undergo annual CRC screening with fecal occult blood testing (FOBT), but adherence to these guidelines is poor and screening rates remain low for unclear reasons. Methods We performed semi-structured interviews with 28 Ghanaians including physicians (n = 14) and patients (n = 14) from the Komfo Anokye Teaching Hospital in Kumasi, Ghana, to better understand the factors driving screening adherence and perceived barriers identified in an earlier quantitative study. Results Participants reported sociocultural factors such as reliance on alternative medicine or religion, lack of education, and financial burden as community-level barriers to CRC screening. At the system level, screening was limited by insufficient access to FOBT as well as a perceived lack of national prioritization. This was described as inadequate efforts from the Ministry of Health regarding national education as well as lack of incorporation of CRC screening into the National Health Insurance Scheme. Conclusion Several community- and system-level barriers exist to widespread screening of CRC in Ghana. A multi-level approach will be required to improve rates of CRC screening and ultimately reduce the burden of CRC in Ghana. Keywords Colorectal Gastrointestinal Oncology Education Global surgery Training http://dx.doi.org/10.13039/100000054 national cancer institute T32CA009672 Lussiez A. ==== Body pmcIntroduction Colorectal cancer (CRC) is the third most commonly diagnosed cancer globally [1, 2]. In 2020, approximately 1.9 million people were diagnosed with colorectal cancer [3]. Traditionally, CRC has been primarily considered a disease of developed countries, but rates are now rising in developing countries [3–6]. In Ghana, the incidence of CRC remains low at 3.9 cases per 100,000 (compared to 25.6 cases per 100,000 in the United States), but has increased several fold since the 1960s. Moreover, the incidence of CRC in Ghana is expected to continue to climb given Ghana’s aging population, rising life-expectancy, and ongoing adoption of Western habits [7–9]. Unfortunately, most patients present late in their disease course resulting in a 16% overall 5-year survival rate [10]. The comparative estimate for 5-year survival in the United States (US) for all stages combined is 65% [3]. Early detection of CRC is critical for Ghanaians given the dramatic decrease in survival rates from Stage I (90%) to Stage IV (0%) [10]. As such, to address the increasing burden of CRC in Ghana, in 2011, the Ministry of Health targeted CRC as an area for improvement and set forth a recommendation that all patients aged 50–70 years old were screened for CRC using fecal occult blood testing (FOBT) with subsequent endoscopy for positive FOBT tests [11]. Despite the national guidelines, adherence is rarely observed. As reported in our earlier work, physicians from Komfo Anokye Teaching Hospital (KATH), Ghana’s second largest tertiary facility, vary widely in their screening habits and only one of 39 surveyed physicians practiced in concordance with the national guidelines [12]. Perceived barriers to CRC screening at the physician level included inadequate time, training, and a lack of equipment and facilities. Separately, among patients, knowledge of cancer and cancer screening in Ghana is generally low, even for more prevalent cancers such as breast, prostate, and cervical cancers [13–15]. Increasing CRC screening rates will require coordinated efforts between communities and health systems and a better understanding of the existing barriers. In this context, we designed an exploratory qualitative study using semi-structured interviews to develop a deeper understanding of community- and system-level barriers to CRC screening in Kumasi, Ghana. The study was designed as a follow-up to our previous quantitative study in which we surveyed physicians at KATH regarding their CRC screening practices and perceived barriers to screening [12]. The results of this study will be able to help inform tailored, multi-level interventions aimed at increasing CRC screening rates in Ghana. Methods Study design In order to understand barriers to CRC screening from multiple perspectives, we interviewed both patients and physicians. The interviews sought to obtain a broader understanding of the perceived barriers to CRC screening in Ghana at the community and system levels. Interview guides were loosely based on the seven domains from the Tailored Implementation in Chronic Disease (TICD) framework [16, 17]. Separate interview guides were created for physicians and patients but were largely similar. Both interview guides were shared with team members who were culturally familiar to ensure relevancy among Ghanaian participants [GK, CKD, DD, BKO]. Approval for the study was obtained from the Institutional Review Board at KATH as the primary site and additionally at the University of Michigan. Interview participants Participants were recruited by convenience sampling in order to not place undue burden on our local physician partners. KATH physician participants were included if they screened patients for CRC as part of their practice. Patient participants were either surgical inpatients or being seen in consultation as outpatients. Demographics outside of gender were not recorded in order to protect participants’ privacy and our local partners’ concern for potential retribution. For example, physicians at KATH are employed by the Ghanaian government and did not want to risk loss of their job if identified which was a possibility given the limited number of physicians at KATH. In total, 14 physicians and 14 patients completed interviews. Participants did not receive any compensation for their participation. Data collection In-person interviews were performed by six resident physicians (EAB, KB, EP, FG, LS, and FY) at KATH between October 2020 and January 2021. Prior to interviewing, interviewers explained the purpose of the study and obtained verbal consent. Interviews lasted 10–15 min and were audio-recorded, transcribed verbatim, and de-identified. Of note, physician interviews were conducted in English and patient interviews were conducted in Twi, the most commonly spoken language in Kumasi. Patient interviews were transcribed from Twi to English. Back-translation was not used due to a limitation of resources. Data analysis Transcripts were coded using open coding informed by the iterative steps used in inductive thematic analysis [18]. During the first round of coding, five members of our research team (MB, AV, AL, LJ, and LH) independently reviewed three physician or three patient interview transcripts and annotated them with initial codes. Given the content similarities between the patient and physician interviews, a single codebook was generated using these preliminary codes. Several additional meetings were held to discuss discrepancies, modify the codebook, and adapt code definitions. Once a final codebook was agreed upon, each transcript was then independently coded by one of three members of the research team (AL, LJ, LH). Transcribed interviews were coded in MAXQDA 2020 qualitative analysis software (VEFBI Software, Berlin, Germany). Results A total of 28 participants were interviewed in this study: 14 physicians (64% male) and 14 patients (43% male). Previous work has found the majority of physicians at KATH to be male [12]. Barriers were categorized into two groups, community level and system level. At the community level, sociocultural factors, lack of education and financial burden were identified as barriers. At the system level, insufficient access and lack of national prioritization were most commonly reported. Perceived barriers: community level Sociocultural Table 1 describes the range of sociocultural barriers to CRC screening described by patients and physicians. These included health care beliefs, the influence of religion and traditional medicine, perception of stool collection for FOBT, and the stigma associated with the diagnosis of cancer. The idea of physician consultation upon initial symptom presentation and for routine cancer screening was not widely accepted as the norm. Respondents also commented on the prevalence of religion and its role in shaping community members’ medical beliefs and habits. Some physicians noted that patients may seek treatment from their church or a spiritualist before coming to see them. One physician explained his frustration with this phenomena:In the village…they start the herbs before the operations are planned, various concoctions and leaves, drinking stuff here and there....sometimes, some of these religious bodies or persons sort of hinder the progress of medicine. And also with the traditional medicine people too, preaching false cures to people, grinding all sorts of leaves and herbs for people, they are sort of compounding the problem. -Physician 05 Table 1 Community-level barriers: sociocultural Theme Respondent Exemplary Sociocultural Health care beliefs Patient 14 "There are some people who have made up their minds. They don’t believe in the existence of many things in the world unless it affects them directly. Even if they heard the announcement and are educated on it, still would not be enough to convince them." Patient 10 "Some people haven’t made up their mind to go get tested, like someone at 50 years. They don’t even have that in mind of going to check for cancer or do the fecal occult blood test. So they might not go." Physician 01 "In my community there isn’t a, um, they aren’t used to the idea of frequent screening and check-ups. Hospitals are usually seen as a last resort, after options have failed to provide." Stigma Patient 01 "Oh, I don’t think they will [look down upon you]. Since it is not like HIV or let us say COVID-19 that will stigmatize you." Patient 10 "I have not seen someone say they won’t get close to someone with cancer." Physician 09 "Well, um, cancer in itself is terrible. It’s terrible from the onset, from when you are told, you know. The stigma and also the unavailability of a lot of cancer facilities, oncology facilities and clinics, so yeah, it’s a terrible thing to be diagnosed of it and to even be aware or have the knowledge that you have it, yes." Perception of FOBT Patient 13 "As human as we are who can say they do not defecate. So why will you laugh at me for taking a stool sample for a test? If they decide to laugh it is their own problem. This is not the only illness that requires a sample of your stool. If your stomach aches and go to the hospital, the doctor can request you go to the lab for your stool to be examined. So, I do not think you they will be laughed at. Even if they do, I do not care." Patient 14 "I can attest to the fact that nothing like this will ever happen. No one will be teased. For instance, if you are unwell and you go to the hospital and the doctor asks you to bring your stool and urine for testing to determine the illness, the person will run to go and bring it to the doctor. " Physician 05 "Generally in the traditional setting, people see fecal matter as some sort of an abomination and I mean, they may not be comfortable having other people looking at their stool. So sometimes they need a certain level of encouragement before they are able to provide stool samples for you." Role of superstition, religion and traditional medicine Patient 13 "Taking a sample of your stool is not a big deal, but the problem is, in a community you have superstitious people. I think with a little training we can create awareness for them to understand the importance of the test." Physician 11 "Okay, in our region or area we find ourselves, I mean people most of the time perceive cancer as a strange illness which may be as a result of a religious or belief, they attend to it at home or in churches but not in the hospitals." Physician 05 "In the village so they start the herbs before the operations are planned, various concoctions and leaves, drinking stuff here and there. So we didn’t even know about it until it was almost too late when he [patient] had lost so much weight before it got too low and started seeking professional medical care….sometimes, some of these religious bodies or persons sort of hinder the progress of medicine. And also with the traditional medicine people too, preaching false cures to people, grinding all sorts of leaves and herbs for people, they are sort of compounding the problem." Willingness to undergo screening Patient 12 "God has given you people [doctors] the knowledge, so I will do as you say and follow any direction I’m given." Patient 14 "When someone is sick, we can not tell what is wrong till we go to the hospital for lab tests to know what is wrong. This is just a way doctors use to find out what is really wrong with us." Patient 03 "It will not bring any issues. If you want to be healthy you have to get tested." This physician explains that while he also believes in God, religion and traditional medicine may “hinder the progress of medicine,” and prohibit timely access to care and effective medical treatment. The stigma of having a diagnosis of cancer was noted a few times but overall, infrequently discussed. One patient mentioned that cancer is not stigmatized, unlike diagnoses of human immunodeficiency virus (HIV) or coronavirus disease 2019 (COVID-19) (Patient 01). More often, patients discussed the collection of fecal matter for FOBT. Patient participants rarely said they would be uncomfortable giving a sample of stool, most instead expressed faith in their physician and willingness to undergo any recommended test in order to facilitate diagnosis and treatment. Moreover, they believed their communities would feel the same way. When asked if people would be laughed at for undergoing FOBT, one patient commented:I can attest to the fact that nothing like this will ever happen. No one will be teased. For instance, if you are unwell and you go to the hospital and the doctor asks you to bring your stool and urine for testing to determine the illness, the person will run to go and bring it to the doctor. – Patient 14 Lack of education Lack of education was one of the most commonly mentioned barriers to CRC from among patients and physicians (Table 2). Participants believed that improved medical education related to CRC and screening awareness would be beneficial for increasing rates of CRC screening in the community. Further, without understanding the importance of CRC screening, patients felt their community would be unlikely to undergo screening.If they do not understand what the test is for, they will not see the importance of participating in the screening.—Patient 13 Table 2 Community-level barriers: lack of education and financial Theme Respondent Exemplary Lack of education Patient 13 "If they do not understand what the test is for, they will not see the importance of participating in the screening. If they are educated on it, they will surely partake in it." Patient 06 "For someone to bring his or her sample stool, the person needs to understand what this is all about before he will come and do it. So, if the briefing is successful, he or she will know the importance of the stool and knows how dangerous the cancer is." Physician 03 "There is a role that we as clinicians for instance…help educate religious leaders use their church platforms and mosque and religious setting to preach and to education the people so that people will come our way, politicians providing the necessary resources and funding for its treatment, education and screening." Financial Patient 06 "When it comes to the money issues, we would prefer the screening being close to us. Likes the hospitals by us or in schools. This is because, when most of us want to come to the hospitals for check-ups, we always assume they will charge us a whole lot. We cannot afford that, so we will be happy if it is brought to our community and the cost is reduced…sometimes when I visit the hospital, the Health Insurance does not cover all my hospital bills and drugs. Sometimes it’s just the hospital card." Patient 08 "People are excited of getting tested. The problem is, they cannot afford it. They don’t have the money to go to the hospital." Physician 11 "The screening process which involves the FBOT and then the colonoscopy as well in our setting in very expensive and so people may not have the ability to afford the kind of screening test….Normally, most of the tests is not in the insurance program so when you write for them and people who normally have the cases are from the rural area so eventually they may not end up going to pay for these tests." Specifically, patients pointed out that education should not only describe what CRC and CRC screening are but emphasize their importance and how cancer is a “dangerous illness” (Patient 10). Separately, both patient and physician respondents acknowledged the power of community and religious leaders to leverage their platforms to educate community members and promote the importance of CRC screening. One physician highlighted the role clinicians could play to encourage this effort:…we all need to play a role. There is a role that we as clinicians for instance…help educate religious leaders use their church platforms and mosque and religious setting to preach and to education the people so that people will come our way. – Physician 03 Here, this physician emphasized the unique role they can play in educating key personnel who have highly influential positions in their communities. Financial burden Several patients and physicians reported cost as a barrier to CRC screening (Table 2). Many alluded to a general state of poverty in their community resulting in lack of funds to cover the costs of screening as well how they “don’t have the money to go to the hospital” (Patient 08). Additionally, a few patients reported that despite having health insurance, not all of their medical bills were covered. Perceived barriers: system-level Access Although the majority of patients reported having a local hospital where they could seek medical care, they were not always able to receive what they needed (Table 3). Some noted local hospitals did not always have physicians available resulting in exceedingly long wait times for general medical care and at times, resources for diagnosis and treatment were frequently lacking. Separately, physicians noted that FOBT was not always available at every hospital and in particular highlighted the need for easy access at all medical system levels: community, sub-district, and district.If these so called fecal blood tests can be made available especially for those of us coming or working in the sub-district. If these test kits are not available and then they always have to come to these CT, or what they are about to get, will be a big problem, they will lose interest in doing it. So I think they, the point should be available right down to the district level and sub-district level, so that it can easily be picked from there.—Physician 03. Table 3 System-level barriers Theme Respondent Exemplary Access Patient 06 "I think the reason is they are understaffed, or their schedule is not planned well, especially on weekends no doctor attends to you. Some doctors may tell you they are not in charge of a specific illness, so they cannot help you. They told us they have a special unit, that special unit should at least have one doctor there on weekends in case of an emergency. Whenever I go to the hospital, I am told the doctor in charge of the sickness I am suffering from is not around. So, I am left helpless. They expect to me to go home and die." Physician 02 "This fecal occult blood test, you don’t see it. I-its-it’s not available actually…..Some of them [patients] too you may have to get the test to them at their doorstep." Physician 03 "If these so called fecal blood tests can be made available especially for those of us coming or working in the sub-district. If these test kits are not available and then they always have to come to these CT, or what they are about to get, will be a big problem, they will lose interest in doing it. So I think they, the point should be available right down to the district level and sub-district level, so that it can easily be picked from there." Lack of National Prioritization Physician 09 "The Ministry of Health needs to really invest a lot of time and create more public awareness of cancers…public education is very necessary. Also…the government and Ministry of Health entirely should also invest in cancer care, invest in providing health facilities resources for the management of these patients, and also if they could factor it in to the national health insurance scheme. " Physician 05 "I don’t know if it is already on the national health insurance, but I think that one of the easiest things to do is put Fecal Occult Blood on the insurance so if national health insurance covers it, it would be easier to get more people to do it, but sometimes if you realize it’s going to add to the patient’s bill, it makes it hard to even request something like that." Physician 06 "Some of the cancers like breast, some of their chemotherapy drugs are covered by NHIS [National Health Insurance System], but currently I don’t know of any NHIS that is covering eh maybe fecal occult blood test…I think we can do more to actually prioritize it as we do compared to breasts and the other cancers." This physician notes that if FOBT is not widely accessible, physicians will lose interest in recommending the screening test. Lack of national prioritization CRC screening was generally not perceived to be a national priority in Ghana (Table 3). To address this, participants suggested that: (1) the Ministry of Health (MOH) increases pertinent cancer education and public awareness of CRC and (2) FOBT financial coverage is incorporated into the National Health Insurance Scheme (NHIS), Ghana’s national insurance policy that was designed to provide basic healthcare services [19]. Although the NHIS was designed to cover over 95% of disease conditions in Ghana, including treatment for breast and cervical cancers, it notably does not over CRC screening or treatment [20]. One physician commented:The Ministry of Health needs to really invest a lot of time and create more public awareness of cancers...public education is very necessary. Also…the government and Ministry of Health entirely should also invest in cancer care, invest in providing health facilities resources for the management of these patients, and also if they could factor it in to the national health insurance scheme. -Physician 09 This physician believed that given the increasing prevalence of CRC in Ghana, investment from the MOH to “create more public awareness,” increase education, and advocate for insurance coverage of FOBT would have significant impact on Ghanaians. Discussion The results of this study highlight several barriers to CRC screening in Ghana at both the community and system levels. At the community level, sociocultural factors included reliance on alternative medicine or religion and limited use of hospitals for medical care. Respondents simultaneously noted that lack of education regarding CRC and CRC screening was widespread and without addressing this issue, patients would be unlikely to undergo screening. At the system level, screening was limited by insufficient access to FOBT and a lack of national prioritization. Efforts from the Ministry of Health to promote awareness and education as well as incorporation of FOBT into the National Health Insurance Scheme were cited as opportunities to improve rates of CRC screening. Our results build on our previous work identifying barriers at the physician level including lack of equipment, personnel shortages, and limited training and, within this context, underscore the need for a multi-pronged approach to improving CRC screening rates in Ghana in order to ultimately reduce the burden of CRC [12]. As our respondents highlighted, improved education, garnering support from the community, and reducing the financial burden of CRC screening will be required to make substantial impact. Mobile health platforms are effective means of communication and education and can help provide timely medical care within the confines of low resource settings [21–23]. As such, mobile health platforms may be an innovative way to address CRC screening and may be particularly effective in Ghana, where WhatsApp accounts for 90% of information flow [24]. Prior studies have shown the success of mobile health platforms in improving the conduction of cervical cancer screening [25]. Our team is currently working on the development and implementation of ConqueringCRCancer, a mobile health platform that leverages pre-existing electronic tools such as WhatsApp [26] and Qualtrics [27] and trains community health workers in Ghana to bring and administer FOBT for patients in their homes. Further, use of this tool will facilitate localization of the nearest endoscopy center should more invasive testing be required and will serve as a longitudinal tracking system for CRC screening. No matter how effective an intervention, sustainable impact will require change at the system level, including improved access and national support. The system-level barriers identified in our study are similar to other work evaluating the barriers to breast and cervical cancer screening and despite increased national attention, screening rates for breast and cervical cancer in Ghana remain low at 12% and 3.4%, respectively [13, 14, 28]. Strategies employed as part of the response to the HIV/AIDS epidemic may be especially helpful in designing efforts aimed at improving screening for all three cancers. Ghana’s multi-sector, coordinated approach to fighting the HIV/AIDS epidemic has allowed for the successful implementation of interventions within the decentralized Ghanaian health care system. This has resulted in improved physician awareness, increased testing and treatment access, prevention of mother-to-child transmission, and universal precautions to prevent infection [29, 30]. Such an approach may additionally encourage communication between the government, hospitals, and physicians and ultimately align priorities to more effectively improve CRC screening rates. Another helpful tactic from Ghana’s management of HIV/AIDS was early creation of the HIV Sentinel Surveillance (HSS) system, and later the Ghana Demographic and Health Survey, which allowed for estimation of the extent of HIV infection and the linkage of HIV results to key demographic, social, and behavioral factors [31]. These data repositories—which are lacking for CRC—were critical in understanding the burden of HIV/AIDS, tracking progress, and identifying where needs were highest. As a marker of Ghana’s success, HIV screening is now widely available at the national, regional, district, and subdistrict levels [30]. Our study must be interpreted in light of its limitations. We used convenience sampling which may limit the generalizability of our results in two ways. First, we sampled from only one institution; however, KATH is the second largest tertiary facility in Ghana and receives referrals from 12 of 16 regions of Ghana [32]. It receives a diverse set of patients representative of almost the whole country. Second, by interviewing individuals who were patients at KATH, our sample may be biased as it is comprised from a population willing to seek care at a hospital. Therefore, these participants may not share the same health beliefs as the Ghanaian population at large. However, several patients acknowledged their community’s hesitancy in seeking hospital care; we believe community views were broadly represented. Lastly, our study was conducted during the COVID-19 pandemic which limited our sample size, resources, and ability to train interviewers as intensively as originally planned. Despite these limitations, we found our collected data to be incredibly informative and our local partners perceive it to be in line with their experiences. Future studies are needed to understand barriers to CRC screening in a broader Ghanaian population and to investigate the most appropriate method for CRC screening in Ghana. The authors are deeply indebted to Dr. Anne Sales and Dr. Jennifer Ervin for their guidance and support in the creation and execution of this study. Author contributions All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by AL, EB, KB, EP, LH, LJ, AV, MB, AD, CV, LM, DD, FA, KG-S, and BO. The first draft of the manuscript was written by AL and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding AL is supported by the National Cancer Institute (T32CA009672). Data availability The dataset generated during this study are not publicly available but are available from the corresponding author on reasonable request. Code availability Not applicable. Declarations Conflict of interest None. Ethics approval Not applicable. Consent to participate All interviewees gave verbal consent prior to participating. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Keum NN Giovannucci E Global burden of colorectal cancer: emerging trends, risk factors and prevention strategies Nat Rev Gastroenterol Hepatol 2019 16 12 713 732 10.1038/s41575-019-0189-8 31455888 2. Fitzmaurice C Abate D Abbasi N Global, regional, and national cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years for 29 cancer groups, 1990 to 2017: a systematic analysis for the global burden of disease study JAMA Oncol 2019 5 12 1749 1768 10.1001/jamaoncol.2019.2996 31560378 3. Colorectal Cancer: Statistics. Published 2020. Accessed April 20, 2021. https://www.cancer.net/cancer-types/colorectal-cancer/statistics# 4. Center MM Jemal A Ward E International trends in colorectal cancer incidence rates Cancer Epidemiol Biomarkers Prev 2009 18 6 1688 1694 10.1158/1055-9965.EPI-09-0090 19505900 5. Favoriti P Carbone G Greco M Pirozzi F Pirozzi REM Corcione F Worldwide burden of colorectal cancer: a review Updates Surg 2016 68 1 7 11 10.1007/s13304-016-0359-y 27067591 6. Arnold M Sierra MS Laversanne M Soerjomataram I Jemal A Bray F Global patterns and trends in colorectal cancer incidence and mortality Gut 2017 66 4 683 691 10.1136/gutjnl-2015-310912 26818619 7. Dakubo J Naaeder S Tettey Y Gyasi R Colorectal carcinoma: an update of current trends in Accra West Afr J Med 2011 10.4314/wajm.v29i3.68218 8. Yeboah FA Yorke J Obirikorang C Patterns and presentations of colorectal cancer at Komfo-Anokye teaching hospital Kumasi Ghana Pan Afr Med J 2017 28 1 10 10.11604/pamj.2017.28.121.12927 9. Global Cancer Observatory. World Health Organization. Published 2022. Accessed March 24, 2022. https://gco.iarc.fr 10. Agyemang-Yeboah F Yorke J Obirikorang C Colorectal cancer survival rates in Ghana: a retrospective hospital-based study PLoS ONE 2018 13 12 1 15 10.1371/journal.pone.0209307 11. MoH. National Strategy for Cancer. Minist Heal Ghana. Published online 2016:1–8. 12. Lussiez A Dualeh SHA Dally CK Colorectal cancer screening in Ghana: physicians’ practices and perceived barriers World J Surg 2021 45 2 390 403 10.1007/s00268-020-05838-y 33145608 13 Binka C Nyarko SH Awusabo-Asare K Doku DT Barriers to the uptake of cervical cancer screening and treatment among rural women in Ghana Biomed Res Int 2019 2019 1 8 10.1155/2019/6320938 14. Opoku SY Benwell M Yarney J Knowledge, attitudes, beliefs, behaviour and breast cancer screening practices in Ghana, West Africa Pan Afr Med J 2012 11 28 22514762 15. Baratedi WM Tshiamo WB Mogobe KD McFarland DM Barriers to prostate cancer screening by men in sub-saharan Africa: an integrated review J Nurs Scholarsh 2020 52 1 85 94 10.1111/jnu.12529 31733043 16 Flottorp SA Oxman AD Krause J A checklist for identifying determinants of practice: a systematic review and synthesis of frameworks and taxonomies of factors that prevent or enable improvements in healthcare professional practice Implement Sci 2013 10.1186/1748-5908-8-35 17. Wensing M The tailored implementation in chronic diseases (TICD) project: introduction and main findings Implement Sci 2017 12 1 10 13 10.1186/s13012-016-0536-x 28148305 18. Terry G, et al (2017) Thematic analysis. The Sage Handbook of Qualitative Research in Psychology 19. Blanchet NJ Fink G Osei-Akoto I The effect of Ghana’s National Health Insurance Scheme on health care utilisation Ghana Med J 2012 46 2 76 84 22942455 20. Benefits Package. National Health Insurance Scheme. Published 2022. https://www.nhis.gov.gh/benefits.aspx. Accessed 27 Feb 2022 21. Woods J Moorhouse M Knight L A descriptive analysis of the role of a WhatsApp clinical discussion group as a forum for continuing medical education in the management of complicated HIV and TB clinical cases in a group of doctors in the Eastern Cape, South Africa South Afr J HIV Med 2019 20 1 1 10 10.4102/sajhivmed.v20i1.982 22. Bervell B Al-Samarraie H A comparative review of mobile health and electronic health utilization in sub-Saharan African countries Soc Sci Med 2019 232 April 1 16 10.1016/j.socscimed.2019.04.024 31035241 23. Amoakoh HB Klipstein-Grobusch K Amoakoh-Coleman M The effect of a clinical decision-making mHealth support system on maternal and neonatal mortality and morbidity in Ghana: study protocol for a cluster randomized controlled trial Trials 2017 18 1 1 11 10.1186/s13063-017-1897-4 28049491 24. Ahiabenu K. Ghana runs on whatsApp. Published 2018. https://www.graphic.com.gh/features/opinion/ghana-runs-on-whatsapp.html. Accessed 30 Mar 2021 25 Asgary R Cole H Adongo P Acceptability and implementation challenges of smartphone-based training of community health nurses for visual inspection with acetic acid in Ghana: mHealth and cervical cancer screening BMJ Open. 2019 9 7 e030528 10.1136/bmjopen-2019-030528 26. WhatsApp. Published 2020. https://www.whatsapp.com/. Accessed 30 Mar 2021 27. Qualtrics. Published 2020. https://www.qualtrics.com. Accessed 1 July 2020 28. Ayanore MA Adjuik M Ameko A Self-reported breast and cervical cancer screening practices among women in Ghana: predictive factors and reproductive health policy implications from the WHO study on global AGEing and adult health BMC Womens Health 2020 20 1 1 10 10.1186/s12905-020-01022-5 31898500 29. WHO. GHANA. Summary Country Profile for HIV/AIDS Treatment Scale-Up.; 2005. https://www.who.int/hiv/HIVCP_GHA.pdf. Accessed 27 Feb 2022 30. AyisiAddo S Abdulai M Yawson A Availability of HIV services along the continuum of HIV testing, care and treatment in Ghana BMC Health Serv Res 2018 18 1 739 10.1186/s12913-018-3485-z 30257660 31. Akwara PA, Fosu GB, Govindasamy P, Alayón S, Hyslop A (2005) An in-depth analysis of HIV prevalence in Ghana: further analysis of demographic and health surveys data. Calverton: ORC Macro. http://dhsprogram.com/publications/publication-FA46-Further-Analysis.cfm 32. Komfo Anokye Teaching Hospital - About Us. Published 2019. http://www.kathhsp.org/about-us/. Accessed 22 Feb 2022
PMC009xxxxxx/PMC9005025.txt	==== Front TechTrends TechTrends Techtrends 8756-3894 1559-7075 Springer US New York 35434728 724 10.1007/s11528-022-00724-x Original Paper Teaching Programming Online: Design, Facilitation and Assessment Strategies and Recommendations for High School Teachers http://orcid.org/0000-0003-1719-7615 Shanley Nicole nshanley@uncc.edu 1 Martin Florence 1 Hite Nicole 1 Perez-Quinones Manuel 1 Ahlgrim-Delzell Lynn 1 Pugalee David 1 Hart Ellen 2 1 grid.266859.6 0000 0000 8598 2218 The University of North Carolina at Charlotte, Charlotte, NC USA 2 grid.10698.36 0000000122483208 North Carolina Virtual Public Schools, Raleigh, NC USA 12 4 2022 2022 66 3 483494 25 3 2022 © Association for Educational Communications & Technology 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Current research surrounding online computer science education emphasizes the need for high-quality professional development opportunities. However, there is a gap in research in the inclusion of online computer science educators to identify needs and strategies that make the online computer science courses effective. Through a Research-to-Practice Partnership (RPP), this paper examines the instructional strategies and recommendations from online Computer Science teachers. This study seeks to better understand (1) What design, facilitation, and assessment strategies do teachers use to teach programming online? and (2) What recommendations do teachers have for those interested in teaching programming online? The feedback teachers provided during the study assisted in identifying the current needs in online AP Computer Science. The participants suggested additional ways the RPP could support teachers in strengthening their practice, which has assisted in the production of high-quality professional development to support novice teachers entering the field of Computer Science. Keywords Computer science Online teaching and learning Instructional strategies Assessment Design Facilitation issue-copyright-statement© Association for Educational Communications & Technology 2022 ==== Body pmcMany high school students currently take computer science courses through virtual schools due to the lack of offerings and unavailability of teacher expertise at their local schools. Goode et al. (2020) consider preparing thousands of teachers with high-quality, accessible professional development as a grand challenge. While most teachers enter the classroom with the general ability and skills necessary to teach, many of those teachers are not content experts nor have been trained to specifically teach online. When combining both a new content area and a teaching platform, many challenges arise. This highlights the issue that teaching computer science online requires the use and implementation of different instructional strategies. In this study, we discuss strategies and recommendations from teachers through a Research to Practice Partnership (RPP) with a State Virtual Public School (SVPS), through which we plan to design and offer online professional development for teachers across the state to teach AP Computer Science advanced courses. This paper discusses the findings from a needs assessment conducted via three focus groups with 14 teachers from SVPS and a collaborative seminar held in the summer of 2020 by the Research to Practice Partnership. It will also address the direct connection between the results and thoughts shared in the focus group sessions and the data collected during a workshop using Jamboard, a collaborative digital whiteboard. Conceptual Framework Using the framework proposed by Martin et al. (2019), we focus on design, facilitation, and assessment strategies teachers use while teaching computer science online (Fig. 1).Fig. 1 Strategies for effective online courses Design Strategies Design strategies include the various instructional strategies that teachers and designers use while designing a course. According to Veletsianos et al. (2016), the National Science Foundation has been encouraging the computing education community to partner with education researchers to support the overall development of computer science education, curriculum, and course design. There are few sources in the literature surrounding online course design to speak to the intricacies of online instructional design specifically as it applies to computer science education. Zendler and Klaudt (2015) identified several instructional methods such as problem-based learning, learning tasks, discovery learning, computer simulation, project work and direct instruction that work well in computer science courses. They propose recommendations on how each of the computer science courses can be designed for each of these instructional methods. McGowan (2016) presented a four-component theory-based design framework that can be used in computer programming eLearning courses. Building on the framework proposed by Dabbagh (2005), McGowan’s framework included “pedagogical model, a body of exemplars, instructional strategies and learning technologies to facilitate meaningful learning of proper CP practices and knowledge building (p.11)”. A study conducted by Proulx (2000) sought to assist instructional designers in streamlining their online courses by creating a framework to help guide course design. Their “goal is to help students focus on mastering reasoning and design skills before the language idiosyncrasies muddy the water. (p.80)”. Similarly, subsequent studies have alluded to the fact that while the course work and content taught within computer science courses can be difficult for novice computer science students to pick up, there is room for improvement in how the course is designed. There is also research conducted to support the development of Massive Online Open Courses (MOOCs) for computer science foundational courses. An article published in 2015 highlighted the design of these MOOCs to include new and improved approaches to computer science course design; the research team implemented a more balanced pedagogical approach, one that worked to assist the overall course design to better address the “cognitive, interpersonal, and intrapersonal” (p.1) needs of students and take into students ``deeper learning” (Grover et al., 2015). When considering design strategies within the context of online computer science courses, another design process frequently appears within the literature, design thinking. This iterative process has been recently explored within the world of K-12 education through the work of Crane et al. (2018) and Li and Fu (2020). Both authors have used design thinking as a framework that guides course design within K-12 education, promotes and builds community within teachers who are acting as course designers, and works to support both physical and technical innovation into the design process. Facilitation Strategies Martin et al. (2020) define facilitation as “how, what, when, and why an online faculty member makes decisions and takes actions to help students meet the learning outcomes (p.36)”. A recent study published in Science Direct, The Effectiveness of Online Learning with Facilitation Method (Zulfikar et al., 2019) evaluates the level of student participation in online discussion forums and other tools useful in online course design. Specifically, the authors reviewed the effects of facilitation methods and teacher involvement in student participation in online discussion forums. Applying these generalizable studies to the field of online computer science could provide a new lens through which we view facilitation as applied to computer science. While there is little research that directly seeks to identify and understand the effectiveness of online facilitation strategies for computer science teachers, there are a few studies that review the effectiveness that online computer science courses have had, and ways that teachers in this learning environment have worked to support their students, virtually. Evidence of this can be found in Huan et al.’s article, Teaching Computer Science Courses in Distance Learning (2011). Throughout this study, the research team highlights the influence of distance learning and its increasing popularity due to both flexibility and convenience of learning and as more recently notes, out of necessity because of the 2020 COVID-19 pandemic. Throughout this study, we see mention of online tools that increase learner engagement such as the inclusion of multiple learning modalities, PowerPoint presentations embedded into the course, PDF documents, and the ability of the course to work across multiple platforms, allowing for accessibility among mobile devices. One of the seminal texts that support the foundational understanding of online facilitation strategies for computer science students, comes from the work of Wilson et al. (1997). In its inception, online facilitation for computer science students had the goal of supporting students in an asynchronous environment, increasing their engagement and overall understanding of the content being taught. During this time, the major question being addressed was “How can we best support such teaching and learning and what aspects of this process work well when compared to face-to-face teaching?” It has been 24 years since its publication, and this same question is being asked across computer science publications with educational researchers working to support the connection between online facilitation strategies and computer science. Assessment Strategies Computer science courses, especially those operating underneath the heading of “Advanced Placement,” are heavily tested with careful consideration and alignment placed on the final Advanced Placement (AP) examination. However, computer science teachers have the autonomy to create and apply assessment strategies throughout the course, with respect to the overarching needs placed by their district or school administration. When considering assessments and the various types that can be utilized within a course, it is important to consider the value of both formative and summative assessment. According to Black and Wiliam (2009), formative assessments are defined as “…evidence about student achievement is elicited, interpreted, and used by teachers, learners, or their peers, to make decisions about the next steps in instruction that are likely to be better, or better founded, than the decisions they would have taken in the absence of the evidence that was elicited” (p. 9). Similarly, Grover (2021) agrees with the work of Black and Wiliam (2009) that the major difference between summative and formative assessment lies within motivation behind the assessment and how teachers respond to the data collected. The goal of summative assessments is to grade students, using commonplace tests or quizzes, typically through several multiple-choice questions (Sorva & Sirkiä, 2015). However, formative assessments are aimed at monitoring students learning and coaching them through the learning process (Grover, 2021). With these differences in mind, granting teachers the autonomy to carefully select their assessment types and tools in a way that supports their students needs and allows them to continue to coach them through the learning process is vital. According to a systematic review conducted by Garcia et al. (2018), many of the available e-tools, such as automated tools (or auto graders) assist instructors in grading large quantities of work and provide students with instant feedback. When instructors include auto graders, they can support students by offering hints or guidance on their assignments by flagging compiler, test case, solution, and style errors (Keuning et al., 2016). As previously mentioned, there is a significant emphasis placed on the College Board Advanced Placement Computer Science examinations, as these are the assessments used to determine a student’s content proficiency. Specifically for the AP Computer Science A (CSA) course, the exam consists of 40 multiple choice questions and four free-response questions that require students to demonstrate their understanding of basic skills related to computer science programing (The College Board, 2006). Of the four Free Response Questions (FRQ), students will be asked to write code that displays mastery of the following skills: Methods and Control Structures, Classes, Array/ArrayList, and 2D Array. An assessment strategy commonly cited in computer science research is incorporating short free-response questions built into computer science courses, especially in an online learning environment. According to Klein et al. (2011), standardized tests and commonplace multiple-choice questions provide a shallow understanding of students’ actual ability. To truly engage students and understand the depth and complexity of their knowledge and evaluate application skills, instructors need to invest in meaningful forms of assessment such as the inclusion of open-ended or free-response questions. The study by Klein and colleagues tested the effectiveness of auto graders when used to grade free-response questions to assist instructors in providing students with more meaningful opportunities to demonstrate their learning. They support the overarching theme of this paper, which aims to support investment and integration of high-quality auto graders into online computer science courses. Purpose of the Study and Research Questions Computer science courses are not offered by all school districts, and therefore some students enroll to complete computer science courses online through the SVPS. Teaching computer science online requires different instructional strategies, and both students and instructors experience challenges teaching and learning online. In this study, we examine the instructional strategies currently used by high school teachers who teach computer science online and analyze the current design, facilitation, and assessment strategies they use to engage with their students. The research question addressed in this study include,What design, facilitation, and assessment strategies do teachers use to teach programming online? What recommendations do teachers have for those interested in teaching programming online? Methods This section describes the instructional context, research participants, data collection, and data analysis. Context The context of this study is based on a Research to Practice Partnership (RPP) currently held between a southeastern public university in the United States and a state Virtual Public High School. As part of a (Foundation) Grant - Computer Science for All, the research team collaborated to create and offer online professional development to teach the AP Computer Science advanced course to high school teachers. We established a RPP to guide the development of professional development for online computer science instruction. Establishing an intentional, long-standing, and collaborative partnership between computer science education researchers and computer science teachers at the State Virtual Public School is critical to addressing the professional development needs of a larger audience of online computer science teachers. Our RPP approach stresses the role of our lead teachers from the State Virtual Public School as key researchers in shaping the design of the professional development. Using participatory research approaches, the project team engaged the lead teachers during the project’s first year to identify critical instructional strategies and resources vital to their success in teaching computer science online. Through the RPP, the lead teachers’ roles as key stakeholders in the design process was reinforced. The leaders were reminded of their roles as experts in the partnership and the critical importance their input plays in the future design and implementation of the professional development program. Teachers were invited to participate in a focus group, followed by a one-week summer workshop, where the research team engaged the participants in online professional development. The teachers were put into the role of “content expert” working to identify best practices for online instruction. A primary goal of this focus group and workshop was to allow teachers to extend their thinking and consider approaches to formative assessment and methods for promoting equity in computer science instruction. The use of formative assessment within this context refers to the use of a “low stakes” assessment as an ongoing way to monitor student learning (Black & Wiliam, 2009). This participatory research approach allowed the project team to capture ideas and outcomes from the teachers that will guide the professional development design. An ongoing process of sharing and refining establishes a synergistic partnership that will continue to be the foundation of this RPP project. Participants Focus Group Participants Purposive sampling was used to select participants for this focus group. The Instructional Director at the SVPS High School facilitated the recruitment of teachers who teach computer science from within that high school. The teachers were then sent invitations to participate in the study. Interested teachers completed the consent form to participate in a focus group. Three focus groups were scheduled with ten teachers. The focus groups included two groups of 3 and one group of 4 participants, facilitated by members of the research team. The teachers who participated in the focus groups varied in their background and experience but taught a computer science course for the SVPS. Workshop Participants The research team then recruited the online Computer Science teachers from SVPS to participate in a summer workshop. These same teachers had identified themselves as interested in participating in a workshop created to identify large-scale needs and provide support to online Computer Science teachers. The participants engaged in a weeklong seminar where the research team presented topics such as Approach, Challenges, Solutions to Online APCSA, Online Teaching Strategies, Engagement within Online Learning, Auto graders, and Other Online Tools, and Culturally Relevant Computing and Social Impact. In addition to presentations, participants actively engaged in discussions surrounding these topics and connected their experiences teaching online Computer Science courses to the research presented. Data Collection Online Focus Groups on Zoom The researchers conducted three semi-structured focus groups using the breakout room functionality in Zoom. Each interview averaged about 26 min. The sessions were audio-recorded and then transcribed using Otter machine transcription, followed by human transcription. Two focus group questions were discussed and finalized by the research team. The focus group questions were directly aligned to the research questions of this study and were (1) What design, facilitation, and assessment strategies do teachers use to teach programming online? And (2) What recommendations do teachers have for those interested in teaching programming online? The responses from an additional four questions are not included in this study. Online Collaboration on Jamboard during Summer Workshop Following the participation in the focus group sessions, a subset of participants volunteered to participate in the summer workshop held in 2021. This workshop provided the opportunity to discuss questions asked during the focus group interviews to support a deeper understanding of teachers’ thoughts, experiences, and perceptions related to our research questions. The question for the online collaborative Jamboard activity was, “What design, facilitation, and assessment strategies are helpful to include in an AP Computer Science advanced course?” Data Analysis Focus Group Data The researchers used an inductive coding process (Miles et al., 2013) to analyze the data. Two researchers analyzed the data from each research question using the same process. The transcribed interviews were initially coded using an open coding process. These were color-coded to form different categories and grouped to develop themes. Once the coding was completed, the larger research team met to discuss the codes and categories generated. Workshop Jamboard Data Throughout the summer workshop, the participants were asked to engage with a Jamboard on an online collaborative activity responding to specific questions. Jamboard (Google Workspace, n.d.) is a digital interactive whiteboard developed by Google to work within the Google Workspace. This tool allows for collaboration by using a digital whiteboard, making it easy to create and share ideas in real-time, regardless of distance. The posts on the Jamboard were grouped to identify common themes. Results The results section presents the findings from the digital collaborative activity data collected during the summer workshop and the online focus groups. Design Strategies During the summer workshop, the participants were asked to engage with and reflect on the topics presented, and to share their experiences and expertise related to the research questions within this study. They were asked, “What design strategies are helpful to include in APCSA?” Fig. 2 includes a screenshot of the Design Jamboard.Fig. 2 Design Jamboard The following themes emerged: go to resources, examples, assessments, and making real-world connections. This included purposeful exposure to common misconceptions and resources to support these errors, a bank of high-quality resources (such as access to high-frequency vocabulary words related to the content area, and short videos created for students that align to the computer science curriculum). Additionally, multiple participants expressed the need for access to superiorly designed questions with answers that are not located using search engines (i.e., Google). However, it should also be noted that many teachers expressed that while courses can be adapted and additional resources may be included, there is hesitancy to make significant adjustments to the course shell as the school aims to provide continuity among its courses. There was a significant misconception surrounding the use of auto graders and other feedback tools. Many participants stated that they believed auto graders to be the automatic grading function in Canvas and were unaware of the potential impact of auto graders when applied to their course design. The following themes on computing and pedagogical tools and resources used emerged from the focus groups regarding design strategies. Computing and Pedagogical Tools and Resources Several teachers proposed the theme, online resources, as an essential instructional strategy and included both computing/programming resources and pedagogy tools. For the purpose of this study, we are operating under the shared understanding that a pedagogical tool is that which enhances a student’s understanding of the content or a support for a teacher who teaches in an online space. Furthermore, a computing tool or resource is a content specific application or software that supports students’ understanding strictly as it relates to computer science. Some of the online resources used by the teachers are mentioned in Table 1.Table 1 Computing and Pedagogical tools/Resources Computing tools/Resources Pedagogical tools/Resources GitHub Blue J Replit different types of online compilers for Java W3Schools Azura Visual Studio Gmetrix auto grader in code HS new certify Code.org java.org Code HS Kahoot Jam board Snap Microsoft Teams Collaboratory Video resources Facilitation Strategies Online facilitation is the ability of an instructor to promote learning in an online environment by fostering a positive learning experience and engaging with students in a way that supports personal growth. During the summer workshop, the teachers were presented with the question, “What facilitation strategies are helpful to include in APCSA?” Many of the summer workshop and focus group participants have various levels of experience teaching simultaneously online and in traditional face-to-face settings, requiring the instructor to be both effective facilitators virtually and in person. With limited time and resources, many teachers struggle when asked to transition between the two, as the digital divide has impacted both students and teachers. The following themes emerged from the participants’ responses on the facilitation Jamboard: AP CSA specific Free Response Questions (FRQ) examples, video resources and feedback. This included the need for a bank of free-response questions (without published answers), purposeful video resources that isolate skills (designed for students), and a way to support students by providing more detailed and meaningful feedback. Figure 3 includes a screenshot of the Facilitation Jamboard.Fig. 3 Facilitation Jamboard During the focus group interviews, many teachers mentioned that they struggled with online facilitation and described frustrations when they experienced a lack of student engagement or felt that their whole group communication was limited due to no synchronous learning sessions built into the course. These elements of course facilitation that teachers identified resulted in teachers having a limited understanding of students’ abilities, or lack thereof, until it was too late in the semester to provide additional support. Many participants emphasized that teachers must take an active role in their virtual learning environment and demonstrate their engagement in the course by promptly responding to students and their questions, promptly providing meaningful feedback, and incorporating additional resources to support errors made on an individual basis. While some of the teachers who participated in the focus groups were involved in course design, several were only tasked with teaching and facilitating the course. From the focus groups, the following facilitation themes emerged. Weekly Announcements Some of the instructional strategies mentioned as part of the course facilitation included “weekly announcements” and “we can add materials to our announcements.” One teacher commented, “we don’t really have flexibility in designing the courses. They’re structured for us, and the teachers get a Canvas shell, but we do have the flexibility to add supplemental material.” Live Synchronous Sessions Live synchronous sessions were also mentioned as part of the facilitation. A teacher added, “a few kids that would come in and ask questions, she would always record our sessions and make them available as archives so that students could then go back and view them.” A teacher added that the live sessions might not have worked for all students, but they conducted a live session for each topic. More Practice Videos Teachers thought it was essential to include more practice videos as part of course facilitation. They noted that providing students with various videos for each standard or concept provided similar explanations in slightly different ways to allow students multiple opportunities for enhanced clarity. Providing additional practice videos was a course facilitation strategy the teachers implemented to assist students in an asynchronous online setting. Assessment Strategies In the summer workshop, when the participants were asked to respond to the question “What assessments are helpful to include in AP CSA?”, the following themes emerged: the need for both a larger bank of programming questions, and access to shorter formative assessments. Additionally, teachers mentioned the need for supplemental assessments to be created, with emphasis placed on alignment to mastering specific computer science concepts. Figure 4 includes a screenshot of the Jamboard.Fig. 4 Assessment Jamboard Furthermore, in the responses to the need for summative assessments, participants identified the need to create assessments unique to the course each semester and situate the assignments within the “real world” context. Instructors also included the need to support students using computer science programs. Additionally, an overarching theme that was identified during the focus group interviews and reinforced during the workshop was that many of the participants were either unfamiliar with auto graders or had a limited understanding of what they were or, if they were familiar, they had little knowledge of what they were or their capabilities. The most popular and commonplace assessment strategies are quizzes, tests, state-administered standardized tests, and essays. While each of these relatively traditional forms of assessment has its place in a curriculum, it is becoming increasingly common within the field of education to be critical of these assessments as they limit students’ ability to demonstrate their knowledge on a specific topic or within a content area. Similarly, many of the participants in this study agreed that assessment strategies need to be carefully selected before applying them to a course. Assessments need to be both meaningful and carefully aligned with the course objectives and content standards. From the focus groups, the following assessment strategies emerged. Connection to College Board Teachers mentioned several instructional strategies exercised in their classrooms to align with the College Board examination. They used College Board materials, videos in the AP Classroom, and AP free-response style questions to prepare students for the AP Classroom. One teacher commented, “we’ve added things that have made it a much better course. We’ve added structure to it to make it seem more realistic, as far as testing is concerned with the AP exam.” Variety of Assessments and Feedback Teachers mentioned utilizing a variety of assessments in their online computer science course. Some of the teachers’ assessments included checkpoints to ensure students are prepared, tests including multiple-choice questions, projects, and timed free-response questions. They also emphasized the importance of providing feedback. In addition, teachers mentioned the importance of including an evaluation in the end. Evaluation is used to collect student feedback on the course to make improvements before the following implementation. Overall Instructional Strategies While the data collected were categorized by design, facilitation, and assessment, there were some strategies that included various aspects. Collaboration in Design and Teaching The interviewed teachers discussed the collaborative aspect of both design and facilitation used by this virtual public school. The course was assigned a course lead and included a large team of teachers. A teacher commented, “…have a team of the content experts develop the course, lay out the outline, and actually develop the content for the course.” While every teacher’s opinion is considered, changes are made based on the consensus. Also, one teacher noted, “Typically, we don’t take them away unless it’s a group decision….” Student Engagement A few of the teachers discussed the importance of student engagement. While getting the content on time is essential, it is also crucial to embed engaging and collaborative activities. One teacher commented, “A major platform that we started using to facilitate our content, which allowed the students to be more engaging, more engaged in the course as well as access to those tools.” Teachers discussed the importance of including short videos about 10 min in length to engage the students. Evidence-Based Teaching Practices A few teachers described using evidence-based practices such as modeling, guided practice, tutorials explaining how something is done, and scaffolding as instructional strategies in their online computer science course. Recommendations for Teachers Who Are Interested in Teaching Online The participants within this study provided valuable feedback and insight to support those interested in teaching computer science online. Teacher Preparation While several themes and pieces of advice emerged during the focus group discussion, the most prevalent theme was preparation. Among all three focus groups, roughly half of the responses spoke to the need for high-quality teacher preparation by attending a technical college specifically for computer science or attending ongoing professional development sessions. The professional development workshop offered by the technical college provides intensive support designed to prepare instructors of all levels, especially those who have not taught or studied computer science. Teacher Commitment The second most prevalent theme that resulted from this research question was the need to inform new teachers about the demands of the course, more specifically, the demand on the instructor’s time. Most participants spoke to the commitment that the course requires, both in providing feedback to students (as described as a feedback loop due to its continuous nature and the revisions students need to make to improve), the communication required to maintain relationships with students in a virtual, asynchronous environment and the most considerable demand stated was the time spent grading assignments. It was noted that while the grading takes significant time and effort from the instructor, the result is worth the work as the student implements the changes and improves both their product and understanding. Adapting Instruction The final theme addressed during the focus group discussion was the need to adjust expectations and adapt to the students within the classroom, albeit virtual. Many instructors noted that students and their prior knowledge vary significantly from semester to semester. Adjusting instruction to meet that baseline understanding of computer science is necessary to set up students for success. However, it is also important to note that students will be expected to still take the College Board assessment at the end of their course regardless of course entry knowledge. With this in mind, a participant shared the following, “.... I’ve taught computer science face to face and teaching it online, and different methods to approach, you know, the same objective, same units and just be very flexible. You can’t do it the same way.” Based on the conversations highlighted within this study among the instructors at the SVPS, it is evident that adjusting teaching, course design, and facilitation are necessary to meet the learners’ needs. However, when discussing assessment strategies, the teachers seem to agree that the presence of the AP examination weighs heavily on both students and instructors. Due to this pressure, there is less flexibility with assessment practices than with other strategies associated with course design. This has resulted in the necessity to have near-perfect alignment between practice problems, free response questions, and course assignments to mirror the possible questions students will be expected to answer on the AP exam providing opportunities for exposure. Discussion The phrase instructional strategies is being used here in an overarching way to encompass the three main strategies discussed within this study; design, facilitation, and assessment strategies. Each of these strategies has been identified as necessary, and through the assistance of our Research to Practice Partnership, several examples of each strategy have been provided. The facilitation strategies outlined in this paper also align with Berge’s (1995) roles of the Online Facilitator as the instructor engages in a Pedagogical, Managerial, Social and Technical Role simultaneously. Based on the examples that were shared by our participants, we can see a clear alignment between the needs and experiences that these teachers identified, and the similarities shared with the current body of research. As noted by Grover et al. (2015), design and facilitation strategies are essential to the overall success of an online course, such as incorporating active learning components into computer science courses. Similarly, by creating a course that challenges how our students encounter, engage, and reflect on their learning (Fink, 2013), teachers work to create a course that fosters active engagement and increases both student engagement and the overall effectiveness of the course. As highlighted in our literature review, while there is a large body of research that supports the design and facilitation of online course creation, there is limited evidence to support which facilitation strategies best support students within computer science courses. The few content-specific studies that were included that evaluated facilitation and course design (Grover et al., 2015; Huan et al., 2011 & Proulx, 2000) were able to provide an insight into traditional course design strategies and their effectiveness within online computer science courses. Similarly, the studies included about design thinking, while relevant to the K-12 educational sector, were not specific to the world of computer science education. However, when viewing this study through the lens of design thinking and its phases, there is clear alignment between the processes in which we completed this study and the phases that the design thinking process follows. With the significant emphasis that design thinking places on identifying problems and creating possible solutions, the basis of this study, follows these general principles. Additionally, throughout the duration of this study, it became apparent that there is a significant need to further develop high-quality resources that are available to online computer science teachers. Teachers expressed the need to have access to well-designed question banks that aligned with the course standards and assisted in preparing students to take the College Board examination, directly tied to the completion of the online course. Furthermore, the teachers expressed the desire to have these question banks inaccessible to the students, as they have experienced the negative effects of students using search engines to simply identify the answer, which limits the instructor’s understanding of their student’s knowledge. Similar recommendations were made by Klein et al. (2011) as they supported the addition and integration of auto graders as a potential solution for this challenge. Implications The findings of this study have implications for teachers who currently teach or wish to teach computer science online in the future. The various strategies used by these teachers will be beneficial when teaching computer science online. Teachers Computer Science teachers must use strong course design, facilitation and assessment strategies. Specifically, in their design teachers could use, “go-to resources’ ‘, examples, assessments, and assignments based within real-world situation. During facilitation, teachers could integrate AP CSA specific Free Response (FRQ) examples as well as video resources for practice and feedback. In addition they should communicate by sending weekly announcements, and live synchronous sessions. During assessment, teachers could choose questions from a larger bank of programming questions and also include shorter formative assessments in the course to provide a variety of assessments. Teachers should provide regular feedback to the students on these various assessments. Additionally, teachers must integrate methods for assessment and evaluation, promote student engagement, through evidence-based teaching practices. We see specific examples of this in the Jamboard and focus group findings as teachers identified the importance of maintaining a positive online presence through providing continuous feedback, meeting the needs of students through various communication methods, aligning course assignments and assessments to meet the demand of the course while embedding the task within a real-world context. Administrators and Instructional Designers The findings also have implications for administrators and instructional designers who support teachers in designing and delivering online courses. Instructional designers could use all the instructionalstrategies discussed above for design, facilitation, and assessment in the design of the course. Administrators can also benefit from these research focused findings and support the instructional designers and teachers to use these strategies in the Computer Science courses. The participants in this study explained the process in which changes, or edits can be made to their online course frameworks. Alterations would not be granted without stakeholder approval, which is built into the course design process to ensure that all online courses within the same state high school are held to the same standard. Participating in a study structured similarly to ours, allowed for the participants to share their thoughts and experiences openly and honestly with the current framework in place, which allowed for direct communication between teachers and their administration without fear or repercussions. This open forum and exchanging of ideas have directly benefited the stakeholders involved because when high quality edits are made the AP CSA course shell, teachers are supplied with a more supportive foundation, students are equipped with more online supports that are directly related to their course standards, and school leaders view their teachers as both content experts and advocates for their students, which should ultimately result in more students successfully passing their AP CSA exam. Students Finally, the study has implications for online students who will benefit from various instructional strategies used in the courses. Limitations There are a few methodological limitations to this study. This study only included teachers from one virtual public school from one state, and data was collected in three online focus groups and three Jamboards during a collaborative activity. This data may not be generalizable to non-virtual school settings. Teachers may have responded differently to the online facilitation of the focus groups and online workshop through Zoom and Jamboard than they might with face-to-face focus groups or interviews. Accessing the meeting with a phone instead of a computer, or only some teachers turning on their video may have impacted how they participated in the focus group. Additionally, the subset of participants who engaged in the summer workshop met synchronously for more extended periods (approximately five hours per day for five days). During this time, the data was collected in a group-like setting, which may have resulted in conformity among participants. While there was a significant benefit to conducting this portion of the study in a collaborative seminar setting, the largest of which was the sharing and melting of ideas and past experiences, social pressure was a likely natural consequence. In this unavoidable limitation of social pressure, participants change their beliefs or behavior to fit in with others, creating the possibility for swayed responses leading that might have influenced data. Future Directions While this study was conducted using interviews and from a digital whiteboard from online teachers at one virtual public school, this could be extended to teachers teaching online in various settings nationwide. Also, a large-scale survey will assist in collecting data on teacher perceptions regarding instructional strategies they use and teacher and student challenges. It is recommended that further research be conducted to directly identify and determine which course tools or programs, and student engagement techniques are best suited to support online computer science courses. In addition, interviewing administrators, parents and students will help us understand successful online teaching and learning strategies and challenges identified from various perspectives. Acknowledgements This project has been funded by the National Science Foundation under Grant No. 2031496. Declarations Conflict of Interests/Competing Interests This paper and the research conducted to support its conclusions was the result of a grant awarded by the (foundation) to three of the authors. The support provided by the grant allowed for the research to be conducted but had no influence on the data collected or the interpretation of the findings. There are no conflicts of interest to report. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Berge Z Facilitating computer conferencing: Recommendations from the field Technology 1995 35 1 22 30 Black P Wiliam D Developing the theory of formative assessment Educational Assessment, Evaluation and Accountability 2009 21 1 5 31 10.1007/s11092-008-9068-5 Crane, A., Chairperson, R., Aust, Y.-J., Lee, J., Rury, R., Branham, M., & Eckersley, M. (2018). Exploring best practices for implementing design thinking processes in K12 education by Ó 2018 (pp. 1–126) [dissertation]. Dabbagh N Pedagogical models for E-Learning: A theory-based design framework International journal of technology in teaching and learning 2005 1 1 25 44 Fink D Creating significant learning experiences: An integrated approach to designing college courses 2013 Jossey-Bass Garcia R Falkner K Vivian R Systematic literature review: Self-regulated learning strategies using e-learning tools for computer science Computers & Education 2018 123 150 163 10.1016/j.compedu.2018.05.006 Goode, J., Skorodinsky, M., Hubbard, J., & Hook, J. (2020). Computer science for equity: Teacher education, agency, and statewide reform. Frontiers in Education, 4(162). 10.3389/feduc.2019.00162 Google Workspace. (n.d.). Google Jamboard: Interactive Business Whiteboard \| Google Workspace. Workspace.google.com. https://workspace.google.com/products/jamboard/ Grover, S. (2021). Toward A framework for formative assessment of conceptual learning in K-12 computer science classrooms. Proceedings of the 52nd ACM Technical Symposium on Computer Science Education. 10.1145/3408877.3432460. Grover S Pea R Cooper S Designing for deeper learning in a blended computer science course for middle school students Computer Science Education 2015 25 2 199 237 10.1080/08993408.2015.1033142 Huan, X., Shehane, R., & Ali, A. (2011). Teaching computer science courses in distance learning. In Journal of Instructional Pedagogies Teaching computer science (p. 1). Keuning, H., Jeuring, J., & Heeren, B. (2016). Towards a systematic review of automated feedback generation for programming exercises. Proceedings of the 2016 ACM Conference on Innovation and Technology in Computer Science Education. 10.1145/2899415.2899422 Klein, R., Kyrilov, A., & Tokman, M. (2011). Automated assessment of short free-text responses in computer science using latent semantic analysis general terms (pp. 158–162). Li, Y., & Fu, Z. (2020). Creativity initiative: Design thinking drives K12 education from a future thinking. Cross-Cultural Design, 325–337. 10.1007/978-3-030-49913-6_28 Martin F Ritzhaupt A Kumar S Budhrani K Award-winning Faculty Online Teaching Practices: Course design, Assessment and Evaluation, and Facilitation The Internet and Higher Education 2019 42 34 43 10.1016/j.iheduc.2019.04.001 Martin, F., Wang, C., & Sadaf, A. (2020). Facilitation matters: Instructor perception of helpfulness of facilitation strategies in online courses. Online Learning, 24(1). 10.24059/olj.v24i1.1980 McGowan IS Towards a theory-based design framework for an effective E-learning computer programming course 2016 International Association for Development of the Information Society Miles MB Huberman AM Saldaña J Qualitative data analysis: A methods sourcebook 2013 Sage Proulx VK Programming patterns and design patterns in the introductory computer science course ACM SIGCSE Bulletin 2000 32 1 80 84 10.1145/331795.331819 Sorva, J., & Sirkiä, T. (2015). Embedded questions in ebooks on programming. Proceedings of the 15th Koli Calling Conference on Computing Education Research. 10.1145/2828959.2828961. The College Board. (2006). AP Computer Science A Exam. AP Central. https://apcentral.collegeboard.org/courses/ap-computer-science-a/exam. Veletsianos G Beth B Lin C Russell G Design principles for thriving in our digital world Journal of Educational Computing Research 2016 54 4 443 461 10.1177/0735633115625247 Wilson T Whitelock D Open University. Computers And Learning Research Group Facilitation of on-line learning environments: What works when teaching distance learning computer science students 1997 Open University, Computers, And Learning Research Group Zendler A Klaudt D Instructional methods to computer science education as investigated by computer science teachers Journal of Computer Science 2015 11 8 915 10.3844/jcssp.2015.915.927 Zulfikar AF Muhidin A Pranoto Suparta W Trisetyarso A Abbas BS Kang CH The effectiveness of online learning with facilitation method Procedia Computer Science 2019 161 161 32 40 10.1016/j.procs.2019.11.096
PMC009xxxxxx/PMC9005026.txt	==== Front J Nonverbal Behav J Nonverbal Behav Journal of Nonverbal Behavior 0191-5886 1573-3653 Springer US New York 35431380 398 10.1007/s10919-022-00398-2 Original Paper Two Means Together? Effects of Response Bias and Sensitivity on Communicative Action Detection http://orcid.org/0000-0002-6122-7845 Piejka Aleksandra apiejka@psych.pan.pl 1 Piaskowska Liwia 2 Okruszek Łukasz 1 1 grid.413454.3 0000 0001 1958 0162 Social Neuroscience Lab, Institute of Psychology, Polish Academy of Sciences, Jaracza 1, 00-378 Warsaw, Poland 2 grid.12847.38 0000 0004 1937 1290 Department of Neuropsychology, Faculty of Psychology, University of Warsaw, Warsaw, Poland 12 4 2022 2022 46 3 281298 3 3 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Numerous lines of research suggest that communicative dyadic actions elicit preferential processing and more accurate detection compared to similar but individual actions. However, it is unclear whether the presence of the second agent provides additional cues that allow for more accurate discriminability between communicative and individual intentions or whether it lowers the threshold for perceiving third-party encounters as interactive. We performed a series of studies comparing the recognition of communicative actions from single and dyadic displays in healthy individuals. A decreased response threshold for communicative actions was observed for dyadic vs. single-agent animations across all three studies, providing evidence for the dyadic communicative bias. Furthermore, consistent with the facilitated recognition hypothesis, congruent response to a communicative gesture increased the ability to accurately interpret the actions. In line with dual-process theory, we propose that both mechanisms may be perceived as complementary rather than competitive and affect different stages of stimuli processing. Keywords Communicative actions Signal detection theory Communicative intentions Response bias Point-light http://dx.doi.org/10.13039/501100004281 Narodowe Centrum Nauki 2016/23/D/HS6/02947 Okruszek Łukasz issue-copyright-statement© Springer Science+Business Media, LLC, part of Springer Nature 2022 ==== Body pmcIntroduction The ability to build a representative model of the surrounding environment is a key function of the human brain (Fletcher & Frith, 2009). Given the importance and complexity of social interactions and the increased salience of social signals, it has been proposed that social information modifies both bottom-up processing of sensory evidence and top-down expectancies (Brown & Brüne, 2012). While many lines of social cognition research have focused on the ability to read others’ intentions from their faces (Hugenberg & Wilson, 2013), it has been suggested that the capacity to accurately infer others’ intentions from their body movements lies at the core of human communication (Yovel & O’Toole, 2016)—misreading intentions when the other agent is not close enough to observe their face could have dire consequences on survival. Thus, it is believed that communicative actions (COM) of other people are the primary source of interpersonal information when the face is obscured and are crucial for adequately interpreting and engaging in social encounters (Centelles et al., 2013; Puglia & Morris, 2017). COM can be defined as body motions intended to convey specific information to the observer, e.g., greeting someone by hand-waving or drawing attention to something by pointing one’s finger at it. It has been proposed that these types of actions are processed differently than individual actions (IND), i.e., body motions without communicative intentions, such as using tools or exercising (Ding et al., 2017; Redcay & Carlson, 2015). In line with this notion, communicative gestures are easily recognized from motion kinematics by healthy individuals (Becchio et al., 2012), even after dramatically obscuring the availability of visual information using point-light motion displays (Zaini et al., 2013). High proficiency at detecting communicative interactions in third-party encounters (Manera et al., 2016) is associated with widespread activity of the main social brain networks (Centelles et al., 2013). Neuroimaging studies have also documented that detection of communicative intentions elicits activation of both mentalizing and action observation networks, as well as increased coupling between them (Ciaramidaro et al., 2014; Trujillo et al., 2020), and may be observed as rapidly as 100 ms after the presentation of an agent (Redcay & Carlson, 2015). These robustly replicated effects have been linked to semantic coupling within a dyad, i.e., the communicative gestures of one agent carry significant information about the expected response of the other (Manera, Becchio, et al., 2011). In line with this suggestion, which can be framed as the facilitated recognition hypothesis, the communicative gesture of one agent has been shown to increase the accuracy of detecting a masked agent responding to the gesture in a stereotypical and time-congruent way (Manera et al., 2013; Manera, Becchio, et al., 2011). Manera, Becchio, et al. (2011) showed that, for example, observing a hand gesture of an agent who invites someone to sit facilitates spotting an agent who sits down and is masked with additional noise-dots. Furthermore, dyadic social interactions were found to be stored in the working memory as a single chunk of information (Ding et al., 2017). It has been proposed that dyadic social interactions are bounded in unified representations providing an “initial perceptual framework” (Fedorov et al., 2018; Vestner et al., 2019), and thus the additional cues derived from temporal and behavioral congruence between observed individuals allow for increased sensitivity to COM presented in a dyadic context. This phenomenon, termed interpersonal predictive coding, is believed to stem from coupling between social perception and action observation networks and top-down expectations of specific responses to communicative gestures that shape bottom-up perceptual processes (Manera, Becchio, et al., 2011). It is also possible that, as in the case of other salient signals, the mere presence of two agents alters perceptual thresholds due to the importance of communicative signals. The very presence of a potential interaction partner may elicit communication bias, i.e., a tendency to infer communicative intentions from observed dyadic displays. In line with this hypothesis, it has been previously shown that observing the communicative gesture of one agent may prompt the perception of a second agent, even when the second agent is not present (Manera, Becchio, et al., 2011). Dyadic displays suggesting typical interaction patterns (i.e., two faces or silhouettes facing each other) were found to elicit more spontaneous attention in both adults (Roché et al., 2013) and preverbal children (Augusti et al., 2010). Abassi and Papeo (2020) demonstrated that body-selective visual cortex (extrastriate body area; EBA) responds more strongly to face-to-face bodies than identical bodies presented back-to-back and that the basic function of the EBA is enhanced when a single body is presented in the context of another facing body. These findings provide initial evidence that observing two agents (whose spatial proximity and mutual accessibility enable communicative interaction) may automatically alter expectations about potential interactions. These two hypotheses, namely facilitated recognition and dyadic communicative bias, offer different perspectives on the crucial mechanisms underlying preferential processing and the accuracy of detecting COM; the former highlights the importance of additional information provided by congruent responses, while the latter focuses on the initial expectations elicited by the context itself. However, whether the presence of a second agent enables more accurate discriminability between communicative and individual intentions or if it biases viewers to perceive third-party encounters as interactive has not been directly investigated. While the available literature supports the facilitated recognition hypothesis (e.g., only time-locked responses of the other agent facilitate recognition of communicative interactions; Manera et al., 2013), to our knowledge, no previous study has examined whether the second agent changes the response criterion by increasing the likelihood of reporting COM. As discussed by Stanislaw and Todorov (1999), separating the effects of the response criterion and sensitivity on hit rates and false alarms is crucial for accurately grasping the processes underlying classification and decision-making. Thus, the current study uses a signal detection theory (SDT) framework to investigate the implicit processes associated with either facilitated recognition or dyadic communicative bias based on the pattern of explicit responses during a COM detection task. Pilot Study Method To explore the hypothesized distinct effects of response bias and sensitivity on detecting COM, we retrospectively pooled data from previous projects performed in our lab. Each project utilized two tasks examining detection of communicative intentions from single (Gestures Task; (Jaywant et al., 2016) or dyadic (Communicative Interaction Database—5 Alternative Forced Choice; Manera et al., 2016) point-light walkers, which are described below. The point-light displays (PLD) technique was used to limit the availability of visual information about the presented person. Initially proposed by Johansson (1973), PLD provides an effective way to convey salient social information, such as the communicative or individual meaning of an action, in the absence of peripheral visual cues (Manera et al., 2010). The tasks were designed to measure two levels of action recognition: simple action classification (communicative vs. individual) and action interpretation (interpretation of the specific content of the presented actions). However, due to the large discrepancy between the format of action interpretation (5-alternative choice vs. verbal description), only action classification scores are included in the current analysis as they were collected in a similar manner. The sample consisted of 284 participants self-reporting no neurological or psychiatric diagnoses. All participants were below the age of 50, as the processing of PLDs may be impaired in older adults (Billino et al., 2008). Each participant provided informed consent prior to participation and was reimbursed upon completion of the study. All study protocols were approved by the Ethics Committee of the Institute of Psychology, Polish Academy of Sciences. The study took place at the Institute of Psychology in Warsaw. The tasks were presented in a counterbalanced order: 147 participants completed the CID-5 task first, while 126 participants completed the Gesture task first. Eleven participants were excluded from further analysis due to outlying scores (defined as COM or IND classification accuracy scores more than three standard deviations (SD) from the mean). Thus, the final sample consisted of 273 participants (126 males; age: M = 26.6, SD = 6.9). Requests for the data or materials used in all the studies can be sent via email to the corresponding author. Neither of the experiments reported in this article was preregistered. Communicative Interaction Database – 5AFC Format (CID-5) The task was originally created by Manera et al. (2016). It consists of 14 vignettes depicting two point-light agents interacting with each other and seven control vignettes presenting two point-light agents performing activities independently. Each point-light agent had 13 markers indicating head, shoulders, elbows, wrists, hips, knees, and feet. Every animation was presented twice, with a separation cross between both displays. The duration of each animation varied between 2.5 and 8 s, depending on the action presented. The second presentation was followed by two questions. First, participants were asked to indicate whether the vignette had depicted an interaction (COM) or independent actions (IND). After the first question, five response alternatives were presented, which included the correct action description and four incorrect response alternatives (two COM and two IND). Participants were asked to choose the correct description of the animation. More detailed information about the stimuli and the procedure can be found in Table 1 and in Manera et al. (2016). The task was validated for Polish sample and effectively utilized in previous studies carried out on clinical populations (Bala et al., 2018; Okruszek et al., 2018).Table 1 Stimuli from CID-5 task from Manera et al. (2016) Type Item Description COM Choose which one A asks B to choose between two objects. B takes an object COM Come closer A asks B to come closer. B moves forward COM Go out of the way A asks B to go out of the way. B moves over COM Imitate me A squats down, and asks B to imitate him. B squats down COM Look at the ceiling A asks B to look at something behind him on the ceiling. B turns around COM Look at the ground A asks B to look at something on the ground. B squats down COM Move this down A asks B to move something. B moves something COM No A says ‘No’. B stops COM Pick this up A points to B something to pick up. B picks something up COM Sit down A asks B to sit down. B sits down COM Squat down A asks B to squat down. B squats down COM Stand up A asks B to stand up. B stands up COM Stop A asks B to stop. B stops COM Walk away A asks B to walk away. B takes some steps IND Drink A drinks. B sits down IND Jump A jumps. B picks something up IND Lateral steps A makes some lateral steps. B takes something and eats it IND Look under the foot A looks under his foot. B moves something IND Sneeze A sneezes. B turns around IND Stretch A stretches. B moves something IND Turn over A turns over. B squats down Gestures from BioMotion The task consisted of 26 animations derived from a set provided by Zaini et al. (2013). Each animation depicted a single point-light agent made of 13 markers indicating head, shoulders, elbows, wrists, hips, knees, feet, and, additionally, 10 markers for finger joints of both hands. Thirteen animations presented an agent performing a communicative gesture and 13 presented an agent performing an object-oriented gesture. A similar task was previously utilized in Parkinson's disease research (Jaywant et al., 2016). A list of the stimuli for each category is presented in Table 1. The stimuli ranged in duration from 2 to 5.5 s. Each display was presented twice and followed by a question, regarding whether the agent was communicating something or privately using an object. Participants were instructed to press the assigned key depending on their answer and were awarded one point for each stimulus they correctly identified as communicative or non-communicative. After choosing a response, participants were asked to verbally describe the content of the presented animation, which was recorded by the examiner. Subsequently, the responses were rated by two independent judges to see how well they matched the correct responses given in Table 2. There was 95% scoring agreement for communicative actions and 96% for individual actions.Table 2 Stimuli from dataset of Zaini et al. (2013) used in Gestures task Communicative Individual Blowing a kiss Call (phone) “Calm down” Combing hair “Come here” Drinking water “Enough” Eating “Get out” Fishing “Good job” Hammering a nail “I'm cold” Juggling “I’m not listening” Opening a bottle “I'm sleepy” Shoveling “I’m watching you” Stirring “Over there” Sweeping the floor Shrug Washing Rubbing tummy Writing Statistical Analysis To examine response bias and discriminability between the signal (COM) and the noise (IND), two SDT parameters were extracted for the pilot study and the subsequent studies: criterion and sensitivity scores. The criterion parameter is a measure of the bias—a higher value suggests a more conservative response threshold, or a tendency to classify stimuli as noise, whereas lower values indicate a more liberal threshold, or a tendency to classify stimuli as the signal. Criterion equal to zero indicates no bias in any direction. The sensitivity parameter measures the ability to differentiate between the signal and the noise, and higher values reflect better discriminability. Both parameters were calculated with IBM SPSS Statistics 26 using a syntax proposed by Stanislaw and Todorov (1999), i.e., criterion = − (PROBIT(H) + PROBIT(F))/2 and sensitivity = PROBIT(H) − PROBIT(F), where H and F stand for hit rate and false-alarm rate, respectively. Accuracy for COM animations during the detection task was used as a hit rate, while the false alarm rate was calculated as 1—accuracy for IND. Moreover, following the recommendations of Macmillan and Creelman (2004), proportions of 0 and 1 were replaced with 1/(2 N) and 1 – 1/(2 N), respectively. In order to investigate whether facilitated recognition and dyadic communicative bias were observed during the processing of COM from dyadic displays, we compared SDT parameters associated with sensitivity for COM vs. non-COM actions (facilitated recognition) and response criterion (dyadic communicative bias) observed for the detection of COM actions from single and dyadic PLDs. Facilitated recognition should be linked to increased sensitivity when detecting COM from dyadic vs. single-agent displays, whereas dyadic communicative bias should be linked to a lower response criterion during dyadic vs. single-agent tasks, as it signalizes bias towards recognizing ambivalent stimuli as the signal. In order to examine differences in SDT parameters between the tasks, two paired t-tests were used to compare the criterion and sensitivity scores. One-sample t-tests were used to examine response bias (i.e., criterion significantly different from zero). Additionally, two mixed ANOVAs with criterion or sensitivity parameters as a within-subject factor and order of tasks as a between-subject factor were used in order to assess whether the order of displays (single-agent first vs dyadic first) could have an impact on participants’ scores. Results Mean hit and false-alarm rates for dyadic displays were 92% (SD = 7%) and 16% (SD = 13%) respectively, and 89% (SD = 9%) and 10% (SD = 13%) for single-agent displays (the values were based on the accuracy scores corrected in accordance with Macmillan & Creelman, 2004 recommendations). Criterion differed from zero significantly in both single-agent (M = 0.05; SD = 0.33) and dyadic displays (M = -0.19; SD = 0.32), suggesting bias in participants responses, single-agent: t(272) = 2.78, p < 0.01; dyadic: t(272) = − 9.77, p < 0.001. Criterion was found to be significantly lower for dyadic displays than for single-agent displays: t(272) = 8.49, p < 0.001, d = 0.514, 95% CI [0.19, 0.30]. Sensitivity was significantly lower for dyadic displays (M = 2.58; SD = 0.55) than for single-agent displays (M = 2.77; SD = 0.66): t(272) = 3.63, p < 0.001, d = 0.220, 95% CI [0.08, 0.28]. Additional analyses revealed no effects of order of the task for criterion, F(1, 271) < 0.001, p = 0.991, η2 < 0.001, or sensitivity, F(1, 271) < 0.001, p = 0.957, η2 < 0.001. Interaction effects were also found not significant, criterion: F(1, 271) = 0.87, p = 0.350, η2 = 0.003; sensitivity: F(1, 271) = 1.27, p = 0.261, η2 = 0.005. The results are presented in Fig. 1.Fig. 1 Results for SDT parameters in Pilot study. Error bars indicate standard deviation. **p < .001 Discussion The results of the first study support the dyadic communicative bias hypothesis, given the finding of a significantly lower response criterion for two agent vs single-agent displays. Interestingly, sensitivity to COM was lower for dyadic vs. single-agent presentations, which is contrary to the facilitated recognition hypothesis. These results are limited by several factors. The PLDs used in this study came from two available datasets (Jaywant et al., 2016; Manera et al., 2016) that differ in their basic visual characteristics, such as the number of markers (additional markers for hands in single-agent displays), perspective (second-person perspective in single-agent displays and third-person perspective in dyadic displays), and proportion of IND and COM displays (2:1 ratio for COM and IND in the CID-5 task). Thus, the increased sensitivity in single-agent displays might be associated with the additional information provided by the extra markers, while the decreased response criterion in dyadic displays could arise from the higher probability of encountering COM than IND. To address these methodological limitations, a follow-up experiment using modified dyadic displays was designed to investigate the replicability of the main findings. Study 1 Method To address the limitations of the pilot study, we modified the CID-5 task to create CID-Removed (CID-R). In CID-R, the side of the original stimuli that entailed the second agent response was masked to limit the visibility of the cues solely to the first agent actions. Participants were asked to assess whether the agent performed an individual gesture or communicated something to another person, invisible on the display. One animation was excluded due to visual overlap of the presented agents, resulting in 13 COM and 7 IND displays. Thus, the same set of first agent actions was used in the single (CID-R) and dyadic (CID-5) versions of the task. In each trial, the PLDs were presented twice. Upon the second presentation, participants were asked to distinguish whether the presented action was COM or IND. Participants were then asked to choose which of five descriptions best matched the presented animation. For CID-R, the original descriptions from CID-5 were modified by omitting the description of the second agent response (an example can be found in Table 3). The response format was the same for both tasks, which allowed for an unbiased comparison of the scores from each version of the task. A between-subject design was chosen due to the fact that the single- and two-agent displays consisted of the same actions, and thus showing both sets to participants would strongly influence their recognition of the repeated stimuli and not allow for counterbalancing.Table 3 Example of response alternatives for a communicative action "choose which one" for CID-5 and CID-R CID-5 CID-R Alternative Response COM (correct) A asks B to choose between two objects. B takes an object Asks (someone) to choose between two objects COM—1 A offers something to B. B takes an object Offers something (to someone) COM—2 A squats down and asks B to imitate him. B takes an object Squats down and asks (someone) to imitate him IND—1 A lifts something. B takes an object Lifts something IND—2 A weighs something in his hands. B takes an object Weighs something in his hands A sample of 80 participants (19–46 years old, with no self-reported history of neurological or psychiatric diagnoses) was recruited for Study 1. One participant was excluded due to highly outlying scores (sensitivity < − 1, 0% accuracy in single-agent IND detection), thus the final sample consisted of 79 participants of which 39 completed CID-5 (14 males, age: M = 24.97, SD = 5) and 40 completed CID-R (18 males, age: M = 24.3, SD = 5.66). There was no difference in gender, Χ2(1) = 0.68, p = 0.41, or age, t(77) = 0.56, p = 0.576, between the two groups. We originally planned to gather a sample of 96 participants, as calculated using GPower software (Faul et al., 2007) for a one-tailed comparison of two independent means, given effect size d = 0.514 (based on Cohen’s d for criterion scores difference in the pilot study) and power = 0.80. However, due to the COVID-19 pandemic, we had to suspend examination of participants. For a one-tailed comparison of two independent means with power = 0.80, the sample size of 79 should be sufficient to detect an effect size d = 0.564, which falls a little above the effect size observed in the pilot study. As Study 1 was not an exact replication of the pilot study and thus the effect size for a between-subject design could differ, we decided to analyze the gathered data. Statistical Analysis Two independent-sample t-tests were used to compare the criterion and sensitivity scores between CID-R and CID-5 tasks. Moreover, mixed ANOVA was used to measure the accuracy of interpretation of given actions. The type of action (COM vs. IND) was used as a within-subject factor, and the task (CID-5 vs. CID-R) as a between-subject factor. Results SDT Parameters Mean hit and false-alarm rates for CID-5 were 89% (SD = 13%) and 34% (SD = 29%) respectively, and 62% (SD = 14%) and 12% (SD = 8%) for CID-R (values were corrected as in the pilot study). Similarly to the pilot study, criterion differed from zero both for CID-R (M = 0.45; SD = 0.27) and CID-5 (M = − 0.48; SD = 0.58) conditions: CID-R: t(39) = 10.416, p < 0.001; CID-5: t(38) = − 5.08, p < 0.001. Criterion was found to be significantly lower in CID-5 condition than in CID-R condition: t(77) = − 9, p < 0.001, d = 2.034, 95% CI [− 1.14, − 0.72]. Sensitivity did not differ significantly between the CID-5 (M = 1.9; SD = 1) and CID-R (M = 1.57; SD = 0.54): t(77) = 1.79, p = 0.079, d = 0.403, 95% CI [0.18, − 0.04]. The results are illustrated in Fig. 2.Fig. 2 Results for SDT parameters in Study 1. Error bars indicate standard deviation. p < .001 Action Interpretation There was a significant main effect of action type, F(1, 77) = 164.03, p < 0.001, η2 = 0.681, with IND (M = 0.76; SD = 0.2) being interpreted more accurately than COM (M = 0.52; SD = 0.24). The main effect of task was also significant, F(1, 77) = 8.77, p < 0.010, η2 = 0.102, such that participants in the CID-5 group (M = 0.69; SD = 0.17) were more successful at identifying specific actions than participants in the CID-R group (M = 0.52; SD = 0.13). The interaction effect between type of action and task was significant, F(1, 80) = 166.54, p < 0.001, η2 = 0.684. Bonferroni-corrected post-hoc comparisons showed that COM actions were more successfully recognized in the CID-5 group (M = 0.69; SD = 0.17) than the CID-R group (M = 0.35; SD = 15%; 95% CI [0.27, 0.42], p < 0.001), while the reverse pattern was present for IND actions, with more successful recognition in the CID-R group (M = 0.83; SD = 0.14) than the CID-5 group (M = 0.69; SD = 0.22; 95% CI [0.6, 0.22], p = 0.001). The results are illustrated in Fig. 3.Fig. 3 Results for action interpretation in Study 1. Error bars indicate standard deviation. p < .001, p < .01 Discussion Again, in line with the dyadic communicative bias hypothesis, participants showed more liberal decision thresholds for classifying actions as COM in the CID-5 task than in the CID-R task. Although the SDT analyses do not support the facilitated recognition hypothesis, the presentation of dyadic displays increased recognition of specific COM actions but decreased recognition of IND actions. Several limitations of this study should be pointed out: The unequal number of IND and COM trials, significant variance in animation length (2.5–8 s), and between-subject design preclude inferences whether the observed differences can be attributed solely to dyadic vs. single action presentation or to any between-subject individual differences. To address these issues, we conducted a third study employing a novel set of PLDs with a counterbalanced, within-subject design. Study 2 Method Thirty-seven participants (20–49 years old, with no self-reported history of neurological or psychiatric diagnoses) participated in Study 2. The minimum sample size of 32 was determined using GPower software (Faul et al., 2007) for a comparison of two dependent means, given effect size d = 0.514 (based on the pilot study) and power = 0.80. The experimental stimuli were prepared using the Social Perception and Interaction Database database (Okruszek & Chrustowicz, 2020). The database consists of a set of PLDs created to facilitate examination of COM and IND from single-agent and dyadic animations. All of the actions recorded for the database can be freely combined, enabling the generation of well-matched vignettes. The database has been validated in two studies involving samples of healthy participants (Okruszek & Chrustowicz, 2020). The stimuli prepared for this study comprised 20 vignettes divided into 4 categories (single-agent COM, dyadic COM, single-agent IND, dyadic IND). The length of the stimuli ranged from 3 to 4.5 s. We chose 10 COM and 10 IND vignettes and split them into 2 counterbalanced versions of the task that were presented to 2 groups of participants (version A: N = 19, 10 males, age: M = 27.89, SD = 7.26; version B: N = 18, 5 males, age: M = 29.28, SD = 7.59). No differences in age, t(36) = − 0.57, p = 0.575, or gender, Χ2(1) = 2.37, p = 0.124, were found between the two groups. Single-agent COM animations from version A (e.g., an agent gives a signal to come closer) were presented as dyadic COM animations in version B by being matched with a gesture-congruent action of second agent (e.g., second agent comes closer). Similarly, single-agent IND from version A (e.g., an agent chops wood) were presented as dyadic in version B, matched with an incongruent action (e.g., second agent comes closer). Exemplary animations are presented in Fig. 4, and the exact scheme of the study can be found in Table 4.Fig. 4 Exemplary animations generated from Study 2. Panels A and B present dyadic and single-agent COM (“Come closer”), panels C and D present dyadic and single-agent IND (“Chopping wood”) Table 4 The scheme of the task in Study 2 Version Type Single Dyadic A COM Look there Cheer up Pick it up Come closer Sit down Give me that Squat down Go over there Stand up Look at this IND Digging Arm waving/Cheer up Skip A Drinking/Give me that Hammering a nail Sawing/Look there Lateral step Chopping wood/Come closer Shoveling Lateral kick/Pick it up B COM Cheer up Look there Come closer Pick it up Give me that Sit down Go over there Squat down Look at this Stand up IND Arm waving Digging/Stand up Drinking Skip A/Stand up Sawing Hammering a nail/Look at this Chopping wood Lateral Step/Squat Down Lateral kick Shoveling/Go over there Due to the COVID-19 pandemic, the experiment took place on an online platform. The task was divided into two blocks (single- and dyadic displays), each preceded by brief instructions (based on the instructions for the CID-5 task). Participants were asked to classify the animations in each block as COM or IND and choose from one of four potential descriptions, i.e., two COM and two IND alternatives for each vignette. The order of presentation was fixed: the block consisting of single-agent displays always preceded the dyadic displays. Starting with the more ambiguous (single-agent) stimuli prevented possible bias based on the cues given in the dyadic displays. Statistical Analysis In order to examine differences in SDT parameters between the tasks, two paired-sample t-tests were used to compare the criterion and sensitivity scores. Criterion scores were again compared to zero in order to examine response bias. Mixed ANOVA was used to measure the accuracy of interpretation of given actions with the type of action (COM vs. IND) and type of display (single vs dyadic) as within-subject factors and version of the task (A vs. B) as a between-subject factor. Results SDT Parameters Mean hit and false-alarm rates for dyadic displays were 86% (SD = 8%) and 21% (SD = 14%) respectively, and 60% (SD = 17%) and 11% (SD = 5%) for single-agent displays (values were corrected as in previous studies). Again, criterion differed from zero both for single-agent (M = 0.49; SD = 0.25) and dyadic displays (M = − 0.12; SD = 0.27): single-agent: t(36) = 11.99, p < 0.001; dyadic: t(36) = − 2.7, p < 0.05. Criterion was significantly lower for dyadic displays than for single-agent displays: t(36) = 11.09, p < 0.001, d = 1.82, 95% CI [0.50, 0.72]. In contrast, sensitivity was found to be significantly higher for dyadic displays (M = 2; SD = 0.53) as compared with single-agent displays (M = 1.51; SD = 0.52): t(36) = − 3.6, p < 0.001, d = 0.59, 95% CI [− 0.76, − 0.21]. The results are illustrated in Fig. 5.Fig. 5 Results for SDT parameters in Study 2. Error bars indicate standard deviation. *p < .001 Action Interpretation Both main effects of the within-subject factors were significant. Again, dyadic animations (M = 0.86; SD = 0.11) were interpreted more accurately than single-agent animations (M = 0.66; SD = 0.11), F(1, 35) = 87.60, p < 0.001, η2 = 0.715, and so were IND (M = 0.89; SD = 0.11) compared with COM (M = 0.63; SD = 0.12), F(1, 35) = 123, p < 0.001, η2 = 0.778. The interaction effect between the type of display and type of action was also significant, F(1, 35) = 122.26, p < 0.001, η2 = 0.777. Bonferroni-corrected post-hoc comparisons showed that COM was interpreted more accurately in dyadic displays (M = 0.86; SD = 0.16) than in single-agent displays (M = 0.4; SD = 0.18; 95% CI [0.39, 0.54], p < 0.001), while IND was interpreted more accurately in single-agent displays (M = 0.93; SD = 0.11) than in dyadic displays (M = 0.86; SD = 0.16; 95% CI [0.01, 0.13], p = 0.015). We additionally investigated the impact of versions of the tasks. While the analysis revealed a significant interaction between the type of action and version, F(1, 35) = 4.47, p = 0.042, η2 = 0.113, and a significant 3-way interaction, F(1, 35) = 5.22, p = 0.029, η2 = 0.130, the difference between scores in the two versions was found to be caused by one item. After its exclusion, the significance of the main effects and the interaction effect remained intact, while the effects of version became insignificant. The results are illustrated in Fig. 6.Fig. 6 Results for action interpretation in Study 2. Error bars indicate standard deviation. p < .001, p < .05 Discussion Study 2 provides further evidence for dyadic communicative bias, in line with the pilot study and Study 1, as participants had a more liberal response threshold when detecting COM actions for dyadic displays than single-agent displays. Furthermore, this study supports facilitated recognition, as participants showed a higher sensitivity to COM vs. IND displays when presented with dyadic vs. single-agent animations and, in line with Study 1, were more accurate when interpreting COM in dyadic (vs. single-agent) displays and IND in single-agent (vs. dyadic) displays. General Discussion The aim of the current study was to investigate the mechanisms associated with processing COM from single and dyadic displays. Although there is a large body of research focusing on the facilitated recognition hypothesis, which emphasizes the role of congruency between dyadic actions observed during dyadic communicative interactions for stimuli perception, we propose that the presentation of two agents itself is sufficient to elicit dyadic communicative bias, i.e., an increased likelihood of reporting COM when observing dyadic actions. Analysis of the SDT parameters linked to dyadic communicative response and facilitated recognition provides robust evidence for the former. A decreased response threshold for COM was observed for dyadic vs. single-agent animations across all three studies, regardless of their methodological differences. This consistency supports the notion that third-party encounters are salient social signals that may alter the observer’s perceptual thresholds. The opposite effect—i.e., an increased likelihood of perceiving the masked second agent after perceiving the communicative gesture of the first agent—has been previously shown using psychophysics methodology in both the general population (Manera, Del Giudice, et al., 2011) and clinical samples (Okruszek et al., 2019). Observations from the current study are complemented by mounting behavioral (Papeo et al., 2017) and neuroimaging (Abassi & Papeo, 2020) evidence of strong visual preference for dyadic cues in interaction-enabling configurations. Analysis of the sensitivity parameters throughout the studies provides mixed support for the facilitated recognition hypothesis. Higher sensitivity for COM in dyadic displays was only observed in the well-matched, within-subject design (Study 2). It is plausible that, in Study 1, the sensitivity scores may have been biased by individual differences in the participants of the two groups. Moreover, the pilot study utilized high-resolution PLDs with additional markers representing hands and fingers, which may have facilitated the recognition of COM actions and resulted in higher sensitivity to single-agent vs. dyadic actions. Although the SDT analyses do not provide congruent support for the facilitated recognition hypothesis, the presentation of dyadic displays increased recognition of specific COM actions—but decreased recognition of IND actions—in Studies 2 and 3. This pattern of findings suggests that, in line with the facilitated recognition hypothesis, the coupling of meaningful actions between two agents increases the ability to accurately interpret them. Notably, the reverse pattern of findings was found for IND actions: in the case of uncoupled actions, the presence of the second agent actually increases the processing effort and decreases the ability to correctly grasp both actions, which is congruent with the chunk-storage hypothesis (Ding et al., 2017). Taken together, our findings extend the literature by providing evidence that both communicative bias and facilitated recognition affect the processing of COM from dyadic vs. single-agent displays. Based on our results, these two mechanisms may be seen as complementary rather than competitive. In line with two-systems theory (Satpute & Lieberman, 2006), dyadic communicative bias and facilitated recognition may be linked with different stages of stimuli processing. Dyadic communicative bias may be more related to early orienting toward salient stimuli and preattentive processes (Papeo & Abassi, 2019), while more detailed analysis of specific information carried by the stimuli is needed to elicit facilitated recognition. In line with these proposals, a recent functional neuroimaging study reported increased coupling between the amygdala, which supports bottom-up orienting toward salient stimuli, and the medial prefrontal cortex, which supports top-down mentalizing abilities, during communicative vs. individual action processing from PLDs (Zillekens et al., 2019). Given the previously observed double dissociation in overt and covert mechanisms of communicative interactions processing in patients with schizophrenia (Okruszek Piejka et al., 2018) and high-functioning autism spectrum conditions (von der Lühe et al., 2016), investigation of these processes may provide vital clinical insights and extend previous conceptualization of the processes underlying one of the most basic and crucial human capacities. While the current article presents robust multi-study evidence for the communicative bias, some limitations of the current findings should be pointed out. First, there were significant discrepancies between the methodologies of the three presented studies. Even though different designs and paradigms were utilized, communicative bias has been consistently observed across studies. However, the similar consistency was lacking for sensitivity scores, thus no similar conclusions can be drawn about the impact of the second agent’s response on participants’ ability to process communicative gestures. Second, the unequal number of IND and COM stimuli and differences in timings of particular animations might have impacted the results in the pilot study and Study 1. However, Study 2 addressed those issues while still replicating the effect of dyadic presentation on response bias. Finally, the number of trials across paradigms is on the lower edge of SDT implementation. At the same time, a similar approach has been previously successfully implemented in CID studies (Manera et al., 2016). As the main finding of the study was repeatedly revealed across paradigms, participants groups, and study designs, the effect of communicative bias in dyadic displays can be considered as a consistent and significant insight into the mechanism of action perception and interpretation. Funding This work was supported by the National Science Centre, Poland (Grant No: 2016/23/D/HS6/02947). Declarations Conflict of interest The authors declare no conflicts of interest. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Abassi E Papeo L The representation of two-body shapes in the human visual cortex The Journal of Neuroscience 2020 40 4 852 863 10.1523/JNEUROSCI.1378-19.2019 31801812 Augusti E-M Melinder A Gredebäck G Look who’s talking: Pre-verbal infants’ perception of face-to-face and back-to-back social interactions Frontiers in Psychology 2010 1 161 10.3389/fpsyg.2010.00161 21833226 Bala A Okruszek Ł Piejka A Głębicka A Szewczyk E Bosak K Marchel A Social perception in mesial temporal lobe epilepsy: Interpreting social information from moving shapes and biological motion The Journal of Neuropsychiatry and Clinical Neurosciences 2018 30 3 228 235 10.1176/appi.neuropsych.17080153 29621926 Becchio C Manera V Sartori L Cavallo A Castiello U Grasping intentions: From thought experiments to empirical evidence Frontiers in Human Neuroscience 2012 6 117 10.3389/fnhum.2012.00117 22557961 Billino J Bremmer F Gegenfurtner KR Differential aging of motion processing mechanisms: Evidence against general perceptual decline Vision Research 2008 48 10 1254 1261 10.1016/j.visres.2008.02.014 18396307 Brown EC Brüne M Evolution of social predictive brains? Frontiers in Psychology 2012 3 414 10.3389/fpsyg.2012.00414 23181028 Centelles L Assaiante C Etchegoyhen K Bouvard M Schmitz C From action to interaction: Exploring the contribution of body motion cues to social understanding in typical development and in autism spectrum disorders Journal of Autism and Developmental Disorders 2013 43 5 1140 1150 10.1007/s10803-012-1655-0 23008056 Ciaramidaro A Becchio C Colle L Bara BG Walter H Do you mean me? Communicative intentions recruit the mirror and the mentalizing system Social Cognitive and Affective Neuroscience 2014 9 7 909 916 10.1093/scan/nst062 23620602 Ding X Gao Z Shen M Two equals one: Two human actions during social interaction are grouped as one unit in working memory Psychological Science 2017 28 9 1311 1320 10.1177/0956797617707318 28719763 Faul F Erdfelder E Lang A-G Buchner A G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences Behavior Research Methods 2007 39 2 175 191 10.3758/BF03193146 17695343 Fedorov LA Chang D-S Giese MA Bülthoff HH de la Rosa S Adaptation aftereffects reveal representations for encoding of contingent social actions Proceedings of the National Academy of Sciences of the United States of America 2018 115 29 7515 7520 10.1073/pnas.1801364115 29967149 Fletcher PC Frith CD Perceiving is believing: A Bayesian approach to explaining the positive symptoms of schizophrenia Nature Reviews. Neuroscience 2009 10 1 48 58 10.1038/nrn2536 19050712 Hugenberg K Wilson JP Faces are central to social cognition Oxford University Press 2013 10.1093/oxfordhb/9780199730018.013.0009 Jaywant A Wasserman V Kemppainen M Neargarder S Cronin-Golomb A Perception of communicative and non-communicative motion-defined gestures in Parkinson’s disease Journal of the International Neuropsychological Society 2016 22 5 540 550 10.1017/S1355617716000114 27055646 Johansson G Visual perception of biological motion and a model for its analysis Perception & Psychophysics 1973 14 2 201 211 10.3758/BF03212378 Macmillan NA Creelman CD Detection theory: A user's guide 2004 2 Psychology Press Manera V Becchio C Schouten B Bara BG Verfaillie K Communicative interactions improve visual detection of biological motion PLoS ONE 2011 6 1 e14594 10.1371/journal.pone.0014594 21297865 Manera V Del Giudice M Bara BG Verfaillie K Becchio C The second-agent effect: Communicative gestures increase the likelihood of perceiving a second agent PLoS ONE 2011 6 7 e22650 10.1371/journal.pone.0022650 21829472 Manera V Schouten B Becchio C Bara BG Verfaillie K Inferring intentions from biological motion: A stimulus set of point-light communicative interactions Behavior Research Methods 2010 42 1 168 178 10.3758/BRM.42.1.168 20160297 Manera V Schouten B Verfaillie K Becchio C Time will show: Real time predictions during interpersonal action perception PLoS ONE 2013 8 1 e54949 10.1371/journal.pone.0054949 23349992 Manera V von der Lühe T Schilbach L Verfaillie K Becchio C Communicative interactions in point-light displays: Choosing among multiple response alternatives Behavior Research Methods 2016 48 4 1580 1590 10.3758/s13428-015-0669-x 26487054 Okruszek Ł Chrustowicz M Social perception and interaction database—A novel tool to study social cognitive processes with point-light displays Frontiers in Psychiatry 2020 11 123 10.3389/fpsyt.2020.00123 32218745 Okruszek Ł Piejka A Wysokiński A Szczepocka E Manera V Biological motion sensitivity, but not interpersonal predictive coding is impaired in schizophrenia Journal of Abnormal Psychology 2018 127 3 305 10.1037/abn0000335 29369645 Okruszek Ł Piejka A Wysokiński A Szczepocka E Manera V The second agent effect: Interpersonal predictive coding in people with schizophrenia Social Neuroscience 2019 14 2 208 213 10.1080/17470919.2017.1415969 29227757 Papeo L Abassi E Seeing social events: The visual specialization for dyadic human-human interactions Journal of Experimental Psychology. Human Perception and Performance 2019 45 7 877 888 10.1037/xhp0000646 30998069 Papeo L Stein T Soto-Faraco S The two-body inversion effect Psychological Science 2017 28 3 369 379 10.1177/0956797616685769 28140764 Puglia MH Morris JP Neural response to biological motion in healthy adults varies as a function of autistic-like traits Frontiers in Neuroscience 2017 11 404 10.3389/fnins.2017.00404 28769743 Redcay E Carlson TA Rapid neural discrimination of communicative gestures Social Cognitive and Affective Neuroscience 2015 10 4 545 551 10.1093/scan/nsu089 24958087 Roché L Hernandez N Blanc R Bonnet-Brilhault F Centelles L Schmitz C Martineau J Discrimination between biological motion with and without social intention: A pilot study using visual scanning in healthy adults International Journal of Psychophysiology 2013 88 1 47 54 10.1016/j.ijpsycho.2013.01.009 23376597 Satpute AB Lieberman MD Integrating automatic and controlled processes into neurocognitive models of social cognition Brain Research 2006 1079 1 86 97 10.1016/j.brainres.2006.01.005 16490183 Stanislaw H Todorov N Calculation of signal detection theory measures Behavior Research Methods, Instruments, & Computers : A Journal of the Psychonomic Society Inc 1999 31 1 137 149 10.3758/BF03207704 Trujillo JP Simanova I Özyürek A Bekkering H Seeing the unexpected: How brains read communicative intent through kinematics Cerebral Cortex 2020 30 3 1056 1067 10.1093/cercor/bhz148 31504305 Vestner T Tipper SP Hartley T Over H Rueschemeyer S-A Bound together: Social binding leads to faster processing, spatial distortion, and enhanced memory of interacting partners Journal of Experimental Psychology: General 2019 148 7 1251 1268 10.1037/xge0000545 30652892 von der Lühe T Manera V Barisic I Becchio C Vogeley K Schilbach L Interpersonal predictive coding, not action perception, is impaired in autism Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 2016 371 1693 10.1098/rstb.2015.0373 Yovel G O’Toole AJ Recognizing people in motion Trends in Cognitive Sciences 2016 20 5 383 395 10.1016/j.tics.2016.02.005 27016844 Zaini H Fawcett JM White NC Newman AJ Communicative and noncommunicative point-light actions featuring high-resolution representation of the hands and fingers Behavior Research Methods 2013 45 2 319 328 10.3758/s13428-012-0273-2 23073730 Zillekens IC Brandi M-L Lahnakoski JM Koul A Manera V Becchio C Schilbach L Increased functional coupling of the left amygdala and medial prefrontal cortex during the perception of communicative point-light stimuli Social Cognitive and Affective Neuroscience 2019 14 1 97 107 10.1093/scan/nsy105 30481356
PMC009xxxxxx/PMC9005028.txt	==== Front Arab J Geosci Arabian Journal of Geosciences 1866-7511 1866-7538 Springer International Publishing Cham 10041 10.1007/s12517-022-10041-5 Original Paper Drought and flood dynamics of Godavari basin, India: A geospatial perspective Sarkar Soma ssarkar.delhi@gmail.com grid.8195.5 0000 0001 2109 4999 Department of Geography, Indraprastha College for Women, University of Delhi, Delhi, 54 India Responsible Editor: Amjad Kallel 12 4 2022 2022 15 8 77227 9 2021 1 4 2022 © Saudi Society for Geosciences 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Assessing incidence and trends of droughts and floods can provide essential information for better water resource management, particularly in the context of ongoing climate change. The present study has examined the prevalence of drought and flood events of Godavari basin, India, for the last four decades using the Standardized Precipitation Evapotranspiration Index (SPEI) 3 months and 9 months, and their trends using Mann–Kendall’s test and Sen’s slope. The spatial distribution, frequency, and intensity of drought and flood episodes in the basin are not uniform. A 20-year breakdown, 1980–1999 and 2000–2019, shows a drastic increase in drought frequency in Manjra and Pranhita sub-basins, while Godavari Upper, Indravati, and northern part of Weinganga have registered increased flood frequency. The monthly trend of SPEI-3 shows that most of the sub-basins have registered a negative trend (increasing dryness) only during the winter months, while under the SPEI-9 scenario both Manjra and Pranhita registered strong negative trend throughout the year. Sen’s slope estimations detected an increasing trend in severity from SPEI-3 to SPEI-9. Dryness severity trend is highest in Pranhita, followed by Manjra, Wardah, and Lower Godavari sub-basins, and the increasing wetness severity trend is in Upper Godavari, while rest of the sub-basins do not show any significant trend. The interior districts of the basin are more prone to the high drought conditions in near future if the current trends persist. To sustain the future climate change challenges, the vulnerable districts need urgent structural and non-structural changes in their current water resource management practices. Supplementary Information The online version contains supplementary material available at 10.1007/s12517-022-10041-5. Keywords Godavari basin SPEI Droughts Floods Water resource management practices issue-copyright-statement© Saudi Society for Geosciences 2022 ==== Body pmcIntroduction Drought is the second-most geographically extensive climate-related hazards after floods covering nearly one-third of disaster-affected population globally (CRED 2018). Droughts/floods may cause substantial economic or social damage to individuals, property, or the environment (Wilhite 2005). On the basis of the degree of severity, the length of event, areal extent, total loss of life, economic loss, social effect, long-term impact, suddenness, and occurrence of associated hazards, drought was ranked first among all hazards (Bryant 1991; EM-DAT., n.d). Implications of climate change on the nature of these disasters are clearly visible with the fact that each year is surpassing their previous years in disaster graph. Year 2020, besides being ravaged by ongoing COVID-19 pandemic, had 23 per cent more floods than the annual average of 163 events and 15 per cent more economic losses in droughts than the annual average of 6.5 billion US$ around the globe (CRED 2021). The 2020 edition of the World Water Development Report (WWDR 2020) emphasizes that climate change will affect the availability, quality, and quantity of water in the planet, which will threaten the basic human rights to water and sanitation for potentially billions of people. With climate variability, the surface water, which acts as the principal freshwater supply globally, will become less reliable and predictable and the importance of groundwater will increase (Kundzewicz and Döll 2009), and over exploitation may lead to groundwater stress. Approximately 53 per cent of agriculture in India is rain-fed (DoES 2016). Thus, droughts create severe water crisis problems for India’s rain-dependent farmers. The National Drought Mitigation Centre, USA, has defined drought under three categories, namely meteorological, agricultural, and hydrological. A fourth category namely, socio-economic drought, incorporating the concept of water supply and demand has also been included in the classification of drought (Wilhite and Glantz 1985). Nearly one-third of the 328 million hectares geographical area of India is affected by droughts, which includes about 39 per cent of cultivable land and about 29 per cent of the population (Das et al. 2007). The 2015/2016 drought in India had affected 330 million people (CRED 2021). Studies on precipitation data reveal that India has recorded an increase in moderate drought’s intensity and its areal spread and also increased frequency of multi-year drought occurrences during recent decades (Niranjan et al. 2013; Mallya et al. 2016). Under climate change scenario, frequency and intensity of drought events are likely to increase (NIC 2009). The ‘Special Report on Drought, 2021’ (UNDRR 2021) has estimated that the effect of severe droughts is at 2–5% on India’s gross domestic product (GDP). Unlike drought that is slow growing and long-term prevailing, floods are instantaneous and may cause extensive damage in short time. Between 1998 and 2017, globally more than 2 billion people are affected due to flood hazards (CRED 2018). In India, the plains are affected by floods almost every year. According to Disaster Management in India report 2010, out of 40 million hectares of the flood prone area in the country, on an average, floods affect an area of around 7.5 million hectares per year. In 2020, flooding was responsible for the 3rd deadliest event of the year costing 1,922 lives, with an estimated economic loss of 7.5 billion US$ (CRED 2021). The impacts of a flood are strongly influenced by the hydraulic characteristics of the flood itself, the physical characteristics of the flooded area, and the preparedness for evacuation. The floods may be classified into coastal floods, flash floods, river floods, drainage floods (aka urban flooding), tsunamis, and tidal waves/bore (Jonkman 2005). Several drought indices were developed regionally and globally, based on a range of climatic and socio-economic variables and parameters, in the past. Some of them are: rainfall anomaly index (Van Rooy 1965), crop moisture index (Palmer 1968), surface water supply index (Shafer and Dezman 1982), Palmer hydrological drought index (Karl 1986), Palfai aridity index (Pálfai 2002; Mezὄsi et al. 2016), aggregate drought index (Keyantash and Dracup 2004), decile index (Morid et al. 2006), and atmospheric crop moisture index (Uang-aree et al. 2017). Standardized precipitation index (SPI) (McKee et al. 1993) and Palmer drought severity index (PDSI) (Palmer 1965) are the most popular drought indices among all. However, both have certain limitations (Mishra and Singh 2010). Fixed time scale, inability to represent different drought types (Vicente-Serrano et al. 2010a), complex computation, and limited application where observations are scarce (Smakhtin and Hughes 2004) do not make PDSI a good choice to represent drought conditions. SPI, on the other hand, though considers precipitation and captures the multi-scalar nature of droughts, completely ignores temperature, an indicator highly correlated with the severity of drought (Abramopoulos et al. 1988; Fischer et al. 2007) and global warming scenarios. Recently, the Standardized Precipitation Evapotranspiration Index (SPEI), which considers both temperature and precipitation, has been proposed to quantify the drought conditions (Begueria et al. 2010; Vicente-Serrano et al. 2010a, 2010b). SPEI is capable of depicting the multi-temporal nature of hydrological, meteorological, and agricultural drought based on the probability of precipitation and potential evapotranspiration (P-PET) differences (Vicente-Serrano et al. 2010a) and therefore allows a more holistic approach to explore the effects of climate change on droughts. SPEI has been applied in many regional studies (Li et al. 2015; Bezdan et al. 2019; Ayugi et al. 2020; Qaisrani et al. 2021) and was found satisfactory to capture the multi-scalar characteristic of drought conditions. The index is also applied in India at the country level (Kumar et al. 2013; Dhangar et al. 2019) and regional levels (Alam et al. 2017; Malik and Kumar 2021). River Godavari, the longest of peninsular rivers, accounts for nearly 9.5 per cent of the total geographical area of India. The basin’s lower reaches and coastal areas are prone to cyclone and flood hazards, while the rain-shadow and plateau areas are prone to droughts. This basin has witnessed numerous extreme events of droughts and floods in last 100 years. A local climate modeling study has projected decreasing rainfall trends in the basin (Hengade and Eldho 2019). Yaduvanshi et al. (2020) have observed that extreme rain indices like consecutive dry days, highest one-day precipitation, rainfall > 10 mm, and rainfall > 20 mm show significant increase trend in recent decades. With less access to surface water due to frequent droughts, the dependency on groundwater has increased significantly in last few decades. According to ‘Groundwater Scenario Report’ of Govt. of Andhra Pradesh, in Andhra Pradesh and Telangana states alone, the well population has increased from 0.8 million to 2.5 million, and area irrigated through groundwater increased from 1million to 3.4 million hectares in just last three decades. As per Central Ground Water Commission report, with insufficient-cum-erratic rainfall and increased extraction, the groundwater level is falling in the basin. With increasing vulnerability and inefficient water resource management, climate change will further worsen the situation in most cases. The ‘Special Report on Drought, 2021’ (UNDRR 2021) has highlighted that in developing countries since the institutions treat drought as an episodic and outlier event and choose to respond only when drought emergencies arise, it perpetuates the drought vulnerabilities and its related crisis. In recent times, most of the scholars have investigated either the climate variability or projected drought scenarios of the basin at various scales (Bhavani et al. 2017; Masoor et al. 2020; Dixit et al. 2021; Kumar et al. 2021), but ignored the need to study the drought and flood dynamics of this huge basin under a single frame for better understanding the water resource management equations. The creation of a better knowledge base on the spatial and temporal variability of droughts and floods is necessary to improve hazard mitigation and preparedness. Water resource management helps a community to adapt the climate variability by reducing vulnerability and increasing resilience through implementing pro‐active measures (Estrela and Vargas 2012; Mortazavi et al. 2012). Therefore, the objectives of the paper are: 1) to assess frequency and intensity of drought and flood events of Godavari Basin using SPEI and 2) to examine the trend of these events at sub-basin level using Mann–Kendall’s test during last four decades. Materials and methods Study area The Godavari basin, the second largest basin of India, spreads over 312,812.00 Sq. Km of peninsular India. The longest river of the Deccan Plateau, River Godavari, originates near Trimbakeshwar of Nashik district at the height of 1067 m amsl, virtually bisects the plateau of Maharashtra, and flows for a length of about 1465 km before draining into the Bay of Bengal. The upper reaches of the basin are occupied by the Deccan Traps; the middle part of the basin is principally Archean granites and Dharwars; and the downstream part is occupied mainly by the Cuddapah and Vindhyan metasediments and rocks of the Gondwana group. The basin is very dissected and rugged in the north-eastern part too. The entire basin slopes toward the floodplains to the east that are very flat (0 to 3%) and often experiencing inundation. It covers 55 districts from the states of Maharashtra (48.7 per cent), Telangana (18.8 per cent), Chhattisgarh (12.4 per cent), Madhya Pradesh (7.8 per cent), Orissa (5.7 per cent), Andhra Pradesh (4.9 per cent), Karnataka (1.4 per cent), and Puducherry (0.01 per cent) (India-WRIS 2014). The term ‘district’ represents an administrative unit of an Indian state. The overall catchment comprises of 466 watersheds clustered into 8 sub-basins: Upper, Middle and Lower Godavari, Weinganga, Pranhita, Manjra, Indravati and Wardah. Around 37 per cent of the total basin area lies in the elevation zone of 500–1500 m (Fig. 1a). The Godavari basin has a tropical climate and is divided into three climatic regions arid, semiarid and humid (Fig. 1b). The average annual maximum and minimum temperature recorded was 32.85 °C in and 20.53 °C, respectively, in the basin (India-WRIS 2014). The average annual rainfall is 1096.92 mm, 85 per cent of it happens between the months of June and September from south-west monsoon (Fig. 1c). The basin frequently experiences droughts and floods. As per 1986 ‘Drought Area’ map prepared by National Atlas Thematic Mapping Organization, NATMO, Nasik, Ahmednagar, Bir, and Osmanabad were the drought prone districts. East Godavari, West Godavari, Nagpur, Pune, Nasik and Rangareddy are the most densely populated districts of this basin. There are about 59,293 tanks, most of them are small, nearly 870 reservoirs and nearly 286 major / medium irrigation projects in the basin.Fig. 1 Godavari Basin: a) Physiography (SRTM DEM), b) Aridity Index (World Atlas of Desertification, Office of the European Union, 2018), c) Box and whisker plot of monthly rainfall (1980–2019) Standardized Precipitation Evapotranspiration Index (SPEI) This study has extracted the 1° × 1° global gridded SPEI dataset developed by Vicente-Serrano et al. (2010a) for the Godavari basin from 1980 to 2019. The SPEI evaluates monthly difference between precipitation (P) and Potential Evapotranspiration (PET) over multiple time scales up to 48 months (Vicente-Serrano et al. 2010b). SPEI calculation was done based on the Thortnthwaite equation for estimating PET, using mean temperature data obtained from the NOAA NCEP CPC GHCN_CAMS gridded dataset and monthly precipitation data obtained from the 'first guess' Global Precipitation Climatology Centre (GPCC). With a value for PET, the difference between the P and PET for the month i is calculated:Di=Pi-PETi Vicente-Serrano et al. (2010a) expounded the details on the mathematical equation for computing the SPEI. SPEI can indicate the severity of dry/wet conditions using a (-/ +) scale, where a negative (positive) value represents dry (wet) conditions. The scale can be classified into four categories of dry/wet events: extreme (SPEI > -/ + 2.00), severe (-/ + 1.50 > SPEI > -/ + 1.99), moderate (-/ + 1.00 > SPEI > -/ + 1.49), and mild (0 > SPEI > -/ + 0.99). In this study, a threshold of SPEI ≤ -1.0 was considered to signify dry events and SPEI ≥ 1.0 as wet events over the Godavari basin. Here ‘dry’ represents ‘drought’ episodes and ‘wet’ represents ‘flood’ episodes. This study has considered SPEI-3 and SPEI-9 for the drought and flood analysis of the Godavari basin. SPEI-3 reflects a seasonal estimation of dry/wet conditions. SPEI-9 can be effective in showing the long term dryness/wetness patterns over distinct seasons, further reflecting the agricultural-cum-hydrological consequences of drought. It compares the dry/wet conditions for nine consecutive months with that recorded in the same nine consecutive months in all previous years of available data. It may be the delayed and prolonged effects of meteorological drought (Nicholson 2000; Wu et al. 2002). Theory of Runs In this study, drought/flood duration and severity were extracted from the SPEI-3 and SPEI-9 series as the drought characteristics factors by using the ‘Theory of Runs.’ Proposed by Yevjevich (1967) the theory is recognized as one of the most effective methods for analyzing time series, and was applied in numerous hydrological and meteorological studies (Griffiths 1990; Peel et al. 2004; Martínez et al. 2010; Razmkhah 2017). ‘Theory of Runs’ is a method to identify drought/flood episodes, estimate the average drought length over the desired time period and study their properties for their characterization (Moyé et al. 1988). ‘‘Run’’ denotes a series of the same phenomena that satisfies certain threshold level. This threshold level helps to identify the beginning, continuation or end of dry/wet events. In this study, a threshold of SPEI ≤ -1.0 was considered to signify dry events and SPEI ≥ 1.0 as wet events over the Godavari basin. Dry/wet events duration (d) denotes the period from the beginning to the end of the dry/wet event. The severity, intensity, and frequency for both dry and wet events over the study area were defined by Eqs. (1)–(3): Severity (S) for dry/wet condition is the cumulative sum of the SPEI ≥ -/ + 1.0 value based on the duration extent (Eq. 1).1 S=∑i=1480Index Intensity (I) of an event is the average severity during the study period (Eq. 2). Events with shorter duration and higher severity will have larger intensities.2 I=S/d Frequency (Fg) of the drought/flood event occurrence during the study period is defined in Eq. (3):3 Fg=ngNgx100 where ng is the number of months with dry/wet events (SPEI < -/ + 1.0), Ng is the total of the months in the study period, and g is a grid cell. Assessment of dryness/wetness trends This study has used Mann–Kendall (MK) test to detect the existence of a short-term and long term-trend in any data against the null hypothesis of no trend. Nonparametric MK test (Mann 1945; Kendall 1975) has been used widely (Araghi et al. 2016; Yaduvanshi and Ranade 2017; Ayugi and Tan 2019), as it does not require the data to conform to any specific probability distribution, and works well even with missing or abnormal values. The significance of the dry/wet trend was tested at three different significance levels α: 0.1, 0.05 and 0.01. MK test exploits the normal approximation (Z statistics). A Z-score at a given significance level denotes a monotonic trend of drought/flood events. The Mann–Kendall test statistic S can be obtained by Eq. (4):4 S=∑k=1n-1∑j=k+1nsgn(xj-xk) where n is the length of the sample, xk and xj are from k = 1, 2, …, n-1 and j = k + 1, …, n, and5 sgn(xj-xk)=1ifxj-xk>00ifxj-xk=0-1ifxj-xk<0 If n is bigger than 10, statistic S approximates to normal distribution. The variance of S can be acquired as follows:6 VAR(S)=118nn-1(2n+5) The values of S and VAR(S) are used to compute the test statistic Z as follows:7 Z=S-1VAR(S)ifS>00ifS=0S+1VAR(S)ifS<0 The Sen’s slope estimator, another non-parametric procedure, uses a linear model to detect the magnitude of the trend (Sen 1968).The Sen’s method is not greatly affected by gross data errors or outliers, and also it can be computed when data are missing. This method has been widely used in several hydro-meteorological time series (Lettenmaier et al. 1994; Yue and Hashino 2003; Partal and Kahya 2006; Gocic and Trajkovic 2013). The magnitude of the time series trend is calculated by:8 β=Medianxj-xkj-k,j>k where β is Sen’s slope estimate. β > 0 indicates upward trend in a time series, otherwise downward during the time period. The methodological framework of the present study is represented in Fig. 2.Fig. 2 Methodological framework of the study Result and Discussion Dry/Wet episodes in Godavari Basin Figure 3 provides an overview of historical analysis for SPEI 3- and 9-month for the years 1980 to 2019 over the Godavari basin. The distribution of drought and wet events were categorized under mild, moderate, severe, and extreme frequencies. SPEI-3 has identified extreme drought episode only during 2009, while 1992, 2002–03, 2016 and 2019 as strong episode. 2002–03 are found to be consecutive years that experienced meteorological droughts. Strong wetness episodes are identified only during 1990–91. Under SPEI-9 scenario, 2009–10 and 2015–16 were found to be two consecutive drought episodes of extreme nature. Similarly 1999–2000 and 2018–19 were found to be the two consecutive droughts of strong nature, along with isolated drought years 1991, 1996, and 2009. The main cause of strong to extreme nature of droughts during the mentioned years is the below normal precipitation (IMD). It is a matter of concern that frequency of the consecutive droughts has increased in the recent decades.Fig. 3 Godavari SPEI-3 and SPEI-9: dryness and wetness episodes between 1980 and 2019. SPEI-3 detects 2009 as year of extreme dryness with no extreme wetness years, while SPEI-9 detects 2009/2010 and 2015/2016 as years of extreme dryness and 1990 as extreme wetness Frequencies: drought (DF) and flood (FF) The spatial distribution of drought and flood episodes in the basin is not uniform. Figure 4 shows the drought and flood frequency under SPEI 3- and 9-month scenarios in last four decades. For SPEI-3, the drought frequency (DF) varies between 10 and 30 per cent, while for SPEI-9 DF varies between 10 and 40 per cent. Manjra and Pranhita sub-basins have experienced highest frequency of drought episodes, and Godavari Upper the least. A 20-year breakdown of the frequency distribution shows that a drastic increase in DF has taken place between 1980–1999 and 2000–2019 time frames for Manjra and Pranhita. Under SPEI-3, the entire basin has registered an increase in DF during 2000–2019 than earlier two decades. In SPEI-9, except Indravati, the rest of the sub-basins have experienced rise in the DF between 2000 and 2019. More than two fold increase in DF has been recorded in Pranhita sub-basin in recent decades compared to 1980–1999. The return time of drought in the interior Godavari basin was 5 years during 1980–1999 that has decreased to 3 years in recent decades. Regarding the flood frequency (FF), decreased frequency in the central part of the basin has been recorded for both SPEI 3 and SPEI 9-months. Godavari Upper, Indravati, and northern part of Weinganga have registered an increase in FF during the recent decades compared to 1980 -1999. FF in Upper Godavari has increased by more than double during recent decades under SPEI-9 scenario. The return period of flood in the Upper Godavari sub-basin was 5.5 years during 1980–1999 that has decreased to 3.5 years in recent decades.Fig. 4 Flood and Drought Frequency: SPEI 3- and 9-months during 1980–1999 and 2000–2019. Major increase in DF is detected in SPEI-9 for interior parts of the basin than in SPEI-3 during the study period. Moderate and high increase in FF is observed in Upper Godavari sub-basin for SPEI-3 and SPEI-9, respectively Intensities: drought (DI) and flood (FI) Figure 5 shows the drought and flood intensities under SPEI 3- and 9-month scenarios in last four decades. Severe DI prevails over central part of the basin for both SPEI 3 and SPEI 9-months. Similarly, the Godavari Upper and Lower sub-basins have severe FI for both SPEI 3 and SPEI 9-months scenarios. The central parts of the basin experience moderate FI under SPEI-3 but a severe FI under SPEI-9. Although FF has reduced in the last four decades in central Godavari basin, the intensity of wetness index is quite high for the years that have encountered the flood events.Fig. 5 Flood and Drought Intensity: SPEI 3- and 9-months. DI of severe nature is visible in major parts of Pranhita, Manjra, Warda and Weinganga sub-basins. Though Upper Godavari has recorded severe FI in both SPEI time frames, but for other sub-basins the spatial distribution of severe FI is mixed Temporal trends of Dry/Wet episodes in Godavari sub-basins The basin-wise short-term and long-term dryness/wetness trends for each month is important from agriculture point of view. It helps to pro-actively adapt appropriate mechanism to overcome the future challenges of droughts and floods. Figure 6 shows the sub-basin-wise monthly SPEI 3- and 9-month trend results based on the application of MK test (Z parameter) for different time series. In this figure, color code represents the negative/positive trend at three significance levels, α: 0.01, 0.05 and 0.1. Both negative/downward (increasing dryness) and positive/upward (increasing wetness) trends have been represented along with the no trend months. In SPEI-3 except Indravati, rest of the sub-basins show a negative trend during the winter months (January/ February/ December). Manjra and Pranhita sub-basins have registered negative trend during the monsoon months (June – August). A positive trend (increasing wetness) has been recorded only in Upper Godavari sub-basin during July to September months (α: ≤ 0.05), while Indravati sub-basin has registered increased wetness during the month of October (α = 0.1). Annually no significant trend is found for Indravati, Godavari Middle, and Godavari Upper sub-basins, while rest of the basins have shown significant downward trend (Fig. 7). Under SPEI-9 scenario, both Manjra and Pranhita register strong negative (increasing dryness) trend throughout the year. Wardah sub-basin has April and August months, and Godavari Lower has November and December months that have recorded increasing dryness trend. Godavari Upper is the only sub-basin that has recorded strong positive trend during monsoon and post-monsoon months. Godavari Middle, Indravati and Weinganga have shown no dryness/wetness trend in any of the months. Annually, Manjra and Pranhita registered strong negative trend (α = 0.01); Godavari Upper has recorded positive trend (α = 0.1); and rest have no trends (Fig. 8).Fig. 6 Sub-basin-wise monthly SPEI-3 and SPEI-9 trends. At α = 0.01, January month of 4 out of 8 sub-basins have recorded increasing dryness under SPEI-3, and Manjra and Pranhita have recorded increasing dryness under SPEI-9 for July and August months under SPEI-9. Upper Godavari is only sub-basin with increasing wetness at α = 0.05 during SW monsoon months under both SPEI-3 and 9 months Fig. 7 Sub-basin-wise annual trend of SPEI-3. Pranhita, Manjra, Wardah and Godavari lower have recorded significant negative trends Fig. 8 Sub-basin-wise annual trend of SPEI-9. Godavari Middle, Indravati and Weinganga show no significant trends; Godavari Upper shows significant positive trend; while rest shows significant negative trends Sen’s slope estimations detected an increasing trend in severity from SPEI-3 to SPEI-9. Dryness severity trend is highest in Pranhita, followed by Manjra, Wardah and Lower Godavari sub-basins, and the increasing wetness severity trend is only visible in Upper Godavari, while rest of the sub-basins do not show any significant trend. If the dryness trend persists in near future, then Bidar, Madak, Ranga Reddy, Osmanabad, Warangal, Nizamabad, Karimnagar, Khammam, Adilabad, Yavatmal, Chandrapur, Wardha, Bhandara, West Godavari and Bijapur districts will be worst affected due to drought situations. Discussion Drought and flood studies at basin level in India have mainly considered rainfall anomaly (Gore and Sinha Ray 2002). This study has, therefore, considered SPEI 3- and 9-months using precipitation and temperature data for drought and flood occurrence and trend analysis in Godavari basin in smaller spatial scale. The SPEI 3 and SPEI 9- months have appropriately identified the distribution of short- and long-term drought situations of the Godavari basin over last 40 years. Prolonged SPEI-3, representing the short-term dryness for a few consecutive months, leads to long-term dryness was captured by SPEI-9. Since the agricultural drought is commonly related to the deficit in soil moisture and has direct implications on crop yield, increasing dryness trend during monsoon and winter months under SPEI-9 is a major concern. Moreover, the episodes of drought of moderate to severe nature have increased by 2 to 3 folds in many parts of the basin in last 20 years than before. The basin falls in six agro-ecological zones, where interior part of the basin is characterized by moderate rainfall (500 − 1000 mm annually) with high variability (coefficient of variation greater than 20 per cent) and dry spell in winter and hot summer. The increasing dryness may have severe implications on agricultural output in near future. As per 2010 estimation, the per capita water availability of Godavari basin was 1486.01cubic m, which was less than the national average (1608.26 cubic m). And it was estimated that by 2050 it will be 1053 cubic m (CWC 2013). As per global standards, a region having per capita water availability less than 1700 cubic m is considered as water stressed, and less than 1000 as water scarce. Thus, Godavari is already a water stressed region and increasing dryness trend may push it to the verge of water scarce. Therefore, when the NATMO’s 1986 ‘Drought Atlas’ captured the drought situations being concentrated in Upper Godavari and Manjra sub-basins, the present study has found that dryness has extended toward interior parts of the basin comprising the districts Bidar, Madak, Ranga Reddy, Osmanabad, Warangal, Nizamabad, Karimnagar, Khammam, Adilabad, Yavatmal, Chandrapur, Wardha, and Bhandara. Water resources mismanagement, both surface and ground water, will further increase the drought risk of these districts. Large part of Godavari is under agricultural land utility (59 per cent), and tank irrigation in the basin had a great bearing to the development of its agricultural economy and livelihoods of diverse communities on the whole in past. However, in last few decades, wells and bore-wells assumed primacy in the irrigation systems of the basin due to the availability of credit facilities and electricity at subsidized rates, deteriorating conditions of the tanks and the change in land use in the tank catchments (ISEC 2006; Pingle 2011; Kumar and Vedantam 2016). Thus, it is necessary to immediately undertake the work of streamlining the usage of surface and ground water for irrigation, industry and municipal supply. A closer look into the relationship between the rainfall and ground water depth of the basin may give better insights toward managing the water resources. In past three decades, Upper Godavari, Lower Godavari, Pranhita and Indravati sub- basins have experienced positive trend in annual rainfall, while rest have recorded negative trend (Fig. 9). Similarly, groundwater situation is highly variable in the basin where in some of the sub-basins the ground water depth is greater than 7 m below ground level (m bgl) throughout the year. Manjra sub-basin has registered highest fall in ground water levels in last three decades. The correlation between rainfall and ground water depth shows that except Manjra and Weinganga, rest of the sub-basins experiences reduction in ground water depth with increase in rainfall. Weinganga does not show any significant correlation. Manjra that already receives very less rainfall and have geo-morphometric constraints to groundwater recharge, at one point of time groundwater exploitation was rampant. As per the 4th minor irrigation census data (2006–07), in a single district of Beed located in Manjra sub basin, area irrigated by ground water was 1463.24 sq. km., whereas the surface water accounted for about 82.43 sq.km; thus, ground water accounted for 94% of net irrigated area (Dhonde 2014). Such regions need systematic creation of surface water resource management and ground water recharge infrastructures.Fig. 9 Sub-basin-wise rainfall and ground water fluctuation trend in last three decades (WRIS-India, https://indiawris.gov.in/wris/#/groundWater) Note: Groundwater data is available between 1993 and 2019; thus, the correlation has considered the said time frame only Recently, understanding the significance of the tanks to climate adaptation and water management, state governments of Maharashtra and Telangana have initiated the tank rejuvenation and farm pond creation programs namely, Gaalmukt Dharan, Gaalyukt Shivar Yojana (GDGS) 2017, ‘Magel Tyala Shet Tale (Farm Pond on Demand)’ scheme 2016 and Mission Kakatiya (MK) 2015. However, no fund has been allotted for the maintenance of the renovated tanks. Instead of heavily investing on developing new infrastructure, restoring traditional water tanks is more viable (Gujja et al. 2009) with community participation. A comparative study conducted on the semi-arid regions of Marathawada, Vidarbha and Saurashtra has found that the groundwater levels in Saurashtra has increased immensely compared to other two regions during the same study period, and attributed it to the community-led construction of rainwater harvesting and artificial recharge structures in the region (Patel et al. 2020). It has further mentioned that groundwater recharge situation in Marathawada and Vidarbha has not equally flourished even after efforts by the Government as expected because it involved little community participation and focused more on regulation. While the highly urbanized districts like Thane, Nasik and Ahmadnagar are experiencing increased number of heavy rainy days (IMD 2020) and subsequently increased occurrences of flood situations; Parbhani, Hingoli, and Nanded, lying downstream, have suffered extreme water stress situation during the same period. It’s high time that the tendency to view flood and drought risks separately be avoided, and must consider them to be the part of an integrated multi-hazard risk reduction strategy. Revisiting the indigenous water management practices with community participation like ‘Jal Khet’ – rainwater harvesting structures in Diwas districts of Malwa region; the Ramtek model, traditional water harvesting structures of Ramtek in Maharashtra, may provide an insight to the need for encouraging such practices. Conclusion Increased climate variability has great implications on the incidence of extreme hazards like droughts and floods. Occurrences of these hazards are not new to Godavari basin, but the present study has shown that their intensity and frequency have increased in recent decades. Moreover, their trends are not uniform throughout the basin; when upper Godavari sub-basin registered increased wetness, Manjra and Pranhita sub-basins recorded increased dryness. UN-Water, 2021 under SDG Target 6.5 suggests that implementation of integrated water resources management at all levels, including through transboundary cooperation as appropriate, is necessary to balance competing water demands from across society and the economy. Understand the ongoing climate variability and weather extremes, it is highly essential that regions of water excess and regions of water stress of a basin should be studied under the single framework of water resource management for optimality and sustainability. With this objective present study has attempted to map the trend, intensity and frequency of both drought and flood episodes of the basin. Use of high resolution climate data may provide better insights to the drought and flood severity at watershed level, though such data availability are limited in open source domain. Even at the coarser resolution SPEI-3 and 9 months have well captured the spatio-temporal dimensions of drought/flood events at sub-basin level within Godavari basin. Such mapping methods can help regional planners and disaster managers to access the prevailing water resource management practices and accordingly recommend both structural and non-structural changes required to sustain the future climate change challenges. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (XLSX 16 KB) Supplementary file2 (XLS 707 KB) Funding Not available. Declarations Conflict of interest The author declares no competing interests. ==== Refs References Abramopoulos F Rosenzweig C Choudhury B Improved ground hydrology calculations for global climate models (GCMs): soil water movement and evapotranspiration J Clim 1988 1 921 941 10.1175/1520-0442(1988)001<0921:IGHCFG>2.0.CO;2 Alam NM Sharma GC Moreira E Jana C Mishra PK Sharma NK Mandal D Evaluation of drought using SPEI drought class transitions and log-linear models for different agro-ecological regions of India Phy Chem of the Earth Parts a/b/c 2017 100 31 43 10.1016/j.pce.2017.02.008 Araghi A Mousavi-Baygi M Adamowski J Detection of trends in days with extreme temperatures in Iran from 1961 to 2010 Theor Appl Climatol 2016 125 213 225 10.1007/s00704-015-1499-6 Ayugi BO Tan G Recent trends of surface air temperatures over Kenya from 1971 to 2010 Meteorol Atmos Phys 2019 131 1401 1413 10.1007/s00703-018-0644-z Ayugi B Tan G Niu R Dong Z Ojara M Mumo L Babaousmail H Ongoma V Evaluation of Meteorological Drought and Flood Scenarios over Kenya East Africa Atmosphere 2020 11 307 Begueria S Vicente Serrano SM Martínez MA A multiscalar global drought dataset: the SPEIbase: A new gridded product for the analysis of drought variability and impacts Bull Am Meteorol Soc 2010 91 10 1351 1356 10.1175/2010BAMS2988.1 Bezdan J Bezdan A Blagojević B Mesaroš M Pejić B Vranešević M Pavić D Nikolić-Đorić E SPEI-Based Approach to Agricultural Drought Monitoring in Vojvodina Region Water 2019 11 1481 10.3390/w11071481 Bhavani P, Chakravarthi V, Roy PS, Joshi PK, Chandrasekar K (2017) Long-term agricultural performance and climate variability for drought assessment: a regional study from Telangana and Andhra Pradesh states, India. Geomatics, Natural Hazards and Risk 8(2). 10.1080/19475705.2016.1271831 Bryant EA Natural Hazards 1991 Cambridge Cambridge University Press CRED (2018) Economic Losses, Poverty & Disasters 1998–2017. Centre for Research on the Epidemiology of Disasters and UN Office for Disaster Risk Reduction. CRED (2021) CRED & UNDRR. 2020: The Non-COVID Year in Disasters. Centre for Research on the Epidemiology of Disasters and UN Office for Disaster Risk Reduction. Brussels. CWC (2013) Water and Related Statistics. Water Resources Information System Directorate, Information System Organisation, Water Planning & Project Wing, Central Water Commission, GoI. Das SK, Gupta RK, Varma HK (2007) Flood and Drought Management through Water Resources Development in India. WMO Bull 56(3) Dhangar N Vyas S Guhathakurta P Mukim S Tidke N Balasubramanian R Chattopadhyay N Drought monitoring over India using multi-scalar standardized precipitation evapotranspiration index Mausam 2019 70 4 833 840 10.54302/mausam.v70i4.277 Dhonde UV (2014) Ground Water Information, Beed District, Maharashtra. Central Ground Water Board, Ministry of Water Resources, Government of India, Nagpur. Dixit S Tayyaba S Jayakumar KV Spatio-temporal variation and future risk assessment of projected drought events in the Godavari River basin using regional climate models Journal of Water and Climate Change 2021 12 7 3240 3263 10.2166/wcc.2021.093 DoES (2016) State of Indian Agriculture 2015–16. Directorate of Economics & Statistics, Ministry of Agriculture & Farmers Welfare. EM-DAT. (n.d.) The International Disaster Database; Centre for Research on the Epidemiology of Disasters CRED. https://public.emdat.be Estrela T Vargas E Drought management plans in the European Union: The case of Spain Water Resour Manage 2012 26 1537 1553 10.1007/s11269-011-9971-2 Fischer EM Seneviratne SI Vidale PL Lüthi D Schär C Soil moisture–atmosphere interactions during the2003 European summer heatwave J Clim 2007 20 5081 5099 10.1175/JCLI4288.1 Gocic M Trajkovic S Analysis of changes in meteorological variables using Mann-Kendall and Sen's slope estimator statistical tests in Serbia Glob Planet Change 2013 100 172 182 10.1016/j.gloplacha.2012.10.014 Gore PG Sinha Ray KC Variability in drought incidence over districts of Maharashtra Mausam 2002 53 4 533 542 10.54302/mausam.v53i4.1668 Griffiths GA Rainfall deficits: distribution of monthly runs J Hydrol 1990 115 219 229 10.1016/0022-1694(90)90205-C Gujja B Dalai S Shaik H Goud V Adapting to climate change in the Godavari River basin of India by restoring traditional water storage systems Clim Dev 2009 1 3 229 240 10.3763/cdev.2009.0020 Hengade N Eldho TI Relative impact of recent climate and land cover changes in the Godavari river basin India J Earth Syst Sci 2019 128 94 10.1007/s12040-019-1135-4 IMD (2020) Observed Rainfall Variability and Changes Over Maharashtra State; Met Monograph No.: ESSO/IMD/HS/Rainfall Variability/16(2020)/40 India-WRIS (2014) Godavari Basin; Central Water Commission (CWC) and National Remote Sensing Centre (NRSC), Indian Space Research Organisation (ISRO), Ministry of Water Resources, Government of India. ISEC (2006) Andhra Pradesh Community Based Tank Project: Environmental and Social Assessment Study; Report no. E1559, Centre for Ecological Economics and Natural Resources, Institute for Social and Economic Change, Bangalore. Jonkman SN Global Perspectives on Loss of Human Life Caused by Floods Nat Hazards 2005 34 151 175 10.1007/s11069-004-8891-3 Karl TR (1986) The sensitivity of the Palmer Drought Severity Index and Palmer”s Z-Index to their calibration coefficients including potential evapotranspiration; J Appl Meteorol Climatol. 25:77–86 Kendall MG Rank Correlation Methods 1975 London Griffin Keyantash JA Dracup JA An aggregate drought index: Assessing drought severity based on fluctuations in the hydrologic cycle and surface water storage Water Resour Res 2004 40 W09304 10.1029/2003WR002610 Kumar KN Rajeevan M Pai DS Srivastava AK Preethi B On the observed variability of monsoon droughts over India Weather Clim Extremes 2013 1 42 50 10.1016/j.wace.2013.07.006 Kumar KS AnandRaj P Sreelatha K Sridhar V Regional analysis of drought severity-duration-frequency and severity-area-frequency curves in the Godavari River Basin India Int J Climatol 2021 41 12 5481 5501 10.1002/joc.7137 Kumar MD, Vedantam N (2016) Groundwater Use and Decline in Tank Irrigation? Analysis from erstwhile Andhra Pradesh. In: Kumar MD, James AJ, Kabir Y (eds) Rural Water Systems for Multiple Uses and Livelihood Security, Elsevier. 10.1016/C2015-0-00906-6 Kundzewicz ZW Döll P Will groundwater ease freshwater stress under climate change? Hydrol Sci J 2009 54 4 665 675 10.1623/hysj.54.4.665 Lettenmaier DP Wood EF Wallis JR Hydro-climatological trends in the continental United States, 1948–88 J Clim 1994 7 586 607 10.1175/1520-0442(1994)007<0586:HCTITC>2.0.CO;2 Li X He B Quan X Liao Z Bai X Use of the Standardized Precipitation Evapotranspiration Index (SPEI) to Characterize the Drying Trend in Southwest China from 1982–2012 Remote Sens 2015 7 8 10917 10937 10.3390/rs70810917 Malik A Kumar A Application of standardized precipitation index for monitoring meteorological drought and wet conditions in Garhwal region (Uttarakhand) Arab J Geosci 2021 14 800 10.1007/s12517-021-07158-4 Mallya G Mishra V Niyogi D Tripathi S Govindaraju RS Trends and variability of droughts over the Indian monsoon region Weather Clim Extremes 2016 12 43 68 10.1016/j.wace.2016.01.002 Mann HB Nonparametric tests against trend Econometrica 1945 13 245 259 Martínez MD Lana X Burgueño A Long-term rainfall monthly shortage in Spain: spatial patterns, statistical models and time trends Int J Climatol 2010 30 11 1668 1688 10.1002/joc.2017 Masoor Md, Rehman S, Avtar R, Sahana M, Ahmed R, Sajjad H (2020) Exploring climate variability and its impact on drought occurrence: Evidence from Godavari Middle sub-basin, India. Weather Clim Extremes 30. 10.1016/j.wace.2020.100277 McKee TB, Doesken NJ, Kliest J (1993) The relationship of drought frequency and duration to time scales. Proceedings of the 8th Conference on Applied Climatology, American Meteorological Society, Boston, MA, 179–184 Mezὄsi G Blanka V Ladanyi Z Bata T Urdea P Frank A Meyer BC Expected mid-and long-term changes in drought hazard for the South-Eastern Carpathian Basin Carpathian J Earth Environ Sci 2016 11 2 355 366 Mishra AK Singh VP A review of drought concepts, Review paper J Hydrol 2010 391 202 216 10.1016/j.jhydrol.2010.07.012 Morid S Smakhtin V Moghaddasi M Comparison of seven meteorological indices for drought monitoring in Iran Int J Climatol 2006 26 7 971 985 10.1002/joc.1264 Mortazavi M Kuczera G Cui L Multiobjective optimization of urban water resources: Moving toward more practical solutions Water Resour Res 2012 48 W03514 10.1029/2011WR010866 Moyé LA Kapadia AS Cech IM Hardy RJ The theory of runs with applications to drought prediction J Hydrol 1988 103 127 137 10.1016/0022-1694(88)90010-8 NIC (2009) India: The Impact of Climate Change to 2030; A Commissioned Research Report: NIC 2009–03D, Joint Global Change Research Institute and Battelle Memorial Institute, Pacific Northwest Division, National Intelligence Council. Nicholson S Land surface processes and Sahel climate Rev of Geophys 2000 38 117 139 10.1029/1999RG900014 Niranjan K Rajeevan M Pai DS Srivastava AK Preethi B On the observed variability of monsoon droughts over India Weather Clim Extremes 2013 1 42 50 10.1016/J.WACE.2013.07.006 Pálfai I Probability of drought occurrence in Hungary Quarterly J Hungarian Meteorological Service 2002 106 3–4 265 275 Palmer WC Keeping Track of Crop Moisture Conditions, Nationwide: the New Crop Moisture Index Weatherwise 1968 21 4 156 161 10.1080/00431672.1968.9932814 Palmer WC (1965) Meteorological Drought; Research Paper No. 45, US Department of Commerce, Weather Bureau, Washington, DC, 59. Partal T Kahya E Trend analysis in Turkish precipitation data Hydrol Process 2006 20 2011 2026 10.1002/hyp.5993 Patel PM, Saha D, Saha T (2020) Sustainability of groundwater through community-driven distributed recharge: An analysis of arguments for water scarce regions of semi-arid India. J Hydro (Regional Studies) 29:100680 Peel MC Pegram GG Mcmahon TA Global analysis of runs of annual precipitation and runoff equal to or below the median: run length Int J Climatol J Royal Meteorol Soc 2004 24 7 807 822 Pingle G (2011) Irrigation in Telangana: The Rise and Fall of Tanks. Econ Political Wkly 46(26–27) Qaisrani ZN Nuthammachot N Asadullah TK Drought monitoring based on Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index in the arid zone of Balochistan province Pakistan Arab J Geosci 2021 14 11 10.1007/s12517-020-06302-w Razmkhah H Comparing threshold level methods in development of stream flow drought severity-duration-frequency curves Water Resour Manag 2017 31 13 4045 4061 10.1007/s11269-017-1587-8 Sen PK Estimates of the regression coefficient based on Kendall's tau J Am Stat Assoc 1968 63 1379 1389 10.1080/01621459.1968.10480934 Shafer BA, Dezman LE (1982) Development of surface water supply index to assess the severity of drought condition in snowpack runoff areas. Proceedings of the Western Snow Conference, Colorado State University, Fort Collins, 164–175 Smakhtin VU Hughes DA (2004) Automated Estimation and Analyses of Drought Indices in South Asia. Working Paper 83, International Water Management Institute. Uang-aree P Kingpaiboon S Khuanmar K The development of atmospheric crop moisture index for irrigated agriculture Russ Meteorol Hydrol 2017 42 11 731 739 10.3103/S1068373917110073 UNDRR (2021) GAR Special Report on Drought 2021. The UN Global Assessment Report on Disaster Risk Reduction (GAR), United Nations Office for Disaster Risk Reduction. Van Rooy MP A rainfall anomaly index independent of time and space Notos 1965 14 43 6 Vicente-Serrano SM Beguería S López-Moreno JI A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index J Clim 2010 23 1696 1718 10.1175/2009JCLI2909.1 Vicente-Serrano SM Begueria S Martínez MA Angulo M El Kenawy A A new global 0.5 gridded dataset (1901–2006) of a multiscalar drought index: comparison with current drought index datasets based on the Palmer Drought Severity Index J Hydrometeorol 2010 11 1033 1043 10.1175/2010JHM1224.1 Wilhite DA Drought and water crises—Science, technology and management issues 2005 Boca Raton Florida, CRC Press 432 Wilhite DA Glantz MH Understanding the drought phenomenon: the role of definitions Water Int 1985 10 111 120 10.1080/02508068508686328 Wu W Geller MA Dickinson RE The response of soil moisture to long-term variability of precipitation J Hydrometeorol 2002 3 5 604 613 10.1175/1525-7541(2002)003<0604:TROSMT>2.0.CO;2 WWDR (2020) Water and Climate Change. The United Nations World Water Development Report 2020, UN Water. Yaduvanshi A Ranade A Long-term rainfall variability in the eastern Gangetic Plain in relation to global temperature change Atmos Ocean 2017 55 2 94 109 10.1080/07055900.2017.1284041 Yaduvanshi A Kulkarni A Bendapudi R Haldar K Observed changes in extreme rain indices in semiarid and humid regions of Godavari basin, India: risks and opportunities Nat Hazards 2020 103 685 711 10.1007/s11069-020-04006-8 Yevjevich VM Objective approach to definitions and investigations of continental hydrologic droughts, An (Doctoral dissertation 1967 Libraries) Colorado State University Yue S Hashino M Temperature trends in Japan: 1900–1996 Theor Appl Climatol 2003 75 15 27 10.1007/s00704-002-0717-1
PMC009xxxxxx/PMC9005041.txt	==== Front J Radiol Nurs J Radiol Nurs Journal of Radiology Nursing 1546-0843 1555-9912 Association for Radiologic & Imaging Nursing. Published by Elsevier Inc. S1546-0843(22)00044-X 10.1016/j.jradnu.2022.03.003 Just Pediatrics Behind the Mask: Reflections on the Impact in Pediatrics Reilly Lorie MSN, CRNP, CPNP-AC ∗ Sedation/Radiology Nurse Practitioner, The Children’s Hospital of Philadelphia, Philadelphia, PA ∗ Corresponding author: Lorie Reilly, MSN, CRNP, CPNP-AC, Sedation/Radiology Nurse Practitioner, The Children’s Hospital of Philadelphia, 3401 Civic Center Boulevard, Philadelphia, PA 19104. 12 4 2022 6 2022 12 4 2022 41 2 7273 © 2022 Association for Radiologic & Imaging Nursing. Published by Elsevier Inc. All rights reserved. 2022 Association for Radiologic & Imaging Nursing Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcIt is 0630 on a Monday morning in the Sedation Unit, and the first patient, a 5-year-old male, has arrived for a sedated brain MRI. As he walks confidently ahead of his parents past the nursing station, he pauses for a few seconds and stares at the masked nurses and providers who greet him with a lot of OOHs and AAHs and comments about how cute he is while admiring his Paw Patrol mask. The patient immediately runs behind his father’s legs and starts to whimper. Still whimpering, he is then taken to his room, where another masked and goggled staff member takes his height, weight, and vital signs. The patient’s whimpers soon turn into a constant cry, and he becomes uncooperative. Staff attempts to calm him and wonder if the reaction from this patient is appropriate for his age or if he was simply overwhelmed by the masked staff. For the past 2 years, the COVID-19 global pandemic has created many challenges for everyone. But for the children who are either hospitalized or who are in contact with healthcare workers, the impact has been even more significant. These children now have limited access to other children because of visitation policy changes, closed playrooms, and restricted services that are vital for their health and well-being. We all miss seeing each other’s faces! Facial expressions coupled with body language are critical in helping us gauge emotions, as well as evaluating the degrees of health and pain. The adage by Ralph Waldo Emerson that “the eyes indicate the antiquity of the soul” is never more relevant today. Now we must rely on people’s eye movements, body language, and alterations in the pitch and tone of their voices to understand and convey our feelings and intentions. Our ability to engage a child has been impacted by our mask-wearing. Many children become fearful when you walk into a room, while others have become accustomed to it over the past 2 years. They cannot see the smiles on our faces. Our experiences have shown that children overall are very compliant with mask-wearing and look toward their parents for approval when we ask them to remove their masks. Praising children for wearing their masks and commenting about the colors or designs on the mask is so important. Some children even arrive in healthcare settings with “masked” dolls or stuffed animals. Children with hearing loss are even more disadvantaged since they cannot read lips under the mask. “The Communicator” is a face mask with a see-through window opening across the mouth, which makes it easier to communicate with this population (Figure 1 ). It enables patients to see our mouths and thus, further aids in communicating a smile by removing the barrier of a traditional face mask. Children with developmental disabilities faced even more challenges along with their fear and apprehension in medical settings. Our requirement for COVID testing prior to sedation or general anesthesia has been stressful for these children as they had to go through drive-through testing centers or come in early for their procedure to have a rapid COVID testing at the hospital.Figure 1 Communicator mask. Over the past 2 years, we have also become better hand washers, and children even refer to hand sanitizer as “hanitizer.” Staff wearing masks and goggles are all too familiar to our oncology patients, and parents may be fearful of our compliance with the COVID protocols as we go from one patient room to the next. We have all spent so much more time using communications technology, including the hospitalized children, facetiming their friends and family who were unable to visit in person. Visitors were seen outside the hospital waving up to the hospitalized children through their decorated patient room windows and welcoming their friendly faces. Even our furry friends from Pet Therapy were absent along with our hospital volunteers. Hospital events that used to normalize the holiday times were canceled, donations to the hospital declined, and donors had to use Amazon wish lists vs face to face giving. Fewer donations during these times resulted in a certain lack of resources for our Child Life Specialists to share with the children. We have done our best to keep the children, families, and staff safe throughout the pandemic, but it has created new challenges in our relationships. Our attempts to normalize these relationships have resulted in finding alternate ways to communicate. Let us continue on as we do our best to communicate with and support our youngest patients, who have been so resilient, keeping in mind the impact that the pandemic has had on their mental health. Children are social creatures, and they depend on us to keep them safe and at the same time engaged. Conflict of interest: I am the primary author of the manuscript: Behind the Mask: Reflections on the Impact in Pediatrics, and I have no conflict of interest.
PMC009xxxxxx/PMC9005042.txt	==== Front Arch Bronconeumol Arch Bronconeumol Archivos De Bronconeumologia 0300-2896 1579-2129 SEPAR. Published by Elsevier España, S.L.U. S0300-2896(22)00168-5 10.1016/j.arbres.2022.01.021 Letter to the Director [Translated article] Pleuropericardial Effusion and Systemic Inflammatory Syndrome Secondary to the Administration of the mRNA-1273 Vaccine for SARS-CoV-2 Derrame pleuropericárdico y síndrome inflamatorio sistémico secundario a la administración de vacuna ARNm-1273 para coronavirus del síndrome respiratorio agudo grave de tipo 2 (SARS-CoV-2)Boira Ignacio Torba Anastasiya Castelló Carmen Esteban Violeta Vañes Sandra Chiner Eusebi ⁎ Servicio de Neumología, Hospital Universitario San Juan de Alicante, Alicante, Spain ⁎ Corresponding author. 12 4 2022 4 2022 12 4 2022 58 4 T372T373 © 2022 SEPAR. Published by Elsevier España, S.L.U. All rights reserved. 2022 SEPAR Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcTo the Director, SARS-CoV-2 infection has caused more than 5 million deaths worldwide and has had a major impact on the economy and healthcare. Vaccination has changed the management of these patients by preventing transmission and reducing mortality.1 Side effects associated with SARS-CoV-2 mRNA vaccines, some serious, have been reported. One of the major adverse events associated with the Moderna vaccine (mRNA-1273), myocarditis, seen primarily in young adults, occasionally occurs concomitantly with heart disease, while another rarely reported event is multisystem inflammatory syndrome (MIS).2 We report a clinical case of MIS following administration of the Moderna vaccine. To our knowledge, this is the only published case of pleural exudate in this context. Our patient was a 33-year-old woman, with no drug allergies, smoker of 5 pack-years, with no medical or surgical history of interest. She was not receiving any regular treatment, and worked at home as a mother of 2 children with a dog and a cat as pets. She attended the emergency room with a 3-week history of sternal pain that worsened with postural changes. The symptoms had developed 5 days after administration of the second dose of the Moderna vaccine. She also reported moderate dyspnea on exertion, low-grade fever, and arthralgias, with no cough or other symptoms. On physical examination, she was in good general condition, with normal coloring, normal hydration, and normal breathing at rest. She had no supraclavicular lymphadenopathies or nail clubbing. Blood pressure was 101/70 mmHg, heart rate 74 bpm, respiration rate 12 breaths/m, and SpO2 (room air) 98% Cardiac auscultation was normal and pulmonary auscultation showed decreased breath sounds in the lower third of both lungs, with decreased transmission of vocal fremitus, dullness on percussion, and signs of pleural effusion. The rest of the examination was normal. Antigen and polymerase chain reaction (PCR) for COVID-19 were negative. Chest X-ray in the emergency room showed blunting of both costophrenic angles consistent with bilateral pleural effusion (Fig. 1A).Fig. 1 (A) Anterior-posterior chest X-ray at admission showing right pleural effusion and left costophrenic angle blunting. (B) Chest computed axial tomography showing bilateral pleural effusion and pericardial effusion (upper arrow indicates pericardial effusion and the right arrow indicates pleural effusion). The blood count showed 10.1 × 109/l leukocytes with 5.9% eosinophils (600 eosinophils) and hemoglobin 14.2 g/dl. Biochemistry showed total proteins 6.69 g/dl, LDH 257 U/l, CRP 4.03 mg/dl. Coagulation tests showed D-dimer 2656 ng/ml. Pneumococcus and Legionella antigens in urine and parallel respiratory serology for Legionella, Mycoplasma and respiratory viruses were negative. A chest computed tomography angiogram conducted to rule out pulmonary thromboembolism confirmed moderate bilateral pleural effusion with bibasilar and middle lobe atelectasis (Fig. 1B). This was confirmed on chest ultrasound that also ruled out vascular involvement. Echocardiography showed mild pericardial effusion not affecting cavity filling. Abdominal ultrasound was normal. Thoracentesis showed exudate (proteins 4.25 g/dl and LDH 191 U/l) with predominantly polymorphonuclear cells (73.2%), normal ADA (6.7 U/l), cholesterol 75 mg/dl, CEA < 1.7 ng/ml, and negative antinuclear antibodies. Cytology was negative for malignancy, culture showed no microorganisms, and Ziehl-Neelsen and culture in Löwenstein-Jensen media were negative. Azithromycin 500 mg/24 h and methylprednisolone 40 mg/12 h were administered for 6 days with favorable clinical and radiological evolution, and the patient was discharged with prednisone 30 mg in a tapering regimen and azithromycin 500 mg/24 h for a further 4 days, and was referred for monitoring in the outpatient clinic. Follow-up chest CT at 15 days showed a marked decrease in pleural and pericardial effusion. Follow-up clinical laboratory tests at 23 days after discharge showed normal antinuclear antibodies and anti-neutrophil cytoplasm antibodies (P-ANCA and C-ANCA). Serology was positive for SARS-CoV-2 IgG. The Moderna mRNA-1273 vaccine is given in 2 doses and has demonstrated 94.7% efficacy and very good safety. Most of the adverse effects are mild, and while serious events, such as myalgias, arthralgias and asthenia, have been reported, particularly after the second dose, pleural and pericardial effusion are not listed in the Summary of Product Characteristics or reported in the literature.3 Our patient, who had recently received the Moderna vaccine, developed bilateral and pericardial pleural effusion. After ruling out other causes, her clinical picture was consistent with MIS caused by the Moderna vaccine, resulting in bilateral pleural and pericardial exudate. MIS after vaccination (usually the second dose) is exceptional, and very few cases have been published, particularly in children and adolescents. It consists of a systemic clinical syndrome (diarrhea, dyspnea, abdominal pain, skin rash, and hypotension) involving at least 2 organs. Progress is favorable after administration of corticosteroids and, in some cases, intravenous immunoglobulins.4 This entity may have clinical similarities with active SARS-CoV-2 infection, so assessment of the vaccination history is essential when taking the medical history. Most cases are reported at 2–4 weeks after administration and involve gastrointestinal symptoms. The only cases of pleural effusion reported have been associated with heart failure, and none of these patients were studied for pleural fluid. Long-term functional consequences are unknown.5 The SARS-CoV-2 vaccine – generally the Janssen vaccine – has been associated with some rare reports of MIS in the literature, and in no case has pleural fluid been studied to complete the diagnosis. Exudative pleural-pericardial effusion without accompanying myocarditis should be considered among the complications caused by the mRNA-1273 vaccine. ==== Refs References 1 Castells M.C. Phillips E.J. Maintaining safety with SARS-CoV-2 vaccines N Engl J Med 384 2021 643 649 10.1056/NEJMra2035343 33378605 2 Perez Y. Levy E.R. Joshi A.Y. Virk A. Rodriguez-Porcel M. Johnson M. Myocarditis following COVID-19 mRNA vaccine: a case series and incidence rate determination Clin Infect Dis 2021 ciab926 10.1093/cid/ciab926 3 Baden L.R. El Sahly H.M. Essink B. Kotloff K. Frey S. Novak R. Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine N Engl J Med 384 2021 403 416 10.1056/NEJMoa2035389 33378609 4 Uwaydah A.K. Hassan N.M.M. Abu Ghoush M.S. Shahin K.M.M. Adult multisystem inflammatory syndrome in a patient who recovered from COVID-19 postvaccination BMJ Case Rep 14 2021 e242060 10.1136/bcr-2021-242060 5 Tenforde M.W. Morris S.B. Multisystem inflammatory syndrome in adults: coming into focus Chest 159 2021 471 472 10.1016/j.chest.2020.09.097 33285106
PMC009xxxxxx/PMC9005044.txt	==== Front Arch Bronconeumol Arch Bronconeumol Archivos De Bronconeumologia 0300-2896 1579-2129 SEPAR. Published by Elsevier España, S.L.U. S0300-2896(22)00167-3 10.1016/j.arbres.2022.01.020 Letter to the Director [Translated article] Reply to “Absence of Relevant Clinical Effects of SARS-CoV-2 on the Affinity of Hemoglobin for O2 in Patients With COVID-19” Respuesta a «Ausencia de efectos clínicos destacables del SARS-CoV-2 sobre la afinidad de la hemoglobina por el O2 en pacientes con COVID-19»Pascual-Guàrdia Sergi a Ferrer Antoni a Diaz Oscar b Caguana Antonio O. a Tejedor Elvira b Rodríguez-Chiaradia Diego A. a Gea Joaquim a⁎ a Servicio de Neumología, Hospital del Mar – IMIM, Departamento MELIS, Universitat Pompeu Fabra, CIBERES, ISCiii, BRN, Barcelona, Spain b Laboratorio de Referencia de Cataluña, El Prat de Llobregat, Barcelona, Spain ⁎ Corresponding author. 12 4 2022 4 2022 12 4 2022 58 4 T374T375 © 2022 SEPAR. Published by Elsevier España, S.L.U. All rights reserved. 2022 SEPAR Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcTo the Director, We read with interest the scientific letter from Prof. Böning et al.1 commenting on our article on the affinity of hemoglobin for oxygen in patients with COVID-19.2 The aim of our study was to reply from a clinical point of view to the hypothesis generated by two basic research papers that suggested that the virus decreases this affinity by way of different mechanisms.3, 4 We, like previous authors such as Gille et al.5 and Laredo et al.,6 found that the virus appears to have little effect on the interaction of hemoglobin with oxygen. However, these results contrast with those published by other authors, such as Vogel et al.7 It is important for the management of patients that these discrepancies are clarified, since the clinical impact of this affinity affects the real value of PaO2 as an indicator of blood oxygenation. The dispersion of our data, which included hospitalized patients with COVID-19 of heterogeneous severity, prompted Prof. Böning's group of physiologists to suggest that a reduced affinity between hemoglobin and oxygen may be a particular problem in more serious patients. They also suggest that this may be caused by shifts in the oxygen–hemoglobin dissociation curve, due, for example, to changes in the concentration of 2–3 diphosphoglycerate (2.3-BPG). Unfortunately, we did not determine blood levels of this 1.3-diphosphoglycerate isomer, which, together with several other variables that were analyzed in our study (PaCO2, pH, temperature), affects the shape of the above-mentioned curve. We reanalyzed our data, including only the group of patients admitted to the intensive care unit for lung involvement that led to respiratory failure (n = 75, with a total of 343 arterial and 220 venous samples), but this procedure did not reveal any relevant discrepancies between measured and calculated saturation (Fig. 1 ), so we conclude that there does not appear to be a loss of normal affinity of hemoglobin for oxygen. The possible decrease in hemoglobin itself, another parameter that varies considerably among clinical studies,7, 8 could be due to different causes, such as the viral disease itself, nutritional deficiencies, or the proinflammatory state of critical patients. Another cause could be the frequent repetition of laboratory tests in intensive care units,9 a practice that was particularly common for both healthcare and for research reasons in severe COVID-19 patients. However, the rate of anemia in our severe patients was 27.8% (mean hemoglobin 12.1 ± 1.9 g/dl), even lower than in critical patients with other diseases.10 It should be noted here that our hospital was very restrictive with regard to blood draws for research: a single sample was drawn from each patient and competitively allocated among the various studies.Fig. 1 Conflict of interests The authors state that they have no conflict of interests. ==== Refs References 1 Böning D. Bloch W. Kuebler W.M. Sobre «Ausencia de efectos clínicos destacables del SARS-CoV-2 sobre la afinidad de la hemoglobina por el O2 en pacientes con COVID-19» Arch Bronconeumol 2022 10.1016/j.arbres.2021.12.011 2 Pascual-Guàrdia S. Ferrer A. Díaz O. Caguana A.O. Tejedor E. Bellido-Calduch S. Absence of relevant clinical effects of SARS-CoV-2 on the affinity of hemoglobin for O2 in patients with COVID-19 Arch Bronconeumol 57 2021 757 763 3 Liu W. Li H. COVID-19: attacks the 1-beta chain of hemoglobin and captures the porphyrin to inhibit human heme metabolism ChemRxiv 2021 Available from: https://chemrxiv.org/articles/COVID19_Disease_ORF8_and_Surface_Glycoprotein_Inhibit_Heme_Metabolism_by_Binding_to_Porphyrin/11938173 [consulted December 2021] 4 Thomas T. Stefanoni D. Dzieciatkowska M. Issaian A. Nemkov T. Hill R.C. Evidence for structural protein damage and membrane lipid remodeling in red blood cells from COVID-19 patients J Proteome Res 19 2020 4455 4469 33103907 5 Gille T. Sesé L. Aubourg E. Fabre E.E. Cymbalista F. Ratnam K.C. The affinity of hemoglobin for oxygen is not altered during COVID-19 Front Physiol 12 2021 578708 33912067 6 Laredo M. Curis E. Masson-Fron E. Voicu S. Mégarbane B. Does COVID-19 alter the oxyhemoglobin dissociation curve?—An observational cohort study using a mixed-effect modelling Clin Chem Lab Med 59 2021 e416 e419 34144638 7 Vogel D.J. Formenti F. Retter A.J. Vasques F. Camporota L. A left shift in the oxyhaemoglobin dissociation curve in patients with severe coronavirus disease 2019 (COVID-19) Br J Haematol 191 2020 390 393 33037620 8 Guan W.J. Ni Z.Y. Hu Y. Liang W.H. Ou C.Q. He J.X. Clinical characteristics of Coronavirus disease 2019 in China N Engl J Med 382 2020 1708 1720 32109013 9 O’Malley P. Hidden anemias in the critically ill Crit Care Nurs Clin North Am 29 2017 363 368 28778295 10 Retter A. Wyncoll D. Pearse R. Carson D. McKechnie S. Stanworth S. Guidelines on the management of anaemia and red cell transfusion in adult critically ill patients Br J Haematol 160 2013 445 464 23278459
PMC009xxxxxx/PMC9005085.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 35200195 00023 10.1097/CCM.0000000000005504 3 Editorials Do Tracheostomy and Gastrostomy Confer Immortality in COVID-19?* Layba Cathline MD, FACS 1 Wallace David J. MD, MPH 2 1 Division of Trauma and Acute Care Surgery, Department of Surgery, Wright State University, Fairborn, OH 2 Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA 11 2 2022 5 2022 11 2 2022 50 5 891893 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. acute respiratory distress syndrome COVID-19 gastrostomy tracheostomy ==== Body pmcThe elective tracheostomy procedure is one of the oldest surgical interventions, with an early description attributed to the Greek physician Asclepiades (1). Although Asclepiades is principally known for his emphasis on the relationships between nutrition, exercise, light, and hydrotherapy with health, he also performed a version of today’s elective tracheostomy procedure for the treatment of cynanche, a group of afflictions involving the floor of the mouth and the throat (1). Today the indications for tracheostomy have expanded beyond diseases of the upper airway, with over 100,000 percutaneous tracheostomy procedures performed annually in the setting of acute respiratory failure alone in the United States (2). Importantly, percutaneous tracheostomy performed for patients with acute respiratory distress syndrome is considered an elective procedure and should be reserved for patients who are expected to survive their acute illness (3). Durable access to the stomach for the purpose of nutrition and medication administration using a percutaneous approach is a newer procedure, first described in a collaboration between a gastroenterologist and a pediatric surgeon in the late 1970s (4). Despite its more recent arrival to medical care, the number of percutaneous endoscopically placed gastrostomy tubes has grown to more frequently than 100,000 procedures per year in the United States—making it one of the most performed invasive surgeries in patients with critical illness. Like tracheostomy, percutaneous endoscopic gastrostomy is considered an elective procedure and is reserved for patients who are expected to survive their acute illness. In this issue of Critical Care Medicine, Kiser et al (5) present an analysis of 30- and 90-day outcomes from a retrospective observational cohort study of patients with COVID-19 pneumonia who received both percutaneous tracheostomy and gastrostomy tube placement. The team collected data from four hospitals within a single healthcare system using a clinical medical record registry, including patients who received both procedures between February 2020 and August 2020. They reported short-term outcomes and sought inform clinical decision-making related to both procedures for patients with COVID-19. The overall outcomes in this cohort are encouraging. The results tell us that among selected patients referred for both procedures in a four-hospital network in Boston, through 6 early months in the COVID-19 pandemic, 90-day mortality was very low and many patients even returned to home. The results are consistent with other recent studies of tracheostomy in patients with COVID-19, including a similarly sized and timed cohort from New York City, where mortality was 7.5% (6), and the interim results of the larger National Health Service COVIDTrach cohort, which reported a 12% mortality during the first follow-up period (7). Based on these findings, it would seem that combined tracheostomy and gastrostomy can be safe in patients with COVID-19. Applying the results of the study by Kiser et al (5) to clinical decision-making, however, is challenging. While we know that a decision was made by the treating team to perform tracheostomy and gastrostomy, there is insufficient information reported on how that decision was made and what factors went into making it. Complicating matters further, there is evidence that tracheostomy timing varies substantially across hospitals in the United States, a finding that was present even before the COVID-19 pandemic (8). Additionally, while several publications have reported encouraging outcomes for COVID-19 patients following tracheostomy, others are less optimistic, including one study reporting 30-day mortality of nearly 30% in a combined cohort of patients who underwent either percutaneous or open surgical tracheostomy (9) and another reporting 31% mortality at 1 year (10). Without knowing more about how timing and severity of illness factored into the decision to proceed with tracheostomy and gastrostomy in the current study by Kiser et al (5), it is hard to say more than that the clinicians appeared to have made good choices in retrospect. How are we to reconcile varying reported outcomes after tracheostomy and gastrostomy in COVID-19? Likely culprits are that, across studies, the patients are different, the treating teams are different, or some combination of these factors is true. Additionally, given that tracheostomy and gastrostomy are elective procedures performed at varying decision thresholds, intended for patients with anticipated survival, and performed at the very earliest several days after starting mechanical ventilation, observational studies of outcomes for patients who receive these procedures can be impacted by a form of selection bias known as survivorship or immortal time bias. Classically, immortal time bias occurs in an observational study when the exposure definition includes time when the outcome of interest is not possible. In the case of studies evaluating outcomes following of tracheostomy and gastrostomy, it would be an error to include time prior the procedure in a survival outcome, as patients in the exposure group survived this period by virtue of the fact that they had the procedures performed. On the other hand, any patient who died prior to consideration of tracheostomy and gastrostomy tube would be automatically included in the nonexposed group, resulting in additional bias in favor of the procedures. To their credit, the authors of the current study by Kiser et al (5) did not make this analytic error, but immortality bias is evident in other recent studies evaluating the impact of these procedures on outcomes in the setting of COVID-19. Interpretation of clinical outcomes following tracheostomy and gastrostomy is also complicated by the potential for reverse causality. Reverse causality refers to the situation when the directional association between an exposure and outcome is reversed—the outcome leads to the exposure. In the cases of tracheostomy and gastrostomy, the elective procedures should generally be performed in patients are considered to have promising survival chances (i.e., the anticipated outcome leads to the exposure). Although optimal timing has not been established, there is also the hope that these procedures improve the quality of care for patients with prolonged respiratory failure (i.e., the exposure leads to an improved outcome). On the other hand, poorly timed tracheostomy and gastrostomy could negatively impact quality of care, with examples including procedural complications resulting from very early tracheostomy in patients with little physiologic reserve, tracheostomy dislodgment in a patient who is receiving prone ventilation, or complications arising from prolonged endotracheal intubation (11). Untangling the magnitude and directionality of these relationships is a challenge, and yet it is essential when the outcomes experienced in one setting are hoped to be deployed in another. A preventable tragedy of COVID-19 would be to “not” learn while doing. There has been incredible innovation in trial design, data sharing, and models of research collaboration since very early in the pandemic—and this must continue. We applaud the authors collecting, analyzing, and sharing their institutional results, as this is the only way that we will collectively address critical knowledge gaps and ultimately improve the care of our patients. It is fitting that over 2,000 years ago, the originator of the elective tracheostomy advocated for medical care that was “cito tuto jucunde”—quick, safe, and joyful. While the optimal timing of tracheostomy and gastrostomy placement in COVID-19 remains to be determined, it is encouraging that there is growing evidence that the procedures can be performed safely in some patients and that many patients even return to their lives. *See also p. 819. The authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Borman J Davidson JT : A history of tracheostomy: Si spiritum ducit vivit (Cicero). Br J Anaesth. 1963; 35 :388–390 14013998 2. Mehta AB Walkey AJ Curran-Everett D : One-year outcomes following tracheostomy for acute respiratory failure. Crit Care Med. 2019; 47 :1572–1581 31397716 3. Hosokawa K Nishimura M Egi M : Timing of tracheotomy in ICU patients: A systematic review of randomized controlled trials. Crit Care. 2015; 19 :424 26635016 4. Ponsky JL : The development of PEG: How it was. J Interv Gastroenterol. 2011; 1 :88–89 21776432 5. Kiser SB Sciacca K Jail N : A Retrospective Observational Study Exploring 30- and 90-Day Outcomes for Patients With COVID-19 After Percutaneous Tracheostomy and Gastrostomy Placement. Crit Care Med. 2022; 50 :819–824 6. Chao TN Harbison SP Braslow BM : Outcomes after tracheostomy in COVID-19 patients. Ann Surg. 2020; 272 :e181–e186 32541213 7. COVIDTrach Collaborative: COVIDTrach; the outcomes of mechanically ventilated COVID-19 patients undergoing tracheostomy in the UK: Interim report. Br J Surg. 2020; 107 :e583–e584 32940347 8. Mehta AB Cooke CR Wiener RS : Hospital variation in early tracheostomy in the United States: A population-based study. Crit Care Med. 2016; 44 :1506–1514 27031382 9. Botti C Lusetti F Neri T : Comparison of percutaneous dilatational tracheotomy versus open surgical technique in severe COVID-19: Complication rates, relative risks and benefits. Auris Nasus Larynx. 2021; 48 :511–517 33143935 10. Vallejo-Díez J Peral-Cagigal B García-Sierra C : Percutaneous tracheostomy in COVID patients. Experience in our hospital center after one year of pandemic and review of the literature. Med Oral Patol Oral Cir Bucal. 2022; 27 :e18–e24 34415000 11. Stauffer JL Olson DE Petty TL : Complications and consequences of endotracheal intubation and tracheotomy. A prospective study of 150 critically ill adult patients. Am J Med. 1981; 70 :65–76 7457492
PMC009xxxxxx/PMC9005086.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 35120038 00016 10.1097/CCM.0000000000005486 3 Editorials Prone Position and COVID-19: Mechanisms and Effects* Gattinoni Luciano MD, FRCP 1 Camporota Luigi MD, PhD 2 Marini John J. MD 3 1 Department of Anesthesiology, University Medical Center Göttingen, Göttingen, Germany 2 Department of Adult Critical Care, Guy’s and St Thomas’ NHS Foundation Trust, Health Centre for Human and Applied Physiological Sciences, London, United Kingdom 3 Department of Pulmonary and Critical Care Medicine, University of Minnesota and Regions Hospital, St. Paul, MN 07 2 2022 5 2022 07 2 2022 50 5 873875 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. coronavirus disease 2019 mechanical ventilation lung recruitment prone position ==== Body pmcProne positioning was first applied in critically ill patients by Piehl and Brown (1) in 1976 who reported a marked oxygenation improvement in five patients with acute respiratory failure. This observation remained essentially a curiosity until the first CT scan of acute respiratory distress syndrome (ARDS) patients showed that parenchymal densities were disproportionately distributed in the dorsal lung regions. This finding provided the anatomical hypothesis for the improvement in gas exchange and led to an increase in the clinical use of prone position (2). The assumption was that the normally ventilated lung located in the “ventral” regions (i.e., the anatomical concept of “baby lung”) would be better perfused if placed in a gravitationally dependent position, improving the overall ventilation/perfusion (V/Q) ratio, and hence gas exchange. However, further studies of CT scans taken with patients in the prone position disproved the mechanistic hypothesis that change in perfusion determined its effects on oxygenation (3). On the contrary, it became apparent that the main effect following prone position was the redistribution of lung densities from the dorsal to the ventral lung regions. This discovery led to the formulation of the “sponge” model of the lung in which the rapid resolution and formation of atelectasis in different regions of the heavily edematous ARDS lung are primarily due to changes of superimposed hydrostatic pressure which follow the change in gravitational axis (4). The improvement of oxygenation, consistently found both clinically and experimentally, prompted the parallel implementation of clinical trials. Over the ensuing years, diverse studies of prone positioning led on one hand to better understanding of the mechanisms that improve oxygenation; on the other hand, observational and clinical trials that progressively refined the indications for its use in ARDS. Significant survival benefits were demonstrated when prone position was applied for 12–16 consecutive hours in the more severe forms of respiratory failure (Pao2/Fio2 < 150 mm Hg) (5), whereas its application does not provide convincing survival advantage in patients with milder disease (6). Therefore, what we understood regarding “typical” ARDS might be summarized as follows: prone positioning in ARDS, a condition characterized by extensive inflammatory edema, leads to decreased frontal chest wall compliance, to partial clearing of dorsal atelectasis, and to the development of new ventral atelectasis. In most patients, the balance favors clearing of dorsal atelectasis, thus increasing the net amount of well-aerated tissue. Regarding perfusion, evidence from both experimental models and clinical studies indicated little change with the shift from supine to prone position (7, 8). As we learned in other landmark trials, however, better oxygen exchange does not satisfactorily explain the survival advantage attributable to prone positioning. The improved survival is more likely to result from the greater homogeneity in alveolar dimensions, the reduction in the maximal tissue stretch, and a more uniform distribution of stress and strain throughout the lung parenchyma associated with prone position (9). This possibility is supported by data from human and experimental animals, which show less variation in the size of the individual pulmonary units along the vertical axis in the prone as opposed to in supine position, due to a better anatomical matching of lung and chest wall shapes and compliances (9). During the COVID-19 pandemic, the use of prone position increased exponentially to reverse hypoxemia not only in patients receiving mechanical ventilation but also in awake patients who breathe spontaneously or receiving noninvasive ventilation (10). A series of epidemiological studies has confirmed that oxygenation improves with prone positioning in 60–80% of COVID-19 patients (11), but little physiologic data are available to understand the relationship between improvement in gas exchange and patient outcome. Indeed, we should understand why some patients do not respond to prone positioning and determine whether this technique is just a gas exchange cosmetic or, conversely, a driver of improved clinical outcomes. Several mechanisms may account for improved oxygenation: 1) global alveolar recruitment, 2) increase of the nondependent lung mass in prone position (60% vs 40% of the supine position), and 3) decrease of the total chest wall compliance due to the functional stiffening of the ventral chest wall. These mechanisms may be discussed in the light of the new information provided by Fossali et al (12), published in this issue of Critical Care Medicine. Global Alveolar Recruitment In “typical” ARDS, a net recruitment, with unmodified perfusion, is the widely accepted mechanism for improved oxygenation (13). However, in this series of COVID-19 patients, studied on the first week after intubation, the recruitment was relatively modest (6% vs a median of 16–20% in “typical” ARDS) and, more relevant, was completely dissociated from the improvement of gas exchange. These data undercut the importance of recruitment in the early stages of the disease, particularly as prone position has been found to improve oxygenation even in early stages of COVID-19 pneumonia, when atelectasis and potential for lung recruitment are negligible compared with “typical” ARDS with similar Berlin-defined severity. This strongly implies that the mechanisms of hypoxemia and the responses to positive end-expiratory pressure also differ in these COVID-19 patients. Despite the relatively low global recruitment and regardless of the gas exchange, it must be noted that in the study by Fossali et al (12), the dorsal aeration did improve lung homogeneity and may have reduced the stimulus for atelectotrauma—as evidenced by electrical impedance tomography. Increase of the Nondependent Lung Mass In fully supine (0°) position, approximately 60% of the total lung mass is dependent, that is, in the lower 50% of the sternovertebral axis. In COVID-19 pneumonia, the lack of regulatory control of perfusion promotes greater perfusion of these dependent regions, leading to a decrease in the V/Q ratio. As the final arterial Po2 depends on the weighted mean of the Po2 of blood flowing from diverse pulmonary units, the greater number of atelectatic units in the dependent lung regions, the more severe will be the hypoxemia. In contrast, in prone position, only 40% of the tissue mass is in the dependent position, that is, fewer pulmonary units will be hyperperfused, resulting in better oxygenation. In summary, the distribution of the lung mass and the regional perfusion may play a significant role in determining oxygenation. Decrease of the Total Chest Wall Compliance The normal mechanical response to prone position is a decrease in the total chest wall compliance due to the functional stiffening of the anterior chest wall. Consequently, regional ventilation becomes less unevenly allocated, resulting in more homogeneous V/Q distribution. Indeed, Fossali et al (12) showed a reduction in dead space in the ventral areas and decrease in shunt in the dorsal areas. The lack of decrease of total respiratory system compliance during prone in position, as observed in the study by Fossali et al (12), suggests that the overall lung compliance improved. It must be noted that the same effect may be achieved by compressing the anterior chest-wall in supine position (i.e., making the anterior chest wall as stiff as the dorsal one), a maneuver which may result in “paradoxical” improvements of gas exchange (13). In summary, the improvement (or the lack of improvement) in oxygenation in COVID-19 patients depends on the interplay among the mechanisms described. At this stage of the disease (within 1 wk from intubation), the recruitment is one of the possibilities, but likely, not the most important, as strongly suggested by the lack of correlation between recruitment and oxygenation, a finding similar to what recently found with the CT scan in COVID-19 patients when studied at the same stage of the disease (14, 15). Although the cure of COVID-19 patients in ICU is often confused with a cure of Pao2, we must remember that increased survival, when related to mechanical ventilation, may not be directly attributable to the Pao2/Fio2 ratio but instead with a better distribution of alveolar stress and strain. Indeed, the possible decrease of mortality due to a supportive therapy such as prone position is solely due to a decreased harm of mechanical ventilation. In early stages of COVID-19 disease where the compliance is usually high and the lungs are full of gas, the hypoxemia is dictated by alteration in perfusion. It is difficult to imagine that under such circumstances, mechanical ventilation produces dangerous levels of stress and strain. In this condition, prone positioning is not strictly necessary. However, COVID-19 pneumonia is an evolving disease. As shown in this article by Fossali et al (12) lung weight may increase with passing time causing more atelectasis to develop. At this more advanced stage, prone position may find its place. Finally, in late stages of the disease, the likelihood for oxygenation to improve with prone positioning becomes extremely low. We recently found that this phenomenon may be due to the progression of lung consolidation toward organizing fibrotic pneumonia (14). We thank Fossali et al (12), who contributed to a better understating of COVID-19 disease by performing an impressive physiologic study under extremely difficult pandemic conditions. When all the data are taken into consideration, we believe that prone position lessens the damage delivered by the mechanical ventilator, regardless of gas exchange. Therefore, given the safety of the procedure and if staffing is available, prone position should be a strong consideration in the care of critically ill ventilated patients. *See also p. 723. Supported, in part, by departmental sources and two grants for research in respiratory medicine, proportioned by Sartorius AG (Otto-Brenner-Straße 20, 37079, Göttingen, Germany). Dr. Gattinoni received funding from SIDAM, General Electrics, Mindray, ESTOR, Masimo, Medtronic, Grifols, and GSK. The remaining authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Piehl MA Brown RS : Use of extreme position changes in acute respiratory failure. Crit Care Med. 1976; 4 :13–14 1253612 2. Langer M Mascheroni D Marcolin R : The prone position in ARDS patients. A clinical study. Chest. 1988; 94 :103–107 3383620 3. Gattinoni L Pelosi P Vitale G : Body position changes redistribute lung computed-tomographic density in patients with acute respiratory failure. Anesthesiology. 1991; 74 :15–23 1986640 4. Bone RC : The ARDS lung. New insights from computed tomography. JAMA. 1993; 269 :2134–2135 8468771 5. Guérin C Reignier J Richard JC ; PROSEVA Study Group: Prone positioning in severe acute respiratory distress syndrome. N Engl J Med. 2013; 368 :2159–2168 23688302 6. Gattinoni L Carlesso E Taccone P : Prone positioning improves survival in severe ARDS: A pathophysiologic review and individual patient meta-analysis. Minerva Anestesiol. 2010; 76 :448–454 20473258 7. Nyrén S Mure M Jacobsson H : Pulmonary perfusion is more uniform in the prone than in the supine position: Scintigraphy in healthy humans. J Appl Physiol (1985). 1999; 86 :1135–1141 10194194 8. Lamm WJ Graham MM Albert RK : Mechanism by which the prone position improves oxygenation in acute lung injury. Am J Respir Crit Care Med. 1994; 150 :184–193 8025748 9. Gattinoni L Taccone P Carlesso E : Prone position in acute respiratory distress syndrome. Rationale, indications, and limits. Am J Respir Crit Care Med. 2013; 188 :1286–1293 24134414 10. Coppo A Bellani G Winterton D : Feasibility and physiological effects of prone positioning in non-intubated patients with acute respiratory failure due to COVID-19 (PRON-COVID): A prospective cohort study. Lancet Respir Med. 2020; 8 :765–774 32569585 11. Camporota L Sanderson B Chiumello D : Prone position in coronavirus disease 2019 and noncoronavirus disease 2019 acute respiratory distress syndrome: An international multicenter observational comparative study. Crit Care Med. 2021 12. Fossali T Pavlovsky B Ottolina D : Effects of Prone Position on Lung Recruitment and Ventilation-Perfusion Matching in Patients With COVID-19 Acute Respiratory Distress Syndrome: A Combined CT Scan/Electrical Impedance Tomography Study. Crit Care Med. 2022; 50 :723–732 13. Marini JJ Gattinoni L : Improving lung compliance by external compression of the chest wall. Crit Care. 2021; 25 :264 34321060 14. Rossi S Palumbo MM Sverzellati N : Mechanisms of oxygenation responses to proning and recruitment in COVID-19 pneumonia. Intensive Care Med. 2022; 48:56-66 15. Lang M Som A Mendoza DP : Hypoxaemia related to COVID-19: Vascular and perfusion abnormalities on dual-energy CT. Lancet Infect Dis. 2020; 20 :1365–1366
PMC009xxxxxx/PMC9005089.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 34605777 00008 10.1097/CCM.0000000000005374 3 Clinical Investigations Coronavirus Disease 2019 and Out-of-Hospital Cardiac Arrest: No Survivors* Baert Valentine PhD 12 Beuscart Jean-Baptiste MD, PhD 1 Recher Morgan MD 1 Javaudin François MD 3 Hugenschmitt Delphine BScN 4 Bony Thomas MD 4 Revaux François MD 5 Mansouri Nadia MD 5 Larcher Fanny MD 6 Chazard Emmanuel PhD 1 Hubert Hervé PhD 12 the French National OHCA Registry (RéAC) Study Group 1 University of Lille, CHU Lille, ULR 2694 - METRICS: Évaluation des technologies de santé et des pratiques médicales, F-59000, Lille, France. 2 French National Out-of-Hospital Cardiac Arrest Registry, Lille, France. 3 Department of Emergency Medicine, Nantes University Medical Center and University of Nantes, Microbiotas Hosts Antibiotics and bacterial Resistances (MiHAR), University of Nantes, Nantes, France. 4 SAMU 69, Edouard Herriot Hospital, Lyon University Medical Center, Lyon, France. 5 Assistance Publique - Hôpitaux de Paris (AP-HP), Henri Mondor University Hospital, SAMU 94, Créteil, France. 6 SAMU 59, Dron Hospital, Tourcoing, France. For information regarding this article, E-mail: valentine.baert@univ-lille.fr 04 10 2021 5 2022 04 10 2021 50 5 791798 Copyright © 2021 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2021 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. OBJECTIVES: To describe and compare survival among patients with out-of-hospital cardiac arrest as a function of their status for coronavirus disease 2019. DESIGN: We performed an observational study of out-of-hospital cardiac arrest patients between March 2020 and December 2020. Coronavirus disease 2019 status (confirmed, suspected, or negative) was defined according to the World Health Organization’s criteria. SETTING: Information on the patients and their care was extracted from the French national out-of-hospital cardiac arrest registry. The French prehospital emergency medical system has two tiers: the fire department intervenes rapidly to provide basic life support, and mobile ICUs provide advanced life support. The study data (including each patient’s coronavirus disease 2019 status) were collected by 95 mobile ICUs throughout France. PATIENTS: We included 6,624 out-of-hospital cardiac arrest patients: 127 cases with confirmed coronavirus disease 2019, 473 with suspected coronavirus disease 2019, and 6,024 negative for coronavirus disease 2019. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The “confirmed” and “suspected” groups of coronavirus disease 2019 patients had similar characteristics and were more likely to have suffered an out-of-hospital cardiac arrest with a respiratory cause (confirmed: 53.7%, suspected coronavirus disease 2019: 56.5%; p = 0.472) than noncoronavirus disease 2019 patients (14.0%; p < 0.001 vs confirmed coronavirus disease 2019 patients). Advanced life support was initiated for 57.5% of the confirmed coronavirus disease 2019 patients, compared with 64.5% of the suspected coronavirus disease 2019 patients (p = 0.149) and 70.6% of the noncoronavirus disease 2019 ones (p = 0.002). The survival rate at 30-day postout-of-hospital cardiac arrest was 0% in the confirmed coronavirus disease 2019 group, 0.9% in the suspected coronavirus disease 2019 group (p = 0.583 vs confirmed), and 3.5% (p = 0.023) in the noncoronavirus disease 2019 group. CONCLUSIONS: Our results highlighted a zero survival rate in out-of-hospital cardiac arrest patients with confirmed coronavirus disease 2019. This finding raises important questions with regard to the futility of resuscitation for coronavirus disease 2019 patients and the management of the associated risks. coronavirus disease 2019 mobile intensive care units out-of-hospital cardiac arrest registry resuscitation zero survival ==== Body pmcThe past year has resulted in the pandemic (COVID-19) having both a direct and an indirect impact on health outcomes. These include morbidity, increase in severity of chronic diseases, increase in psychiatric illnesses (1, 2), and an increase in out-of-hospital arrests (OHCAs) (3, 4). We have previously reported that some post-OHCA deaths are directly related to COVID-19 (5). The survival rate 30 days after an OHCA is typically very low—10.3% in Europe, for example (6, 7). We have previously reported that the survival rate in France fell during the COVID-19 pandemic (from 5.3% to 3.1%) (5). To date, published studies of COVID-19 and OHCA pooled confirmed and suspected cases of COVID-19 or compared prepandemic and per-pandemic cohorts (5, 8). Furthermore, none of the studies described the survival rate and other characteristics in a specific cohort of OHCA patients with confirmed COVID-19. In this context, our primary objective of the present study was to describe the survival rate 30 days after OHCA among confirmed COVID-19 patients. Our secondary objective was to compare the confirmed COVID-19 patients with suspected COVID-19 patients and non-COVID-19 patients having experienced OHCA during the same period. MATERIALS AND METHODS Study Setting Data were extracted from the French National OHCA Registry (RéAC). The RéAC records OHCAs managed by mobile ICUs (MICUs) in France and has been described elsewhere (9). The French prehospital emergency medical system has two tiers: the fire department acts as the first professional responder and intervenes rapidly to provide basic life support (BLS), whereas MICUs provide advanced life support (ALS). An MICU comprises at minimum an ambulance driver, a nurse, and a senior emergency medicine physician. The RéAC data entry form meets the requirements of the French emergency medical services and complies with the Utstein Resuscitation Registry’s template (10). Study Population and Data We analyzed cases of OHCA recorded in the RéAC between March 1, 2020, and December 31, 2020. Data were gathered by 95 centers in France. The investigating physicians filled out the patients’ COVID-19 status in the RéAC database. We excluded patients with prolonged downtime and unwitnessed arrest with signs of rigor mortis and those whose COVID-19 status was not known. We separated the OHCA study population into three groups (confirmed COVID-19, suspected COVID-19, and non-COVID-19), according to the World Health Organization (WHO)’s definition. Hence, confirmed cases were defined as patients with laboratory-confirmed COVID-19 (after inhospital or outpatient screening) and who were allowed to return home or remain at home (because of nonseverity). Suspected cases in our study were defined as patients who: 1) had consulted a family physician before the OHCA, 2) were suspect cases according to the WHO definitions A, B, or C, and 3) did not have a laboratory confirmation of COVID-19 (11). Statistical Analysis The normality of the data distribution for categorical variables was assessed using the Kolmogorov-Smirnov test. Quantitative variables were described as the median and first and third quartiles (Q1–Q3). Qualitative variables were described as the frequency (percentage), and 95% CIs were computed. Bivariate analyses were assessed using Pearson chi-square test or Fisher exact test for categorical variables and the nonparametric Mann-Whitney test for continuous variables. All tests were two-sided, and the threshold for statistical significance was set to p < 0.05. Ethics The study was approved by the French Advisory Committee on Information Processing in Material Research in the Field of Health (“Comité Consultatif sur le Traitement de l’Information en Matière de Recherche dans le Domaine de la Santé”) and registered with the French National Data Protection Commission (“Commission Nationale de l’Informatique et des Libertés”: reference number: 910946). RESULTS Population Between March 1, 2020, and December 31, 2020, the participating MICUs registered 9,255 patients in the RéAC registry. A total of 6,624 of these patients were included in our study (Fig. 1). There were 127 (1.9%) confirmed cases of COVID-19, 473 (7.1%) suspected cases, and 6,024 (90.9%) non-COVID-19 patients. Figure 1. Study flowchart. COVID-19 = coronavirus disease 2019, MICU = mobile ICU, OHCA = out-of-hospital cardiac arrest, RéAC = French National Out-of-Hospital Cardiac Arrest registry. Intergroup Comparisons The confirmed and suspected COVID-19 groups did not differ significantly with regard to sex, age, location of the OHCA, medical history, OHCA etiology, receipt of BLS, receipt of ALS (if a European Resuscitation Council ALS algorithm was implemented), the time between the call to emergency services and the arrival of the first professional responder, “no-flow” status, and “low-flow” status (Table 1). In the confirmed COVID-19 group, as soon as an ALS was implemented, intubation was performed. No difference was observed between the confirmed and suspected COVID-19 patients regarding intubation (p = 0.085). The median (Q1–Q3) time between the emergency call and the MICU’s arrival was shorter in the confirmed COVID-19 group (19 min [12–25 min]) than that in the suspected COVID-19 group (20 min [14–30 min]; p = 0.026). TABLE 1. Comparison of the Confirmed Coronavirus Disease 2019, Suspected Coronavirus Disease 2019, and Noncoronavirus Disease 2019 Groups Variables COVID-19 Status p (vs Confirmed COVID-19) Confirmed (n = 127) Suspected (n = 473) Non-COVID-19 (n = 6,024) Suspected Non-COVID-19 Sex (% men) 76 (59.8) 292 (61.7) 4,092 (67.9) 0.758 0.153 Age 70 (60–84) 71 (59–82) 68 (55–80) 0.541 0.024 Location of OHCA Home 102 (86.5) 418 (89.9) 4,372 (77.9) 0.059 0.006 Public place 5 (4.2) 28 (6.0) 820 (14.7) Other location 11 (9.3) 19 (4.1) 417 (7.4) Medical history Cardiovascular disease 54 (42.5) 214 (45.2) 2,549 (42.3) 0.616 0.999 Respiratory disease 26 (20.5) 107 (22.6) 738 (12.3) 0.633 0.009 Diabetes 21 (16.5) 84 (17.8) 786 (13.0) 0.794 0.234 Other disease 39 (30.7) 146 (30.9) 1,705 (28.3) 0.999 0.551 No disease 6 (4.7) 22 (4.7) 451 (7.5) 0.999 0.304 Etiology of the OHCA < 0.001 Medical 121 (95.3) 463 (97.9) 4,918 (81.6) 0.103 If medical, % respiratory 65 (53.7) 266 (56.5) 690 (14.0) 0.472 < 0.001 Traumatic 0 (0.0) 2 (0.4) 479 (8.0) Other 6 (4.7) 8 (1.7) 627 (10.4) BLS Witness to the patient’s collapse 84 (66.1) 304 (64.3) 3,656 (60.7) 0.754 0.233 BLS by the witness: 0.349 0.192 CC only 44 (34.6) 142 (30.0) 2,233 (37.1) CC + ventilation 17 (13.4) 85 (18.0) 906 (13.5) No BLS 66 (52.0) 246 (52.0) 2,975 (49.4) BLS by the first responder 104 (81.9) 401 (84.8) 5,126 (85.1) 0.415 0.315 Automated external defibrillator used before the MICU’s arrival 10 (7.9) 48 (10.1) 1,031 (17.1) 0.503 0.004 ALS First recorded rhythm 0.947 0.286 Asystole 109 (85.8) 409 (86.5) 4,876 (80.9) Pulseless electrical activity 9 (7.1) 33 (7.0) 419 (7.0) Ventricular fibrillation/ventricular tachycardia 4 (3.1) 17 (3.6) 461 (7.7) Return of spontaneous circulation due to BLS 5 (3.9) 14 (3.0) 268 (4.4) ALS initiated 73 (57.5) 305 (64.5) 4,255 (70.6) 0.149 0.002 Intubation 73 (57.5) 284 (60.0) 3,900 (64.7) 0.085 0.002 Times T0—first responder’s arrival 10 (5–13) 10 (5–16) 10 (6–15) 0.114 0.113 T0—MICU’s arrival 19 (12–25) 20 (14–30) 19 (13–27) 0.026 0.282 No flow 10 (1–19) 12 (3–20) 11 (2–19) 0.305 0.833 Low flow 27 (15–40) 12 (3–20) 28 (15–40) 0.409 0.635 ALS = advanced life support, BLS = basic life support, CC = chest compression, MICU = mobile ICU, T0 = time of the call to the emergency services. Data are quoted as the frequency (percentage) for qualitative variables or the median (first quartile–third quartile) for quantitative variables. The confirmed COVID-19 patients and the non-COVID-19 patients did not differ significantly with regard to sex, diabetes, a history of cardiovascular disease, a history of another disease, and the provision of BLS (except for defibrillator use, which was less frequent in the confirmed COVID-19 group: 7.9%, versus 17.1% in the non-COVID-19 group; p = 0.009). No differences were observed with regard to the first cardiac rhythm recorded by the MICU or other timings. The confirmed COVID-19 patients were more likely to have a history of respiratory disease (20.5% versus 12.3% in the non-COVID-19 group; p = 0.009), and the OHCA was more likely to have a medical cause (cardiac, neurologic, respiratory, or other medical cause) (95.3% vs 81.6%; p < 0.001). More than half the OHCA with a medical cause in the confirmed COVID-19 group were due to respiratory disease (53.7%, vs 14.0% in the non-COVID-19 group; p < 0.001). ALS provision by the MICU was less frequent (57.5%, vs 70.6% in the non-COVID-19 group; p = 0.002). The intubation was also less frequent in the confirmed COVID-19 group compared with the non-COVID-19 one (57.5% vs 64.7%; p = 0.002). Survival The D30 survival rate (95% CI) in the confirmed COVID-19 group was 0.00% (0.00–2.93), which was significantly lower than that in the non-COVID-19 group (3.5%; 95% CI [3.10–4.06]; p = 0.023) and lower (albeit not significantly) than in the suspected COVID-19 group (0.9%; 95% CI [0.34–2.20]; p = 0.583). There were no significant differences between the confirmed and suspected COVID-19 patients in terms of return of spontaneous circulation (ROSC: 17.3% vs 16.1%, respectively; p = 0.787) and survival at hospital admission (D0 survival: 11.8% vs 11.0%; p = 0.753) (Fig. 2). Likewise, there were no significant differences between the confirmed COVID-19 patients and non-COVID-19 patients with regard to ROSC (17.3% vs 19.8%, respectively; p = 0.573) and D0 survival (11.8% vs 16.6, respectively; p = 0.183) (Fig. 2). Figure 2. Survival. D0 = at hospital admission, D30 = 30 d after out-of-hospital cardiac arrest or at hospital discharge, ROSC = return of spontaneous circulation. DISCUSSION To the best of our knowledge, the present study is the first to have specifically described patients with confirmed COVID-19 (according to the WHO definition [11]) having experienced OHCA. The study’s main finding was that none of the OHCA patients with confirmed COVID-19 were alive 30 days after the event. In our study, we observed a difference in survival rates between the confirmed COVID-19 patients and non-COVID ones. Even though the post-OHCA survival rate has fallen markedly during the COVID-19 pandemic, survivors were always observed. It is extremely rare to observe a survival rate of zero (95% CI, 0.00–2.93) in a specific cohort of OHCA patients. However, researchers working in the state of Georgia (United States) did not observe any survivors among a cohort of 63 patients affected by the coronavirus who experienced inhospital cardiac arrest (survival rate [95% CI], 0.00% [0.00–5.69]) (12). This finding raises questions about the futility of resuscitation for confirmed COVID-19 patients. In our study, we noticed that a low proportion (57.5%) of the patients known to have COVID-19 received ALS. This proportion is much lower than that for the non-COVID-19 patients in our study. This could partially explain the difference in survival. Overall, patients with suspected COVID-19 were treated in the same way as those with confirmed COVID-19. This lower level of ALS initiation during the COVID-19 era has been observed previously (5, 13). It has been suggested that resuscitation procedures can generate aerosols and, thus, risks for healthcare professionals, although the evidence has a very low degree of certainty (14). The WHO listed cardiopulmonary resuscitation (CPR) as an aerosol-generating procedure, and the international liaison committee on resuscitation confirmed this hypothesis (10). Initially, the scientific literature advised rescuers to consider their own safety before resuscitation or to change the ALS algorithm (15–17). International guidelines on resuscitation of COVID-19 patients came very late. The European Resuscitation Council COVID-19 guidelines suggested considering defibrillation before chest compression and ventilation while wearing personal protective equipment (PPE) (14). Professionals who provide ALS must take account of the patient’s context and medical history when assessing personal risks associated with treatment (14). Furthermore, the ethic in resuscitation suggests to take into account the prognostication of patient when to start ALS. Hence, regarding ethic, the futility of the resuscitation based on a zero survival rate is questioned. In this context, the high ALS initiation rate (57.5%) observed here with systematic intubation for each of the resuscitated patients testifies to MICU team members’ level of commitment. We observed that medical OHCAs in confirmed COVID-19 patients were mainly due to respiratory distress (53.7%). Furthermore, confirmed COVID-19 patients were significantly more likely to have a history of respiratory disease than the other OHCA patients studied here. Hence, it is important to follow up COVID-19 patients carefully, especially when a history of respiratory disease is known or if the patients have few or only mild symptoms and have not been hospitalized. Indeed, acute respiratory distress syndrome (ARDS) can even occur in patients without comorbidities and who do not receive expert, individual medical follow-up. ARDS can rapidly lead to multiple organ failure and cardiac arrest (18). The present study had a number of strengths. It was based on a large, national registry fed by MICUs throughout France (including both rural and urban areas). However, participation by the MICUs was voluntary, and some French MICUs did not participate in the study. The study also had some limitations. First, this study might be not generalizable to some other countries. The present study population was predominantly Caucasian and had some specific characteristics that prevented us from generalizing our results further. In addition, this study was carried out on a “stay and play” emergency system model and then may be not fully generalizable to countries applying a “scoop and run” model. Second, some OHCAs may have been misclassified with regard to their COVID-19 status. Indeed, some of the “non-COVID-19” cases might have been false-negatives, and we did not have access to postmortem information. Furthermore, suspected cases of COVID-19 could have been misclassified. However, we did not observe significant differences in the characteristics of suspected COVID-19 patients and confirmed COVID-19 patients, and so the error level was probably low. Third, our knowledge of inhospital data was limited. Hence, some of these cares, as a withdrawal of care in the confirmed COVID-19 group, might explain the absence of difference in ROSC and D0 survival and the presence of difference at D30. Finally, we performed a cohort study and excluded 1,245 patients because their COVID-19 status was unknown. Thus, their COVID group and vital status were not observed, and this could have changed some results. However, we worked on a large sample of 6,624 patients, which allows us to observe some effects. CONCLUSIONS Our results highlighted a zero survival rate in OHCA patients with confirmed COVID-19. The current resuscitation guidelines suggest that professional emergency responders use PPE and assess the risk before to starting CPR in suspected or confirmed cases of COVID-19. The risk-benefit balance for resuscitating confirmed COVID-19 patients should be investigated. The survival of OHCA patients with confirmed COVID-19 should be analyzed in other countries. ACKNOWLEDGMENTS The members of the French National Out-of-Hospital Cardiac Arrest Registry (RéAC) study group are: Drs. Oganov Kirill, Babin Clement, Vasseur Laurene, Agostinucci Jean Marc, Pernot Thomas, Guery Carole, Fritsch Emmanuelle, Harel David, Guillon Alain, Lorge Sarah, Halbout Laurent, Levrard Pascaline, Narcisse Sophie, Hugenschmitt Delphine, Potriquier Stephane, Chassin Coralie, Deslais Benoit, Courcoux Hubert, Larcher Fanny, Bourg Arthur, Laot Melanie, Beaka Placido, De schlichting Marie Alix, Duperron Yann, Simeon Isabelle, Bargain Philippe, Morel Marechal Emanuel, Conio Alice, Le Beuan Celine, Dattin Alix, Andriamirado Florian, Dussoulier Sebastien, Cohen Rudy, Baina Anne, Morel Jean-Charles, Ballet Celine, Lafitte Blandine, Klimas Stephane, Robart Jean-Christophe, Simonnet Bruno, Busi Olivia, Maurel Marion, Decker Sandra, Remond Annabelle, Dubois Camille, Ahui Terence, Dubeaux Josephine, Ginoux Lucie, Sciacca Christelle, Fuseau Celine, Fiani Nasri, Aubert Raphael, Nenert Eloi, Prudor Florence, Goulvin Virginie, Bonhomme Cecile, Peixoto Brandon, Theurey Odile, Grave Eric, Marrakchi Faycal, Guillet Aline, Megard Marie fleur, Lamarche-Vadel Yacine, Sussat Myriam, Goulois Nathalie, Plenier Cecile, Montagnon Francois-Xavier, Laborne Francois xavier, Piboule Ludovic, Grua David, Thibaud Eric, Jeanmasson Yoann, Sauvaget Geraldine, Muteaud Margaux, Guillaumee Frederic, Jaeger Deborah, Genuyt Benoit, Naud Julien, Baudin Marine, Abarrategui Diego, Altervain Yohan, Gay Julien, Maroteix Paul, Weyer Claire Marie, Thiriez Sylvain, Ramaherison Thierry, Mesli Adil, Massacrier Sylvie, Ovtcharenko Mariane, Duchier Caroline, Gaillard Nancy, Ursat Cecile, Jardel Benoit, Suhas Pauline, Rallu Martin, Picot Jessica, Pancher Agathe, Handwerk Tom, Bardelay Romain, Lagadec Steven, Lespiaucq Christine, Joliet Geraldine, Maigre Olivier, Mur Sebastien, Longo Celine, Li Crapi Raffaello, Barberis Christophe, Ngoyi Natacha, Serre Patrice, Mansouri Nadia, Leroy Antoine, Sanchez Caroline, Chevrier Guillaume, Jonquet Sebastien, Boutin Celia, Vanderstraeten Carine, Colson Camille, Barbery Adele, Meunier Juliette, Bertrand Philippe, Watrelot Olivier, Hiller Pascale, Guinard Sollweig, Javaudin Francois, Lepeve Alexandra, Labarrere Franck, Pes Philippe, Fromont Isaure, Carle Olivier, Crusoe Clarysse, Bernigaud Emmanuel, Dubernat Manon, Hoff Julie, Garcia Carolina, Huet Loic, Petitprez Martin, Lambert Julia, Trouvain Helene, Jung Mathilde, Gentilhomme Angelie, Durieux Emilie, Kamara Mariam, Agbemebia Fabrice, David Sandy, Savu Alexandru, Outrequin Maud, Garcia Lea, Delprat Adrien, Segard Lionel, Sebai Salim, Leduc Aurelien, Orange Rodolphe, Le Pennetier Olivier, Arnaudet Idriss, Prouve Christina, Boulanger Chloe, Trogoff Bruno, Nicolats Ophelie, Rouet Christelle, Guenier Pierre-Alban, Branche Fabienne, Ferreira Justine, Carruesco Chloe, Letur Gregory, Parsis Pierre, Hebrard Manon, Vermersch Celine, Yvetot Quentin, Besserve Paulin, Dyani Mohamed, Bujon Cecile, Loquet Lea, Le Normand Thomas, Miquelestorena Julie, Costa Aurelie, Hsing Priscilia, Desclefs Jean-Philippe, Roucaud Nicolas, Negrello Florian, Lafay Marina, Jubert Ignasi, Jeziorny Alexandre, Besnier Sylvie, Alba Pierre, Bouilleau Guillem, Tellier Eric, Hamdan David, Roux Nicolas, Pradeau Catherine, Petitdemange Olivia, Yahiaoui Samraa, Pradignac Nicolas, Torreton Florian, Goldstein Adrien, Letourneur Emilie, Bouhey Emmanuel, Dumont Nathalie, Marinoni Heloise, Mozzi Valerie, Verge Guillaume, Galtier Veronique, Nussbaum Camille, Lamourere Charles, Majour Gilles, Gevrey Vincent, Chatrenet Arthur, Laville Edouard, Gress Gauthier, Simon Benoit, Pascalon Clemence, Vara Paul, Basty-Ghuysen Marielle, Tabary Romain, Chesnoy Marine, Majoufre Gwenaelle, Lougnon Jean-Paul, Antouard Jerome, Pic Daniel, Noel Magaly, Bokobza Romain, Clauw Emmelyne, Robert Helene, Guerin Magali, Blain Stephane, Charney Alexandre, Durand Sandrine, Rakotonirina Jean-Louis, Lafon Jean, Pretalli Jean-Baptiste, Guigon Victor, Lory Thomas, Evain Yoann, Emonet Anne, Capuano Elvira, Macabre Yannick, Moine Linda, Foudi Lahcene, Andre Antoine, Bertille Hedi, Oliveira Larissa, Gornet Marion, Laggia Kelly, Edange Claire, Martinage Arnaud, Leroux Pierre, Bonnet Lucie, Rojas Jerome, Beunas Veronique, Millot Audrey, Beayni Zaher, Robert Frederique, Masson Loic, Segard Julien, Lannes Mathilde, Billier Laurianne, Barandiaran Patrick, Ferri Anne Laure, Parisot Sarah, Mazzoldi Margherita, Roudiak Nathalie, Laboure Anne, Lucas Remy, Chomono Hendricks, Metzger Jacques, Rapp Jocelyn, Ferquel Martin, Sylvie Roux, Echeikh Malek, Lamazou Elisa, Lahmar Sana, Herpin Aurelien, and Viratelle Clelia. *See also p. 883. The French National Out-of-Hospital Cardiac Arrest Registry (RéAC) study group are listed in the Acknowledgments. The French National Cardiac Arrest Registry is supported by the French Society of Emergency Medicine; a patient foundation, the Fédération Française de Cardiologie; the Mutuelle Générale de l’Education Nationale, and Lille University. The registry is funded by the Hauts-de-France Regional Council and the European Union via the European Regional Development Fund. The authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Guessoum SB Lachal J Radjack R : Adolescent psychiatric disorders during the COVID-19 pandemic and lockdown. Psychiatry Res. 2020; 291 :113264 32622172 2. Li LZ Wang S : Prevalence and predictors of general psychiatric disorders and loneliness during COVID-19 in the United Kingdom. Psychiatry Res. 2020; 291 :113267 32623266 3. Baldi E Sechi GM Mare C ; Lombardia CARe researchers: COVID-19 kills at home: The close relationship between the epidemic and the increase of out-of-hospital cardiac arrests. Eur Heart J. 2020; 41 :3045–3054 32562486 4. Hubert H Baert V Beuscart JB : Use of out-of-hospital cardiac arrest registries to assess COVID-19 home mortality. BMC Med Res Methodol. 2020; 20 :305 33317467 5. Baert V Jaeger D Hubert H ; GR-RéAC: Assessment of changes in cardiopulmonary resuscitation practices and outcomes on 1005 victims of out-of-hospital cardiac arrest during the COVID-19 outbreak: Registry-based study. Scand J Trauma Resusc Emerg Med. 2020; 28 :119 33339538 6. Berdowski J Berg RA Tijssen JG : Global incidences of out-of-hospital cardiac arrest and survival rates: Systematic review of 67 prospective studies. Resuscitation. 2010; 81 :1479–1487 20828914 7. Gräsner JT Wnent J Herlitz J : Survival after out-of-hospital cardiac arrest in Europe - results of the EuReCa TWO study. Resuscitation. 2020; 148 :218–226 32027980 8. Sultanian P Lundgren P Strömsöe A : Cardiac arrest in COVID-19: Characteristics and outcomes of in- and out-of-hospital cardiac arrest. A report from the Swedish Registry for Cardiopulmonary Resuscitation. Eur Heart J. 2021;42 :1094–1106 33543259 9. Javaudin F Penverne Y Montassier E : Organisation of prehospital care: The French experience. Eur J Emerg Med. 2020; 27 :404–405 33105295 10. Perkins GD Jacobs IG Nadkarni VM ; Utstein Collaborators: Cardiac arrest and cardiopulmonary resuscitation outcome reports: Update of the Utstein Resuscitation Registry Templates for Out-of-Hospital Cardiac Arrest: A statement for healthcare professionals from a task force of the International Liaison Committee on Resuscitation (American Heart Association, European Resuscitation Council, Australian and New Zealand Council on Resuscitation, Heart and Stroke Foundation of Canada, InterAmerican Heart Foundation, Resuscitation Council of Southern Africa, Resuscitation Council of Asia); and the American Heart Association Emergency Cardiovascular Care Committee and the Council on Cardiopulmonary, Critical Care, Perioperative and Resuscitation. Circulation. 2015; 132 :1286–1300 25391522 11. World Health Organization: Global Surveillance for COVID-19 Caused by Human Infection With COVID-19 Virus: Interim Guidance, 20 March 2020. Geneva, Switzerland, World Health Organization. 2020. Available at: https://apps.who.int/iris/handle/10665/331506. Accessed May 28, 2021 12. Shah P Smith H Olarewaju A : Is cardiopulmonary resuscitation futile in coronavirus disease 2019 patients experiencing in-hospital cardiac arrest? Crit Care Med. 2021; 49 :201–208 33093278 13. Fothergill RT Smith AL Wrigley F : Out-of-hospital cardiac arrest in London during the COVID-19 pandemic. Resusc Plus. 2021; 5 :100066 33521706 14. Nolan JP Monsieurs KG Bossaert L ; European Resuscitation Council COVID-Guideline Writing Groups: European Resuscitation Council COVID-19 guidelines executive summary. Resuscitation. 2020; 153 :45–55 32525022 15. Edelson DP Sasson C Chan PS ; American Heart Association ECC Interim COVID Guidance Authors: Interim guidance for basic and advanced life support in adults, children, and neonates with suspected or confirmed COVID-19: From the emergency cardiovascular care committee and get with the guidelines-resuscitation adult and pediatric task forces of the American Heart Association. Circulation. 2020; 141 :e933–e943 32270695 16. Kapoor I Prabhakar H Mahajan C : Cardiopulmonary resuscitation in COVID-19 patients - To do or not to? J Clin Anesth. 2020; 65 :109879 32450476 17. Malysz M Dabrowski M Böttiger BW : Resuscitation of the patient with suspected/confirmed COVID-19 when wearing personal protective equipment: A randomized multicenter crossover simulation trial. Cardiol J. 2020; 27 :497–506 32419128 18. Dhont S Derom E Van Braeckel E : The pathophysiology of ‘happy’ hypoxemia in COVID-19. Respir Res. 2020; 21 :198 32723327
PMC009xxxxxx/PMC9005090.txt	==== Front Am J Phys Med Rehabil Am J Phys Med Rehabil AJPMR American Journal of Physical Medicine & Rehabilitation 0894-9115 1537-7385 Lippincott Williams & Wilkins 35067551 AJPMR_220316 10.1097/PHM.0000000000001969 00001 3 CME Article . 2022 Series . Number 5 Musculoskeletal and Neurological Pain Symptoms Among Hospitalized COVID-19 Patients Jena Debasish MD debasish.mbbs@gmail.com Sahoo Jagannatha DNB pmr_jagannath@aiimsbhubaneswar.edu.in Barman Apurba MD, DNB pmr_apurba@aiimsbhubaneswar.edu.in Gupta Anil MD, DNB dranilaiims@yahoo.co.in Patel Vikas MBBS vikaiims2012@gmail.com From the Department of Physical Medicine and Rehabilitation, AIIMS, Bhubaneswar, India (DJ, JS, AB, VP); and Department of Physical Medicine and Rehabilitation, King George’s Medical University, Lucknow, India (AG). All correspondence should be addressed to: Debasish Jena, MD, Department of PMR, AIIMS Bhubaneswar, Sijua, Patrapada, PO-Dumduma, Bhubaneswar, Odisha, India, 751019. 5 2022 24 1 2022 24 1 2022 101 5 411416 Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved. 2022 Wolters Kluwer Health, Inc. All rights reserved. This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. Objective As the coronavirus disease 2019 pandemic continues to grow, its clinical manifestations are still emerging and are being widely investigated. However, the pain symptoms, including neurological and musculoskeletal pain symptoms, are still poorly understood. Design In this cross-sectional study, we investigated the prevalence of musculoskeletal and neurological pain symptoms among hospitalized coronavirus disease 2019 patients. Furthermore, the association of clinical and demographic factors with the prevalence of pain symptoms was also investigated. Result We included 182 hospitalized coronavirus disease 2019 patients with a mean age of 48.86 ± 13.98 yrs. Pain symptoms were reported by 61.54% patients (n = 112). Most common symptoms reported were generalized myalgia (n = 60, 32.96%), headache (n = 50, 27.47%), and low back pain (n = 41, 22.53%). Interestingly, neuropathic pain was present in 14 participants (7.69%). Logistic regression analysis revealed an association of pain symptoms with coronavirus disease 2019 severity, male sex, higher body mass index, and a history of addiction. Conclusions Pain symptoms are common manifestation of coronavirus disease 2019. Generalized myalgia, headache, and low back pain are the three most common new-onset pain symptoms in hospitalized coronavirus disease 2019 patients. Further investigation of pain symptoms and their predictive factors are recommended, which may guide healthcare workers and policymakers to plan in this direction. To Claim CME Credits Complete the self-assessment activity and evaluation online at http://www.physiatry.org/JournalCME CME Objectives Upon completion of this article, the reader should be able to: (1) Understand common musculoskeletal and neurological pain symptoms among hospitalized COVID-19 patients; (2) Understand the basic etiopathogenesis of COVID-19 associated pain; and (3) Identify factors associated with presence of COVID-19 pain symptoms. Level Advanced Accreditation The Association of Academic Physiatrists is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The Association of Academic Physiatrists designates this Journal-based CME activity for a maximum of 1.0 AMA PRA Category 1 Credit(s)™. Physicians should only claim credit commensurate with the extent of their participation in the activity. Key Words Coronavirus Disease 2019 COVID-19 Musculoskeletal Pain Neuropathic Pain Pain Symptoms SDCT CMECME ==== Body pmc What Is Known The literature so far has displayed that musculoskeletal pain is among the most common symptoms of COVID-19 with myalgia, headache, and limb pain being the most frequent pain symptoms. What Is New This study demonstrated that generalized myalgia, headache, and low back pain are the most frequently reported pain symptoms of COVID-19. In addition, neuropathic pain, though rare, was reported in 7.69% of participants. We also revealed a direct association of new-onset pain symptoms with COVID-19 severity, male sex, body mass index, and a history of addiction. The coronavirus disease 2019 (COVID-19) outbreak originated in Wuhan, China, and was declared as a global pandemic on March 11, 2020, by the World Health Organization.1 As of August 10, 2021, the disease has spread across more than 220 countries and territories, infecting more than 202 million people across the world with more than 4.2 million deaths.2 The COVID-19 has dramatically changed the way of living and posed a significant threat to public health worldwide. This has drawn unprecedented interest from public health researchers around the world, promoting extensive research on disease characteristics and strategies for management. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the novel coronavirus that causes COVID-19, can affect nearly every organ system, causing respiratory, cardiovascular, gastrointestinal, musculoskeletal (MSK), and neurologic symptoms. Besides the major clinical manifestations indicating a respiratory infection, such as fever, cough, sore throat, and nasal stuffiness, pain is a commonly reported symptom accompanying COVID-19.3,4 Pathophysiology of COVID-19–associated pain can be multifactorial and may include an autoimmune response to the virus, increased cytokine production, and widespread tissue damage (such as muscles and joints).5 Furthermore, a neurotropic property of the virus has been described, which may explain the neurologic pain symptoms of COVID-19 including headache and neuropathic pain.6 If untreated, pain may become chronic and substantially affect activities of daily living, which may further contribute to disease-related disability and reduced quality of life. Acute pain symptoms associated with COVID-19 can be grouped as localized (e.g., pharyngalgia or sore throat), remote pain (e.g., headache), or generalized (e.g., myalgia, arthralgia, or limb pain). Headache and myalgia are among the most common acute pain symptoms reported so far and can occur in up to 71% of patients.6–8 A recent retrospective study has reported that the head and limbs were the most frequently complained painful sites.9 Neuropathic pain has also been reported in COVID-19 patients,10 although there are very scant data regarding its prevalence and clinical characteristics. Therefore, we aimed to explicitly evaluate the prevalence of MSK and neurologic pain symptoms in the acute phase of COVID-19 and their relationship with demographic and clinical characteristics. METHODS This cross-sectional study was conducted at a tertiary care center on hospitalized patients diagnosed with COVID-19 between May 10, 2021, and July 10, 2021. Inclusion criteria were confirmed cases of COVID-19 and age of 18 yrs or greater. Exclusion criteria were known history of chronic pain disorders, known psychiatric illness, serious concurrent illness preventing participation, and unwillingness for participation. A confirmed case was defined as a person with laboratory confirmation of COVID-19 infection, irrespective of clinical signs and symptoms. Real-time reverse transcription–polymerase chain reaction was used to confirm the SARS-CoV-2 infection. Sample size was calculated using a standard formula based on the prevalence from a recent similar study.9 Clearance for the study was taken from institutional ethical committee (reference number: T/IM-NF/PMR/21/19). Patients who satisfied the inclusion and exclusion criteria were approached with the study proposal, and its future implications were explained. The participants signed an informed written consent that was approved by the institutional ethical committee. This study complies with Strengthening the Reporting of Observational Studies in Epidemiology guidelines (see Supplemental Checklist, Supplemental Digital Content 1, http://links.lww.com/PHM/B530). A physical medicine and rehabilitation resident posted in the COVID-19 dedicated ward collected the data, which included age, sex, height, weight, COVID-19 severity, history of addiction, and known comorbidities. Severity classification was done according to the guidelines issued by the Ministry of Health and Family Welfare, Government of India.11 Mild disease was defined as uncomplicated upper respiratory tract infection without the evidence of breathlessness or hypoxia (with normal saturation). Moderate disease was defined as presence of clinical features of dyspnea and/or hypoxia, fever, cough, including Spo2 90% to 94% on room air and a respiratory rate of 24 breaths/min or greater. Severe disease was identified as presence of clinical signs of pneumonia plus one of the following; respiratory rate of greater than 30 breaths/min, severe respiratory distress, and Spo2 of less than 90% on room air. Patients were asked to report whether they were experiencing any new-onset pain symptoms after the diagnosis of COVID-19. We included only MSK and neurological pain symptoms. If pain was present, its location, character, and severity were noted. For MSK pain and headache, a pain severity of two or more (of 10) on Numeric Rating Scale was considered for inclusion. Furthermore, the Douleur Neuropathique 4 questionnaire score of four or more was used for the confirmation of the diagnosis of neuropathic pain. Pain localized to a specific part of the body was named according to the anatomical location (e.g., low back pain, knee pain, foot pain, etc). Neck and shoulder pain were considered together as most patients were unable to differentiate between the two. Generalized myalgia was defined as diffuse muscle pain and polyarthralgia was identified as pain involving four or more joints. Headache and neuropathic pain were included under neurologic pain. Pain described as electric shock–like, burning, shooting, prickling, or crawling sensations were identified as neuropathic pain. Statistical analysis was performed using IBM SPSS version 20.0. Continuous variables are reported as mean ± SD and categorical demographic and clinical data are reported using numbers of participants and percentage. The significance of the two means was compared using an unpaired t test, and a comparison of categorical data is reported using Yate corrected χ2 test. A binary logistic regression analysis was performed to identify the predictors of incidence of the new-onset pain symptoms. The significance level was presented as “P value,” with P < 0.05 considered as significant. Randomization, logistic regression, and odds ratio were used to control for the confounders. RESULTS A total of 372 hospitalized patients with laboratory confirmed COVID-19 were screened for the study. Thirty-four patients were of age less than 18 and were excluded from the study. Forty-four patients were having severe illness with either, ventilator support, reduced consciousness, or communication issues. Seventy-nine patients gave history of chronic pain and 12 patients were having concurrent psychiatric illness. Last, 21 patients did not consent to participate. Thus, after exclusion of the previously mentioned patients, a total of 182 patients were included in the final analysis after they satisfied the inclusion and exclusion criteria and consented for participation (Fig. 1). The mean age of the participants was 48.86 ± 13.98 yrs with a predominantly male population (n = 132, 72.53%). Seventy participants (38.46%) had their body mass index (BMI) in either the overweight or obese category with the rest (61.54%) falling into the normal category. The severity of COVID-19 was distributed as 102 mild cases (56.04%), 62 moderate cases (34.07%), and 18 severe cases (9.89%). Type 2 diabetes mellitus was the most commonly reported comorbidity, followed by hypertension. Other comorbidities in decreasing order of prevalence were coronary artery disease, chronic respiratory disease, hypothyroidism, neoplasm, chronic kidney disease, chronic liver disease, and others. Multiple comorbidities (≥2) were present in 36 participants (19.78%). History of addiction with either alcohol or smoking was similar (n = 20, 10.99%) and addiction to both of these was found in 15 participants (8.24%). No addiction history was found in 127 patients (69.78%). The baseline characteristics of all participants have been summarized in Table 1. FIGURE 1 Participant flow diagram. TABLE 1 Demographic and clinical characteristics of participants (N = 182) Mean age, yr 48.86 ± 13.98 Sex Male 132 (72.53) Female 50 (27.47) BMI, kg/m2 Normal 112 (61.54) Overweight 61 (33.52) Obese 9 (4.94) Severity of COVID-19 Mild 102 (56.04) Moderate 62 (34.07) Severe 18 (9.89) Comorbidities 109 (59.89) Type 2 diabetes mellitus 66 (36.26) Hypertension 42 (23.08) Coronary artery disease 14 (7.69) Chronic respiratory disease 11 (6.04) Hypothyroidism 10 (5.49) Neoplasm 8 (4.39) Chronic kidney disease 7 (3.85) Chronic liver disease 6 (3.30) Others 11 (6.04) Multiple (≥2 comorbidities) 36 (19.78) None 73 (40.11) Addiction 55 (30.22) Alcohol 20 (10.99) Smoking 20 (10.99) Both 15 (8.24) None 127 (69.78) Data are presented as mean ± SD or n (%). One or more pain symptoms were reported by 61.54% of patients (n = 112). Three most common symptoms were generalized myalgia (n = 60, 32.96%), headache (n = 50, 27.47%), and low back pain (n = 41, 22.53%). Other pain symptoms in decreasing order of incidence were as follows: leg/calf pain, polyarthralgia, knee pain, hand pain, neck/shoulder pain, neuropathic pain, and foot pain. Neuropathic pain was present in 14 participants (7.69%). The distribution of neuropathic pain was as follows: only upper limb in five patients, only lower limbs in six patients, and diffuse (all limbs and trunk) in three patients. Two patients presented pain (described as tingling sensation) in the median nerve distribution similar to that of carpal tunnel syndrome. Diffuse neuropathic pain (in three patients) was described as burning and/or pin and needles in character involving the whole body below the neck. Three patients presented pain (electric shock–like or tingling sensation) in the distribution of lateral femoral cutaneous nerve. In rest seven patients, no specific nerve distribution could be identified. Objective assessment of neuropathic pain including electrodiagnostic studies and radiological investigations could not be performed because of the COVID-19 regulations. Seventy participants (38.46%) reported no MSK and/or neurologic pain symptoms. Simultaneous presence of four or more pain symptoms was reported by 26 patients (14.29%). In mild cases, generalized myalgia (n = 34, 33.33%) was the most common pain symptom, whereas in moderate and severe cases, headache (n = 25, 40.32%, and n = 12, 66.67%, respectively) was the most commonly reported pain symptom. The distribution of pain symptoms among participants is depicted in Table 2 and Figure 2. TABLE 2 Distribution of pain and associated symptoms among the participants (N = 182) Musculoskeletal pain symptoms Generalized myalgia 70 (38.46) Low back pain 41 (22.53) Leg pain 19 (10.44) Polyarthralgia 17 (9.34) Knee pain 16 (8.79) Hand pain 16 (8.79) Neck/shoulder pain 15 (8.24) Foot pain 8 (4.39) Neurological pain symptoms A. Headache 50 (27.47) B. Neuropathic pain 14 (7.69) 1. Upper limb 5 (2.75) 2. Lower limb 6 (3.30) 3. Diffuse 3 (1.65) Multiple pain symptoms 1. Two symptoms 25 (13.74) 2. Three symptoms 15 (8.24) 3. Four or more symptoms 26 (14.29) No pain symptoms 70 (38.46) Data are presented as n (%). FIGURE 2 Distribution of pain symptoms among participants (N = 182). Association of demographic characteristics with COVID-19 severity was analyzed. COVID-19 severity was associated with increased age, presence of comorbidities, and history of addiction (P < 0.05), but no significant association was found with sex and BMI of participants. A logistic regression analysis revealed an association of pain symptoms with male sex, BMI, COVID severity, and history of addiction. No clear association could be demonstrated between the presence of pain symptoms with age and presence of comorbidities (Table 3). TABLE 3 Logistic regression analysis demonstrating association of participant variables with pain symptoms (N = 182) Variable Predictor Coefficient Standard Error P Odds Ratio 95% Confidence Interval Age, yr ≥60 (vs. <60) 0.4386 0.4859 0.3667 1.5505 0.5982–4.0187 Sex Male (vs. female) 1.4688 0.4532 0.0012a 4.3442 1.7871–10.5603 BMI, kg/m2 ≥25 (vs. <25) 0.9798 0.4168 0.0187a 2.6641 1.1770–6.0299 COVID-19 severity Moderate to severe (vs. mild) 1.2594 0.3836 0.0010a 3.5233 1.6612–7.4728 Comorbidity Present (vs. absent) 0.2618 0.3771 0.4875 1.2993 0.6205–2.7208 Addiction Present (vs. absent) 1.2408 0.4676 0.0080a 3.4585 1.3831–8.6481 a P < 0.05 (statistical significance). DISCUSSION This study demonstrated a substantial proportion (61.54%) of hospitalized COVID-19 patients presenting with acute pain symptoms. There has been wide variation in reports of COVID-19–associated pain with very few studies evaluating pain symptoms in particular. This discrepancy in the incidence of pain symptoms may be ascribed to a difference in study objectives and surveillance methods.9,12,13 In a recent retrospective study conducted by Şahin et al.,9 the reported prevalence of pain symptoms was 40.7% before the infection that increased to 82.5% during the infection. Knox et al.12 conducted a prospective observational study in which 38.5% of COVID-19 patients presented with an active pain symptom. The higher prevalence of pain in our study can be attributed to several factors. First, memory bias could be avoided by cross-sectional evaluation of symptoms. Second, we evaluated pain locations in a more precise manner (e.g., hand pain or foot pain) in comparison with previous studies where a more generalized approach was undertaken (e.g., upper limb or lower limb pain).9 Third, the inclusion of neuropathic pain might be another determinant for this high prevalence. Last, a difference in baseline demographic characteristics and virus properties are few other important factors that might have resulted in the difference in the pain incidence. Our finding is consistent with the fact that myalgia (either localized or generalized) is one of the most frequently encountered pain symptoms among COVID-19 patients. In a recent review of symptoms during COVID-19, myalgia was encountered as the fifth most common symptom.14 Arthralgia is a relatively uncommon symptom of COVID-19 and a recent study reported an incidence of 15.1%. The incidence rate of myalgia or arthralgia has been inconsistent across studies and reported in up to 61.0% of cases.15 These MSK pain symptoms can be a direct result of tissue damage by the virus or secondary to several other factors. Angiotensin-converting enzyme 2 (ACE-2) receptors are known to be widely distributed in the musculoskeletal system.16 These receptors are considered to play a central role in the entry of SARS-CoV-2 into the cells of MSK tissue. This direct invasion through ACE-2 receptors may result in MSK injury, pain, weakness, and fatigue, all of which are known to occur in COVID-19.17–19 In addition, SARS-CoV-2 is known to cause cytokine storm that involves interleukin 6, interleukin 10, and tumor necrosis factor α. This cytokine storm may further induce or aggravate damage in various tissues, such as muscle and joints triggering pain and related symptoms.5 Mao et al.20 reported that COVID-19 patients displaying muscle symptoms demonstrated higher levels of creatine kinase and lactate dehydrogenase than those without muscle symptoms.20 Other potential mechanisms known to cause MSK pain include virus-associated myositis or myopathy, rhabdomyolysis, and steroid-induced myopathy.21–23 In addition, central nervous system involvement by the virus can also manifest as MSK tissue damage and pain.20 Headache is another commonly encountered pain symptom of COVID-19, and the prevalence rate in past studies ranged from 6% to 21%.7 Similar to myalgia/arthralgia, headache can also occur as a result of a direct viral invasion of the central nervous system or because of indirect effects of the virus. The virus has been known to directly invade the trigeminal nerve ending inside the nasal or oral cavity, potentially causing the headache.24 Penetration and damage to central nervous system have been confirmed by isolating SARS-CoV-2 from cerebrospinal fluid by genome sequencing.25 Furthermore, the neurotropism of the virus has been described as a potential mechanism of direct involvement of the nervous system. Besides the virus-specific factors, several triggers may contribute to the occurrence of the headache. These triggers may include psychological stress due to social isolation and/or disease-related anxiety, an adverse effect due to multiple drug prescriptions, or prolonged use of a mask during hospitalization. In a web-based study by Uygun et al.,24 COVID-19–associated headache was frequently long-lasting, bilateral, with a male predominance, and had resistance to analgesics. In our study, a higher prevalence of headache in moderate to severe disease supports the fact that neurological complications are more frequent in severe COVID-19.20 Neuropathic pain is a rare pain symptom of COVID-19 and has been explored in lesser detail so far. Similar to headaches, neuropathic pain may result from direct invasion of the nervous system (central nervous system or peripheral nervous system) or due to virus-mediated immune reactions. SARS-CoV-2 can cause an imbalance between ACE-2 and angiotensin II in the spinal cord that might result in neuropathic pain. Moreover, cytokines and chemokines can lead to activation of nociceptive sensory neurons resulting in pain.25 In addition, prolonged prone positioning deployed to improve oxygenation is known to cause direct injury of peripheral nerves.26 Other well-known neurological complications of COVID-19 that might result in neuropathic pain include transverse myelitis, Guillain-Barre syndrome, etc. Moreover, there are also reports of mixed sensory-motor neuropathy in COVID-19 patients leading to neuropathic pain.10 Nonetheless, the incidence and clinical characteristics of neuropathic pain need further investigation. On careful analysis, we revealed a direct association of the pain symptoms with male sex, history of addiction, and COVID-19 severity. The predominance of pain symptoms among the male population may be attributable to an increased incidence of COVID-19 among the male population and a more efficient immune activity among females during other viral illnesses. In addition, a sex difference regarding the level of expression of ACE-2 and a protective role of female hormones or the location of ACE-2 on the X chromosome might explain this discrepancy.24 Association of pain symptoms with disease severity is by the fact that cytokine storm is more frequently displayed by severe COVID-19 patients. In the past, addiction history (especially smoking), higher BMI, and medical comorbidities (e.g., depression) have all been linked to the prevalence of pain symptoms.27 Nonetheless, an association of these factors with COVID-19–associated pain is yet to be established. To our knowledge, there is no published literature at this point providing a clear analysis of pain symptoms and their association with patient and disease characteristics. This study is not without limitations. First, data from a single tertiary care center prevent the generalization of the results. Second, pain severity and their persistence beyond the acute phase were not assessed. Third, a lower proportion of severe COVID-19 patients were included in this study. Nonetheless, the percentage of severe patients is relatively higher than that reported in a similar recent study.9 Fourth, we could not evaluate few important factors including the duration since the diagnosis of COVID-19 to onset of symptoms, presence of systemic commemoratives, and current use of pain medications. Last, lack of a control population can draw less meaningful conclusions from the study. The strength of our study is its cross-sectional nature with a face-to-face interview of participants. Most of the past studies did a retrospective assessment that might have resulted in memory bias and hence underreporting of the symptoms. In addition, a more explicit evaluation of symptoms including evaluation of neuropathic pain adds to the novelty of this study. Considering the extensive spread of the COVID-19 across the globe and a significantly high prevalence of pain symptoms, healthcare workers must screen and detect such symptoms at the earliest. Chronic pain may lead to psychological distress, sleep impairment, and negatively impact social life. With prompt and efficient treatment, chronic pain and its complications can be prevented, thereby improving health-related quality of life. Furthermore, policy makers should consider reallocation of resources in this direction, which may help prevent a potential pain pandemic in the recent future. CONCLUSIONS Generalized myalgia, headache, and low back pain are the most common pain symptoms associated with COVID-19. The cause of the pain is multifactorial. Without treatment, pain can become chronic and add substantially to disease burden and healthcare costs. Importantly, chronic pain can potentially interfere with the rehabilitation process and delay recovery. Therefore, the potential burden of pain associated with COVID-19 cannot be ignored, especially considering the extent of the affected population. Understanding and early detection of these diverse pain symptoms may guide prompt management. The authors encourage future prospective and systematic studies evaluating COVID-19–associated pain further elaborating their risk factors and persistence beyond the acute phase of the disease. Vikas Patel is in training. Financial disclosure statements have been obtained, and no conflicts of interest have been reported by the authors or by any individuals in control of the content of this article. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.ajpmr.com). ==== Refs REFERENCES 1 WHO: WHO Director-General’s opening remarks at the media briefing on COVID-19–11 March 2020; 2020. Available at: https://www.who.int/dg/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19---11-march-2020. Accessed August 10, 2021 2 Johns Hopkins University: Coronavirus resource center. Available at: https://coronavirus.jhu.edu. Accessed August 10, 2021 3 Drozdzal S Rosik J Lechowicz K , : COVID-19: pain management in patients with SARSCoV-2 infection-molecular mechanisms, challenges, and perspectives. Brain Sci 2020;10 :465 4 WHO: Report of the WHO-China Joint Mission on Coronavirus Disease 2019 (COVID-19). Geneva, Switzerland, World Health Organization, 2020 5 Chen G Wu D Guo W , : Clinical and immunological features of severe and moderate coronavirus disease 2019. J Clin Invest 2020;130 :2620–9 32217835 6 Liu M He P Liu HG , : Clinical characteristics of 30 medical workers infected with new coronavirus pneumonia [in Chinese]. Zhonghua Jie He He Hu Xi Za Zhi 2020;43 :209–14 32164090 7 Tolebeyan AS Zhang N Cooper V , : Headache in patients with severe acute respiratory syndrome coronavirus 2 infection: a narrative review. Headache 2020;60 :2131–8 33017479 8 Tostmann A Bradley J Bousema T , : Strong associations and moderate predictive value of early symptoms for SARS-CoV-2 test positivity among healthcare workers, the Netherlands, March 2020. Euro Surveil 2020;25 :2000508 9 Şahin T Ayyıldız A Gencer-Atalay K , : Pain symptoms in COVID-19. Am J Phys Med Rehabil 2021;100 :307–12 33480608 10 Bureau BL Obeidat A Dhariwal MS , : Peripheral neuropathy as a complication of SARS-CoV-2. Cureus 2020;12 :e11452 33214969 11 Ministry of Health and Family Welfare: Available at: https://www.mohfw.gov.in/pdf/ClinicalManagementProtocolforCOVID19dated27062020.pdf. Accessed August 10, 2021 12 Knox N Lee CS Moon JY , : Pain manifestations of COVID-19 and their association with mortality: a multicenter prospective observational study. Mayo Clin Proc 2021;96 :943–51 33722397 13 Trigo J García-Azorín D Planchuelo-Gómez Á , : Factors associated with the presence of headache in hospitalized COVID-19 patients and impact on prognosis: a retrospective cohort study. J Headache Pain 2020;21 :94 32727345 14 Zhu J Ji P Pang J , : Clinical characteristics of 3062 COVID-19 patients: a meta-analysis. J Med Virol 2020;92 :1902–14 32293716 15 Mo P Xing Y Xiao Y , : Clinical characteristics of refractory coronavirus disease 2019 in Wuhan, China. Clin Infect Dis 2021;73 :e4208–13 32173725 16 Li MY Li L Zhang Y , : Expression of the SARS-CoV-2 cell receptor gene ACE2 in a wide variety of human tissues. Infect Dis Poverty 2020;9 :45 32345362 17 Ferrandi PJ Alway SE Mohamed JS : The interaction between SARSCoV-2 and ACE2 may have consequences for skeletal muscle viral susceptibility and myopathies. J Appl Physiol 2020;129 :864–7 32673162 18 Huang C Wang Y Li X , : Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. Lancet 2020;395 :497–506 31986264 19 Motta-Santos D Dos Santos RA Oliveira M , : Effects of ACE2 deficiency on physical performance and physiological adaptations of cardiac and skeletal muscle to exercise. Hypertens Res 2016;39 :506–12 27053009 20 Mao L Jin H Wang M , : Neurologic manifestations of hospitalized patients with coronavirus disease 2019 in Wuhan, China. JAMA Neurol 2020;77 :683–90 32275288 21 Fan CK Yieh KM Peng MY , : Clinical and laboratory features in the early stage of severe acute respiratory syndrome. J Microbiol Immunol Infect 2006;39 :45–53 16440123 22 Wang JT Sheng WH Fang CT , : Clinical manifestations, laboratory findings, and treatment outcomes of SARS patients. Emerg Infect Dis 2004;10 :818–24 15200814 23 Jin M Tong Q : Rhabdomyolysis as potential late complication associated with COVID-19. Emerg Infect Dis 2020;26 :1618–20 32197060 24 Uygun Ö Ertaş M Ekizoğlu E , : Headache characteristics in COVID-19 pandemic—a survey study. J Headache Pain 2020;21 :121 33050880 25 Baig AM Khaleeq A Ali U , : Evidence of the COVID-19 virus targeting the CNS: tissue distribution, host-virus interaction, and proposed neurotropic mechanisms. ACS Chem Neurosci 2020;11 :995–8 32167747 26 Malik GR Wolfe AR Soriano R , : Injury-prone: peripheral nerve injuries associated with prone positioning for COVID-19–related acute respiratory distress syndrome. Br J Anaesth 2020;125 :e478–80 32948295 27 Shi Y Hooten MW Roberts RO , : Modifiable risk factors for incidence of pain in older adults. Pain 2010;151 :366–71 20696524
PMC009xxxxxx/PMC9005091.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 35200194 00001 10.1097/CCM.0000000000005450 3 Feature Articles Effects of Prone Position on Lung Recruitment and Ventilation-Perfusion Matching in Patients With COVID-19 Acute Respiratory Distress Syndrome: A Combined CT Scan/Electrical Impedance Tomography Study* Fossali Tommaso MD 1 Pavlovsky Bertrand MD 2 Ottolina Davide MD 1 Colombo Riccardo MD 1 Basile Maria Cristina MD 1 Castelli Antonio MD 1 Rech Roberto MD 1 Borghi Beatrice MD 1 Ianniello Andrea MD 3 Flor Nicola MD 3 Spinelli Elena MD 2 Catena Emanuele MD 1 Mauri Tommaso MD 24 1 Department of Anesthesiology and Intensive Care, ASST Fatebenefratelli Sacco, Luigi Sacco Hospital, University of Milan, Milan, Italy. 2 Department of Anesthesia, Critical Care and Emergency, Fondazione Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS) Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy. 3 Department of Radiology, ASST Fatebenefratelli Sacco, Luigi Sacco Hospital, University of Milan, Milan, Italy. 4 Department of Pathophysiology and Transplantation, University of Milan, Milan, Italy. For information regarding this article, E-mail: tommaso.fossali@asst-fbf-sacco.it 11 4 2022 5 2022 11 4 2022 50 5 723732 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. OBJECTIVES: Prone positioning allows to improve oxygenation and decrease mortality rate in COVID-19–associated acute respiratory distress syndrome (C-ARDS). However, the mechanisms leading to these effects are not fully understood. The aim of this study is to assess the physiologic effects of pronation by the means of CT scan and electrical impedance tomography (EIT). DESIGN: Experimental, physiologic study. SETTING: Patients were enrolled from October 2020 to March 2021 in an Italian dedicated COVID-19 ICU. PATIENTS: Twenty-one intubated patients with moderate or severe C-ARDS. INTERVENTIONS: First, patients were transported to the CT scan facility, and image acquisition was performed in prone, then supine position. Back to the ICU, gas exchange, respiratory mechanics, and ventilation and perfusion EIT-based analysis were provided toward the end of two 30 minutes steps (e.g., in supine, then prone position). MEASUREMENTS AND MAIN RESULTS: Prone position induced recruitment in the dorsal part of the lungs (12.5% ± 8.0%; p < 0.001 from baseline) and derecruitment in the ventral regions (–6.9% ± 5.2%; p < 0.001). These changes led to a global increase in recruitment (6.0% ± 6.7%; p < 0.001). Respiratory system compliance did not change with prone position (45 ± 15 vs 45 ± 18 mL/cm H2O in supine and prone position, respectively; p = 0.957) suggesting a decrease in atelectrauma. This hypothesis was supported by the decrease of a time-impedance curve concavity index designed as a surrogate for atelectrauma (1.41 ± 0.16 vs 1.30 ± 0.16; p = 0.001). Dead space measured by EIT was reduced in the ventral regions of the lungs, and the dead-space/shunt ratio decreased significantly (5.1 [2.3–23.4] vs 4.3 [0.7–6.8]; p = 0.035), showing an improvement in ventilation-perfusion matching. CONCLUSIONS: Several changes are associated with prone position in C-ARDS: increased lung recruitment, decreased atelectrauma, and improved ventilation-perfusion matching. These physiologic effects may be associated with more protective ventilation. atelectrauma electrical impedance tomography prone position pulmonary perfusion recruitment SDCT ==== Body pmcThe COVID-19 pandemic has already caused the death of more than 4 million people. In most severe cases, the acute respiratory infection leads to severe pneumonia. In around 20% of hospitalized patients, pneumonia worsens to progressive hypoxemia and acute respiratory distress syndrome (ARDS) (1), which mortality rate is higher than 50% (2). The overwhelming number of intubated patients with ARDS associated with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (COVID-19–associated ARDS [C-ARDS]) and the severity of their disease warrant urgent implementation of simple and effective therapies to decrease mortality. Prolonged sessions in the prone position represent a simple and effective intervention to decrease mortality in patients with ARDS (3). Prone position has been widely adopted for the treatment of C-ARDS, both before intubation and during invasive ventilation (4, 5). Interestingly, despite time and staff constraints due to the pandemic, the proportion of patients with C-ARDS turned prone is significantly higher than patients with ARDS from other etiologies (6, 7), and the prone position is now indicated as a cornerstone for the ventilatory management of C-ARDS (8). Pilot observational studies showed that the prone position in intubated patients with C-ARDS may decrease hospital mortality (9). From a physiologic standpoint, prone position improves oxygenation in patients with C-ARDS, while respiratory mechanics appear unaffected (10). Thus, the physiologic pathway leading to decreased mortality in patients with C-ARDS undergoing pronation needs further exploration, especially since there is no correlation with oxygenation (11) or respiratory mechanics (7). In “classical ARDS,” lung recruitment with nearly constant airway pressure is a key mechanism of lung protection associated with pronation. However, this feature may be unlikely in C-ARDS, as a large proportion of patients presents quasi-normal respiratory system compliance, which may indicate preserved lung inflation and limited recruitability (12). A pilot study suggested a dead space reduction in patients with C-ARDS turned prone (13), which may decrease the risk of deleterious regional hypocapnia (14). Finally, the potential effect of prone position on atelectrauma has, at our knowledge, never been investigated in C-ARDS. Aim of this study was to further characterize the physiologic effects of prone position on key mechanisms of regional lung protection, namely: recruitment, reduced atelectrauma, and improved ventilation-perfusion matching, by CT scan and electrical impedance tomography (EIT). METHODS Study Population We conducted a prospective physiologic study in patients admitted to the dedicated COVID-19 ICU of Luigi Sacco Hospital (ASST Fatebenefratelli Sacco), Milan, Italy. We enrolled intubated patients admitted to the ICU with confirmed infection by novel COVID-19 (SARS-CoV-2) and moderate or severe ARDS according to the Berlin definition. Decision for pronation was reached if the Pao2/Fio2 ratio was measured below 150 mm Hg or as an emergent rescue therapy in patients with Spo2 less than 85% with Fio2 100%. Exclusion criteria were: age less than 18 years old, pregnancy, intubation more than 7 days, confirmed diagnosis of hospital-acquired bacterial pneumonia, contraindications to the prone position or to EIT monitoring (e.g., thoracic wounds), and clinical severity (e.g., need for extracorporeal membrane oxygenation [ECMO] therapy). The institutional Ethical Committee approved the study (Comitato Etico Milano Area 1; protocol n. 2020/ST/388) and informed consent was obtained according to local regulations. Data Collection The following patients’ characteristics were recorded at enrollment: age, gender, body mass index, history of hypertension or diabetes mellitus, plasma C-reactive protein (CRP) and d-dimers level, Sequential Organ Failure Assessment (SOFA) score (15), the number of hours spent in prone position before enrollment and days from onset of symptoms, intubation, and pronation. Respiratory system compliance, Pao2/Fio2, and ventilatory ratio were measured at enrollment (see Online Supplement, http://links.lww.com/CCM/G986). Study Protocol Patients were deeply sedated, paralyzed, and mechanically ventilated on pressure-regulated volume-controlled (PRVC) mode. Ventilator settings were standardized for all patients during all study measures: tidal volume (Vt) 6–8 mL/kg of predicted body weight, respiratory rate to maintain pH between 7.35 and 7.45, and positive end-expiratory pressure (PEEP) 10 cm H2O (Table 1). Fio2 was set to obtain a Spo2 value between 94 and 98% and kept stable during all the study protocol. TABLE 1. Patients Main Characteristics Variable All Patients, n = 21 Patients characteristics Age, yr 67 (61–72) Comorbidities, n (%) Hypertension 12 (57) Diabetes mellitus 5 (24) Male, n (%) 17 (81) Body mass index, kg/m2 28.6 (26.3–32.0) Sequential Organ Failure Assessment score 6 (3–7) C-reactive protein, mg/L 179 (81–211) d-dimers, µg/L 1,360 (815–5,333) Severe acute respiratory distress syndrome, n (%) 11 (52) Days from onset of symptoms, d 12 (8–17) Days from intubation, d 2 (1–4) Days from first pronation, d 1 (1–2) Hours spent in prone position before enrollment, hr 36 (16–72) Ventilator settings Positive end-expiratory pressure, cm H2O 10 (± 1) Fio2, % 83 (± 16) Tidal volume, mL/kg predicted body weight 7.5 (± 0.8) Respiratory rate, breaths/min 19 (± 2) Gas exchange and mechanics in supine at enrollment Pao2/Fio2, mm Hg 105 (84–121) Ventilatory ratio 1.74 (1.50–2.25) Respiratory system compliance, mL/cm H2O 39 (32–52) After enrollment, patients were initially transported to the CT scan facility in the prone position. Whole thorax scans were performed in the prone and supine position during an end-expiratory occlusion at PEEP 10 cm H2O (time between scans 15 and 20 min). After the scan, patients were transported back to the ICU and connected to a ventilator. EIT monitoring was started, and measurements of distribution of ventilation and perfusion were recorded in the supine position and 20–30 minutes after pronation. Immediately before each EIT measurement, arterial and central venous blood gas analyses were obtained, and respiratory mechanics were measured. Further information on data and statistical computation are provided in the Online Supplement (http://links.lww.com/CCM/G986). CT Scan Analysis CT scans were performed by an experienced team, then centralized and analyzed offline, using a standard software (Maluna v3.7, Mannheim, Germany), to provide a quantitative analysis of lung tissue aeration (16). Further information are available in the Online Supplement (http://links.lww.com/CCM/G986). Ventral and dorsal regions were defined as the upper and lower parts, respectively, of an axis from the sternum to the vertebrae (16). This choice allowed us to obtain more superimposable regions of interest between CT scan and EIT images. Recruitment (or derecruitment) between supine and prone position at global and regional levels was computed as the respective decrease or increase in nonaerated weight, divided by the global lung weight in the supine position (16). Electrical Impedance Tomography EIT data were acquired by standard device (PulmoVista; Draeger, Lubeck, Germany), with a sample rate of 50 Hz. The EIT belt was positioned directly below armpits, between the third and fifth intercostal spaces. The EIT belt was kept in the same position during both supine and prone position. We measured a so-called impedance curve concavity index based on a similar concept to the stress index (17). Time-impedance curve was fitted to a power equation to assess its concavity. It is assumed that, during ventilation with constant inspiratory pressure, the concavity of the impedance curve may be an acceptable surrogate for the pressure-volume curve, where upper concavity represents ongoing recruitment of collapsed alveoli/small airways (Fig. E1, http://links.lww.com/CCM/G986) (18). This index was measured at global and regional scales. Ventilation-perfusion matching was measured by using the hypertonic saline bolus method (see Online Supplement, http://links.lww.com/CCM/G986) (19, 20). Statistical Analysis For each variable, Gaussian distribution was assessed by Shapiro-Wilk normality test. After checking for normality, results were expressed as a number (percentage) for qualitative variables and with median (interquartile range) or mean (± sd) for quantitative variables. A paired t test or Wilcoxon signed-rank test, as appropriate, were used to compare between variables measured in the supine and prone position. Based on previous studies on CT scan analysis in C-ARDS patients (21, 22), we hypothesized relatively low lung recruitment induced by the prone position of 5% ± 5%; this, with a type I error of 0.05 and statistical power of 90%, lead to a minimum calculated sample size of 21 patients. A secondary analysis was also performed to identify subgroups with larger recruitment. Patients were grouped according to: 1) severe versus moderate ARDS and 2) higher or lower respiratory system elastance (< 2 vs > 2 cm H2O/kg × mL). p value of less than 0.05 was considered significant. Spearman correlations were used to explore the association between global and regional recruitment and the ΔPaO2/Fio2 (defined by Pao2/Fio2 prone minus supine, divided by the value in supine position). All statistical analysis were performed by using Prism (GraphPad Prism v9.0, La Jolla, CA). RESULTS Patients’aCharacteristics Twenty-one patients were enrolled in the study. Twenty-three consecutive patients were screened for enrollment, two patients were excluded due to their clinical severity and indication for ECMO support. Median age was 67 years old (61–72 yr old) and 17 (81%) were men (Table 1). Clinical severity and level of inflammatory markers were elevated, as suggested by median SOFA score of 6 (3–7) and plasmatic CRP of 179 mg/L (81–211 mg/L) (Table 1). Time between start of symptoms and intubation was 12 days (8–17 d) (Table 1), and all patients were enrolled within 5 days from intubation. Patients underwent 1 day (1–2 d) of pronation before enrollment (Table 1). As per protocol, mechanical ventilation settings were standardized with a Vt of 6–8 mL/kg, PEEP of 10 cm H2O, and fixed respiratory rate targeted for pH greater than 7.25. Settings remained unchanged during the study (Table 1). In supine position at the time enrollment, Pao2/Fio2 was 105 mm Hg (84–121 mm Hg), with a maximal value of 149 mm Hg (Table 1). Respiratory system compliance was 39 mL/cm H2O (23–52 mL/cm H2O). CT Scan Analysis Quantitative CT scan showed that the nonaerated lung weight decreased significantly in the prone position (p = 0.001) (Table 2; and Fig. E2, http://links.lww.com/CCM/G986). Prone position also induced an increase of the normally aerated lung weight (p = 0.004), along with a significant decrease of the hyperinflated tissue (p = 0.008) (Table 2; and Fig. E2, http://links.lww.com/CCM/G986). Regional distribution of lung tissue aeration is reported in Table E1 (http://links.lww.com/CCM/G986). TABLE 2. Regional Quantitative CT Scan and Electrical Impedance Tomography Analysis Between the Supine and Prone Positions Variable Supine, n = 21 Prone, n = 21 p CT scan global analysis Total lung weight, g 1,466 (± 378) 1,394 (± 381) 0.007 Hyperinflated lung weight, g 14 (± 12) 12 (± 9) 0.008 Normally aerated lung weight, g 356 (± 132) 400 (± 164) 0.004 Poorly aerated lung weight, g 525 (± 192) 505 (± 173) 0.335 Nonaerated lung weight, g 571 (± 294) 477 (± 249) 0.001 CT scan recruitment analysis Recruitment, % Baseline 6.0 (± 6.7) < 0.001 Ventral derecruitment, % of lung weight Baseline –6.9 (± 5.2) < 0.001 Dorsal recruitment, % of lung weight Baseline 12.5 (± 8.0) < 0.001 Electrical impedance tomography Vt distribution ventral, % 53 (± 8) 40 (± 11) < 0.001 Vt distribution dorsal, % 47 (± 9) 60 (± 11) < 0.001 TIC concavity index 1.41 (± 0.16) 1.30 (± 0.16) 0.001 Ventral TIC concavity index 1.40 (± 0.16) 1.35 (± 0.16) 0.186 Dorsal TIC concavity index 1.45 (± 0.20) 1.25 (± 0.19) < 0.001 Only perfused units, % 5 (1–12) 8 (4–19) 0.105 Only perfused units, ventral, % 2 (0–5) 7 (1–11) 0.023 Only perfused units, dorsal, % 2 (0–8) 2 (0–10) 0.742 Only ventilated units, % 28 (16–36) 22 (15–31) 0.301 Only ventilated units, ventral, % 14 (12–22) 8 (3–12) < 0.001 Only ventilated units; dorsal, % 11 (4–15) 14 (9–22) 0.133 Dead space/shunt ratio 5.1 (2.3–23.4) 4.3 (0.7–6.8) 0.035 Dead space/shunt ratio, ventral 11.3 (3.7–19.0) 1.5 (0.4–6.0) < 0.001 Dead space/shunt ratio, dorsal 4.3 (0.8–14.8) 8.6 (0.6–21.5) 0.404 TIC = time-impedance curve, Vt = tidal volume. Considering the whole lung, recruitment induced by prone position was significant (p < 0.001) (Table 2) and only two patients (9.6%) experienced derecruitment (Fig. 1). Regional response to prone position was dissociated: ventral areas were characterized by derecruitment (p < 0.001), while significant recruitment characterized the dorsal regions (p < 0.001) (Table 2 and Fig. 1). These changes were associated with an increase in mean Hounsfield Units in the ventral regions and a decrease in the dorsal parts of the lungs (both p < 0.001; Table E2, http://links.lww.com/CCM/G986). Figure 1. Recruitment measured by CT scan expressed as % of total lung weight and Electrical impedance tomography-based time-impedance curve (TIC) concavity index in the supine and prone position. Recruitment induced by the prone position was significant at the global level (A), but ventral lung regions were characterized by derecruitment (B) and only dorsal lung was recruited (C). The TIC concavity index improved at the global and dorsal regional level (D and F) without worsening in the ventral derecruited region (E). Red bars represent mean values. p < 0.01 versus supine. ns = not significant. Figure 2 shows a representative patient with large fraction of recruitment in the dorsal lung when turned prone. Figure 2. Effects of prone position on recruitment and ventilation-perfusion matching in a representative study patient. Top: CT scan images performed in the supine (left) and prone position (right). Note the recruitment in the dorsal regions and the derecruitment in the ventral part of the right lung. Bottom: Electrical impedance tomography assessment of ventilation (blue) and perfusion (red). Note the large fraction of only-ventilated units (dead space) in the ventral lung regions during supine position (left), largely decreased by prone position (right); only perfused units (shunt), instead, increased in the same ventral region. Ventilation and Perfusion by EIT Data from EIT indicate that recruitment in the dorsal region induced significantly increased regional ventilation, while the ventral derecruited lung was characterized by reduced ventilation (p < 0.001 for both) (Table 2). The concavity index significantly decreased in the prone position only in the dorsal regions of the lung (p < 0.001) (Table 2 and Fig. 1). EIT-based measure of pulmonary perfusion was of acceptable quality in 16 patients (76%). Considering the whole lung, prone position did not affect the fraction of mismatched units (i.e., only ventilated and only perfused) (Table 2), but it induced significant decrease of the dead space/shunt ratio (p = 0.035) (Table 2). At the regional level, the fraction of only ventilated units and the dead space/shunt ratio significantly decreased in the ventral region (p < 0.001 for both), together with a slight increase of the only perfused units (p = 0.023) (Table 2; and Fig. E3, http://links.lww.com/CCM/G986). The dorsal region did not show any significant change in the ventilation-perfusion matching after pronation. Figure 2 shows EIT-based pulmonary ventilation and perfusion in supine and prone position in a representative study patient. Respiratory Mechanics and Gas Exchange Prone positioning did not induce any change in respiratory mechanics, while oxygenation improved and calculated pulmonary shunt significantly decreased (p < 0.01) (Table 3; and Fig. E4, A and B, http://links.lww.com/CCM/G986). There was no difference in Pao2/Fio2 nor ventilatory ratio between their values at enrollment and during the study in the supine position after the cycle of pronation (p = 0.618, p = 0.101, and p = respectively). Respiratory system compliance instead improved after the cycle of prone positioning (p = 0.020). TABLE 3. Respiratory Mechanics and Gas Exchange Between the Supine and Prone Positions Variable Supine, n = 21 Early Prone, n = 21 p Respiratory mechanics Plateau pressure, cm H2O 23 (± 3) 23 (± 4) 0.294 Driving pressure, cm H2O 12 (± 3) 12 (± 4) 0.456 Respiratory system compliance, mL/cm H2O 45 (± 15) 45 (± 18) 0.957 Oxygenation Pao2, mm Hg 85 (± 21) 142 (± 90) < 0.001 Pao2/Fio2, mm Hg 108 (± 41) 176 (± 100) 0.002 Arterial dioxygen saturation, % 95 (± 4) 97 (± 3) 0.003 Alveolo-arterial difference in dioxygen partial pressure, mm Hg 441 (± 124) 379 (± 134) 0.003 Measured venous admixture, % 49 (39–55) 35 (27–46) 0.007 Central venous dioxygen saturation, % 81 (± 6) 81 (± 10) 0.973 CO2 clearance Paco2, mm Hg 53 (± 7) 53 (± 8) 0.542 pH 7.38 (± 0.07) 7.37 (± 0.06) 0.134 Corrected minute ventilation, L/min 11.9 (± 2.3) 12.2 (± 2.6) 0.369 Ventilatory ratio 2.03 (± 0.41) 2.06 (± 0.44) 0.477 During prone position, hemodynamics remained stable, as indicated by central venous dioxygen saturation (Table 3), and there was no modification of CO2 clearance by the lungs (Table 3; and Fig. E4C, http://links.lww.com/CCM/G986). Of note, there was no association between global, ventral, or dorsal recruitment, and the ΔPaO2/Fio2 between the supine and prone positions (rho = 0.091, p = 0.703; rho = 0.317, p = 0.173; and rho = 0.184, p = 0.436, respectively) (Fig. E5, http://links.lww.com/CCM/G986). Subgroups Analysis To identify patients more likely to respond to prone position in terms of recruitment, we compared the effect of pronation between patients with severe versus moderate ARDS (recruitment: 7% ± 7% vs 5% ± 6%; p = 0.593) (Fig. E6, http://links.lww.com/CCM/G986) and with lower versus higher compliance (recruitment: 6% ± 8% vs 6% ± 6%; p = 0.802) (Fig. E6, http://links.lww.com/CCM/G986) but found no difference. DISCUSSION This study describes the lung protective effects of prone position in patients with C-ARDS, when performed in the first days after intubation. The main findings can be summarized as follows: 1) despite mild derangement of respiratory mechanics and relatively preserved lung aeration in the supine position, prone position induces extensive alveolar recruitment in the dorsal regions; 2) alveolar derecruitment occurs in the ventral lung regions, albeit by far smaller extent than dorsal recruitment; 3) dorsal recruitment reduces the risk of regional atelectrauma in comparison to the supine position; and 4) ventral lung regions, after pronation, are characterized by decreased fraction of ventilated nonperfused units and reduced dead space/shunt ratio. Study patients, as previously described for C-ARDS (21, 22), had relatively preserved respiratory system compliance in the supine position, and a low amount of collapsed lung tissue (16). Despite this, prone position induced regional recruitment in the collapsed dorsal regions when they were turned from a gravitationally dependent to nondependent position, similarly to “classical” ARDS (23). Interestingly, the corresponding ventral derecruitment was smaller, likely due to differences in the shape of the chest in the two positions (24). This so-called “sponge-lung” phenomenon (25) may decrease dorsal lung strain and ventral overdistension (26), leading to more protective ventilation (25, 26). The fraction of dorsal recruitment obtained by prone position at constant airway pressure was large with higher values than those obtained by the application of a 45 cm H2O inspiratory pressure (16) in unselected ARDS patients. A higher PEEP reduces nonaerated lung tissue (22) and improves the recruitment to inflation ratio (27) in patients with C-ARDS, but the prone position may be regarded as more physiologically safe since the maneuver does not increase the inspiratory and driving pressures. Interestingly, in this study, the amount of recruitment was not associated with disease severity nor with oxygenation improvement. These findings highlight the complexity of hypoxemia mechanisms in C-ARDS and may be a reason to expend criteria for pronation, even to patients with moderate hypoxemia (3, 6). In the present study, significant recruitment induced by the prone position was not associated with an increase in respiratory system compliance. As this finding could indicate that the fraction of ventilated units did not change between positions, we hypothesized that these units might be subject to cyclical opening and closing in the supine position, which is then reduced by turning the patients to prone. The EIT technology allowed us to assess the global and regional dynamics of intra-tidal ventilation by analyzing the slope of the time-impedance curve (18, 28). We assumed that, everything being equal and with a constant pressure (as in PRVC), the time-impedance curve was an acceptable surrogate for the pressure-volume curve (18). Following this assumption, concavity of the impedance curve to values closer to 1 in the prone position would likely be due to a reduced fraction of alveolar units (or of small airways) opening along inspiration and to reduced atelectrauma. These data would be coherent with previous results in ARDS from other etiologies (29). Although chest wall and lung compliance were not measured in this study, decreased atelectrauma potentially could have been determined by an increase in chest wall stiffness coupled with decreased lung elastance (29, 30). Interestingly, in the ventral regions, the impedance curve concavity remained stable, while derecruitment occurred. This result could be explained by complete collapse of alveoli and airways in these regions secondary to increased superimposed lung weight, which might have determined also a decrease of the regional chest wall compliance. Prone position was also associated with changes in ventilation-perfusion matching. First, ventral fraction of ventilated nonperfused units (i.e., pure dead space) decreased, while perfused nonventilated units (i.e., pure shunt) from the same region slightly increased. These data confirm previous results from animal models (31, 32) and pilot clinical studies in C-ARDS patients (13). Interestingly, decreased dead space could be regarded as protective due to decreased risk of regional hypocapnia (14), while minimal increase in shunt should not affect lung protection. Second, the dead space/shunt ratio decreased with prone position. This ratio is elevated in patients with C-ARDS (19, 33) and a recent prospective study in “classical” ARDS showed a correlation with outcome (33). Thus, its decrease could be regarded as another marker of improved lung protection by prone position. Finally, calculated venous admixture significantly decreased with pronation (Table 2), while pure shunt measured by EIT did not change. These results could indicate decreased areas with low ventilation/perfusion ratio after pronation, as these contribute only to calculated and not to pure shunt, further increasing lung protection by reduced lung inhomogeneities (33). The recent clinical study reporting a correlation between improved oxygenation during early pronation and survival of patients with C-ARDS may confirm a causal relationship between changes in ventilation-perfusion matching in the prone position and outcome (34). There are limitations to this study: 1) the sample size was limited, albeit larger than most physiologic studies on this topic (13, 19); 2) patients were enrolled early in the course of C-ARDS and findings may change with clinical evolution; 3) partitioned respiratory mechanics by use of esophageal pressure (especially chest wall and lung compliances) were not measured, leaving the explanation of the effects of chest wall properties in the prone position remained speculative; 4) the EIT characterization of pulmonary regional perfusion by EIT is limited to three-compartment model (ventilated nonperfused, perfused nonventilated, and normal units), while exploration of the larger spectrum of ventilation-perfusion defects might be more accurate for understanding the physiologic effects of prone position; and 5) central hemodynamics, and especially cardiac output were not measured in this cohort, even though they can impact pulmonary perfusion and ventilation-perfusion matching (35). However, central venous saturation remained stable, suggesting unchanged oxygen delivery. CONCLUSIONS Prone position in patients with C-ARDS is associated with lung recruitment, decreased risk of atelectrauma, and improved indexes of ventilation-perfusion matching when performed early after intubation. These physiologic mechanisms may represent the causal link between prone position, lung protection, and improved clinical outcomes in C-ARDS. Supplementary Material See also p. 873. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal). Drs. Fossali, Spinelli, and Mauri designed the study. All authors participated to data collection and analysis. All authors participated to article redaction and revision and approved the final version of this article. The present study was supported by institutional funding of the Department of Anesthesia, Critical Care and Emergency, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy (ricerca corrente 2020). Dr. Mauri received funding from the Department of Anesthesia, Critical Care and Emergency, Fondazione Istituto di Ricovero e Cura a Carattere Scientifico Ca’ Granda Ospedale Maggiore Policlinico, Milan, Italy (ricerca corrente 2020). Dr. Mauri’s institution received funding from Fisher and Paykel and Draeger; he received funding from Fisher and Paykel, Draeger, and BBraun. The remaining authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Richardson S Hirsch JS Narasimhan M ; the Northwell COVID-19 Research Consortium: Presenting characteristics, comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the New York City Area. JAMA. 2020; 323 :2052–2059 32320003 2. Grasselli G Greco M Zanella A ; COVID-19 Lombardy ICU Network: Risk factors associated with mortality among patients with COVID-19 in intensive care units in Lombardy, Italy. JAMA Intern Med. 2020; 180 :1345–1355 32667669 3. Guérin C Reignier J Richard JC ; PROSEVA Study Group: Prone positioning in severe acute respiratory distress syndrome. N Engl J Med. 2013; 368 :2159–2168 23688302 4. Coppo A Bellani G Winterton D : Feasibility and physiological effects of prone positioning in non-intubated patients with acute respiratory failure due to COVID-19 (PRON-COVID): A prospective cohort study. Lancet Respir Med. 2020; 8 :765–774 32569585 5. COVID-ICU Group on behalf of the REVA Network and the COVID-ICU Investigators: Clinical characteristics and day-90 outcomes of 4244 critically ill adults with COVID-19: A prospective cohort study. Intensive Care Med. 2021; 47 :60–73 33211135 6. Guérin C Beuret P Constantin JM ; investigators of the APRONET Study Group, the REVA Network, the Réseau recherche de la Société Française d’Anesthésie-Réanimation (SFAR-recherche) and the ESICM Trials Group: A prospective international observational prevalence study on prone positioning of ARDS patients: The APRONET (ARDS Prone Position Network) study. Intensive Care Med. 2018; 44 :22–37 29218379 7. Langer T Brioni M Guzzardella A ; PRONA-COVID Group: Prone position in intubated, mechanically ventilated patients with COVID-19: A multi-centric study of more than 1000 patients. Crit Care. 2021; 25 :128 33823862 8. Poston JT Patel BK Davis AM : Management of critically ill adults with COVID-19. JAMA. 2020; 323 :1839–1841 32215647 9. Mathews KS Soh H Shaefi S ; Study of the treatment and outcomes in critically ill patients with Coronavirus Disease (STOP-COVID) Investigators: Prone positioning and survival in mechanically ventilated patients with coronavirus disease 2019-related respiratory failure. Crit Care Med. 2021; 49 :1026–1037 33595960 10. Weiss TT Cerda F Scott JB : Prone positioning for patients intubated for severe acute respiratory distress syndrome (ARDS) secondary to COVID-19: A retrospective observational cohort study. Br J Anaesth. 2021; 126 :48–55 33158500 11. Spinelli E Mauri T : Why improved PF ratio should not be our target when treating ARDS. Minerva Anestesiol. 2021; 87 :752–754 33688707 12. Gattinoni L Chiumello D Caironi P : COVID-19 pneumonia: Different respiratory treatments for different phenotypes? Intensive Care Med. 2020; 46 :1099–1102 32291463 13. Perier F Tuffet S Maraffi T : Effect of positive end-expiratory pressure and proning on ventilation and perfusion in COVID-19 acute respiratory distress syndrome. Am J Respir Crit Care Med. 2020; 202 :1713–1717 33075235 14. Kiefmann M Tank S Tritt MO : Dead space ventilation promotes alveolar hypocapnia reducing surfactant secretion by altering mitochondrial function. Thorax. 2019; 74 :219–228 30636196 15. Vincent JL Moreno R Takala J : The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med. 1996; 22 :707–710 8844239 16. Gattinoni L Caironi P Cressoni M : Lung recruitment in patients with the acute respiratory distress syndrome. N Engl J Med. 2006; 354 :1775–1786 16641394 17. Grasso S Terragni P Mascia L : Airway pressure-time curve profile (stress index) detects tidal recruitment/hyperinflation in experimental acute lung injury. Crit Care Med. 2004; 32 :1018–1027 15071395 18. Wrigge H Zinserling J Muders T : Electrical impedance tomography compared with thoracic computed tomography during a slow inflation maneuver in experimental models of lung injury. Crit Care Med. 2008; 36 :903–909 18431279 19. Mauri T Spinelli E Scotti E : Potential for lung recruitment and ventilation-perfusion mismatch in patients with the acute respiratory distress syndrome from coronavirus disease 2019. Crit Care Med. 2020; 48 :1129–1134 32697482 20. Bluth T Kiss T Kircher M : Measurement of relative lung perfusion with electrical impedance and positron emission tomography: An experimental comparative study in pigs. Br J Anaesth. 2019; 123 :246–254 31160064 21. Chiumello D Busana M Coppola S : Physiological and quantitative CT-scan characterization of COVID-19 and typical ARDS: A matched cohort study. Intensive Care Med. 2020; 46 :2187–2196 33089348 22. Ball L Robba C Maiello L ; GECOVID (GEnoa COVID-19) group: Computed tomography assessment of PEEP-induced alveolar recruitment in patients with severe COVID-19 pneumonia. Crit Care. 2021; 25 :81 33627160 23. Gattinoni L Pelosi P Vitale G : Body position changes redistribute lung computed-tomographic density in patients with acute respiratory failure. Anesthesiology. 1991; 74 :15–23 1986640 24. Gattinoni L Taccone P Carlesso E : Prone position in acute respiratory distress syndrome. Rationale, indications, and limits. Am J Respir Crit Care Med. 2013; 188 :1286–1293 24134414 25. Gattinoni L Pesenti A Carlesso E : Body position changes redistribute lung computed-tomographic density in patients with acute respiratory failure: Impact and clinical fallout through the following 20 years. Intensive Care Med. 2013; 39 :1909–1915 24026295 26. Galiatsou E Kostanti E Svarna E : Prone position augments recruitment and prevents alveolar overinflation in acute lung injury. Am J Respir Crit Care Med. 2006; 174 :187–197 16645177 27. Beloncle FM Pavlovsky B Desprez C : Recruitability and effect of PEEP in SARS-Cov-2-associated acute respiratory distress syndrome. Ann Intensive Care. 2020; 10 :55 32399901 28. Grasso S Stripoli T De Michele M : ARDSnet ventilatory protocol and alveolar hyperinflation: Role of positive end-expiratory pressure. Am J Respir Crit Care Med. 2007; 176 :761–767 17656676 29. Cornejo RA Díaz JC Tobar EA : Effects of prone positioning on lung protection in patients with acute respiratory distress syndrome. Am J Respir Crit Care Med. 2013; 188 :440–448 23348974 30. Pelosi P Tubiolo D Mascheroni D : Effects of the prone position on respiratory mechanics and gas exchange during acute lung injury. Am J Respir Crit Care Med. 1998; 157 :387–393 9476848 31. Lamm WJ Graham MM Albert RK : Mechanism by which the prone position improves oxygenation in acute lung injury. Am J Respir Crit Care Med. 1994; 150 :184–193 8025748 32. Richter T Bellani G Scott Harris R : Effect of prone position on regional shunt, aeration, and perfusion in experimental acute lung injury. Am J Respir Crit Care Med. 2005; 172 :480–487 15901611 33. Spinelli E Kircher M Stender B : Unmatched ventilation and perfusion measured by electrical impedance tomography predicts the outcome of ARDS. Crit Care. 2021; 25 :192 34082795 34. Scaramuzzo G Gamberini L Tonetti T ; ICU-RER COVID-19 Collaboration: Sustained oxygenation improvement after first prone positioning is associated with liberation from mechanical ventilation and mortality in critically ill COVID-19 patients: A cohort study. Ann Intensive Care. 2021; 11 :63 33900484 35. Lai C Adda I Teboul JL : Effects of prone positioning on venous return in patients with acute respiratory distress syndrome. Crit Care Med. 2021; 49 :781–789 33590997
PMC009xxxxxx/PMC9005095.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 35120039 00024 10.1097/CCM.0000000000005475 3 Editorials The Puzzles of Ventilator-Associated Pneumonia and COVID-19: Absolute Knowns and Relative Unknowns* Ryder Jonathan H. MD 1 Kalil Andre C. MD, MPH, FACP, FIDSA, FCCM 1 Both authors: Department of Internal Medicine, Division of Infectious Diseases, University of Nebraska Medical Center, Omaha, Nebraska 07 2 2022 5 2022 07 2 2022 50 5 894896 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. COVID-19 incidence ventilator-associated pneumonia ==== Body pmcThe development of ventilator-associated pneumonia (VAP) (1) in ICU patients with severe COVID-19 is common and associated with higher VAP rates compared with non-COVID-19 patients in several retrospective studies (2–6). Additionally, there is an increased risk of shock and blood stream infections associated with VAP in COVID-19 (5). A challenge in understanding the reason for this increased rate is the lack of a clear definition of VAP in COVID-19. Due to the frequent concomitant presence of fever, leukocytosis, and lung infiltrates due to the viral infection, differentiating bacterial colonization from a new bacterial secondary infection is challenging in the presence of these symptoms and imaging findings of an acute viral pneumonia. In part due to these challenges, the term ventilator-associated lower respiratory tract infection (VA-LRTI) has been used: the rate of VA-LRTI has also been found to be higher in ICU patients with COVID-19 compared with ICU patients with influenza and those without viral infection (7). A key remaining question is why VAP and VA-LRTI occur at higher rates in COVID-19 patients. Proposed reasons for increased rate of VAP and VA-LRTI in COVID-19 include viral immunomodulation, prolonged mechanical ventilation and hospital stay, steroids use, acute respiratory distress syndrome (ARDS), prone positioning, sedating and neuromuscular blocking agents, vasopressor use, extracorporeal mechanical oxygenation utilization, and increased demands of the healthcare system including patient volume, healthcare worker shortages, and use of personal protective equipment (8). Prior studies have not systematically evaluated how VA-LRTI has changed throughout the pandemic. In this issue of Critical Care Medicine, Hedberg et al (9) expand on the epidemiology of VA-LRTIs in COVID-19 using a retrospective analysis of ventilated adults in a large Swedish ICU from January 2011 to December 2020. Their primary goal was to evaluate changes in rate of VA-LRTI in COVID-19 compared with non-COVID-19 patients over time. Specifically, they compared COVID-19 patients with non-COVID-19 patients (including influenza and the 10 most common International Classification of Diseases, 10th Edition ICU diagnoses during and prior to the pandemic) and then performed a second analysis comparing the first wave of the COVID-19 pandemic (March–July 2020) with the second wave (October–December 2020). The cohort included 479 COVID-19 ICU episodes and 19,744 non-COVID-19 ICU episodes. Patients with COVID-19 were younger and had more comorbidities but were less likely to be immunosuppressed or have cancer. The median ICU length of stay was significantly longer for COVID-19 patients (14 vs 2 d; p < 0.001) with a higher 30-day mortality (24% vs 15%; p < 0.001). Markedly, the longest median duration of ventilation in non-COVID-19 diagnoses was 5 days for ARDS but reached 10 days in COVID-19. The proportion of VA-LRTI in ICU patients with COVID-19 was higher at 30% compared with 18% in patients without COVID-19. However, this difference was likely due to the longer duration of ventilation in COVID-19, a known risk factor for VAP (1). When adjusted for number of ventilator days at risk, the rate of VA-LRTI for patients with COVID-19 (31/1,000 ventilator-days) compared similarly to patients without COVID-19 (34/1,000 ventilator-days). Interestingly, this rate of VA-LRTI was higher in COVID-19 patients than that in other infectious causes of ICU stay (11/1,000 ventilator-days), namely, bacterial pneumonia, influenza, and severe sepsis. Comparing the first wave of the pandemic with the second wave, the second wave included older patients with more comorbidities, steroid use, and prone positioning, although a shorter length of mechanical ventilation. The VA-LRTI proportion in the COVID-19 group increased from 29% (99/381) to 38% (30/93) from the first to the second wave with a minimal change of 19% (37/567) to 21% (28/324) in the non-COVID-19 group. After multivariable analyses, the second wave had a significantly higher adjusted cause-specific hazard ratio (1.86 [95% CI, 1.15–3.01]) and adjusted subdistribution hazard ratio (1.81 [95% CI, 1.17–2.79]) compared with the first wave. How do we reconcile the various previous studies that have reported a significant increase in VAP rates in patients with COVID-19 (2–8, 10–14) with this finding of no increase from Hedberg et al (9)? The first explanation may be related to the very high rate of VA-LRTI in the control arm of Hedberg et al (9) study: 34 per 1,000 ventilation-days, compared with lower control arm rates from previous studies: 13 per 1,000 ventilation-days (3) and 15 per 1,000 ventilation-days (2), which indicates that Hedberg et al (9) reported more than double the control baseline rate of other studies; thus, if patients without COVID-19 were already at such a high risk for VAP (9), it might have become not feasible to detect any VAP rate difference compared with patients with COVID-19. A second explanation is that the infection definitions (e.g., VAP and VA-LRTI) used in these studies were different and not amenable to comparison. Another surveillance definition, ventilator-associated events (VAEs), has also been evaluated and shown higher VAEs per 100 episodes of ventilation, but similar VAEs rates per 1,000 ventilator-days in patients with COVID-19 versus those without COVID-19 (15). A third explanation is regarding the multiple limitations associated with all these studies done during the pandemic: no reporting and analysis accounting for standardized infection control measures, no collection of local baseline rates of VAP, no data on availability and turnover of hospital and ICU beds, adequacy of the number of healthcare providers needed for each hospital and ICU size, access to and training of personal protective equipment, and use of concomitant medications that may cause further immunosuppression (e.g. steroids), thus increasing the risk of VAP or antivirals that may limit further progression of the viral disease and reduce the length of hospital stay (e.g., remdesivir), thus decreasing the risk of VAP. All of the above variables, individually or combined, may have biased these studies either way, that is, increasing or decreasing the detection of any potential differences in the VAP rates between patients with and without COVID-19. A fourth explanation may stem from the use of absolute versus relative rates of VAP, both of which can produce discrepant interpretations even though reflecting the same actual frequency of infection; for example, although the absolute rate of VAP may be increased with COVID-19, the rate relative to baseline disease severity, age, comorbidities, and hospital length of stay may not show the same increase. Finally, the different methods for statistical adjustment (linear, logistic, Cox, Fine-Gray, and subdistribution regression models) may produce different VAP rates. One thing we can say with certainty: patients hospitalized with COVID-19 are undoubtedly requiring longer hospital/ICU stay and prolonged mechanical ventilation duration, are more frequently proned, and are receiving more immunosuppressive drugs than any other respiratory viral infection ever before. These three variables above have been well-known risk factors for VAP for a long time, even before the COVID-19 pandemic. Thus, it should come as no surprise that we are actually seeing an increased absolute rate of VAP during this pandemic; however, the question that remains is after adjusting for all relevant variables, whether the relative rate of VAP is increased. Until we have prospective cohort studies with more granular and systematic data collection regarding infection control measures, baseline hospital rates of VAP/VA-LRTI/VAE, hospital and ICU capacity, noninvasive and invasive respiratory support availability and utilization, baseline disease severity, concomitant use of antiviral and immunosuppressive drugs, and statistical methods for the appropriate adjustments, the “relative rate puzzle” remains unknown, but the “absolute rate puzzle” is already known. *See also p. 825. The authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Kalil AC Metersky ML Klompas M : Management of adults with hospital-acquired and ventilator-associated pneumonia: 2016 clinical practice guidelines by the Infectious Diseases Society of America and the American Thoracic Society. Clin Infect Dis. 2016; 63 :e61–e111 27418577 2. Vacheron CH Lepape A Savey A : Increased incidence of ventilator-acquired pneumonia in coronavirus disease 2019 patients: A multicentric cohort study. Crit Care Med. 2021 Sep 22. [online ahead of print] 3. Maes M Higginson E Pereira-Dias J : Ventilator-associated pneumonia in critically ill patients with COVID-19. Crit Care. 2021; 25 :25 33430915 4. Grasselli G Scaravilli V Mangioni D : Hospital-acquired infections in critically ill patients with COVID-19. Chest. 2021; 160 :454–465 33857475 5. Rouyer M Strazzulla A Youbong T : Ventilator-associated pneumonia in COVID-19 patients: A retrospective cohort study. Antibiotics (Basel). 2021; 10 :988 34439038 6. Luyt CE Sahnoun T Gautier M : Ventilator-associated pneumonia in patients with SARS-CoV-2-associated acute respiratory distress syndrome requiring ECMO: A retrospective cohort study. Ann Intensive Care. 2020; 10 :158 33230710 7. Rouzé A Martin-Loeches I Povoa P ; coVAPid study Group: Relationship between SARS-CoV-2 infection and the incidence of ventilator-associated lower respiratory tract infections: A European multicenter cohort study. Intensive Care Med. 2021; 47 :188–198 33388794 8. Povoa P Martin-Loeches I Nseir S : Secondary pneumonias in critically ill patients with COVID-19: Risk factors and outcomes. Curr Opin Crit Care. 2021; 27 :468–473 34321415 9. Hedberg P Ternhag A Giske CG : Ventilator-Associated Lower Respiratory Tract Bacterial Infections in COVID-19 Compared to Non-COVID-19 Patients. Crit Care Med. 2022; 50 :825–836 10. Ripa M Galli L Poli A ; COVID-BioB study group: Secondary infections in patients hospitalized with COVID-19: Incidence and predictive factors. Clin Microbiol Infect. 2022; in press 11. Blonz G Kouatchet A Chudeau N : Epidemiology and microbiology of ventilator-associated pneumonia in COVID-19 patients: A multicenter retrospective study in 188 patients in an un-inundated French region. Crit Care. 2021; 25 :72 33602296 12. Martínez-Martínez M Plata-Menchaca EP Nuvials FX : Risk factors and outcomes of ventilator-associated pneumonia in COVID-19 patients: A propensity score matched analysis. Crit Care. 2021; 25 :235 34229747 13. Suarez-de-la-Rica A Serrano P De-la-Oliva R : Secondary infections in mechanically ventilated patients with COVID-19: An overlooked matter? Rev Esp Quimioter. 2021;34 :330–336 33764004 14. Pickens CO Gao CA Cuttica MJ ; NU COVID Investigators: Bacterial superinfection pneumonia in patients mechanically ventilated for COVID-19 pneumonia. Am J Respir Crit Care Med. 2021; 204 :921–932 34409924 15. Weinberger J Rhee C Klompas M : Incidence, characteristics, and outcomes of ventilator-associated events during the COVID-19 pandemic. Ann Am Thorac Soc. 2022; 19 :82–89 34170781
PMC009xxxxxx/PMC9005097.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 34974498 00020 10.1097/CCM.0000000000005411 3 Editorials Resuscitation in Out-of-Hospital Cardiac Arrest Patients With COVID? Never Tell Me the Odds!* Barnicle Ryan N. MD, MS Ed 1 Wright Brian Joseph MD, MPH, FACEP 2 1 Department of Emergency Medicine, Yale University School of Medicine, New Haven, CT 2 Departments of Emergency Medicine and Neurosurgery, Renaissance School of Medicine at Stony Brook, Stony Brook, NY 03 1 2022 5 2022 03 1 2022 50 5 883885 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. cardiopulmonary resuscitation COVID medical ethics resuscitation ==== Body pmcIncluding the words “no survivors” in the title of a critical care research study effectively grabs the attention of readers. It is also likely to elicit strong reactions from those that have been on the frontlines of the COVID pandemic for the past 2 years. It may reinforce the frustration and exasperation felt by those that know first hand how deadly this infectious disease can be. Alternatively, it may offer evidence to those considering the ethical dilemma of attempting resuscitation for out-of-hospital cardiac arrest (OHCA) in acutely infected patients. Regardless of headlines, the study by Baert et al (1) published in this issue of Critical Care Medicine does call into question the futility of resuscitation in COVID patients if the chance of patient survival is zero. Unfortunately, those seeking a definitive answer to the question “Should resuscitation even be attempted in COVID patients with OHCA?” will need to wait for more evidence. Prior to the study by Baert et al (1), there has been plenty of data showing an increase in the number of OHCA during the early pandemic compared with 2019, with a concurrent decrease in survival. This phenomena was seen in Paris (2), New York (3), and Lombardy (4). A meta-analysis of these reports found nearly two times higher odds of admission for OHCA during the pandemic (5). In these studies, it was difficult to discern the direct effect of COVID on the increased mortality. Collectively, these studies allude to a multitude of factors of a strained healthcare system. Delayed emergency medical services (EMS) response from an overworked system, patients delaying care over fear of coming to the hospital, alterations in advanced life support (ALS) protocols and decreased bystander cardiopulmonary resuscitation due to fear of infection, psychologic stress from the pandemic and associated lockdowns, reorganization of healthcare systems, and rationing of care all potentially played a role in increased mortality in OHCA patients. Chan et al (6) explored this further in the United States and showed that rates of sustained return of spontaneous circulation (ROSC) for patients with OHCA fell across all counties during the first wave of the pandemic from 29.8% to 23.0% with corresponding fall in survival to discharge from 9.8% to 6.6%. Interestingly, survival was not significantly affected in areas of otherwise low COVID mortality, supporting the suggestions that there are direct effects from COVID as well as confounding factors beyond the virus itself leading to poor survival rates (6). The objective of the study by Baert et al (1) was to primarily describe the 30-day survival rate of confirmed-COVID patients after experiencing an OHCA between March 2020 and December 2020 in France. Secondarily, the authors compare those known to be infected, those suspected of being infected, and those known to be negative. Data were extracted from the French National OHCA Registry. Six thousand six hundred twenty-four patients were included from this registry with 1.9% confirmed to have COVID and another 7.1% suspected cases. Notably, there was not a significant difference between ROSC and survival to hospital admission between the three groups. However, zero patients with confirmed COVID ultimately survived to day 30 post-ROSC, compared with 3.5% in the non-COVID patients. As the authors point out, it is highly unusual to have a survival rate of zero in a cohort of patients with OHCA. Rightfully so, the authors state that this finding raises the issue of resuscitation futility in patients with COVID experiencing an OHCA. There are findings that support the conclusion that the disease itself was directly responsible for the dismal lack of survival: most of the confirmed-COVID patients had preceding respiratory distress (53.7%) and many had a history of respiratory disease. These results are supported by known evidence that the virus has higher mortality for those with comorbid conditions (7). The authors correctly point out that acute respiratory distress syndrome, of any etiology, can lead to hypoxemic cardiac arrest. Hypoxemia is classically a “reversible” cause of cardiac arrest, even in patients with COVID, but may not be “reversible” in the prehospital setting or in advanced disease states. COVID is also peculiar in its ability to cause silent hypoxemia (8), leading to delayed presentations in some individuals and sudden cardiac arrest. It is also worth noting that other “reversible” etiologies of cardiac arrest such as myocardial infarction and pulmonary embolism can occur at higher rates in patients with COVID (9). Both the authors and the healthcare workers that provided care in these challenging times are to be commended for their efforts in advancing patient care through research and clinical practice. All studies of this nature have limitations, nevermind the challenges of the COVID pandemic. There are prehospital system factors that the authors point out that may contribute to some of this study's findings. Patients with confirmed COVID were less likely to be intubated, less likely to receive ALS, and less likely to have an automated external defibrillator used. As the authors note, this could explain some of the difference in survival in addition to the lethality of COVID. It is not explicitly clear why these intervention rates were different but it can be inferred that some of these differences are potentially related to provider safety concerns and perceived futility. The authors note several other limitations: the French model uses a two-tiered EMS model, with the first tier being basic life support (BLS) providers and the second tier being Emergency Medicine physicians with ALS and intubation capabilities. It is difficult to determine if earlier access to ALS and endotracheal intubation in a different EMS system would make a difference in patient outcomes. ALS versus BLS has not been shown to improve outcomes in OHCA previously (10). Whether this is the case in COVID-related arrests where there is a large burden of respiratory pathology is hard to determine from the study by Baert et al (1). The most important limitation was the lack of knowledge about how patients were managed in the hospital, with regards to withdrawal of care in particular. However, these perceived limitations are real world issues that are reflective of current everyday medical practice in many healthcare systems. The study by Baert et al (1) is important. It is the first to our knowledge to specifically describe the survival rate of patients suffering an OHCA known to be infected with COVID. Although COVID patients had similar rates of ROSC and admission to the hospital, none of these patients with confirmed COVID survived at 30 days. The findings certainly reinforce the known lethality of severe acute respiratory syndrome coronavirus 2. With the healthcare systems across the world stretched to the breaking point, resource utilization and workforce safety remain high priorities that must be considered when weighing the cost versus benefit of these resuscitations. In surge situations where critical care resources are being rationed, the study by Baert et al (1) may inform difficult resource allocation decisions. However, this does not mean that patients with COVID that experience OHCA cannot survive. This is a leap too far—based on the study by Baert et al (1) alone. To conclude that attempting resuscitation is futile based on this data ignores the fact that there are likely confounding factors that also contributed to the higher mortality, many that are related to the indirect effects of this novel pandemic just as much as the disease process at the individual level. Similar results of a near-zero survival (1/471 patients) in OHCA were seen in a study conducted in Detroit, Michigan, in 2002 (11). A focus on modifiable public health factors and EMS initiatives has improved the rate of survival to hospital discharge to 6.4% in OHCA in 2016 (12). Whether a similar improvement can be seen in COVID patients with OHCA remains to be seen. This data need to be replicated in other healthcare systems before resuscitation in OHCA with COVID can be deemed futile. As COVID looks like it will be with us for the foreseeable future, we hope that this futility is not the case. Until further studies prove or disprove the work by Baert et al (1), the decision to perform resuscitation in OHCA from suspected or confirmed COVID should be based on individual patient, arrest, and health system factors similar to non-COVID cardiac arrests with a trial of aggressive critical care when appropriate, timely multimodality prognostication, and palliative care when aggressive critical care is no longer indicated. Healthcare workers should have access to adequate personal protective equipment and vaccinations to perform resuscitations safely. Finally, it is safe to say that preventing OHCA with widespread vaccinations and close follow-up for those known to be infected should remain the primary strategy to avoid this increasingly common tragic scenario for patients, families, and healthcare providers. *See also 791. The authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Baert V Beuscart J-B Recher M ; French National OHCA Registry (RéAC) Study Group: Coronavirus Disease 2019 and Out-of-Hospital Cardiac Arrest: No Survivors. Crit Care Med 2022; 50 :791–798 2. Marijon E Karam N Jost D : Out-of-hospital cardiac arrest during the COVID-19 pandemic in Paris, France: A population-based, observational study. Lancet Public Health. 2020; 5 :e437–e443 32473113 3. Lai PH Lancet EA Weiden MD : Characteristics associated with out-of-hospital cardiac arrests and resuscitations during the novel coronavirus disease 2019 pandemic in New York City. JAMA Cardiol. 2020; 5 :1154–1163 32558876 4. Baldi E Sechi GM Mare C : Out-of-hospital cardiac arrest during the Covid-19 outbreak in Italy. N Engl J Med. 2020; 383 :496–498 32348640 5. Singh S Fong HK Mercedes BR : COVID-19 and out-of-hospital cardiac arrest: A systematic review and meta-analysis. Resuscitation. 2020; 156 :164–166 32946986 6. Chan PS Girotra S Tang Y : Outcomes for out-of-hospital cardiac arrest in the United States during the coronavirus disease 2019 pandemic. JAMA Cardiol. 2021; 6 :296–303 33188678 7. Sanyaolu A Okorie C Marinkovic A : Comorbidity and its impact on patients with COVID-19. SN Compr Clin Med 2020 Jun 25. [online ahead of print] 8. Brouqui P Amrane S Million M : Asymptomatic hypoxia in COVID-19 is associated with poor outcome. Int J Infect Dis. 2021; 102 :233–238 33130200 9. Klok FA Kruip MJHA van der Meer NJM : Incidence of thrombotic complications in critically ill ICU patients with COVID-19. Thromb Res. 2020; 191 :145–147 32291094 10. Stiell IG Wells GA Field B : Ontario Prehospital Advanced Life Support Study Group: Advanced cardiac life support in out-of-hospital cardiac arrest. N Engl J Med. 2004; 351 :647–656 15306666 11. Dunne RB Compton S Zalenski RJ : Outcomes from out-of-hospital cardiac arrest in Detroit. Resuscitation. 2007; 72 :59–65 17113209 12. May S Zhang L Foley D : Improvement in non-traumatic, out-of-hospital cardiac arrest survival in detroit from 2014 to 2016. J Am Heart Assoc. 2018; 7 :e009831 30369308
PMC009xxxxxx/PMC9005099.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 35148524 00012 10.1097/CCM.0000000000005462 3 Clinical Investigations Ventilator-Associated Lower Respiratory Tract Bacterial Infections in COVID-19 Compared With Non-COVID-19 Patients* Hedberg Pontus MD 12 Ternhag Anders MD, PhD 12 Giske Christian G. MD, PhD 34 Strålin Kristoffer MD, PhD 15 Özenci Volkan MD, PhD 34 Johansson Niclas MD, PhD 12 Spindler Carl MD, PhD 12 Hedlund Jonas MD, PhD 12 Mårtensson Johan MD, PhD 67 Nauclér Pontus MD, PhD 13 1 Department of Infectious Diseases, Karolinska University Hospital, Stockholm, Sweden. 2 Department of Medicine Solna, Karolinska Institutet, Stockholm, Sweden. 3 Department of Clinical Microbiology, Karolinska University Hospital, Stockholm, Sweden. 4 Department of Laboratory Medicine, Karolinska Institutet, Stockholm, Sweden. 5 Department of Medicine Huddinge, Karolinska Institutet, Stockholm, Sweden. 6 Department of Physiology and Pharmacology, Karolinska Institutet, Stockholm, Sweden. 7 Department of Perioperative Medicine and Intensive Care, Karolinska University Hospital, Stockholm, Sweden. For information regarding this article, E-mail: pontus.naucler@ki.se 14 2 2022 5 2022 14 2 2022 50 5 825836 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. OBJECTIVES: Ventilator-associated lower respiratory tract infections (VA-LRTIs) are associated with prolonged length of stay and increased mortality. We aimed to investigate the occurrence of bacterial VA-LRTI among mechanically ventilated COVID-19 patients and compare these findings to non-COVID-19 cohorts throughout the first and second wave of the pandemic. DESIGN: Retrospective cohort study. SETTING: Karolinska University Hospital, Stockholm, Sweden. PATIENTS: All patients greater than or equal to 18 years treated with mechanical ventilation between January 1, 2011, and December 31, 2020. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The cohort consisted of 20,223 ICU episodes (479 COVID-19), with a VA-LRTI incidence proportion of 30% (129/426) in COVID-19 and 18% (1,081/5,907) in non-COVID-19 among patients ventilated greater than or equal to 48 hours. The median length of ventilator treatment for COVID-19 patients was 10 days (interquartile range, 5–18 d), which was significantly longer than for all other investigated specific diagnoses. The VA-LRTI incidence rate per 1,000 ventilator days at risk was 31 (95% CI, 26–37) for COVID-19 and 34 (95% CI, 32–36) for non-COVID-19. With COVID-19 as reference, adjusted subdistribution hazard ratios for VA-LRTI was 0.29–0.50 (95% CI, < 1) for influenza, bacterial pneumonia, acute respiratory distress syndrome, and severe sepsis, but 1.38 (95% CI, 1.15–1.65) for specific noninfectious diagnoses. Compared with COVID-19 in the first wave of the pandemic, COVID-19 in the second wave had adjusted subdistribution hazard ratio of 1.85 (95% CI, 1.14–2.99). In early VA-LRTI Staphylococcus aureus was more common and Streptococcus pneumoniae, Haemophilus influenzae, and Escherichia coli less common in COVID-19 patients, while Serratia species was more often identified in late VA-LRTI. CONCLUSIONS: COVID-19 is associated with exceptionally long durations of mechanical ventilation treatment and high VA-LRTI occurrence proportions. The incidence rate of VA-LRTI was compared with the pooled non-COVID-19 cohort, however, not increased in COVID-19. Significant differences in the incidence of VA-LRTI occurred between the first and second wave of the COVID-19 pandemic. artificial respiration bacterial infections COVID-19 critical care severe acute respiratory syndrome coronavirus 2 ventilator-associated pneumonia SDCT ==== Body pmcCOVID-19 continues to exert a tremendous pressure on healthcare systems worldwide. The number of COVID-19 patients in need of ICUs varies between countries and time of the pandemic and is estimated to be around 10–30% of hospital-admitted patients, with 15–20% of these receiving ventilatory support (1–4). Ventilator-associated lower respiratory tract infection (VA-LRTI) encompasses ventilator-associated tracheobronchitis (VAT) and ventilator-associated pneumonia (VAP), where the presence of new or progressive infiltrates on chest radiography distinguishes the two (5, 6). Difficulties preclude accurate diagnosis of VAP, involving subjectivity and interobserver variability with regards to the presence of new or worsening infiltrates and concomitant lung parenchyma invasion (7, 8). In COVID-19, this is further aggravated by the similar clinical presentation, involving fever, leukocytosis, and extensive radiographic infiltrates. Microbiological findings from the lower respiratory tract (LRT) has also been considered the sole reliable criterion to support a VAP diagnosis in COVID-19 patients (9). Previous reports have found an increased risk of VAP and VA-LRTI in critically ill patients with COVID-19, with high reported incidence proportions ranging from 29% to 86% (10–16). The clinical management of COVID-19 has since the beginning of the pandemic undergone major changes, with use of corticosteroid therapy for severely ill patients and an increased use of prone positioning, noninvasive ventilation, and anticoagulants warranting comparison of VA-LRTI between different stages of the pandemic. The aim of this study was to investigate the occurrence of microbiologically defined bacterial VA-LRTI among mechanically ventilated COVID-19 patients during the first 10 months of the pandemic and compare these findings to mechanically ventilated non-COVID-19 patients during and before the pandemic. MATERIALS AND METHODS Patient Population and Study Setting This was a retrospective study in four ICUs at Karolinska University Hospital in Stockholm, Sweden, a tertiary care hospital at two sites with 1,100 beds and a catchment area of 2.3 million inhabitants. Patients greater than or equal to 18 years admitted to the ICU between January 1, 2011, and December 31, 2020, treated with mechanical ventilation during their ICU stay were included in the study cohort. The study was approved by the Swedish Ethical Review Board with a waiver of informed consent (Dnr 2018/1030-31, COVID-19 research amendments Dnr 2020-01385 and Dnr 2020-02145). Data Collection Data were obtained from a data extraction of the Swedish Intensive Care Registry of all patients admitted to Karolinska University Hospital between January 2010 and February 2021, including demographics, admission reasons, descriptives of the ICU stay, and discharge status. Further, a database of electronic health records (EHRs) was used for extraction of International Classification of Diseases, 10th Revision (ICD-10) codes, mortality data, and microbiology. Comorbidities were based on ICD-10 codes recorded from hospital care up to 5 years before admission (Table S1, http://links.lww.com/CCM/G989). For admission vital signs and laboratory parameters, the most deviating value during the first 24 hours of the ICU stay was registered. Use and duration of mechanical ventilation as well as prone positioning were identified using nationally harmonized procedure codes. Definitions Given the extensive bilateral radiographic infiltrates commonly observed in critically ill COVID-19 patients, the presence of new or worsening infiltrates was deemed difficult to assess retrospectively and thus not assessed (9). As such, ascertaining whether the VA-LRTI was a VAP or VAT was not possible, and therefore the outcome was referred to as VA-LRTI only. The definition of VA-LRTI was based strictly on microbiological criteria, with a positive microbiological isolation of at least 105 colony-forming units (CFUs) per mL in tracheal and bronchial secretions, 104 CFU per mL in bronchoalveolar lavage or 103 CFU per mL in protected brush specimens (5, 17). All quantitative microbiological cultures from 48 hours after intubation until extubation were considered. At Karolinska University Hospital, surveillance LRT cultures are not performed, but rather cultures are performed on clinical suspicion of pneumonia. Only significant bacterial pathogens were included in the analyses, excluding organisms rarely causing manifest VA-LRTI (Table S2, http://links.lww.com/CCM/G989) (18, 19). Only new findings were considered, that is, if the same pathogen was detected at significant levels from the time of hospitalization until 48 hours after intubation, it was not considered a VA-LRTI. In a predefined sensitivity analysis, VA-LRTI was restricted to the presence of fever or a leukocyte count greater than 12,000 or less than 4,000 cells per μL, 1 day before up to 1 day after the time of the microbiological sampling (10). Given the retrospective study design, we considered the documentation of purulent sputum to be inadequate and therefore excluded this criterion. Only the first ventilator episode per ICU admission was included, where we considered intubations with more than 48 hours since a preceding extubation as separate ventilator episodes. Statistical Methods The incidence of VA-LRTI in COVID-19 was compared with all non-COVID-19 patients as well as to influenza and the 10 most common mutually exclusive specific main diagnoses given by the ICU physician (Table S1, http://links.lww.com/CCM/G989). For comparisons of different phases of the COVID-19 pandemic, the study time period March 1, 2020, until July 31, 2020, represented the first wave and October 1, 2020, until December 31, 2020, the second wave. Continuous variables were summarized as medians and interquartile ranges (IQRs), and categorical variables were summarized as numbers and percentages, with testing of significance using chi-square test and Wilcoxon signed-rank test. Statistical testing was two-sided, with p value of less than 0.05 considered significant. The incidence of VA-LRTI was studied using a competing-risks analysis, with extubation (dead or alive) as a competing event (10, 20). Follow-up included the entire ICU stay (or up to 30 d for mortality if ICU stay was shorter). Incidence proportions and rates (per 1,000 ventilator days at risk) of first episodes of VA-LRTI were calculated. To avoid immortal time (i.e., the time period where the patient can not develop VA-LRTI), only the time interval from 48 hours after intubation until the occurrence of VA-LRTI or extubation was included in the incidence rate denominator. The cumulative incidence of first episodes of VA-LRTI was estimated using the Aalen and Johansen estimator (21). Cause-specific hazard ratios (CSHRs) were calculated using Cox proportional hazard models for each event (VA-LRTI or extubation), and Fine and Gray models were performed to calculate subdistribution hazard ratios (sHRs) (22). Multivariable analyses were performed adjusting for age, sex, and Charlson Comorbidity Index (CCI) score. For the models comparing VA-LRTI during the first and second wave in COVID-19 episodes, obesity, prone positioning, and steroid use before ICU admission was also included in the multivariable models. In order to address potential time-related drifts in diagnostic procedures and LRT culture sampling strategies as well as different case-mixes in the non-COVID-19 cohort, as well as the potential biases arising by the use of historical controls, the following predefined sensitivity analyses were performed for the incidence of VA-LRTI, restriction to episodes: 1) admitted from 2017 and onwards, 2) admitted during the pandemic, 3) with a LRT culture sampling performed, 4) with a Pao2/Fio2 ratio less than 13.3 kPa at admission, and 5) with fever or a leukocyte count less than 4,000 or greater than 12,000 cells per μL. All statistical analyses were performed in R Version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). RESULTS In total, 28,347 ICU episodes admitted from January 1, 2011, to December 31, 2020, were recorded, with 680 of these being COVID-19 episodes. The final study cohort consisted of 20,223 episodes from 18,674 persons treated with mechanical ventilation, with 19,744 and 479 non-COVID-19 and COVID-19 episodes, respectively. Patient Characteristics, ICU Admission Status, and Clinical Outcomes COVID-19 patient were, compared with non-COVID-19 patients, younger, more often male, had lower CCI score, more often had diabetes, hypertension, and chronic respiratory disease, but less often had malignancy and immunosuppression (Table 1). COVID-19 patients more often received antibiotics and steroids prior to ICU admission and presented with a better SAPS III but worse Pao2/Fio2. Among non-COVID-19 episodes, the five most common diagnoses were cardiac arrest, nonintracranial injuries, nontraumatic intracranial hemorrhage, sepsis, and intracranial injuries. TABLE 1. Baseline Characteristics and Description of ICU Stay Characteristic Ventilated Ventilated ≥ 48 hr COVID-19 (n = 479) Non-COVID-19 (n = 19,744) p COVID-19 (n = 426) Non-COVID-19 (n = 5,907) p Age at ICU admission (yr) 60 (52–68) 64 (51–72) < 0.001 60 (52–68) 62 (49–71) 0.079 18–44 62 (13) 3,315 (17) < 0.001 47 (11) 1,128 (19) 45–64 251 (52) 7,054 (36) 230 (54) 2,165 (37) < 0.001 ≥ 65 166 (35) 9,375 (47) 149 (35) 2,614 (44) Male sex 370 (77) 13,296 (67) < 0.001 330 (77) 3,836 (65) < 0.001 Charlson Comorbidity Index (points) 1 (0–2) 1 (0–2) < 0.001 1 (0–2) 1 (0–3) < 0.001 0–1 345 (72) 11,789 (60) < 0.001 302 (71) 3,088 (52) 2–4 115 (24) 6,227 (32) 108 (25) 2,114 (36) < 0.001 ≥ 5 19 (4) 1,728 (9) 16 (4) 705 (12) Diabetes mellitus 136 (28) 3,369 (17) < 0.001 125 (29) 986 (17) < 0.001 Hypertension 221 (46) 7,700 (39) 0.002 201 (47) 2,098 (36) < 0.001 Chronic lower respiratory disease 79 (16) 1,951 (10) < 0.001 71 (17) 725 (12) 0.010 Chronic kidney disease 16 (3) 1,128 (6) 0.034 15 (4) 448 (8) 0.003 Malignancy 31 (6) 2,809 (14) < 0.001 28 (7) 1,081 (18) < 0.001 Immunosuppression 53 (11) 3,938 (20) < 0.001 48 (11) 1,540 (26) < 0.001 Time in hospital before ICU admission (d)a 3 (1–6) 1 (0–2) < 0.001 3 (1–6) 1 (0–4) < 0.001 Location before ICU admission Emergency department 56 (12) 3,495 (18) < 0.001 45 (11) 1,349 (23) < 0.001 Ward 299 (62) 3,878 (20) 276 (65) 1,960 (33) Surgical 12 (3) 10,609 (54) 6 (1) 1,567 (27) Other hospital 9 (2) 622 (3) 8 (2) 308 (5) Other ICU 103 (22) 1,140 (6) 91 (21) 723 (12) Antibiotic treatment before ICU admissionb 189 (39) 6,594 (33) 0.006 173 (41) 2,153 (36) 0.095 Steroid treatment before ICU admissionc 175 (37) 2,695 (14) < 0.001 163 (38) 998 (17) < 0.001 Simplified Acute Physiology Score III (points)d 57 (50–65) 62 (50–73) < 0.001 57 (50–65) 66 (55–75) < 0.001 Pao2/Fio2 (kPa)e 12 (9–19) 33 (20–50) < 0.001 12 (10–18) 28 (17–45) < 0.001 > 40 25 (5) 4,439 (38) < 0.001 16 1,684 26.8–40 33 (7) 2,734 (23) 27 1,172 < 0.001 13.4–26.7 151 (32) 3,002 (26) 133 1,657 ≤ 13.3 263 (56) 1,536 (13) 244 900 Prone positioning 264 (55) 116 (1) < 0.001 258 (61) 92 (2) < 0.001 ICU length of stay (d) 14 (8–22) 2 (1–6) < 0.001 15 (9–24) 8 (5–15) < 0.001 Length of ventilation (d) 10 (5–18) 1 (0–3) < 0.001 11 (7–20) 6 (3–11) < 0.001 Ventilated ≥ 48 hr 426 (89) 5,907 (30) < 0.001 — — — Mortality within 48 hr of intubation 13 (3) 1,073 (5) 0.012 — — — ICU mortality 123 (26) 2,176 (11) < 0.001 110 (26) 1,042 (18) < 0.001 30-d mortality 114 (24) 3,055 (15) < 0.001 98 (23) 1,503 (25) 0.289 a Ventilated: 7,521 missing values (COVID-19, n = 1; non-COVID-19, n = 7,520), ventilated ≥ 48 hr: 329 missing values (COVID-19, n = 0; non-COVID-19, n = 329). b Defined as any antibiotic administered (Anatomical Therapeutic Chemical [ATC] codes of J01) from 7 d before up until ICU admission. c Defined as any systemic corticosteroid (ATC codes of H02) administered from 7 d before up until ICU admission. d Ventilated: 7,521 missing values (COVID-19, n = 1; non-COVID-19, n = 7,520), ventilated ≥ 48 hr: 329 missing values (COVID-19, n = 0; non-COVID-19, n = 329). e Ventilated: 8,040 missing values (COVID-19, n = 7; non-COVID-19, n = 8,033), ventilated ≥ 48 hr: 500 missing values (COVID-19, n = 6; non-COVID-19, n = 494). Data are presented as absolute frequency (% of the included episodes) or as median and interquartile range. χ2 test was used for categorical variables and Wilcoxon signed-rank test for continuous variables. COVID-19 patients had a median length of stay (LOS) in the ICU of 14 days (IQR, 8–22 d) compared with 2 days (IQR, 1–6 d) in non-COVID-19 patients. The ICU and 30-day mortality were 26% (123/479) and 24% (114/479) among COVID-19 patients, compared with 11% (2,176/19,744) and 15% (3,055/19,744) in non-COVID-19 patients, respectively. Mechanical Ventilation Characteristics Out of the 20,223 investigated ventilator episodes, 6,333 (31%) were lasting for 48 hours or more, thus being at risk for VA-LRTI (Table 1). Upon comparison of COVID-19 episodes with the 10 most common non-COVID-19 diagnoses as well as influenza, COVID-19 had the longest median ventilator treatment duration, 10 days (IQR, 5–18 d), followed by acute respiratory distress syndrome (ARDS), severe sepsis, and bacterial pneumonia; 5 days (IQR, 2–11 d), 4 days (IQR, 1–7 d), and 4 days (IQR, 2–7 d), respectively (Fig. 1A). Figure 1. Mechanical ventilation treatment durations and ventilator-associated lower respiratory tract infections (VA-LRTIs) in COVID-19 and non-COVID-19 conditions. A, Duration of mechanical ventilation for COVID-19, influenza and the 10 most common specific main diagnoses given by the ICU physician. B, Cumulative risk (incidence) of VA-LRTIs in COVID-19 and non-COVID-19 episodes treated with mechanical ventilation for at least 48 hr, considering extubation (dead or alive) as a competing event. ARDS = acute respiratory distress syndrome. The proportion testing positive for a significant pathogen before intubation as well as during the first 48 hours of intubation did not differ between COVID-19 and non-COVID-19 episodes (Table S3, http://links.lww.com/CCM/G989). Among the 426 COVID-19 and 5,907 non-COVID-19 episodes ventilated greater than or equal to 48 hours, 65% (277/426) and 53% (3,108/5,907) had a LRT culture performed from 48 hours of intubation and onwards, respectively. For both COVID-19 and non-COVID-19, LRT cultures were most often sampled from tracheal secretion 52% (223/426) and 38% (2,261/5,907), respectively. Nontraumatic intracranial hemorrhage and aortic aneurysm and dissection were the conditions having a LRT culture performed most often (Fig. S1, http://links.lww.com/CCM/G989). Incidence of VA-LRTI In COVID-19, 30% (129/426) had a VA-LRTI (LRT culture positive for a new significant pathogen greater than or equal to 48 hr after start of mechanical ventilation), whereas in non-COVID-19, the corresponding proportion was 18% (1,081/5,907) (Table 2). The VA-LRTI incidence rate per 1,000 ventilator days at risk was 31 (95% CI, 26–37) for COVID-19, compared with 34 (95% CI, 32–36) for non-COVID-19. Among specific non-COVID-19 diagnoses, severe sepsis, ARDS, influenza, and bacterial pneumonia had the lowest incidence rate, whereas nontraumatic intracranial hemorrhage, aortic aneurysm and dissection, and heart failure had the highest incidence rate. When pooling specific infectious non-COVID-19 diagnoses, the incidence rate was 11 (95% CI, 8–15), whereas for noninfectious diagnoses, the incidence rate was 47 (95% CI, 43–50). TABLE 2. Incidence of Ventilator-Associated Lower Respiratory Tract Infections in COVID-19 and Non-COVID-19 Cohort Ventilated, n Ventilated ≥ 48 hr, n (%) No. Ventilator-Associated Lower Respiratory Tract Infection, n (%) Ventilator Days at Risk Incidence Rate per 1,000 Days at Risk (95% CI) COVID-19 479 426 (89) 129 (30) 4,160 31 (26–37) Non-COVID-19 19,744 5,907 (30) 1,081 (18) 31,895 34 (32–36) Acute respiratory distress syndrome 381 295 (77) 22 (7) 2,068 11 (7–16) Infectious diseasesa 1,103 760 (69) 52 (7) 4,577 11 (8–15) Bacterial pneumonia 212 147 (69) 13 (9) 771 17 (9–29) Influenza 134 108 (81) 7 (6) 663 11 (4–22) Severe sepsis 757 505 (67) 32 (6) 3,143 10 (7–14) Noninfectious diseasesb 6,242 2,901 (46) 769 (27) 16,529 47 (43–50) Acute renal failure 250 39 (16) 6 (15) 115 52 (19–114) Aortic aneurysm and dissection 563 123 (22) 39 (32) 646 60 (43–83) Cardiac arrest 1,432 698 (49) 101 (14) 2,279 44 (36–54) Heart failure 429 99 (23) 23 (23) 402 57 (36–86) Injuries, intracranial 637 269 (42) 75 (28) 1,561 48 (38–60) Injuries, others 1,300 628 (48) 147 (23) 3,911 38 (32–44) Nontraumatic intracranial hemorrhage 1,152 619 (54) 249 (40) 3,457 72 (63–82) First vs second wave COVID-19 First wave 381 345 (91) 99 (29) 3,586 28 (22–34) Second wave 93 80 (86) 30 (38) 572 52 (35–75) Non-COVID-19 First wave 567 194 (34) 37 (19) 1,013 37 (26–50) Second wave 324 131 (40) 28 (21) 542 52 (34–75) a Included bacterial pneumonia, influenza, and severe sepsis. Acute respiratory distress syndrome (ARDS) was excluded as information on infectious or noninfectious etiology was absent. b Included acute renal failure, aortic aneurysm and dissection, cardiac arrest, heart failure, intracranial injuries, other injuries, and nontraumatic intracranial hemorrhage. ARDS was excluded as information on infectious or noninfectious etiology was absent. The incidence proportion and incidence rate of ventilator-associated lower respiratory tract infection in COVID-19, all non-COVID-19, influenza as well as the 10 most common specific diagnoses given by the ICU physician is presented. The incidence during the first and second week of the COVID-19 pandemic is compared in the COVID-19 and non-COVID-19 group. The higher incidence proportion but lower incidence rates in the COVID-19 cohort compared with the non-COVID-19 cohort were consistent in sensitivity analyses restricted to episodes from 2017 and onwards and during the COVID-19 pandemic, respectively, as well as when only including episodes with LRT culture performed and restricting VA-LRTIs to episodes with fever or leukocyte count alterations (Table S4, http://links.lww.com/CCM/G989). The probability of extubation without VA-LRTI was lower in COVID-19 compared with non-COVID-19, including all specific diagnoses (Fig. 1B and Table 3). No significant difference in the cumulative risk of VA-LRTI was observed between the non-COVID-19 and COVID-19 cohort. These findings were consistent across all sensitivity analyses, besides restriction to episodes with Pao2/Fio2 less than 13.3 kPa at admission, where the risk of VA-LRTI was decreased in non-COVID-19 (adjusted CSHR [aCSHR], 0.54; 95% CI, 0.40–0.73) (Table S5, http://links.lww.com/CCM/G989). Further, the results were consistent when adjusting for the specific comorbidities presented in Table 1 instead of CCI (data not shown). The cumulative risk of VA-LRTI was, compared with COVID-19, decreased for all specific infectious diagnoses (bacterial pneumonia, influenza, and sepsis) as well as ARDS, whereas for aortic aneurysm and dissection, intracranial injuries, and nontraumatic intracranial hemorrhage, the risk was increased. TABLE 3. Ventilator-Associated Lower Respiratory Tract Infections Cause-Specific Hazard Ratios and Subdistribution Hazard Ratios Cohort Possible Endpoints (Competing Events) VA-LRTI Extubation (Without VA-LRTI) CSHR (95% CI) Crude CSHR (95% CI) Adjusteda SHR (95% CI) Crude SHR (95% CI) Adjusteda CSHR (95% CI) Crude CSHR (95% CI) Adjusteda COVID-19 Reference Reference Reference Reference Reference Reference Non-COVID-19 0.96 (0.80–1.15) 0.98 (0.82–1.18) 0.96 (0.81–1.14) 0.98 (0.82–1.17) 1.99 (1.77–2.24) 2.02 (1.79–2.27) Acute respiratory distress syndrome 0.34 (0.22–0.54) 0.33 (0.20–0.53) 0.34 (0.22–0.54) 0.33 (0.20–0.53) 1.88 (1.59–2.21) 1.82 (1.52–2.18) Infectious diseasesb 0.36 (0.26–0.50) 0.34 (0.24–0.48) 0.36 (0.26–0.50) 0.34 (0.24–0.48) 2.13 (1.86–2.44) 2.28 (1.97–2.64) Bacterial pneumonia 0.51 (0.29–0.91) 0.50 (0.28–0.91) 0.51 (0.29–0.90) 0.50 (0.28–0.92) 2.55 (2.07–3.14) 2.63 (2.10–3.30) Influenza 0.33 (0.16–0.71) 0.32 (0.15–0.70) 0.33 (0.17–0.67) 0.32 (0.16–0.66) 2.15 (1.71–2.70) 2.15 (1.69–2.74) Severe sepsis 0.33 (0.22–0.48) 0.29 (0.19–0.45) 0.33 (0.22–0.48) 0.29 (0.19–0.43) 2.10 (1.82–2.44) 2.26 (1.92–2.66) Noninfectious diseasesc 1.41 (1.16–1.70) 1.38 (1.14–1.67) 1.40 (1.18–1.68) 1.38 (1.15–1.65) 1.98 (1.75–2.24) 2.08 (1.83–2.36) Acute renal failure 1.55 (0.68–3.55) 1.49 (0.63–3.53) 1.55 (0.70–3.41) 1.49 (0.65–3.46) 4.56 (3.13–6.64) 4.36 (2.93–6.50) Aortic aneurysm and dissection 1.86 (1.30–2.67) 1.81 (1.21–2.70) 1.86 (1.26–2.74) 1.81 (1.17–2.79) 1.83 (1.43–2.34) 2.12 (1.61–2.80) Cardiac arrest 1.14 (0.87–1.50) 1.14 (0.86–1.50) 1.14 (0.88–1.48) 1.13 (0.87–1.48) 3.30 (2.86–3.82) 3.36 (2.90–3.90) Heart failure 1.69 (1.07–2.65) 1.54 (0.95–2.49) 1.68 (1.07–2.65) 1.54 (0.97–2.45) 2.89 (2.23–3.76) 2.79 (2.11–3.70) Injuries, intracranial 1.45 (1.08–1.93) 1.42 (1.05–1.92) 1.45 (1.08–1.93) 1.42 (1.05–1.92) 1.81 (1.50–2.18) 1.90 (1.57–2.30) Injuries, others 1.11 (0.87–1.41) 1.20 (0.93–1.55) 1.11 (0.88–1.40) 1.20 (0.93–1.55) 1.74 (1.50–2.01) 1.82 (1.56–2.14) Nontraumatic intracranial hemorrhage 2.11 (1.70–2.62) 2.05 (1.63–2.58) 2.11 (1.71–2.60) 2.05 (1.65–2.54) 1.54 (1.32–1.80) 1.60 (1.35–1.89) First vs second wave COVID-19d First wave Reference Reference Reference Reference Reference Reference Second wave 1.85 (1.22–2.79) 1.86 (1.15–3.01) 1.85 (1.23–2.77) 1.85 (1.14–2.99) 1.36 (1.00–1.84) 1.70 (1.19–2.43) Non-COVID-19 First wave Reference Reference Reference Reference Reference Reference Second wave 1.34 (0.81–2.19) 1.37 (0.83–2.25) 1.34 (0.82–2.18) 1.37 (0.84–2.24) 1.17 (0.91–1.51) 1.16 (0.90–1.49) CSHR = cause-specific hazard ratio, SHR = subdistribution hazard ratio, VA-LRTI = ventilator-associated lower respiratory tract infection. a Adjusted for age (continuous), sex (categorical), and Charlson Comorbidity Index score (continuous). b Included bacterial pneumonia, influenza, and severe sepsis. Acute respiratory distress syndrome (ARDS) was excluded as information on infectious or noninfectious etiology was absent. c Included acute renal failure, aortic aneurysm and dissection, cardiac arrest, heart failure, intracranial injuries, other injuries, and nontraumatic intracranial hemorrhage. ARDS was excluded as information on infectious or noninfectious etiology was absent. d Adjusted for age (continuous), sex (categorical), and Charlson Comorbidity Index score (continuous), obesity (categorical), prone positioning (categorical), and steroids before ICU admission (categorical). Two episodes were excluded from the analysis due to missing information about obesity. Competing risk analysis of VA-LRTI and extubation (dead or alive) using Cox proportional hazard models (CSHRs) and Fine and Gray models (SHRs). VA-LRTI During the First and Second Wave of the COVID-19 Pandemic In the COVID-19 cohort, 381 and 93 episodes were registered during the first (March 9, 2020, to July 31, 2020) and second (October 1, 2020, to December 31, 2020) wave, respectively (Table S6, http://links.lww.com/CCM/G989). Patients during the second wave were older, had higher prevalence of diabetes and hypertension, and were more likely to have received systemic corticosteroids before ICU admission and being treated with prone positioning during the ICU stay. The length of mechanical ventilation was shorter during the second wave (8 d; IQR, 5–16 d) compared with the first wave (11 d; IQR, 6–19 d). A higher proportion of mortality within 24 hours of extubation was observed during the second wave (32%, 30/93) compared with the first wave (20%, 76/381). In the non-COVID-19 cohort, 567 and 324 ventilator episodes were registered during the first and second wave, respectively. No major differences were observed for age, sex, ICU LOS, LRT culture performed, and 30-day mortality between the waves. Among COVID-19 episodes ventilated greater than or equal to 48 hours, the VA-LRTI incidence proportion was 29% (99/345) during the first wave and 38% (30/80) during the second wave (Table 2). For non-COVID-19, the VA-LRTI incidence proportion was 19% (37/194) during the first wave and 21% (28/131) during the second wave. For COVID-19, the VA-LRTI incidence rate per 1,000 ventilator days at risk was 28 (95% CI, 22–34) for the first wave and 52 (95% CI, 35–75) for the second wave. For non-COVID-19, the corresponding VA-LRTI incidence were 37 (95% CI, 26–50) and 52 (95% CI, 34–75), respectively. For the COVID-19 cohort, the rate of VA-LRTI in subjects who were still event-free were significantly increased during the second wave (aCSHR, 1.86; 95% CI, 1.15–3.01). The adjusted SHR (aSHR) during the second compared with first wave among COVID-19 patients was 1.85 (95% CI, 1.14–2.99). For the non-COVID-19 cohort, no significant differences were observed for neither extubation nor VA-LRTI. Distribution of Identified VA-LRTI Pathogens Staphylococcus aureus was the most identified pathogen in both COVID-19 and non-COVID-19 patients (Fig. 2). For the COVID-19 cohort, 34% (44/129) of the VA-LRTIs occurred before 5 days of mechanical ventilation, whereas for non-COVID-19 the proportion was 58% (630/1,081). Among the bacterial pathogens identified during less than 5 days of mechanical ventilation, the COVID-19 group had a significantly higher proportion of S. aureus and lower proportion of Haemophilus influenzae and Streptococcus pneumoniae, as well as Escherichia coli compared with the non-COVID-19 group. Among pathogens identified greater than 5 days after mechanical ventilation, COVID-19 patients had a significantly higher proportion of Serratia species identified compared with non-COVID-19 patients. Overall, the three most common pathogens identified among COVID-19 VA-LRTI patients were S. aureus (37%, 48/129), Enterobacter species (15%, 19/129), and Klebsiella species (13%, 17/129) and for non-COVID-19 S. aureus (28%, 299/1,081), H. influenzae (15%, 165/1,081), and Klebsiella species (13%, 142/1,081). Figure 2. Bacterial pathogens identified in COVID-19 and non-COVID-19. Distribution of significant growth of bacterial pathogens in COVID-19 and non-COVID-19 patients for mechanical ventilation day less than or equal to 5 and day greater than 5, respectively. Proportions represent the proportion of all ventilator-associated lower respiratory tract infection (VA-LRTI) per strata (i.e., COVID-19 vs non-COVID-19 and early vs late). All bacterial findings from the first significant culture is presented; thus, the proportion adds up to more than 100%. H. alvei = Hafnia alvei, H. influenzae = Haemophilus influenza, H. parainfluenzae = Haemophilus parainfluenzae, M. catarrhalis = Moraxella catarrhalis, M. morganii = Morganella morganii, S. agalactiae = Streptococcus agalactiae, S. maltophilia = Stenotrophomonas maltophilia, S. mitis = Streptococcus mitis, S. pneumoniae = Streptococcus pneumonia, S. pyogenes = Streptococcus pyogenes, spp. = species. DISCUSSION Herein, we investigated the incidence of bacterial VA-LRTI in COVID-19 as compared with non-COVID-19 patients, and the main findings were: 1) COVID-19 was characterized by the longest mechanical ventilation treatment among all ICU cohorts; 2) the incidence of VA-LRTI in COVID-19 patients, 31 (95% CI, 26–38) per 1,000 ventilator days at risk, was not significantly different compared with the pooled non-COVID-19 cohort, 37 (95% CI, 35–39) per 1,000 ventilator days at risk, but higher than for all investigated infectious diagnoses; 3) the second wave was, compared with the first wave, associated with an increased rate of VA-LRTI in COVID-19, and 4) the microbiological findings differed somewhat between the COVID-19 and non-COVID-19 cohort, yet the pathogen distribution is in line with current empirical treatment recommendations for VAP (23). Previous reports on VA-LRTI and VAP incidence in COVID-19 have observed high incidence proportions, ranging from 29% to 86% (10–16). In studies where incidence rate per 1,000 days of mechanical ventilation have been reported, it has ranged from 18 to 39 per 1,000 ventilator days (11–13, 18). We observed a 30% incidence proportion of VA-LRTI in COVID-19 and an incidence rate of 31 VA-LRTI per 1,000 days of mechanical ventilation at risk. Differences in definitions of VAP and VA-LRTI preclude direct comparison of incident rates from different studies, further impeded by the difficulty in accurately identifying these conditions. A previous prospective, observational study in 114 ICUs before the COVID-19 pandemic reported a 10.2 and 8.8 per 1,000 mechanically ventilated days incidence rate of VAT and VAP, respectively (5). These incidence numbers contrast significantly with the main incidence rates observed in this and other COVID-19–related VAP and VA-LRTI studies, highlighting difficulties in comparison of incidence rates, perhaps in particular during the COVID-19 pandemic. Further, differences in VAP and VA-LRTI incidence differ between different geographical areas and hospitals, with different or no ventilator bundles or decolonization practices in place (19). At Karolinska University Hospital, antimicrobial decolonization is not part of standard procedure. Most clinical phenotypes of COVID-19 in patients admitted to the ICU present with bilateral radiologic infiltrates and severe hypoxemia, leaving the microbiological findings from LRT the most robust criterion to support VAP diagnosis in COVID-19 (9). Further, structural aspects of the provided healthcare and varying pressure on ICU personnel and capacity cannot be excluded as potential reasons for differences in incidence, as well as different uses of prone positioning, systemic corticosteroids, antivirals, and antibiotics. Our finding of an increased VA-LRTI aCSHR as well as aSHR for the COVID-19 cohort during the second wave might be due to several reasons. COVID-19 patients admitted to the ICU during the second wave were older and had more comorbidities. However, a previous multicenter cohort study did not find age to be associated with an increased risk of VAP (24). Further, among COVID-19 patients, prone positioning was more adopted during the second wave of the pandemic. In a study of adults intubated for severe ARDS and allocated either to at least 16 hours of consecutive prone position or standard supine care, there was a higher incidence rate for VA-LRTI among prone positioned patients (25). Further, the use of corticosteroids was more common during the second wave, possibly increasing the risk for VA-LRTI. However, dexamethasone treatment in clinical trials on ARDS was not associated with an increased risk for bacterial pneumonia and the incidence of VA-LRTI has been shown to be significantly lower among immunocompromised patients (6, 26, 27). Further, structural aspects of the provided healthcare and varying pressure on ICU personnel and capacity cannot be excluded as potential reasons. Strengths of our study are the fair sample size with comparison of VA-LRTI between all ventilated ICU patients as well as specific infectious and noninfectious diagnoses with several sensitivity analyses addressing potential confounding factors. Further, we had extensive information about covariates, inclusion of patients from the first and second epidemic waves, detailed information about significant LRT bacterial growth, and consistent results across multiple adjusted analyses. Limitations are the retrospective study design confined to a two-hospital academic center, which might reduce the generalizability of our findings to other hospitals. While we did not use strict criteria of pneumonia, including radiology, symptoms, and laboratory parameters, we only included significant LRT bacterial findings to increase the robustness of our findings (17). Yet, our results were robust in analyses even when restricted to patients with leukocytosis, leukopenia, or fever. Further, LRT microbiological sampling indications might have differed somewhat before and during the pandemic. Due to limited testing rates, we could not investigate the rates of COVID-19–associated pulmonary aspergillosis (CAPA), which in a previous report was observed in 10% of mechanically ventilated COVID-19 patients had CAPA (28). Finally, due to storage in a separate EHR system, we had limited access to data on vital signs and drugs administered during the entire course of the ICU stay, as well as preventive measures, use of sedation, and neuromuscular blocking agents. CONCLUSIONS COVID-19 patients were ventilated for substantially longer durations compared with all other investigated diagnoses. The incidence proportion of VA-LRTI was significantly higher in COVID-19 compared with non-COVID-19 episodes, whereas the incidence rate of VA-LRTI was not increased, although substantial differences between ICU admission diagnoses were observed. Significant differences in the incidence of VA-LRTI occurred between the first and second wave of the COVID-19 pandemic, warranting further investigation with regards to the effect of specific COVID-19 interventions and structural aspects of the provided healthcare. Supplementary Material See also p. 894. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal). Drs. Hedberg and Ternhag shared first authorship. Dr. Hedberg had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Hedberg, Ternhag, Mårtensson, and Nauclér were involved in study concept and design. Dr. Hedberg was involved in acquisition, analysis, or interpretation of data. Drs. Hedberg, Ternhag, and Nauclér were involved in drafting of the article. Dr. Nauclér was involved statistical analysis. All authors involved in critical revision of the article for important intellectual content. Supported, in part, by grants from the Swedish Innovation Agency (Vinnova), Region Stockholm, and Emil och Wera Cornells Stiftelse. Dr. Hedberg received funding from the Swedish Innovation Agency (Vinnova), Region Stockholm, and Emil och Wera Cornell Stiftelse; he was supported by Karolinska Institutet (combined clinical studies and PhD training program). Dr. Ternhag’s institution received funding from the Swedish Innovation Agency (Vinnova). The remaining authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Karagiannidis C Windisch W McAuley DF : Major differences in ICU admissions during the first and second COVID-19 wave in Germany. Lancet Respir Med. 2021; 9 :e47–e48 33684356 2. Strålin K Wahlström E Walther S : Mortality trends among hospitalised COVID-19 patients in Sweden: A nationwide observational cohort study. Lancet Reg Health Eur. 2021; 4 :100054 33997829 3. Karagiannidis C Mostert C Hentschker C : Case characteristics, resource use, and outcomes of 10 021 patients with COVID-19 admitted to 920 German hospitals: An observational study. Lancet Respir Med. 2020; 8 :853–862 32735842 4. Grasselli G Cattaneo E Florio G : Mechanical ventilation parameters in critically ill COVID-19 patients: A scoping review. Crit Care. 2021; 25 :115 33743812 5. Martin-Loeches I Povoa P Rodríguez A ; TAVeM study: Incidence and prognosis of ventilator-associated tracheobronchitis (TAVeM): A multicentre, prospective, observational study. Lancet Respir Med. 2015; 3 :859–868 26472037 6. Moreau AS Martin-Loeches I Povoa P ; TAVeM Study Group: Impact of immunosuppression on incidence, aetiology and outcome of ventilator-associated lower respiratory tract infections. Eur Respir J. 2018; 51 :1701656 29439020 7. Chastre J Luyt CE : Does this patient have VAP? Intensive Care Med. 2016; 42 :1159–1163 26846513 8. Colombo SM Palomeque AC Li Bassi G : The zero-VAP sophistry and controversies surrounding prevention of ventilator-associated pneumonia. Intensive Care Med. 2020; 46 :368–371 31844907 9. François B Laterre PF Luyt CE : The challenge of ventilator-associated pneumonia diagnosis in COVID-19 patients. Crit Care. 2020; 24 :289 32503590 10. Rouzé A Martin-Loeches I Povoa P ; coVAPid study Group: Relationship between SARS-CoV-2 infection and the incidence of ventilator-associated lower respiratory tract infections: A European multicenter cohort study. Intensive Care Med. 2021; 47 :188–198 33388794 11. Blonz G Kouatchet A Chudeau N : Epidemiology and microbiology of ventilator-associated pneumonia in COVID-19 patients: A multicenter retrospective study in 188 patients in an un-inundated French region. Crit Care. 2021; 25 :72 33602296 12. Maes M Higginson E Pereira-Dias J : Ventilator-associated pneumonia in critically ill patients with COVID-19. Crit Care. 2021; 25 :25 33430915 13. Giacobbe DR Battaglini D Enrile EM : Incidence and prognosis of ventilator-associated pneumonia in critically ill patients with COVID-19: A multicenter study. J Clin Med. 2021; 10 :555 33546093 14. Schmidt M Hajage D Demoule A : Clinical characteristics and day-90 outcomes of 4244 critically ill adults with COVID-19: A prospective cohort study. Intensive Care Med. 2021; 47 :60–73 33211135 15. Razazi K Arrestier R Haudebourg AF : Risks of ventilator-associated pneumonia and invasive pulmonary aspergillosis in patients with viral acute respiratory distress syndrome related or not to coronavirus 19 disease. Crit Care. 2020; 24 :699 33339526 16. Luyt CE Sahnoun T Gautier M : Ventilator-associated pneumonia in patients with SARS-CoV-2-associated acute respiratory distress syndrome requiring ECMO: A retrospective cohort study. Ann Intensive Care. 2020; 10 :158 33230710 17. Chastre J Fagon JY : State of the art ventilator-associated pneumonia. Am J Respir Crit Care Med. 2002; 165 :867–903 11934711 18. Grasselli G Scaravilli V Mangioni D : Hospital-acquired infections in critically ill patients with COVID-19. Chest. 2021; 160 :454–465 33857475 19. Papazian L Klompas M Luyt CE : Ventilator-associated pneumonia in adults: A narrative review. Intensive Care Med. 2020; 46 :888–906 32157357 20. Wolkewitz M Palomar-Martinez M Alvarez-Lerma F : Analyzing the impact of duration of ventilation, hospitalization, and ventilation episodes on the risk of pneumonia. Infect Control Hosp Epidemiol. 2019; 40 :301–306 30773159 21. Aalen OO Johansen S : An empirical transition matrix for non-homogeneous Markov chains based on censored observations. Scandinavian J Stat. 1978; 5 :141–150 22. Fine J Gray R : A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999; 94 :496–509 23. Torres A Niederman MS Chastre J : International ERS/ESICM/ESCMID/ALAT guidelines for the management of hospital-acquired pneumonia and ventilator-associated pneumonia. Eur Resp J. 2017; 50 :1700582 24. Blot S Koulenti D Dimopoulos G ; EU-VAP Study Investigators: Prevalence, risk factors, and mortality for ventilator-associated pneumonia in middle-aged, old, and very old critically ill patients. Crit Care Med. 2014; 42 :601–609 24158167 25. Ayzac L Girard R Baboi L : Ventilator-associated pneumonia in ARDS patients: The impact of prone positioning. A secondary analysis of the PROSEVA trial. Intensive Care Med. 2016; 42 :871–878 26699917 26. Villar J Ferrando C Martínez D ; dexamethasone in ARDS network: Dexamethasone treatment for the acute respiratory distress syndrome: A multicentre, randomised controlled trial. Lancet Respir Med. 2020; 8 :267–276 32043986 27. Tomazini BM Maia IS Cavalcanti AB ; COALITION COVID-19 Brazil III Investigators: Effect of dexamethasone on days alive and ventilator-free in patients with moderate or severe acute respiratory distress syndrome and COVID-19: The CoDEX randomized clinical trial. JAMA. 2020; 324 :1307–1316 32876695 28. Permpalung N Chiang TP-Y Massie AB : Coronavirus disease 2019-associated pulmonary aspergillosis in mechanically ventilated patients. Clin Infect Dis. 2022; 74 :83–91 33693551
PMC009xxxxxx/PMC9005100.txt	==== Front Crit Care Med Crit Care Med CCM Critical Care Medicine 0090-3493 1530-0293 Lippincott Williams & Wilkins Hagerstown, MD 35180721 00011 10.1097/CCM.0000000000005451 3 Clinical Investigations A Retrospective Observational Study Exploring 30- and 90-Day Outcomes for Patients With COVID-19 After Percutaneous Tracheostomy and Gastrostomy Placement* Kiser Stephanie B. MD, MPH 1 Sciacca Kate AGACNP-BC 23 Jain Nelia MD, MA 23 Leiter Richard MD, MA 23 Mazzola Emanuele PhD 4 Gelfand Samantha MD 235 Jehle Jonathan NP 23 Bernacki Rachelle MD, MS 23 Lamas Daniela MD 26 Cooper Zara MD 27 Lakin Joshua R. MD 23 1 Division of Palliative Care and Geriatric Medicine, Department of Medicine, Massachusetts General Hospital, Boston, MA. 2 Division of Palliative Medicine, Department of Medicine, Brigham and Women’s Hospital, Boston, MA. 3 Department of Psychosocial Oncology and Palliative Care, Dana-Farber Cancer Institute, Boston, MA. 4 Department of Data Science, Dana-Farber Cancer Institute, Boston, MA. 5 Division of Renal Medicine, Department of Medicine, Brigham and Women’s Hospital, Boston, MA. 6 Division of Pulmonary and Critical Care Medicine, Department of Medicine, Brigham and Women’s Hospital, Boston, MA. 7 Division of Trauma, Burn, and Surgical Critical Care, Department of Surgery, Brigham and Women’s Hospital, Boston, MA. For information regarding this article, E-mail: kate_sciacca@dfci.harvard.edu 21 2 2022 5 2022 21 2 2022 50 5 819824 Copyright © 2022 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. OBJECTIVES: To determine the 30- and 90-day outcomes of COVID-19 patients receiving tracheostomy and percutaneous endoscopic gastrostomy (PEG). DESIGN: Retrospective observational study. SETTING: Multisite, inpatient. PATIENTS: Hospitalized COVID-19 patients who received tracheostomy and PEG at four Boston hospitals. INTERVENTIONS: Tracheostomy and PEG placement. MEASUREMENTS AND MAIN RESULTS: The primary outcome was mortality at 30 and 90 days post-procedure. Secondary outcomes included continued device presence, place of residence, complications, and rehospitalizations. Eighty-one COVID-19 patients with tracheostomy and PEG placement were included. At 90 days post-device placement, the mortality rate was 9.9%, 2.7% still had the tracheostomy, 32.9% still had the PEG, and 58.9% were at home. CONCLUSIONS: More than nine-in-10 patients in our population of COVID-19 patients who underwent tracheostomy and PEG were alive 90 days later and most were living at home. This study provides new information regarding the outcomes of this patient population that may serve as a step in guiding clinicians, patients, and families when making decisions regarding these devices. critical care outcomes gastrostomy severe acute respiratory syndrome coronavirus 2 infection tracheostomy ==== Body pmcSince emerging in December 2019, the severe acute respiratory syndrome coronavirus 2 (COVID-19) virus has caused significant morbidity and mortality (1). A subset of individuals with COVID-19 become critically ill and receive interventions such as tracheostomy and percutaneous endoscopic gastrostomy (PEG) devices (2, 3). We now understand some outcomes of critically ill patients with COVID-19; however, many are largely undiscovered (4–6). As a result, patients/families, and clinicians lack data to aid in decision-making. Prior work has examined outcomes of other critically ill patients receiving tracheostomy. In a prospective cohort study, patients with acute respiratory distress syndrome (ARDS) had a post-tracheostomy 28-day mortality of 23.4% and 90-day mortality of 30.8% (7–10). By oxygenation criteria, most patients with COVID-19 receiving mechanical ventilation meet ARDS criteria (11, 12). However, whether patients with COVID-19 have similar outcomes is unknown. As medical teams guide patients and their caregivers through decisions regarding tracheostomy and PEG placement, they must weigh the risks and benefits by considering mortality and impact on quality of life. These considerations often include whether the patient will be able to live at home, and if they can expect prolonged dependence on the device (13, 14). We examined short-term outcomes of patients with COVID-19 who received tracheostomy and PEG within an integrated healthcare network in Boston, Massachusetts. We hope to contribute to the scaffolding for decision-making regarding placement of these devices. METHODS Study Design This was a multisite retrospective observational chart review study approved by the Mass General Brigham (MGB) Institutional Review Board (Number 2020P002350). Informed consent was waived. Selection of Participants We included patients hospitalized with symptomatic COVID-19 infection confirmed via polymerase chain reaction testing who had subsequent tracheostomy and PEG placement due to primary COVID-related complications at four MGB institutions (Massachusetts General Hospital, Brigham and Women’s Hospital, Faulkner Hospital, and Newton Wellesley Hospital) between February 1, 2020, and August 19, 2020. We identified patients for inclusion retrospectively through our system’s Research Patient Data Repository (RPDR), a clinical medical record registry (15). This registry allowed us to identify all patients hospitalized that met the above criteria by retrieving patients with International Classification of Diseases, 10th Revision codes for device placement and positive COVID test result during the specified time period. Investigators confirmed study eligibility by reviewing the charts of identified patients. Data Collection We obtained demographics and medical comorbidities from RPDR, gathering remaining data through manual chart review. We determined continued device presence, mortality, and patient location by chart review 30 and 90 days from device placement and only counted the outcome if clearly documented. In situations lacking conclusive evidence to indicate the outcome at the time point, reviewers did not extrapolate data from prior or subsequent chart documentation. We determined the number of days a patient spent in the ICU, the number of days between intubation and device insertion and the continued need for ventilatory support at discharge through clinical documentation. We measured complications and rehospitalizations occurring within 90 days post-device placement. Rehospitalizations included admission in the MGB system or any hospital with Care Everywhere, a functionality of the Epic Systems Corporation Electronic Health Record (Epic Systems, Verona, WI) that shares medical records between hospitals. Indication for rehospitalization was determined from the admission history. Discharge summaries were reviewed for complications. Two investigators, a nurse practitioner (K.S.) and a physician (S.B.K.), reviewed all charts independently. Investigators compared data every 10 charts for the first 50 charts and at completion of data collection to track discrepancies. Upon completion of chart review, a larger group of six investigators met and reconciled discrepancies. Data Analysis We used descriptive statistics to analyze baseline demographics, time from intubation to device placement, 30- and 90-day device presence, mortality rate, and place of residence. We calculated a weighted Charlson Comorbidity Index (CCI) for each patient based on individual medical comorbidities (16–18). RESULTS Baseline Characteristics Eighty-one patients with COVID-19 underwent device placement during the study period with tracheostomy and PEG. Table 1 displays baseline characteristics. The mean age was 59.6 years (sd, 12.4 yr) and 72.8% were men. White non-Hispanic patients were most represented (44.4%), followed by Black patients (18.5%). Most were primary English-speaking (54.3%), followed by primary Spanish-speaking (37%). The greatest number of patients had a CCI of 1–2 (40.7%), followed by a score of greater than or equal to 5 (30.9%). Diabetes with (24.7%) and without (54.3%) complications, renal disease (35.8%), chronic lung disease (30.9%), cerebrovascular disease (24.7%), and mild liver disease (21%) were frequently represented in these scores. TABLE 1. Baseline Demographics of Patient Population Characteristic n = 81 Gender, n (%) Male 59 (72.8) Race /ethnicity, n (%) Asian 4 (4.9) Black 15 (18.5) Hispanic 5 (6.2) White Non-Hispanic 36 (44.4) Other 15 (18.5) Missing 6 (7.4) Primary language, n (%) English 44 (54.3) Spanish 30 (37) Other 6 (7.4) Missing 1 (1.2) Age (yr), mean (sd) 59.6 (12.4) Weighted Charlson Comorbidity Score, n (%) 0 8 (9.9) 1-2 33 (40.7) 3-4 15 (18.5) ≥5 25 (30.9) Comorbidities, n (%) Myocardial infarction 8 (9.9) Congestive heart failure 16 (19.8) Peripheral vascular disease 3 (3.7) Cerebrovascular disease 20 (24.7) Dementia 2 (2.5) Chronic pulmonary disease 25 (30.9) Rheumatologic disease 4 (4.9) Mild liver disease 17 (21) Moderate to severe liver disease 2 (2.5) Diabetes without chronic complications 44 (54.3) Diabetes with chronic complications 20 (24.7) Hemiplegia 6 (7.4) Peptic ulcer disease 3 (3.7) Renal disease 29 (35.8) Any malignancy, including leukemia and lymphoma 12 (14.8) Human immunodeficiency virus / acquired immunodeficiency syndrome 0 (0) Number of days in ICU, mean (sd) 35.8 (14.1) Number of days between intubation and tracheostomy, mean (sd) 22.7 (8.7) Time to Device Placement and Discharge Outcomes Patients spent an average of 35.8 days (sd, 14.1 d) in the ICU. On average, patients underwent tracheostomy placement 22.7 days (sd, 8.7 d) after intubation. Thirty-four patients (42%) had tracheostomy removal prior to hospital discharge. Of the patients with tracheostomy still in place at discharge, 7 (17.1%) still required ventilatory support. Thirty- and 90-Day Outcomes At 30 days post-device placement, 35.5% of patients continued to have a tracheostomy in place (Fig. 1). This decreased to 2.7% of patients at 90 days. Information about tracheostomy device presence was missing for 7.9% and 9.6% of patients at 30 and 90 days, respectively. Continued PEG presence was observed in 71 patients (93.4%) at 30 days and 24 patients (32.9%) at 90 days. Six-point 6% and 12.3% of patients were missing information about PEG device presence at 30 and 90 days, respectively. At 30 days post-device placement, five patients had died (6.2%). Among those living, the most frequent patient location was a facility (44.7%) compared with the hospital (35.5%) and home (10.5%). At 90 days post-device placement, eight patients had died (9.9%). Among those living, the most frequent patient location was home (58.9%) compared with a facility (23.3%) and the hospital (5.5%). At 30 and 90 days, 7.9% and 12.3% of patients, respectively, had an unknown location. Figure 1. Ninety-day outcomes in terms of mortality, place of residence, and presence of device, for patients who underwent combined tracheostomy and percutaneous endoscopic gastrostomy (PEG). Characteristics of Deceased Patients and Those Subjects at Home at 90 Days Eight patients died (9.9%) within 90 days of device placement. Most of the deceased were male (75.0%) and 50.0% had a CCI score greater than or equal to 5. The most frequent comorbidities were diabetes (62.5%) and renal disease (62.5%). In comparison, 43 patients were living at home at 90 days. These patients were mostly male (70%), and 21 (46.7%) had a CCI of 1–2. Only 11 patients (24.4%) had a CCI score greater than or equal to 5. The most frequent comorbidities for these patients were diabetes without complications (46.5%), renal disease (30.2%), and mild liver disease (30.2%). Complications and Rehospitalizations Of the 81 study patients, 70 (86.4%) had no documented complications. Nine patients (11.1%) had tracheostomy complications including bleeding (5), pneumothorax (2), tracheal stenosis (1), and infection (1). Six patients (7.4%) had PEG complications including bleeding (5) and infection (1). Eleven patients (13.6%) were rehospitalized within 90 days. Indications for rehospitalization included hypervolemia (1), dislodged tracheostomy (1), multifocal pneumonia (1), urinary tract infection (4), small bowel obstruction (2), gastrointestinal bleed (1), and cholecystostomy tube malfunction (1). Number of rehospitalizations within 90 days for the total study group ranged from 0 to 2, with a mean of 0.2 per patient. DISCUSSION Tracheostomy and PEG placement are common interventions for patients with sustained critical illness. COVID-19 has challenged providers, patients, and their caregivers to make decisions regarding these interventions, with little information regarding outcomes. Considering this, our study presents several notable findings. For this group, mortality rate 90 days after undergoing these interventions was low (9.5–9.9%). Additionally, most patients were living at home at 90 days (58.9%). To date, this is the only study examining 90-day mortality and place of residence in this patient population. Our study demonstrates several unique outcomes. The patients are notably different than the population typically served by our hospital system. Just under 50% identified a language other than English as their primary language. In comparison, for patients admitted at Massachusetts General Hospital in 2018, 9% identified as non-English speaking (19). These findings are consistent with other COVID-19 studies, showing racial and ethnic minority groups disproportionately impacted (20). The burden of chronic critical illness on minority patients, likely due to healthcare inequities, requires further exploration. The study population is younger than expected. The mean age of 59.6 years represents a younger population than other studies of ICU patients with COVID-19 (5, 6). This may reflect a selection bias of the clinical teams for patients likely to benefit from device placement, potentially leading to better outcomes. PEG tube placement is often required for enteral nutrition after tracheostomy. While there was not a specific policy directing these interventions be provided in tandem, we saw that these interventions were almost exclusively being done together during the spring 2020 COVID surge in our setting. While our study was unable to explore the reasoning for this, it may be due to the desire to reduce COVID positive patient operating room time and limit clinician exposure early in the pandemic. During the 2020 spring surge, patients were initially intubated at different clinical thresholds than typical practice, including lower oxygen requirements. This practice could have resulted in prolonged periods of intubation for patients with mild disease, leading to inflated survival statistics. Additionally, calculated mortality could have been artificially low as a result of the lack of any treatment with an effect on survival, leading to early deaths initially and selecting for a population to receive interventions that survived. As clinicians, patients, and caregivers contemplate placement of these devices, they ask how long the device will be needed. Unexpectedly, almost all tracheostomies were removed at 90 days, with only 2.7% remaining. Additionally, the mortality rate for this group was far lower (9.9%) than 90-day mortality after tracheostomy for critically ill patients with non-COVID ARDS (30–40%) (21, 22). This divergence in outcomes may suggest differences in pathophysiology that are currently poorly understood. Notably, while descriptive statistics hint at trends, our sample size was not large enough to draw statistically significant comparisons in terms of comorbidities. We chose to focus on the CCI score rather than measures of acute illness drawing on evidence that preadmission diagnoses are more predicative of mortality than acute measures of illness (23). The strengths of this study include its multi-institution inclusion, robust chart review, and low number of missing data. The retrospective observational study design limited data collection to that documented and reported in the medical record. Findings may be limited in generalizability as they are collected from a single health system. In addition, the data are subject to the selection bias of the critical care teams and surgeons making decisions regarding these interventions. Care practices have evolved since the beginning of the pandemic, possibly influencing these outcomes, and may limit the applicability and reproducibility of these data. A notable limitation of this study is the lack of comparison group. During the time period, these data were gathered early in the COVID-19 pandemic, we did not expect the same frequency of accurately comparable patients with COVID-related ARDS who were intubated for a long period but did not undergo tracheostomy placement to create a comparison; as such, we chose a cross-sectional report of these data. CONCLUSIONS To date, this is the only exploration of 90-day mortality, complications, and place of residence for patients with COVID-19 who receive tracheostomy and PEG. While more studies are needed to clarify the role of critical care interventions for patients with COVID-19, we hope this study serves as a first step in providing clinicians with needed data to assist in decision-making. Additional studies are needed to identify factors that influence these outcomes and to further explore the impact of these interventions on patient quality of life. ACKNOWLEDGMENTS We thank Stacy Duey, MT(ASCP), MCPH, for her support in our Research Patient Data Repository searches. *See also p. 891. Dr. Kiser and Ms. Sciacca are co-first authors. Dr. Leiter received funding from the New England Journal of Medicine, New York University, the New York Times, and the United States Uniformed Services University. Dr. Mazzola received funding from The VeraMedica Institute LLC. The remaining authors have disclosed that they do not have any potential conflicts of interest. ==== Refs REFERENCES 1. Johns Hopkins Coronavirus Resource Center: COVID-19 Map. 2020. Available at: https://coronavirus.jhu.edu/map.html. Accessed January 4, 2021 2. Richardson S Hirsch JS Narasimhan M ; the Northwell COVID-19 Research Consortium: Presenting characteristics, comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the New York City Area. JAMA. 2020; 323 :2052–2059 32320003 3. McCarthy CP Murphy S Jones-O’Connor M : Early clinical and sociodemographic experience with patients hospitalized with COVID-19 at a large American healthcare system. EClinicalMedicine. 2020; 26 :100504 32838244 4. Armstrong RA Kane AD Cook TM : Outcomes from intensive care in patients with COVID-19: A systematic review and meta-analysis of observational studies. Anaesthesia. 2020; 75 :1340–1349 32602561 5. Grasselli G Greco M Zanella A ; COVID-19 Lombardy ICU Network: Risk factors associated with mortality among patients with COVID-19 in intensive care units in Lombardy, Italy. JAMA Intern Med. 2020; 180 :1345–1355 32667669 6. Karagiannidis C Mostert C Hentschker C : Case characteristics, resource use, and outcomes of 10 021 patients with COVID-19 admitted to 920 German hospitals: An observational study. Lancet Respir Med. 2020; 8 :853–862 32735842 7. Siempos II Ntaidou TK Filippidis FT : Effect of early versus late or no tracheostomy on mortality and pneumonia of critically ill patients receiving mechanical ventilation: A systematic review and meta-analysis. Lancet Respir Med. 2015; 3 :150–158 25680911 8. Flaatten H Gjerde S Heimdal JH : The effect of tracheostomy on outcome in intensive care unit patients. Acta Anaesthesiol Scand. 2006; 50 :92–98 16451156 9. Ferraro F Gravina AG d’Elia A : Percutaneous endoscopic gastrostomy for critically ill patients in a general intensive care unit. Acta Gastroenterol Belg. 2013; 76 :306–310 24261024 10. Nobleza COS Pandian V Jasti R : Outcomes of tracheostomy with concomitant and delayed percutaneous endoscopic gastrostomy in the neuroscience critical care unit. J Intensive Care Med. 2019; 34 :835–843 28675111 11. Grasselli G Tonetti T Protti A ; collaborators: Pathophysiology of COVID-19-associated acute respiratory distress syndrome: A multicentre prospective observational study. Lancet Respir Med. 2020; 8 :1201–1208 32861276 12. Bos LDJ : COVID-19-related acute respiratory distress syndrome: Not so atypical. Am J Respir Crit Care Med. 2020; 202 :622–624 32579026 13. Hossein SM Leili M Hossein AM : Acceptability and outcomes of percutaneous endoscopic gastrostomy (PEG) tube placement and patient quality of life. Turk J Gastroenterol. 2011; 22 :128–133 21796547 14. Oeyen SG Vandijck DM Benoit DD : Quality of life after intensive care: A systematic review of the literature. Crit Care Med. 2010; 38 :2386–2400 20838335 15. Nalichowski R Keogh D Chueh HC : Calculating the benefits of a research patient data repository. AMIA Annu Symp Proc. 2006; 2006 :1044 16. Deyo RA Cherkin DC Ciol MA : Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992; 45 :613–619 1607900 17. Quan H Sundararajan V Halfon P : Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005; 43 :1130–1139 16224307 18. Halfon P Eggli Y van Melle G : Measuring potentially avoidable hospital readmissions. J Clin Epidemiol. 2002; 55 :573–587 12063099 19. Betancourt JR : Massachusetts General Hospital Annual Report on Equity in Health Care Quality. 2020. Available at: https://5536401f-20a1-4e61-a28e-914fb5dcef51.filesusr.com/ugd/888d39_6b0f9ebc637443abbb54207cf8dec427.pdf. Accessed August 1, 2021 20. Stokes EK Zambrano LD Anderson KN : Coronavirus disease 2019 case surveillance - United States, January 22-May 30, 2020. MMWR Morb Mortal Wkly Rep. 2020; 69 :759–765 32555134 21. Bellani G Laffey JG Pham T ; LUNG SAFE Investigators; ESICM Trials Group: Epidemiology, patterns of care, and mortality for patients with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA. 2016; 315 :788–800 26903337 22. Scales DC Thiruchelvam D Kiss A : The effect of tracheostomy timing during critical illness on long-term survival. Crit Care Med. 2008; 36 :2547–2557 18679113 23. Kerckoffs M Brinkman A Keizer N : The performance of acute verses antecedent patient characteristics for 1-year mortality prediction during intensive care unit admission: A national cohort study. Crit Care. 2020; 24 :330 32527298
PMC009xxxxxx/PMC9005103.txt	==== Front Circ Res Circ Res RES Circulation Research 0009-7330 1524-4571 Lippincott Williams & Wilkins Hagerstown, MD 00007 10.1161/CIRCRESAHA.122.319954 10071 10173 10176 Stroke and Neurocognitive Impairment Compendium Cerebrovascular Complications of COVID-19 and COVID-19 Vaccination De Michele Manuela 1 Kahan Joshua 3 Berto Irene 1 Schiavo Oscar G. 1 Iacobucci Marta 2 Toni Danilo 1* Merkler Alexander E. 3* Stroke Unit, Emergency Department (M.D.M., I.B., O.G.S., D.T.), Sapienza University of Rome, Italy. Neuroradiology Unit, Department of Human Neurosciences (M.I.), Sapienza University of Rome, Italy. Clinical and Translational Neuroscience Unit, Department of Neurology, Feil Family Brain and Mind Research Institute, Weill Cornell Medicine, New York, NY (J.K., A.E.M.). Correspondence to: Alexander E. Merkler, MD, MS, Department of Neurology, Weill Cornell Medicine, 525 E 68th St, F610, New York, NY 10065. Email alm9097@med.cornell.edu 15 4 2022 15 4 2022 15 4 2022 130 8 11871203 © 2022 American Heart Association, Inc. 2022 This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. The risk of stroke and cerebrovascular disease complicating infection with SARS-CoV-2 has been extensively reported since the onset of the pandemic. The striking efforts of many scientists in cooperation with regulators and governments worldwide have rapidly brought the development of a large landscape of vaccines against SARS-CoV-2. The novel DNA and mRNA vaccines have offered great flexibility in terms of antigen production and led to an unprecedented rapidity in effective and safe vaccine production. However, as mass vaccination has progressed, rare but catastrophic cases of thrombosis have occurred in association with thrombocytopenia and antibodies against PF4 (platelet factor 4). This catastrophic syndrome has been named vaccine-induced immune thrombotic thrombocytopenia. Rarely, ischemic stroke can be the symptom onset of vaccine-induced immune thrombotic thrombocytopenia or can complicate the course of the disease. In this review, we provide an overview of stroke and cerebrovascular disease as a complication of the SARS-CoV-2 infection and outline the main clinical and radiological characteristics of cerebrovascular complications of vaccinations, with a focus on vaccine-induced immune thrombotic thrombocytopenia. Based on the available data from the literature and from our experience, we propose a therapeutic protocol to manage this challenging condition. Finally, we highlight the overlapping pathophysiologic mechanisms of SARS-CoV-2 infection and vaccination leading to thrombosis. COVID 19 pandemics SARS CoV 2 stroke vaccines ==== Body pmcAccording to the World Health Organization, almost 5 million people have died from COVID-19, with >245 million confirmed cases.1 A number of vascular and thromboembolic complications of COVID-19 were noted early in the pandemic,2 and this was soon followed by observations suggesting a heightened risk of stroke and other cerebrovascular complications.3 Comparative meta-analytic studies have since been undertaken to confirm that infection with SARS-CoV-2 increases the risk of ischemic stroke relative to noninfected contemporary or historical controls,4 as well as relative historical controls infected with influenza.5 In addition to ischemic stroke, hemorrhagic stroke,6 cerebral venous sinus thrombosis (CVST),7 and posterior reversible encephalopathy syndrome8 have all been reported as possible complications. Vaccines against SARS-CoV-2 are a milestone in the fight against COVID-19. Response to this global crisis, with devastating health, social, and economic impact, was extraordinary, and thanks to cooperation between companies and governments, within a year, several vaccines against SARS-CoV-2 have shown impressive efficacy in randomized clinical trials that have translated into real-world observations. Unfortunately, extremely rare cases of thrombocytopenia and thromboembolic complications have been reported following administration of the ChAdOx1 nCoV-19 vaccine (Oxford-AstraZeneca) and the Ad26.COV2-S vaccine (Janssen), which has contributed to vaccine hesitancy among the public. The situation, however, is highly nuanced, as the risk of thromboembolic complications from infection with SARS-CoV-2 alone is significant. This is of special relevance to stroke and cerebrovascular complications given the significant morbidity associated with intracranial thromboses and hemorrhage. In what follows, we review the evidence surrounding stroke and cerebrovascular complications of both SARS-CoV-2 infection and SARS-CoV-2 vaccination. In so doing, we review thromboinflammation and the proposed pathophysiology of stroke as a complication of COVID-19, and vaccine-induced immune thrombotic thrombocytopenia (VITT), its typical clinical presentation, and the cases that have presented with stroke and cerebrovascular complications. We conclude by identifying the main pathophysiologic abnormalities common to the two conditions and compare the risk of stroke related to infection and vaccination. Stroke as a Complication of SARS-CoV-2 Infection Early reports of neurological complications of SARS-CoV-2 infection emerged in the pre-peer review literature in March 2020. By April 2020, the first retrospective observational reports from Wuhan were published finding neurological symptoms in as many as 36.4% of the admitted patients, specifically citing both ischemic and hemorrhagic stroke as complications of SARS-CoV-2.3 In this section, we discuss the risk of ischemic stroke and other cerebrovascular disorders, as well as putative pathophysiology for stroke in patients with COVID-19. Ischemic Stroke Oxley at al9 soon reported a series of relatively young patients (<50 years old) presenting with large vessel occlusion ischemic strokes during the first peak in New York City, all of whom tested positive for SARS-CoV-2. As time would tell, the risk of such presentations was not as great as was initially feared. In fact, initial retrospective incidence rates varied considerably; Li et al10 reported ischemic strokes in as many as 4.6% of their Wuhan inpatient cohort (n=219), whereas Yaghi et al11 found that only 0.9% of their patients admitted in New York had stroke diagnosed during their admission (n=3556). Cohorts in Italy,12 France,13 Germany,14 Philadelphia,15 and other New York hospital systems5,16 fell within this range. To date, the largest multinational studies17–19 and meta-analytic4 estimates of risk among hospitalized patients are shown to be between 0.5% and 1.3%. However, there are important caveats to these estimates, most notably that the majority of strokes did not present with typical clinically evident focal neurological deficits. Rather, the events were detected on neuroimaging during hospital admission,5,17,20 leading many to dispute the true incidence given that not all patients undergo neuroimaging.21 From our personal experience in New York, this was especially true during the peak periods of COVID-19, social distancing, and airborne isolation rules. Risk has been shown to vary with clinical severity of COVID-19.4,10,14,22 Consistent with this hypothesis, studies that include mild disease (managed in the outpatient setting) have yielded lower estimates of risk.23,24 In terms of relative risk, patients requiring hospitalization for COVID-19 have a 3- to 4-fold greater risk of stroke compared with noninfected hospitalized historical or contemporary control cohorts.4,5,11,25 Compared with patients with influenza requiring hospitalization, COVID-19 patients have a 7- to 8-fold greater risk of stroke, although the CIs of these estimates are wide and overlap significantly with the risk estimates when compared with historical controls.5 Patients with COVID-19 at the highest risk of ischemic stroke appear to be those with a history of ischemic stroke,26 possibly a history of diabetes4 and other traditional stroke risk factors,27 and higher serum d-dimer levels.28,29 Outcomes in strokes occurring in patients with COVID-19 also appear worse, in terms of initial stroke severity (compared with historical controls),30 functional outcome at discharge (compared with contemporary and historical controls),30,31 discharge destination (compared with historical controls),32 and inpatient mortality (compared with both contemporary and historical controls).4,30–33 With regard to etiologic classification, patients with COVID-19–associated ischemic strokes have been shown to present with more embolic-appearing findings on neuroimaging. Specifically, multiple case series have been published, highlighting an increased rate of strokes with large vessel occlusions.9,16,17,31,34–37 The majority of these strokes were classified as cryptogenic or embolic stroke of undetermined source.38 Consistent with this, the risk of specifically cryptogenic strokes has been found to be disproportionately increased in patients with COVID compared with control cohorts.4,5,11,16 Many of these patients were also found to have other systemic evidence of thromboembolic disease39,40 and visceral infarction—a phenomenon that is known to be associated with cardioembolic and cryptogenic strokes.41 Putative Mechanisms of Ischemic Stroke in Patients With SARS-CoV-2 The mechanism leading to cerebrovascular complications in the setting of SARS-CoV-2 is likely multifactorial. First, patients with COVID-19, especially those with severe disease, frequently have comorbid factors that increase their baseline risk of thromboembolism. These include dehydration, immobilization, chronic cardiovascular risk factors, or prior atherosclerotic diseases (ie, coronary artery disease, cerebrovascular disease, and chronic kidney disease), as well as inherited thrombophilia.42 Patients with severe COVID-19 requiring intensive care unit (ICU) admission have been shown to have significantly higher rates of arterial or venous thromboembolic events,43 presumably due to a combination of factors discussed herein. Interestingly, studies comparing the rates of ischemic stroke between critically ill COVID-19 patients and other acute respiratory distress syndrome patients have not detected a significant difference, suggesting acute respiratory distress syndrome or critical illness itself likely confers some risk for stroke.44 However, more causal mechanisms have also been proposed given that SARS-CoV-2 both increases the risk of cardiac pathology and impacts all 3 factors comprising Virchow triad (endothelial injury, stasis, and hypercoagulable state), ultimately promoting thrombosis. Regarding cardiac complications, SARS-CoV-2 has been shown to increase the risk of developing atrial fibrillation,45 which is a well-established risk factor for ischemic stroke.46 In addition, myocardial infarction,47 myocarditis,48 and Takotsubo cardiomyopathy48 have been reported in hospitalized patients with COVID-19, all of which predispose to the formation of left ventricular thrombi and subsequent cardiac embolism.49 Additionally, bacterial superinfection is common in patients with severe COVID-19,50 which increases the risk of bacteremia and infective endocarditis; both of which increase the risk of ischemic stroke.51 In addition, vascular injury is a recognized hallmark of COVID-19. The precise pathophysiology of this remains unclear, but both ACE2 (angiotensin-converting enzyme 2)-dependent and independent processes have been implicated, with some suggesting direct platelet activation via the SP (spike protein) itself.52,53 Within the pulmonary circulation, postmortem analysis has found severe endothelial injury, disrupted cell membranes, with diffuse vascular thrombosis and occlusion of alveolar capillaries.54 Within the cerebral microvasculature specifically, combined imaging and histopathologic assessments have revealed thinning of the basal lamina of the endothelial cells, capillary congestion with fibrinogen leakage, and perivascular inflammation associated with macrophage infiltrates and CD3+ and CD8+ T cells.55 This mild-to-moderate, nonspecific inflammation without clear evidence of vasculitis is a consistent finding in autopsy studies. In contrast to the pulmonary pathology, however, cerebral histopathology has only rarely revealed frank vascular occlusion55,56 or florid cerebrovascular inflammation as seen in the pulmonary microcirculation.57,58 Despite this, these studies have frequently noted mild-to-moderate hypoxic-ischemic injury and microhemorrhages, as well as a stroke phenotype involving multiple small infarcts, including in the corpus callosum, presumed to be a manifestation of microvascular occlusion secondary to thromboinflammation. There are reports of SARS-CoV-2–like particles being found in the brain and endothelium using ultrastructural analysis of tissue from SARS-CoV-2–infected patients with neurological symptoms59; however, diagnosis is challenging due to similar appearing normal cellular structures.60 Consequent to this vascular endothelial inflammation,61 whether systemically or in the cerebral microcirculation, patients with COVID-19 can demonstrate a coagulopathy with an increased risk of in situ thrombosis.62 Specifically, endothelial release of proinflammatory cytokines (the so-called cytokine storm) has been shown to be associated with a hypercoagulable state as evidenced by deranged levels of VWF (von Willebrand factor), D-dimer, fibrinogen, and factor VIII.63 Specifically, small case series of patients with severe COVID-19 have shown exaggerated interleukin-mediated release of VWF, and suppression of the function of ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13), promoting thrombosis via a thrombotic microangiopathy-like process.64,65 This increases the risk of both arterial and venous thrombosis with or without paradoxical embolism to the cerebral circulation. Other Cerebrovascular Complications Hemorrhagic Stroke Studies evaluating the risk of hemorrhagic stroke, which includes intracerebral, subdural, and subarachnoid hemorrhage, have been more limited. The largest studies to date suggest the prevalence among hospitalized patients is rare (as low as 0.2%) and was more likely to occur in older patients and those receiving therapeutic anticoagulation.6 Those hospitalized with ICH and COVID-19 had worse outcomes than those with COVID-19 alone. Studies with more granular hospital course data have suggested that hemorrhages tend to occur during the hospitalization, and the mechanism of these hemorrhages tends to be related to coagulopathy or supratherapeutic anticoagulation, as opposed to primary causes of intracerebral hemorrhages.66 Cerebral Venous Sinus Thrombosis In a self-controlled case series including 29.1 million people in the United Kingdom comparing rates of thrombotic complications of COVID-19 and vaccinations for COVID-19, testing positive for SARS-CoV-2 was associated with an increased risk of CVST.67 While the point estimate of this risk was high, the CIs were broad, suggesting significant uncertainty of the precise risk. In a meta-analysis of 67 845 patients, the pooled rate of CVST was 0.03%.4 Case series7,68 and reviews have highlighted relatively few (<50) CVST cases reported in the literature to date and suggest it is a relatively rare complication of COVID-19 that is not associated to severe disease as with ischemic stroke. Almost 90% of the reported cases have occurred in women and are mostly found in the transverse sinus. Clinical presentations, however, are subtle, and authors recommend that high suspicion is maintained when encountering patients with COVID-19, headache, and focal neurological deficits. Posterior Reversible Encephalopathy Syndrome Posterior reversible encephalopathy syndrome is another relatively rare neurovascular complication that has been linked to hospitalized patients with COVID-19. Almost all the data on this phenomenon are at the level of case series with relatively few cases reported in the literature in total (<50).69 Cases have mostly presented with prolonged or otherwise unexplained encephalopathy and poor level of consciousness, with some complicated by seizures or a focal neurological deficit.8 Most have been in the setting of critically ill patients who have traditional risk factors for posterior reversible encephalopathy syndrome (acute kidney injury and uncontrolled hypertension), and it remains unclear whether the rate of posterior reversible encephalopathy syndrome in patients with COVID-19 is any greater than in patients with other critical illness or multiorgan failure. Stroke as a Complication of COVID-19 Vaccination To date, the European Medicines Agency has approved 5 vaccines:70 (1) Comirnaty (BNT 162b2 mRNA vaccine) by Pfizer BionTech; (2) Ad26.COV2.S adenovirus vaccine by Johnson & Johnson/Janssen; (3) Spikevax (mRNA-1273 vaccine) by Moderna; (4) Vaxzevria (ChAdOx1 nCoV-19 vaccine) by Oxford-AstraZeneca, and more recently, (5) Nuvaxovid (NVX-CoV2373) by Novovax. The Food and Drug Administration similarly approved these excluding Vaxzevria and Nuvaxovid.71 Both the European Medicines Agency (EudraVigilance) and the Food and Drug Administration are monitoring the safety of authorized COVID-19 vaccines. This enables the detection of any side effects that may emerge during the mass vaccination. At the time of writing, a total of 6 838 727 352 vaccine doses have been administered worldwide.1 Although most of the reported vaccine-related side effects are mild and transient, concerns progressively emerged about post-Vaxzevria embolic and thrombotic events associated with thrombocytopenia (thrombosis with thrombocytopenia syndrome [TTS], renamed VITT). On April 14, 2021, the European Medicines Agency’s Safety Committee (Pharmacovigilance Risk Assessment Committee) concluded that a causal relationship between Vaxzevria vaccination and rare cases of venous thrombosis in unusual sites (ie, CVST and splanchnic vein thrombosis) and less frequently arterial thrombosis was plausible. As of July 31, 2021, 1503 cases of suspected TTS with Vaxzevria out of about 592 million administered doses were globally reported.70 With a much lower incidence, TTS has also been described after Janssen vaccination. In the United States, by July 8, 2021, 38 TTS confirmed cases occurring within 15 days after vaccination were reported to the Vaccine Adverse Event Reporting System, four of which resulted in death. The overall calculated rate was 3.0 TTS cases per million administered doses, with a higher reporting rate of 8.8 TTS cases per million administered doses among women aged 30 to 49 years.72,73 Nevertheless, a population-level risk-benefit analysis has shown a large population benefit of Janssen vaccination as compared with rare occurrence of TTS.72 In Europe, as of June 27, 2021, 21 cases over about 7 million administered doses were spontaneously reported to EudraVigilance, four of which were fatal.70 Though further case ascertainment is required to confirm TTS in these reported cases, the relationship between the administration of DNA vaccines and TTS has now been established. In fact, according to the Bradford-Hill criteria, which are the accepted criteria for assessing causality of an association, the link between the ChAdOx1 nCoV-19 vaccine and TTS has recently been demonstrated.74 However, the precise estimate of this association is not known, since the incidence rates vary from 0.5 to 25 per 100 000 vaccinated individuals, depending on the different countries.74 Nevertheless, benefits of adenoviral vector vaccines, clearly demonstrated in randomized controlled trials,75,76 still outweigh the risks of these rare thrombotic events, especially in subjects >30 years.77 Interestingly, Hippisley-Cox et al67 found an increased risk of ischemic stroke after 15 to 21 days from BNT162b2 mRNA vaccination (Pfizer BionTech), and after a positive SARS-CoV-2 test, but not after ChAdOx1 nCov-19 vaccination (Oxford-AstraZeneca). The same authors found an increased risk of thrombocytopenia after ChAdOx1 nCov-19 vaccination, and of CVST after ChAdOx1 nCoV-19 vaccination (at 8–14 days), after BNT162b2 mRNA vaccination (at 15–21 days), and after a positive SARS-CoV-2 test.67 More data on these adverse events from BNT162b2 mRNA vaccination are needed. COVID-19 Vaccines and Target Proteins All the approved vaccines are based on the full-length homotrimeric SARS-CoV-2 SP. SP plays a key role in viral infection and pathogenesis, since it mediates the entrance of the virus into the host cells via the binding with ACE2. SP, which is located on the viral envelope, comprises 3 S1/S2 heterodimers: S1 harbors the N-terminal domain and the receptor-binding domain.78–80 Interestingly, the SP ectodomain consists of a head where receptor-binding domains are located and a stalk with 3 flexible hinges connecting SP to the viral membrane. This high degree of conformational freedom of SP on the viral surface may interfere with antibody access to the stalk, add strength to the virus, and facilitate the binding of the SP with the host receptor.81 In December 2020, Wajnberg et al82 found that most of the infected individuals with mild-to-moderate COVID-19 had developed a robust IgG antibody response against the viral SP. These authors also showed that titers were long-lasting (several months) and that anti-SP binding titers significantly correlated with neutralization of SARS-CoV-2.82 All these data confirmed the SP as the main target of vaccine development.78 The three vaccines approved by the United States and 4 of the 5 approved by the European Union (EU) are DNA or mRNA vaccines encoding the SARS-CoV-2 SP. In DNA vaccines (Janssen and Oxford-AstraZeneca), the genetic materials need to pass through the nucleus to create mRNA with subsequent transcription of the protein in the cytoplasm.83 In mRNA vaccines (Pfizer and Moderna), the nuclear step is missing, making the process even simpler.84 Pfizer and Moderna vaccines consist of a lipid-enclosed nucleoside-modified mRNA encoding a different mutated SP, whereas the AstraZeneca and Janssen vaccines utilize a chimpanzee nonreplicating adenovirus and a type 26 nonreplicating recombinant adenovirus vector, respectively. Moreover, the AstraZeneca vaccine has the complete coding sequence of SP plus a sequence of a tissue-type plasminogen activator, and the Janssen vaccine has mutations for stabilizing the SP.83 Nuvaxovid is based on the SP produced by recombinant DNA technology using a baculovirus expression system in an insect cell line and is adjuvanted with Matrix-M. Effectiveness and safety of this vaccine have been demonstrated in clinical trials,85 but real-world evidence is still lacking. Efficacy data of DNA- and mRNA-based vaccines against SARS-CoV-2 from clinical trials seem to be consistent with data on vaccine effectiveness from the real world. However, more data are urgently needed, considering both the rapidly emerging appearance of SARS-CoV-2 novel variants and the temporal waning of immunity after vaccination.86 Vaccine-Induced Immune Thrombotic Thrombocytopenia By the end of March 2021, several scientific papers from different countries reported cases of devastating thrombosis in unusual sites, especially CVST, 5 to 30 days after the administration of the first dose of the ChAdOx1 nCoV-19 vaccine. These patients, who were otherwise young and healthy, also presented with thrombocytopenia, elevated D-dimer, sometimes low fibrinogen, and high levels of antibodies against PF4 (platelet factor 4)-heparin.87–89 Similar syndromes have also been reported after Janssen vaccination90 and after Moderna’s mRNA-1273 vaccine.91 The syndrome was named VITT92 since it resembles the heparin-induced immune thrombocytopenia (HIT),93 although in the absence of exposure to heparin. Pathogenetic Hypothesis PF4 is a cationic chemokine consisting of 4 monomers, released from the α-granules of activated platelets as an immune defense mechanism. It is capable of opsonizing negatively charged surfaces of bacteria, ultimately facilitating binding of anti-PF4 antibodies. In HIT, heparin, acting as a polyanion, causes a conformational change of PF4 tetramers and consequently the anti-PF4/heparin antibody induction.93 Things other than heparin, such as chondroitin sulphate, DNA and RNA, bacterial wall components, and high concentration of PF4 per se, can induce the exposure of HIT antigens leading to spontaneous or autoimmune HIT. Sera from patients with autoimmune HIT typically contain high-avidity IgG antibodies, which strongly activate platelets from healthy donors via FcγRIIa, one of the receptors for the Fc domain of IgG antibodies. As a consequence, platelet-derived procoagulant microvesicles are released, resulting in severe thrombocytopenia, leading to an increased frequency of disseminated intravascular coagulation, and atypical thrombotic events.93,94 Similarly to autoimmune HIT, sera from VITT patients contain high levels of PF4-heparin antibodies that activates platelets in the presence of, but also in the absence of, heparin. This activation is greatly enhanced in the presence of PF4.87–89 Notably, a cross-reaction between the anti–SARS-CoV-2 SP antibodies and PF4 or PF4/heparin complexes has been ruled out,95 and no correlation has been found between the anti-PF4 and the anti–SARS-CoV-2 neutralizing antibody levels after ChAdOx1 nCoV-19 vaccination.96 These data exclude the possibility that the anti-PF4 antibodies are a side product of the vaccine immune response. Recently published data suggest that vaccine components, including the adenovirus hexon protein and also the adenovirus per se,97 can generate neoantigen complexes with PF4, thus inducing anti-PF4 antibody production.98 Anti-PF4 antibodies stimulate platelet aggregation. Cross talk of platelets and anti-PF4 antibodies activates neutrophils, leading to the formation of neutrophil extracellular traps and ultimately to the activation of monocytes and endothelial cells, further amplifying the activation of the coagulation cascade. The ChAdOx1 nCov-19 vaccine also contains EDTA, which increases the capillary leakage at the inoculation site, allowing the virus to spread via the bloodstream.98 Data based on an intriguing hypothesis about a possible transcription of spliceosome-mediated soluble SP fragments with thrombogenic properties in DNA vector vaccines have not yet been peer reviewed.99 However, preprinted data from our group seem to support this hypothesis. In fact, a soluble SP has been found in sera from 3 VITT patients and on a platelet-rich thrombus retrieved from middle cerebral artery (MCA) of 1 VITT patient, suggesting that SP could be one of the platelet activation triggers in VITT.100 Undoubtedly, additional experiments are required to fully understand the pathogenesis of this rare, devastating syndrome. Clinical Characteristics, Diagnosis, and Therapy As of April 2021, descriptions of clinical features of patients affected by VITT in case series and case reports from different countries allowed us to delineate precise diagnostic criteria and inform the therapeutic approach.73,87–89,101–103 Guidelines from different medical societies have been developed,87,104–106 although some of the published cases do not strictly meet classification criteria. VITT is an evolving condition, and a low-probability diagnosis of VITT at presentation can rapidly evolve to a fully blown VITT in the subsequent days. Close monitoring of patients is then mandatory to rule out this diagnosis.107 Recently, a pre-VITT syndrome characterized by severe headache without associated CVST or other thrombosis has been described, highlighting the need for clinicians to be aware of different presenting onsets of VITT, and intravenous immunoglobulin (IVIG) therapy promptly.108 Diagnosis of VITT is clinical and radiological. Typical blood test results are needed to confirm the clinical suspicion. Patients with VITT typically present with a classical clinical triad of thrombosis (mainly CVST, pulmonary and splanchnic), thrombocytopenia (<150 000/µL), and elevated D-dimers (>4000 fibrinogen-equivalent units [FEU] or 4–8× the upper limit of normal range). As addressed by Greinacher et al,87 the different combination of these elements leads to 2 different scenarios: VITT likely and VITT unlikely (Table 1). Diagnosis of VITT is confirmed by demonstration of anti-PF4 antibodies by ELISA plate–based PF4/heparin (polyanion) antibody test (but a negative test still does not definitively rule out the diagnosis)87 and functional platelet activation assay.87–89,92,106 Table 1. Criteria to Consider When Risk Stratifying Patients With Suspected VITT Prompt recognition and treatment of this syndrome may reduce mortality. Education of the public and clinicians has reduced mortality of VITT from 50% in the first case series in April 2021 to 22% in June 2021 in the United Kingdom.87 The pillars of VITT therapy are nonheparin anticoagulants (direct oral anticoagulants, danaparoid, argatroban or fondaparinux, continued for at least 3 months) and high-dose IVIG (0.5–1 g/kg of actual body weight for 1 or 2 consecutive days).87,104,105 Steroids may be useful (especially if platelets are <50 000), and plasma exchange can be considered in selected cases.87,104,105 Rituximab can be prescribed in patients who are refractory to repeated doses of IVIG and plasma exchange, although evidence is limited.104 Ischemic Stroke as VITT Atypical Presentation Ischemic stroke can be a rare and challenging symptom onset of VITT or can complicate its course. The real incidence of this serious and life-threatening condition is unknown. We performed a systematic review using MEDLINE, PUBMED, and Google Scholar databases to collect all the published articles related to the development of ischemic stroke after vaccination against SARS-CoV-2. The search process was done on October 27, 2021 by the authors using the following terms in various combinations: “ChAdOx1 ncov19 vaccination,” “stroke,” “vaccine induced immune thrombotic thrombocytopenia,” “infarct,” “VITT,” “AstraZeneca,” “SARS-CoV-2 vaccination,” “COVID-19,” and “PF4.” Overall, 161 published articles were identified, but only 13 were relevant to this review. One article was removed due to insufficient workup to rule out other causes of stroke and for not testing anti-PF4 polyanion antibodies. Consequently, the search and sorting processes were finalized with 12 articles with data on 16 patients (Figure 1). Figure 1. Systematic review flowchart. Table 2. Demographic, Clinical Characteristics, Baseline Blood Samples, Radiological Features, Treatment, and Outcome of Patients With Vaccine-Induced Immune Thrombotic Thrombocytopenia and Ischemic Stroke Described in the Literature In Table 2, we have summarized the demographic information and clinical features of the 16 patients affected by ischemic stroke with VITT confirmed diagnosis, identified from case reports or case series published in peer-reviewed journals. All the patients had received the first dose of the ChAdOx1 ncov19 vaccine. Three cases were reported from Italy, 6 from the United Kingdom, 2 from Germany, 1 from Slovenia, 1 from France, 1 from Denmark, and 2 from Canada. The mean age of the reported patients was 46.6 (SD, ±15.2; range, 21–73) years. Twelve of 16 patients were women (75%). Median time between vaccination and onset of stroke was 10 days. All the individuals were healthy before vaccination, and half of them had no preexisting comorbidities in their medical history. Three of them experienced hypertension and three had hyperlipidemia. Three subjects had thyropathy (1 case of Hashimoto thyroiditis and 2 cases of hypothyroidism). One patient was in follow-up care after breast cancer, and another one had recently received the diagnosis of prostate cancer (not yet staged). Only 1 patient, a 69-year-old man, had multiple vascular risk factors (hypertension, diabetes, obstructive sleep apnea, obesity, and aortic valve replacement) and previous exposure to heparin (9 months earlier). One patient experienced migraine and the another one was on estroprogestative contraceptives. Most of the patients had occlusion of the MCA or its branches (81%), and 7 of them (43.7%) also had thrombotic occlusion of the intracranial internal carotid artery. Five of 11 patients with proximal MCA occlusion (45.4%) developed a malignant MCA infarct involving the whole territory of the MCA with space-occupying cerebral edema and rapid neurological deterioration, successfully treated with hemicraniectomy in 4 cases. In 1 case (case 7), surgical intervention was not performed since the malignant infarct was bilateral (Figure 2).109 Ten of 13 cases in which the data have been reported presented multiple sites of venous thrombosis, particularly of the splanchnic, portal, and hepatic veins and pulmonary embolism, while only 3 of them had CVST. Fifteen patients had low platelet count at admission (mean±SD, 70.33±54×109/L; range, 9–133×109/L), while 1 patient had 217×109/L platelets at admission, with subsequent decrease to a nadir of 152×109/L 2 days after stroke onset. Very high level of D-dimer (mean value, 21 580 μg/L; normal range, 0–550) was present in all 14 patients in which the data have been reported. ELISA for PF4 autoantibodies was positive in 14 of the 16 tested patients. In 1 negative patient at baseline, high levels of antibodies were found at day 15 from admission. Unfortunately, it is not possible to draw a precise follow-up of these patients since outcome at 3 months has only been reported for case 8. Three of the 16 patients died. Patient 6, who survived and was described from our group,109 died 2 months later from an unexpected cardiac arrest while she was hospitalized in a rehabilitation center. Brain computed tomography scan did not show any new vascular events, and platelet counts were in the normal range. Her relatives denied autopsy. Only 1 patient (case 10) with distal MCA occlusion received intravenous thrombolysis with alteplase since platelet count was in the normal range, while 3 patients underwent successful mechanical thrombectomy. Case 6, reported from our group,109 underwent a second mechanical thrombectomy 2 hours after a first successful endovascular procedure (and 3 hours after the symptom onset), due to worsening of the neurological conditions and evidence on brain magnetic resonance imaging of a reocclusion of the same vessel (M1 segment of right MCA), with salvageable penumbra. Unexpectedly, despite a second complete reperfusion of the MCA territory, the patient developed a malignant infarct because of a third reocclusion of the right MCA and extension of the clot to the ipsilateral internal carotid artery terminus 12 hours apart (Figure 3). We hypothesized that the postthrombectomy injured arterial endothelium, combined with the high prothrombotic state and endothelium dysfunction characteristic of VITT, could have been the cause of the repeated arterial occlusions of the same vessel. Unfortunately, at the time the patient was admitted to our hospital, articles on VITT had not still been published and, due to very low baseline platelet count (44×109/L), platelet transfusion was performed before the first thrombectomy, which may have contributed to exacerbate the thrombotic event. At present, guidelines recommend that prophylactic platelet transfusions should be avoided in the context of VITT but should be provided before major surgical interventions (ie, hemicraniectomy) or if life-threatening bleeding is present.87,105 Figure 2. Radiological findings from patient 7. Case reported by De Michele et al.109 A, Computed tomography (CT) demonstrated extensive ischemic changes in the bilateral middle cerebral artery (MCA) distribution with general hypodensity and loss of gray-white matter differentiation. CT angiography showed the occlusion of the right internal carotid artery terminus (white arrow in B) and the proximal M1 segment occlusion of the left MCA (white arrow in C); time-to-maximum (D) and mean transit time (E) in CT perfusion showed hypoperfusion without treatable penumbra; pulmonary artery thrombosis (white arrow in F) with pulmonary consolidation in the right lobe (G). Figure 3. Radiological findings from patient 6. Case reported by De Michele et al.109 A, Computed tomography (CT) showed hyperdensity in the right middle cerebral artery. The CT angiography revealed proximal M1 segment occlusion of the right middle cerebral artery (MCA; white arrow in B and C); CT perfusion maps showed a large area of mismatch indicating salvageable penumbra (D); digital subtraction angiography confirming a proximal MCA occlusion (F) of the MCA occlusion (white arrow in E); MCA reocclusion on M2 segment 2 h after the procedure, 3-dimensional time-of-flight magnetic resonance imaging (MRI) sequence (white arrow in G), with extensive ischemic penumbra (time to peak map in H and cerebral blood flow in I). Second endovascular recanalization, oblique views showed occlusion of M2 segment (white arrow in J) with reopening of the vessel after the mechanical thrombectomy (K). Fourteen-day MRI follow-up after craniectomy showed the extension of ischemia to superficial and deep right MCA territory (M) with occlusion of right internal carotid artery at postcontrast sequences (yellow circle in L). N, Right portal vein thrombosis (black arrow). Additional Information Through the abovementioned systematic review of the literature, we also identified 2 additional studies relevant for the scope of this review, 1 from United Kingdom and 1 from Germany. A prospective multicenter cohort study from the United Kingdom evaluated VITT patients who presented to the hospital between March 22 and June 6, 2021.102 Among 220 patients with diagnosis of VITT classified as definite (ie, all 5 of the following criteria: [1] onset of symptoms 5–30 days after vaccination against SARS-CoV-2; [2] presence of thrombosis; [3] thrombocytopenia [platelet count <150 000 per mm3]; [4] D-dimer level >4000 FEU; [5] positive anti-PF4 antibodies on ELISA) or probable (ie, D-dimer level >4000 FEU but absence of one of the abovementioned criteria or D-dimer level unknown or 2000–4000 FEU and the presence of all other criteria), 17 subjects (7.7%) experienced cerebrovascular accidents. Unfortunately, no more details about these 17 patients have been reported by the authors, since data reported in the article are cumulative and related to all 220 VITT patients. The second study comes from the German Society of Neurology SARS-CoV-2 Vaccination Study Group.101 Nine cases of ischemic stroke (mean age, 55.6; range, 31.0–82.0) of 62 cerebrovascular events (14.5%) within 31 days from a first dose of COVID-19 vaccination have been reported in the December 28, 2020, to April 14, 2021, time period. Eight of these patients had received the ChAdOx1 ncov19 vaccination, and only 1 received the BNT162b2 vaccine. Majority of cases were females (66.7%). Among these 9 ischemic patients, 5 with embolic stroke had a VITT score of >2 (which means a highly probable VITT defined as the presence of the following 2 criteria: [1] time from shot administration between 1 and 16 days; [2] thrombocytopenia, <150×109/L or relative thrombocytopenia, drop of thrombocytes of at least 50%; [3] positive ELISA to detect PF4-polyanion antibodies; [4] positive modified platelet activation assay) without signs of CVST. In four of them, thrombotic occlusion of the MCA or internal carotid artery and recurrent thrombotic material in duplex ultrasound were described. Therapeutic Implications Therapeutic approach of acute ischemic stroke due to large vessel occlusion in VITT is challenging. Based on the available literature89,110–119 and our personal experience,109 we propose the following management protocol for acute stroke patients presenting to the Emergency Department within the time window for reperfusion strategies (Figure 4): Figure 4. Management flowchart of patients with acute ischemic stroke and suspected vaccine-induced thrombotic thrombocytopenia presenting to the Emergency Department within the time windows for reperfusion interventions. CT indicates computed tomography; CTA, computed tomography angiography; DTI, direct thrombin inhibitor; IVIG, intravenous immunoglobulin; MCA, middle cerebral artery; MRI, magnetic resonance imaging; and PF4, platelet factor 4. According to the current international guidelines. Keep in mind that ischemic stroke can be the first presentation symptom at onset of VITT. If a patient has received the first dose of DNA vector vaccination against SARS-CoV-2 within the previous 5 to 30 days, wait for platelet count results before starting thrombolysis. If large vessel occlusion is evident at cerebral computed tomography angiography without signs of malignant MCA infarct, mechanical thrombectomy is indicated according to guidelines from professional medical societies. If low platelet count is evident, avoid platelet transfusion and consider steroid administration (prednisone 1–2 mg/kg per day or dexamethasone 40 mg/day for 4 days) possibly before the endovascular procedure, monitoring blood pressure and blood glucose. Monitor the patient closely after mechanical thrombectomy since the risk of reocclusion and neurological deterioration is high. Schedule a control brain computed tomography scan or magnetic resonance imaging in the next 12 hours to decide the timing for starting anticoagulation. Start early full-dose anticoagulation with oral or parenteral direct thrombin inhibitors, or oral factor Xa inhibitors, or fondaparinux, only if brain infarct is small. If brain infarct is large, start with a reduced dose of anticoagulant (ie, fondaparinux, 2.5 mg daily) and increase the dosage after 2 weeks from stroke onset (ie, fondaparinux, 7.5 mg daily), due to the high risk of hemorrhagic transformation of the ischemic lesion. Thrombocytopenia seems not to be a contraindication to therapeutic dose anticoagulation in VITT, since subjects with the lowest platelet count are at the highest risk of thrombosis.87 However, some of the available current guidelines suggest low-dose anticoagulants if platelet counts are <30 to 50×109/L.87 Consider IVIG treatment (1 g/kg for 2 consecutive days) immediately after reperfusion therapies if VITT diagnosis is probable (thrombosis, thrombocytopenia, high D-dimer after vaccination), without awaiting confirmation from PF4 antibodies ELISA immunoassay. Repeated IVIG may be required. Perform anti-PF4 ELISA immunoassay and functional assay of platelet activation as soon as possible to confirm diagnosis (blood sample for functional assay should be obtained before IVIG administration since IVIG inhibits functional immunoassay).87 Consider plasma exchange (daily for up to ≥5 days) if extensive thrombosis and platelet count is <30×109/L. Consider rituximab for patients who are refractory to repeat doses of IVIG and plasma exchange, although evidence of its efficacy in VITT is scarce. Expert consultation from hematologist is necessary. Pathophysiology of Stroke in COVID-19 and VITT: Similarities and Differences COVID-19 and VITT show some common elements that lead to hypercoagulability and vascular occlusion. The abnormal interaction between platelets, innate immune effectors (neutrophils, macrophages, and complement), and coagulation factors are the key features of both pathological conditions. The ultimate consequence yields clot formation—a phenomenon known as thromboinflammation.120 In contrast to patients with COVID-19 and ischemic stroke, VITT patients do not show the typical elevations in interleukins. Rather, as discussed, the combination between soluble SP, adenovirus, and vaccine excipients probably acts as trigger for platelet activation.99 This distinct mechanism yields very different histopathologic findings postmortem. Specifically, postmortem studies on VITT patients found diffuse vascular thrombosis with endothelial activation, dense recruitment of inflammatory cells, and complement pathway activation in multiple organs.121 While thrombocytopenia is a typical finding of VITT, it is less frequently seen in patients with COVID-19 but is associated with increased risk of serious illness and death.122 Mechanisms of thrombocytopenia in COVID-19 are different, however, and are speculated to be mostly secondary to cytokine release, viral bone marrow infiltration, and increased platelet consumption.123 More rarely, thrombocytopenia in patients with COVID-19 can result from anti-PF4 antibody production as a complication of prolonged exposure to unfractionated heparin. There are also reports of platelet internalization of SARS-CoV-2 inducing platelet apoptosis, release of granular content, reduced platelet functionality, and high prothrombotic and proinflammatory immune response.124 Other common characteristics to both pathologies are the marked endothelial activation with elevated VWF, coagulation abnormalities (which can culminate to disseminated intravascular coagulation), and increased production of neutrophil extracellular traps.123,125 A comparison is summarized in Table 3. Table 3. Comparison Between COVID-19 Critically Ill and VITT Patients Conclusions Stroke and other cerebrovascular complications of SARS-CoV-2 infection are a highly morbid problem, with multifactorial pathophysiology. Given the association between severe SARS-CoV-2 infection and cardiovascular risk factors common to stroke, there is some degree of confounding; however, multiple studies at this point suggest that SARS-CoV-2 infection is an independent risk factor for ischemic stroke.4 Cerebrovascular complications of vaccination, specifically VITT is a rare but devastating syndrome occurring more frequently in young people after inoculation of DNA adenoviral vector vaccines that should be promptly recognized. The VITT variant causing arterial stroke is an even rarer but catastrophic event, whose management in the acute phase is complex and challenging. Some European Union countries have restricted the use of adenovirus vector vaccines to older age groups. In Italy, the Government’s Technical and Scientific Committee has limited the use of Oxford-AstraZeneca vaccines to people over 60 years of age, whereas in the United Kingdom, the Joint Committee on Vaccination and Immunization recommended that the Oxford-AstraZeneca vaccine should not be given to people under 40 years of age. Canada and France have restricted the use of this vaccine to people 55 years of age and over, while Germany has set the bar at 60 and Iceland at 70 years of age. Despite this, to date, studies have demonstrated that the risk of stroke and other prespecified outcomes of interest (thrombocytopenia, venous thromboembolism, arterial thrombosis, CVST, and myocardial infarction) following a SARS-CoV-2 infection were significantly higher than following vaccination with either the Oxford-AstraZeneca or Pfizer vaccines.67 As such, because benefits of mass vaccination against COVID-19 far outweighed the risks of VITT, no age restrictions were announced either by the European Medicines Agency or the Food and Drug Administration. A global immunization campaign is urgently needed, particularly in low-income countries, and all currently available vaccines are approved by emergency authorizations. Nevertheless, more studies about the pathogenesis of VITT are mandatory to ameliorate the risk of adenovirus-based vaccines and to identify those most at risk of VITT. Article Information Acknowledgments We would like to acknowledge Natalie Pacheco LeMoss and Jed Kaiser for their assistance with formatting and editing of this article. Disclosures None. Nonstandard Abbreviations and Acronyms ACE2 angiotensin-converting enzyme 2 CVST cerebral venous sinus thrombosis HIT heparin-induced thrombocytopenia MCA middle cerebral artery PF4 platelet factor 4 SP spike protein TTS thrombosis with thrombocytopenia syndrome VITT vaccine-induced immune thrombotic thrombocytopenia VWF von Willebrand factor D. Toni and A.E. Merkler contributed equally. For Disclosures, see page 1200. ==== Refs References 1. WHO, Others. COVID-19 weekly epidemiological update. Accessed November 12, 2021. https://apps.who.int/iris/bitstream/handle/10665/341525/CoV-weekly-sitrep25May21-eng.pdf?sequence=1 2. Nopp S Moik F Jilma B Pabinger I Ay C . Risk of venous thromboembolism in patients with COVID-19: a systematic review and meta-analysis. Res Pract Thromb Haemost. 2020;4 :1178–1191. doi: 10.1002/rth2.12439 3. Mao L Jin H Wang M Hu Y Chen S He Q Chang J Hong C Zhou Y Wang D . Neurologic manifestations of hospitalized patients with coronavirus disease 2019 in Wuhan, China. JAMA Neurol. 2020;77 :683–690. doi: 10.1001/jamaneurol.2020.1127 32275288 4. Katsanos AH Palaiodimou L Zand R Yaghi S Kamel H Navi BB Turc G Romoli M Sharma VK Mavridis D . The impact of SARS-CoV-2 on stroke epidemiology and care: a meta-analysis. Ann Neurol. 2021;89 :380–388. doi: 10.1002/ana.25967 33219563 5. Merkler AE Parikh NS Mir S Gupta A Kamel H Lin E Lantos J Schenck EJ Goyal P Bruce SS . Risk of ischemic stroke in patients with coronavirus Disease 2019 (COVID-19) vs patients with influenza. JAMA Neurol. 2020;77 :1–7. doi: 10.1001/jamaneurol.2020.2730 6. Leasure AC Khan YM Iyer R Elkind MSV Sansing LH Falcone GJ Sheth KN . Intracerebral hemorrhage in patients with COVID-19: an analysis from the COVID-19 cardiovascular disease registry. Stroke. 2021;52 :e321–e323. doi: 10.1161/STROKEAHA.121.034215 34082576 7. Abdalkader M Shaikh SP Siegler JE Cervantes-Arslanian AM Tiu C Radu RA Tiu VE Jillella DV Mansour OY Vera V . Cerebral venous sinus thrombosis in COVID-19 patients: a multicenter study and review of literature. J Stroke Cerebrovasc Dis. 2021;30 :105733. doi: 10.1016/j.jstrokecerebrovasdis.2021.105733 33743411 8. Parauda SC Gao V Gewirtz AN Parikh NS Merkler AE Lantos J White H Leifer D Navi BB Segal AZ . Posterior reversible encephalopathy syndrome in patients with COVID-19. J Neurol Sci. 2020;416 :117019. doi: 10.1016/j.jns.2020.117019 32679347 9. Oxley TJ Mocco J Majidi S Kellner CP Shoirah H Singh IP De Leacy RA Shigematsu T Ladner TR Yaeger KA . Large-vessel stroke as a presenting feature of Covid-19 in the young. N Engl J Med. 2020;382 :e60. doi: 10.1056/NEJMc2009787 32343504 10. Li Y Li M Wang M Zhou Y Chang J Xian Y Wang D Mao L Jin H Hu B . Acute cerebrovascular disease following COVID-19: a single center, retrospective, observational study. Stroke Vasc Neurol. 2020;5 :279–284. doi: 10.1136/svn-2020-000431 32616524 11. Yaghi S Ishida K Torres J Mac Grory B Raz E Humbert K Henninger N Trivedi T Lillemoe K Alam S . SARS-CoV-2 and stroke in a New York Healthcare System. Stroke. 2020;51 :2002–2011. doi: 10.1161/STROKEAHA.120.030335 32432996 12. Lodigiani C Iapichino G Carenzo L Cecconi M Ferrazzi P Sebastian T Kucher N Studt JD Sacco C Bertuzzi A ; Humanitas COVID-19 Task Force. Venous and arterial thromboembolic complications in COVID-19 patients admitted to an academic hospital in Milan, Italy. Thromb Res. 2020;191 :9–14. doi: 10.1016/j.thromres.2020.04.024 32353746 13. Helms J Kremer S Merdji H Clere-Jehl R Schenck M Kummerlen C Collange O Boulay C Fafi-Kremer S Ohana M . Neurologic features in severe SARS-CoV-2 infection. N Engl J Med. 2020;382 :2268–2270. doi: 10.1056/NEJMc2008597 32294339 14. Siepmann T Sedghi A Simon E Winzer S Barlinn J de With K Mirow L Wolz M Gruenewald T Schroettner P . Increased risk of acute stroke among patients with severe COVID-19: a multicenter study and meta-analysis. Eur J Neurol. 2021;28 :238–247. doi: 10.1111/ene.14535 32920964 15. Rothstein A Oldridge O Schwennesen H Do D Cucchiara BL . Acute cerebrovascular events in hospitalized COVID-19 patients. Stroke. 2020;51 :e219–e222. doi: 10.1161/STROKEAHA.120.030995 32684145 16. Dhamoon MS Thaler A Gururangan K Kohli A Sisniega D Wheelwright D Mensching C Fifi JT Fara MG Jette N ; Mount Sinai Stroke Investigators*. Acute cerebrovascular events with COVID-19 infection. Stroke. 2021;52 :48–56. doi: 10.1161/STROKEAHA.120.031668 33280551 17. Shahjouei S Naderi S Li J Khan A Chaudhary D Farahmand G Male S Griessenauer C Sabra M Mondello S . Risk of stroke in hospitalized SARS-CoV-2 infected patients: a multinational study. EBioMedicine. 2020;59 :102939. doi: 10.1016/j.ebiom.2020.102939 32818804 18. Nogueira RG Abdalkader M Qureshi MM Frankel MR Mansour OY Yamagami H Qiu Z Farhoudi M Siegler JE Yaghi S . Global impact of COVID-19 on stroke care. Int J Stroke. 2021;16 :573–584. doi: 10.1177/1747493021991652 33459583 19. Qureshi AI Baskett WI Huang W Shyu D Myers D Raju M Lobanova I Suri MFK Naqvi SH French BR . Acute ischemic stroke and COVID-19: an analysis of 27 676 patients. Stroke. 2021;52 :905–912. doi: 10.1161/STROKEAHA.120.031786 33535779 20. Katz JM Libman RB Wang JJ Sanelli P Filippi CG Gribko M Pacia SV Kuzniecky RI Najjar S Azhar S . Cerebrovascular complications of COVID-19. Stroke. 2020;51 :e227–e231. doi: 10.1161/STROKEAHA.120.031265 32757751 21. Shtaya A Trippier S Ghatala R Cluckie G Zhang L . Comment on “Stroke in patients with SARS-CoV-2 infection: case series” from a London hospital experience. J Neurol. 2021;268 :424–430. doi: 10.1007/s00415-020-10105-0 32712866 22. Qiu F Wu Y Zhang A Xie G Cao H Du M Jiang H Li S Ding M . Changes of coagulation function and risk of stroke in patients with COVID-19. Brain Behav. 2021;11 :e02185. doi: 10.1002/brb3.2185 33998177 23. Modin D Claggett B Sindet-Pedersen C Lassen MCH Skaarup KG Jensen JUS Fralick M Schou M Lamberts M Gerds T . Acute COVID-19 and the incidence of ischemic stroke and acute myocardial infarction. Circulation. 2020;142 :2080–2082. doi: 10.1161/CIRCULATIONAHA.120.050809 33054349 24. Katsoularis I Fonseca-Rodríguez O Farrington P Lindmark K Fors Connolly AM . Risk of acute myocardial infarction and ischaemic stroke following COVID-19 in Sweden: a self-controlled case series and matched cohort study. Lancet. 2021;398 :599–607. doi: 10.1016/S0140-6736(21)00896-5 34332652 25. Benussi A Pilotto A Premi E Libri I Giunta M Agosti C Alberici A Baldelli E Benini M Bonacina S . Clinical characteristics and outcomes of inpatients with neurologic disease and COVID-19 in Brescia, Lombardy, Italy. Neurology. 2020;95 :e910–e920. doi: 10.1212/WNL.0000000000009848 32444493 26. Shen J Hou Y Zhou Y Mehra R Jehi L Cheng F . The epidemiological and mechanistic understanding of the neurological manifestations of COVID-19: a comprehensive meta-analysis and a network medicine observation. Front Neurosci. 2021;15 :606926. doi: 10.3389/fnins.2021.606926 33732102 27. Annie F Bates MC Nanjundappa A Bhatt DL Alkhouli M . Prevalence and outcomes of acute ischemic stroke among patients ≤50 Years of age with laboratory confirmed COVID-19 infection. Am J Cardiol. 2020;130 :169–170. doi: 10.1016/j.amjcard.2020.06.010 32690214 28. Esenwa C Cheng NT Luna J Willey J Boehme AK Kirchoff-Torres K Labovitz D Liberman AL Mabie P Moncrieffe K . Biomarkers of coagulation and inflammation in COVID-19-Associated ischemic stroke. Stroke. 2021;52 :e706–e709. doi: 10.1161/STROKEAHA.121.035045 34428931 29. Kim Y Khose S Abdelkhaleq R Salazar-Marioni S Zhang GQ Sheth SA . Predicting in-hospital mortality using D-dimer in COVID-19 patients with acute ischemic stroke. Front Neurol. 2021;12 :702927. doi: 10.3389/fneur.2021.702927 34335456 30. Ntaios G Michel P Georgiopoulos G Guo Y Li W Xiong J Calleja P Ostos F González-Ortega G Fuentes B . Characteristics and outcomes in patients with COVID-19 and acute ischemic stroke: the global COVID-19 stroke registry. Stroke. 2020;51 :e254–e258. doi: 10.1161/STROKEAHA.120.031208 32787707 31. Perry RJ Smith CJ Roffe C Simister R Narayanamoorthi S Marigold R Willmot M Dixit A Hassan A Quinn TJ ; SETICOS Collaborators. Characteristics and outcomes of COVID-19 associated stroke: a UK multicentre case-control study. J Neurol Neurosurg Psychiatry. 2021;92 :242–248. doi: 10.1136/jnnp-2020-324927 33154179 32. de Havenon A Ney JP Callaghan B Delic A Hohmann S Shippey E Esper GJ Stulberg E Tirschwell D Frontera J . Impact of COVID-19 on outcomes in ischemic stroke patients in the United States. J Stroke Cerebrovasc Dis. 2021;30 :105535. doi: 10.1016/j.jstrokecerebrovasdis.2020.105535 33310595 33. Yao X Liu S Wang J Zhao K Long X He X Kang H Yang Y Ma X Yue P . The clinical characteristics and prognosis of COVID-19 patients with cerebral stroke: a retrospective study of 113 cases from one single-centre. Eur J Neurosci. 2021;53 :1350–1361. doi: 10.1111/ejn.15007 33052619 34. Hernández-Fernández F Sandoval Valencia H Barbella-Aponte RA Collado-Jiménez R Ayo-Martín Ó Barrena C Molina-Nuevo JD García-García J Lozano-Setién E Alcahut-Rodriguez C . Cerebrovascular disease in patients with COVID-19: neuroimaging, histological and clinical description. Brain. 2020;143 :3089–3103. doi: 10.1093/brain/awaa239 32645151 35. Majidi S Fifi JT Ladner TR Lara-Reyna J Yaeger KA Yim B Dangayach N Oxley TJ Shigematsu T Kummer BR . Emergent large vessel occlusion stroke during New York City’s COVID-19 outbreak: clinical characteristics and paraclinical findings. Stroke. 2020;51 :2656–2663. doi: 10.1161/STROKEAHA.120.030397 32755349 36. Escalard S Chalumeau V Escalard C Redjem H Delvoye F Hébert S Smajda S Ciccio G Desilles JP Mazighi M . Early brain imaging shows increased severity of acute ischemic strokes with large vessel occlusion in COVID-19 patients. Stroke. 2020;51 :3366–3370. doi: 10.1161/STROKEAHA.120.031011 32813602 37. Kihira S Schefflein J Mahmoudi K Rigney B N Delman B Mocco J Doshi A Belani P . Association of coronavirus disease (COVID-19) with large vessel occlusion strokes: a case-control study. AJR Am J Roentgenol. 2021;216 :150–156. doi: 10.2214/AJR.20.23847 32755225 38. Hart RG Catanese L Perera KS Ntaios G Connolly SJ . Embolic stroke of undetermined source: a systematic review and clinical update. Stroke. 2017;48 :867–872. doi: 10.1161/STROKEAHA.116.016414 28265016 39. John S Kesav P Mifsud VA Piechowski-Jozwiak B Dibu J Bayrlee A Elkambergy H Roser F Elhammady MS Zahra K . Characteristics of large-vessel occlusion associated with COVID-19 and ischemic stroke. AJNR Am J Neuroradiol. 2020;41 :2263–2268. doi: 10.3174/ajnr.A6799 32855182 40. McAlpine LS Zubair AS Maran I Chojecka P Lleva P Jasne AS Navaratnam D Matouk C Schindler J Sheth KN . Ischemic stroke, inflammation, and endotheliopathy in COVID-19 patients. Stroke. 2021;52 :e233–e238. doi: 10.1161/STROKEAHA.120.031971 33966492 41. Finn C Hung P Patel P Gupta A Kamel H . Relationship between visceral infarction and ischemic stroke subtype. Stroke. 2018;49 :727–729. doi: 10.1161/STROKEAHA.117.020035 29371436 42. Allegra A Innao V Allegra AG Musolino C . Coagulopathy and thromboembolic events in patients with SARS-CoV-2 infection: pathogenesis and management strategies. Ann Hematol. 2020;99 :1953–1965. doi: 10.1007/s00277-020-04182-4 32671455 43. Piazza G Campia U Hurwitz S Snyder JE Rizzo SM Pfeferman MB Morrison RB Leiva O Fanikos J Nauffal V . Registry of Arterial and Venous Thromboembolic Complications in Patients With COVID-19. J Am Coll Cardiol. 2020;76 :2060–2072. doi: 10.1016/j.jacc.2020.08.070 33121712 44. Helms J Tacquard C Severac F Leonard-Lorant I Ohana M Delabranche X Merdji H Clere-Jehl R Schenck M Fagot Gandet F ; CRICS TRIGGERSEP Group (Clinical Research in Intensive Care and Sepsis Trial Group for Global Evaluation and Research in Sepsis). High risk of thrombosis in patients with severe SARS-CoV-2 infection: a multicenter prospective cohort study. Intensive Care Med. 2020;46 :1089–1098. doi: 10.1007/s00134-020-06062-x 32367170 45. Linschoten M Peters S van Smeden M Jewbali LS Schaap J Siebelink HM Smits PC Tieleman RG van der Harst P van Gilst WH ; CAPACITY-COVID Collaborative Consortium. Cardiac complications in patients hospitalised with COVID-19. Eur Heart J Acute Cardiovasc Care. 2020;9 :817–823. doi: 10.1177/2048872620974605 33222494 46. Wolf PA Kannel WB McGee DL Meeks SL Bharucha NE McNamara PM . Duration of atrial fibrillation and imminence of stroke: the Framingham study. Stroke. 1983;14 :664–667. doi: 10.1161/01.str.14.5.664 6658948 47. Solomon MD McNulty EJ Rana JS Leong TK Lee C Sung SH Ambrosy AP Sidney S Go AS . The Covid-19 pandemic and the incidence of acute myocardial infarction. N Engl J Med. 2020;383 :691–693. doi: 10.1056/NEJMc2015630 32427432 48. Haussner W DeRosa AP Haussner D Tran J Torres-Lavoro J Kamler J Shah K . COVID-19 associated myocarditis: a systematic review. Am J Emerg Med. 2022;51 :150–155. doi: 10.1016/j.ajem.2021.10.001 34739868 49. Ramasamy S Yaghi S Salehi Omran S Lerario MP Devereux R Okin PM Gupta A Navi BB Kamel H Merkler AE . Association between left ventricular ejection fraction, wall motion abnormality, and embolic stroke of undetermined source. J Am Heart Assoc. 2019;8 :e011593. doi: 10.1161/JAHA.118.011593 31057030 50. Hou C Hu Y Yang H Chen W Zeng Y Ying Z Hu Y Sun Y Qu Y Gottfreðsson M . COVID-19 and risk of subsequent life-threatening secondary infections: a matched cohort study in UK Biobank. BMC Med. 2021;19 :301. doi: 10.1186/s12916-021-02177-0 34781951 51. Shao IY Elkind MSV Boehme AK . Risk Factors for stroke in patients with sepsis and bloodstream infections. Stroke. 2019;50 :1046–1051. doi: 10.1161/STROKEAHA.118.023443 30896327 52. Zhang S Liu Y Wang X Yang L Li H Wang Y Liu M Zhao X Xie Y Yang Y . SARS-CoV-2 binds platelet ACE2 to enhance thrombosis in COVID-19. J Hematol Oncol. 2020;13 :120. doi: 10.1186/s13045-020-00954-7 32887634 53. Letarov AV Babenko VV Kulikov EE . Free SARS-CoV-2 spike protein S1 particles may play a role in the pathogenesis of COVID-19 infection. Biochemistry (Mosc). 2021;86 :257–261. doi: 10.1134/S0006297921030032 33838638 54. Ackermann M Verleden SE Kuehnel M Haverich A Welte T Laenger F Vanstapel A Werlein C Stark H Tzankov A . Pulmonary vascular endothelialitis, thrombosis, and angiogenesis in Covid-19. N Engl J Med. 2020;383 :120–128. doi: 10.1056/NEJMoa2015432 32437596 55. Lee MH Perl DP Nair G Li W Maric D Murray H Dodd SJ Koretsky AP Watts JA Cheung V . Microvascular injury in the brains of patients with Covid-19. N Engl J Med. 2021;384 :481–483. doi: 10.1056/NEJMc2033369 33378608 56. Mukerji SS Solomon IH . What can we learn from brain autopsies in COVID-19? Neurosci Lett. 2021;742 :135528. doi: 10.1016/j.neulet.2020.135528 33248159 57. Bryce C Grimes Z Pujadas E Ahuja S Beasley MB Albrecht R Hernandez T Stock A Zhao Z AlRasheed MR . Pathophysiology of SARS-CoV-2: the Mount Sinai COVID-19 autopsy experience. Mod Pathol. 2021;34 :1456–1467. doi: 10.1038/s41379-021-00793-y 33795830 58. Solomon IH Normandin E Bhattacharyya S Mukerji SS Keller K Ali AS Adams G Hornick JL Padera RF Jr Sabeti P . Neuropathological features of Covid-19. N Engl J Med. 2020;383 :989–992. doi: 10.1056/NEJMc2019373 32530583 59. Paniz-Mondolfi A Bryce C Grimes Z Gordon RE Reidy J Lednicky J Sordillo EM Fowkes M . Central nervous system involvement by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2). J Med Virol. 2020;92 :699–702. doi: 10.1002/jmv.25915 32314810 60. Miller SE Goldsmith CS . Caution in identifying coronaviruses by electron microscopy. J Am Soc Nephrol. 2020;31 :2223–2224. doi: 10.1681/ASN.2020050755 32651224 61. Varga Z Flammer AJ Steiger P Haberecker M Andermatt R Zinkernagel AS Mehra MR Schuepbach RA Ruschitzka F Moch H . Endothelial cell infection and endotheliitis in COVID-19. Lancet. 2020;395 :1417–1418. doi: 10.1016/S0140-6736(20)30937-5 32325026 62. Evans PC Rainger GE Mason JC Guzik TJ Osto E Stamataki Z Neil D Hoefer IE Fragiadaki M Waltenberger J . Endothelial dysfunction in COVID-19: a position paper of the ESC Working Group for Atherosclerosis and Vascular Biology, and the ESC Council of Basic Cardiovascular Science. Cardiovasc Res. 2020;116 :2177–2184. doi: 10.1093/cvr/cvaa230 32750108 63. Klok FA Kruip MJHA van der Meer NJM Arbous MS Gommers D Kant KM Kaptein FHJ van Paassen J Stals MAM Huisman MV . Confirmation of the high cumulative incidence of thrombotic complications in critically ill ICU patients with COVID-19: an updated analysis. Thromb Res. 2020;191 :148–150. doi: 10.1016/j.thromres.2020.04.041 32381264 64. Huisman A Beun R Sikma M Westerink J Kusadasi N . Involvement of ADAMTS13 and von Willebrand factor in thromboembolic events in patients infected with SARS-CoV-2. Int J Lab Hematol. 2020;42 :e211–e212. doi: 10.1111/ijlh.13244 32441844 65. South K McCulloch L McColl BW Elkind MS Allan SM Smith CJ . Preceding infection and risk of stroke: an old concept revived by the COVID-19 pandemic. Int J Stroke. 2020;15 :722–732. doi: 10.1177/1747493020943815 32618498 66. Kvernland A Kumar A Yaghi S Raz E Frontera J Lewis A Czeisler B Kahn DE Zhou T Ishida K . Anticoagulation use and hemorrhagic stroke in SARS-CoV-2 patients treated at a New York Healthcare System. Neurocrit Care. 2021;34 :748–759. doi: 10.1007/s12028-020-01077-0 32839867 67. Hippisley-Cox J Patone M Mei XW Saatci D Dixon S Khunti K Zaccardi F Watkinson P Shankar-Hari M Doidge J . Risk of thrombocytopenia and thromboembolism after covid-19 vaccination and SARS-CoV-2 positive testing: self-controlled case series study. BMJ. 2021;374 :n1931. doi: 10.1136/bmj.n1931 34446426 68. Mowla A Shakibajahromi B Shahjouei S Borhani-Haghighi A Rahimian N Baharvahdat H Naderi S Khorvash F Altafi D Ebrahimzadeh SA . Cerebral venous sinus thrombosis associated with SARS-CoV-2; a multinational case series. J Neurol Sci. 2020;419 :117183. doi: 10.1016/j.jns.2020.117183 33075595 69. Hixon AM Thaker AA Pelak VS . Persistent visual dysfunction following posterior reversible encephalopathy syndrome due to COVID-19: case series and literature review. Eur J Neurol. 2021;28 :3289–3302. doi: 10.1111/ene.14965 34115917 70. COVID-19 vaccines [Internet]. Accessed November 12, 2021. https://vaccination-info.eu/en/covid-19/covid-19-vaccines 71. Office of the Commissioner. Learn more about COVID-19 vaccines from the FDA [Internet]. Authored on behalf of The United States Food and Drug Administration. Accessed November 12, 2021. https://www.fda.gov/consumers/consumer-updates/learn-more-about-covid-19-vaccines-fda 72. Rosenblum HG Hadler SC Moulia D Shimabukuro TT Su JR Tepper NK Ess KC Woo EJ Mba-Jonas A Alimchandani M . Use of COVID-19 vaccines after reports of adverse events among adult recipients of Janssen (Johnson & Johnson) and mRNA COVID-19 vaccines (Pfizer-BioNTech and Moderna): update from the advisory committee on immunization practices - United States, July 2021. MMWR Morb Mortal Wkly Rep. 2021;70 :1094–1099. doi: 10.15585/mmwr.mm7032e4 34383735 73. See I Su JR Lale A Woo EJ Guh AY Shimabukuro TT Streiff MB Rao AK Wheeler AP Beavers SF . US case reports of cerebral venous sinus thrombosis with thrombocytopenia after Ad26.COV2.S vaccination, March 2 to April 21, 2021. JAMA. 2021;325 :2448–2456. doi: 10.1001/jama.2021.7517 33929487 74. MacIntyre CR . Using the Bradford-Hill criteria to assess causality in the association between CHADOX1 NCOV-19 vaccine and thrombotic immune thrombocytopenia. Glob Biosecur. 2021;3 . doi: 10.31646/gbio.109 75. Sadoff J Gray G Vandebosch A Cárdenas V Shukarev G Grinsztejn B Goepfert PA Truyers C Fennema H Spiessens B ; ENSEMBLE Study Group. Safety and efficacy of single-dose Ad26.COV2.S vaccine against Covid-19. N Engl J Med. 2021;384 :2187–2201. doi: 10.1056/NEJMoa2101544 33882225 76. Voysey M Clemens SAC Madhi SA Weckx LY Folegatti PM Aley PK Angus B Baillie VL Barnabas SL Bhorat QE ; Oxford COVID Vaccine Trial Group. Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK. Lancet. 2021;397 :99–111. doi: 10.1016/S0140-6736(20)32661-1 33306989 77. Palladino R Ceriotti D De Ambrosi D De Vito M Farsoni M Seminara G Barone-Adesi F . A quantitative risk-benefit analysis of ChAdOx1 nCoV-19 vaccine among people under 60 in Italy. medRxiv. Preprint posted online May 10, 2021. doi: 10.1101/2021.05.07.21256826 78. Yang Y Du L . SARS-CoV-2 spike protein: a key target for eliciting persistent neutralizing antibodies. Signal Transduct Target Ther. 2021;6 :95. doi: 10.1038/s41392-021-00523-5 33637679 79. Song W Gui M Wang X Xiang Y . Cryo-EM structure of the SARS coronavirus spike glycoprotein in complex with its host cell receptor ACE2. PLoS Pathog. 2018;14 :e1007236. doi: 10.1371/journal.ppat.1007236 30102747 80. Lan J Ge J Yu J Shan S Zhou H Fan S Zhang Q Shi X Wang Q Zhang L . Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor. Nature. 2020;581 :215–220. doi: 10.1038/s41586-020-2180-5 32225176 81. Turoňová B Sikora M Schürmann C Hagen WJH Welsch S Blanc FEC von Bülow S Gecht M Bagola K Hörner C . In situ structural analysis of SARS-CoV-2 spike reveals flexibility mediated by three hinges. Science. 2020;370 :203–208. doi: 10.1126/science.abd5223 32817270 82. Wajnberg A Amanat F Firpo A Altman DR Bailey MJ Mansour M McMahon M Meade P Mendu DR Muellers K . Robust neutralizing antibodies to SARS-CoV-2 infection persist for months. Science. 2020;370 :1227–1230. doi: 10.1126/science.abd7728 33115920 83. Silveira MM Moreira GMSG Mendonça M . DNA vaccines against COVID-19: perspectives and challenges. Life Sci. 2021;267 :118919. doi: 10.1016/j.lfs.2020.118919 33352173 84. Abbasi J . COVID-19 and mRNA vaccines-first large test for a new approach. JAMA. 2020;324 :1125–1127. doi: 10.1001/jama.2020.16866 32880613 85. Heath PT Galiza EP Baxter DN Boffito M Browne D Burns F Chadwick DR Clark R Cosgrove C Galloway J ; 2019nCoV-302 Study Group. Safety and efficacy of NVX-CoV2373 Covid-19 vaccine. N Engl J Med. 2021;385 :1172–1183. doi: 10.1056/NEJMoa2107659 34192426 86. Tregoning JS Flight KE Higham SL Wang Z Pierce BF . Progress of the COVID-19 vaccine effort: viruses, vaccines and variants versus efficacy, effectiveness and escape. Nat Rev Immunol. 2021;21 :626–636. doi: 10.1038/s41577-021-00592-1 34373623 87. Greinacher A Langer F Makris M Pai M Pavord S Tran H Warkentin TE . Vaccine-induced immune thrombotic thrombocytopenia (VITT): update on diagnosis and management considering different resources. J Thromb Haemost. 2022;20 :149–156. doi: 10.1111/jth.15572 34693641 88. Schultz NH Sørvoll IH Michelsen AE Munthe LA Lund-Johansen F Ahlen MT Wiedmann M Aamodt AH Skattør TH Tjønnfjord GE . Thrombosis and thrombocytopenia after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;384 :2124–2130. doi: 10.1056/NEJMoa2104882 33835768 89. Scully M Singh D Lown R Poles A Solomon T Levi M Goldblatt D Kotoucek P Thomas W Lester W . Pathologic antibodies to platelet factor 4 after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;384 :2202–2211. doi: 10.1056/NEJMoa2105385 33861525 90. Muir KL Kallam A Koepsell SA Gundabolu K . Thrombotic thrombocytopenia after Ad26.COV2.S vaccination. N Engl J Med. 2021;384 :1964–1965. doi: 10.1056/NEJMc2105869 33852795 91. Sangli S Virani A Cheronis N Vannatter B Minich C Noronha S Bhagavatula R Speredelozzi D Sareen M Kaplan RB . Thrombosis with thrombocytopenia after the messenger RNA-1273 vaccine. Ann Intern Med. 2021;174 :1480–1482. doi: 10.7326/L21-0244 34181446 92. Greinacher A Thiele T Warkentin TE Weisser K Kyrle PA Eichinger S . Thrombotic thrombocytopenia after ChAdOx1 nCov-19 vaccination. N Engl J Med. 2021;384 :2092–2101. doi: 10.1056/NEJMoa2104840 33835769 93. Greinacher A Selleng K Warkentin TE . Autoimmune heparin-induced thrombocytopenia. J Thromb Haemost. 2017;15 :2099–2114. doi: 10.1111/jth.13813 28846826 94. Mullier F Minet V Bailly N Devalet B Douxfils J Chatelain C Elalamy I Dogné JM Chatelain B . Platelet microparticle generation assay: a valuable test for immune heparin-induced thrombocytopenia diagnosis. Thromb Res. 2014;133 :1068–1073. doi: 10.1016/j.thromres.2013.12.009 24360929 95. Greinacher A Selleng K Mayerle J Palankar R Wesche J Reiche S Aebischer A Warkentin TE Muenchhoff M Hellmuth JC ; Immune-Response in COVID-19 Vaccination Study Group. Anti-platelet factor 4 antibodies causing VITT do not cross-react with SARS-CoV-2 spike protein. Blood. 2021;138 :1269–1277. doi: 10.1182/blood.2021012938 34280256 96. Uzun G Althaus K Bakchoul T . No correlation between anti-PF4 and anti–SARS-CoV-2 antibodies after ChAdOx1 nCoV-19 vaccination. N Engl J Med. 2021;385 :1334–1336. doi: 10.1056/NEJMc2111305 34432977 97. Baker AT Boyd RJ Sarkar D Teijeira-Crespo A Chan CK Bates E Waraich K Vant J Wilson E Truong CD . ChAdOx1 interacts with CAR and PF4 with implications for thrombosis with thrombocytopenia syndrome. Sci Adv. 2021;7 :eabl8213. doi: 10.1126/sciadv.abl8213 34851659 98. Greinacher A Selleng K Palankar R Wesche J Handtke S Wolff M Aurich K Lalk M Methling K Völker U . Insights in ChAdOx1 nCoV-19 vaccine-induced immune thrombotic thrombocytopenia. Blood. 2021;138 :2256–2268. doi: 10.1182/blood.2021013231 34587242 99. Kowarz E Krutzke L Külp M Streb P Larghero P Reis J, Bracharz S, Engler T, Kochanek S, Marschalek R . Vaccine-induced COVID-19 mimicry syndrome. e-life. 2022;11:e74974. 10.7554/eLife.74974 100. De Michele M Piscopo P Crestini A Rivabene R d’Amati G Leopizzi M Stefanini L Flego D Pulcinelli F Chistolini A . Vaccine-induced immune thrombotic thrombocytopenia and spike protein. Res Sq. 2021. doi: 10.21203/rs.3.rs-887779/v1 101. Schulz JB Berlit P Diener HC Gerloff C Greinacher A Klein C Petzold GC Piccininni M Poli S Röhrig R ; German Society of Neurology SARS-CoV-2 Vaccination Study Group. COVID-19 vaccine-associated cerebral venous thrombosis in germany. Ann Neurol. 2021;90 :627–639. doi: 10.1002/ana.26172 34288044 102. Pavord S Scully M Hunt BJ Lester W Bagot C Craven B Rampotas A Ambler G Makris M . Clinical features of vaccine-induced immune thrombocytopenia and thrombosis. N Engl J Med. 2021;385 :1680–1689. doi: 10.1056/NEJMoa2109908 34379914 103. van Kammen MS de Sousa DA Poli S Cordonnier C Heldner MR van de Munckhof A Krzywicka K van Haaps T Ciccone A Middeldorp S . Characteristics and outcomes of patients with cerebral venous sinus thrombosis in SARS-CoV-2 vaccine–induced immune thrombotic thrombocytopenia. JAMA Neurol. 2021;78 :1314–1323. doi: 10.1001/jamaneurol.2021.3619 34581763 104. Guidance produced from the Expert Haematology Panel (EHP) focussed on syndrome of thrombosis and thrombocytopenia occurring after coronavirus vaccination [Internet]. Authored on behalf of the Expert Haematology Panel. Accessed November 12, 2021. https://b-s-h.org.uk/about-us/news/guidance-produced-by-the-expert-haematology-panel-ehp-focussed-on-vaccine-induced-thrombosis-and-thrombocytopenia-vitt/ 105. The ISTH releases interim guidance on vaccine-induced immune thrombotic thrombocytopenia (VITT) - International Society on Thrombosis and Haemostasis Inc [Internet]. Authored on behalf of the The International Society on Thrombosis and Haemostasis. Accessed November 12, 2021. https://www.isth.org/news/561406/The-ISTH-Releases-Interim-Guidance-on-Vaccine-Induced-Immune-Thrombotic-Thrombocytopenia-VITT-.htm 106. Oldenburg J Klamroth R Langer F Albisetti M von Auer C Ay C Korte W Scharf RE Pötzsch B Greinacher A . Diagnosis and management of vaccine-related thrombosis following AstraZeneca COVID-19 vaccination: guidance statement from the GTH. Hamostaseologie. 2021;41 :184–189. doi: 10.1055/a-1469-7481 33822348 107. Lavin M Elder PT O’Keeffe D Enright H Ryan E Kelly A El Hassadi E McNicholl FP Benson G Le GN . Vaccine-induced immune thrombotic thrombocytopenia (VITT) - a novel clinico-pathological entity with heterogeneous clinical presentations. Br J Haematol. 2021;195 :76–84. doi: 10.1111/bjh.17613 34159588 108. Salih F Schönborn L Kohler S Franke C Möckel M Dörner T Bauknecht HC Pille C Graw JA Alonso A . Vaccine-induced thrombocytopenia with severe headache. N Engl J Med. 2021;385 :2103–2105. doi: 10.1056/NEJMc2112974 34525282 109. De Michele M Iacobucci M Chistolini A Nicolini E Pulcinelli F Cerbelli B Merenda E Schiavo OG Sbardella E Berto I . Malignant cerebral infarction after ChAdOx1 nCov-19 vaccination: a catastrophic variant of vaccine-induced immune thrombotic thrombocytopenia. Nat Commun. 2021;12 :4663. doi: 10.1038/s41467-021-25010-x 34341358 110. Blauenfeldt RA Kristensen SR Ernstsen SL Kristensen CCH Simonsen CZ Hvas AM . Thrombocytopenia with acute ischemic stroke and bleeding in a patient newly vaccinated with an adenoviral vector-based COVID-19 vaccine. J Thromb Haemost. 2021;19 :1771–1775. doi: 10.1111/jth.15347 33877737 111. Al-Mayhani T Saber S Stubbs MJ Losseff NA Perry RJ Simister RJ Gull D Jäger HR Scully MA Werring DJ . Ischaemic stroke as a presenting feature of ChAdOx1 nCoV-19 vaccine-induced immune thrombotic thrombocytopenia. J Neurol Neurosurg Psychiatry. 2021;92 :1247–1248. doi: 10.1136/jnnp-2021-326984 34035134 112. Bayas A Menacher M Christ M Behrens L Rank A Naumann M . Bilateral superior ophthalmic vein thrombosis, ischaemic stroke, and immune thrombocytopenia after ChAdOx1 nCoV-19 vaccination. Lancet. 2021;397 :e11. doi: 10.1016/S0140-6736(21)00872-2 33864750 113. Kenda J Lovrič D Škerget M Milivojević N . Treatment of ChAdOx1 nCoV-19 vaccine-induced immune thrombotic thrombocytopenia related acute ischemic stroke. J Stroke Cerebrovasc Dis. 2021;30 :106072. doi: 10.1016/j.jstrokecerebrovasdis.2021.106072 34461442 114. Costentin G Ozkul-Wermester O Triquenot A Cam-Duchez VL Massy N Benhamou Y Massardier E . Acute ischemic stroke revealing ChAdOx1 nCov-19 vaccine-induced immune thrombotic thrombocytopenia: impact on recanalization strategy. J Stroke Cerebrovasc Dis. 2021;30 :105942. doi: 10.1016/j.jstrokecerebrovasdis.2021.105942 34175640 115. Walter U Fuchs M Grossmann A Walter M Thiele T Storch A Wittstock M . Adenovirus-Vectored COVID-19 vaccine–induced immune thrombosis of carotid artery. Neurology. 2021;97 :716–719. 116. Bourguignon A Arnold DM Warkentin TE Smith JW Pannu T Shrum JM Al Maqrashi ZAA Shroff A Lessard MC Blais N . Adjunct immune globulin for vaccine-induced immune thrombotic thrombocytopenia. N Engl J Med. 2021;385 :720–728. doi: 10.1056/NEJMoa2107051 34107198 117. Ceschia N Scheggi V Gori AM Rogolino AA Cesari F Giusti B Cipollini F Marchionni N Alterini B Marcucci R . Diffuse prothrombotic syndrome after ChAdOx1 nCoV-19 vaccine administration: a case report. J Med Case Rep. 2021;15 :496. doi: 10.1186/s13256-021-03083-y 34615534 118. Patriquin CJ Laroche V Selby R Pendergrast J Barth D Côté B Gagnon N Roberge G Carrier M Castellucci LA . Therapeutic plasma exchange in vaccine-induced immune thrombotic thrombocytopenia. N Engl J Med. 2021;385 :857–859. doi: 10.1056/NEJMc2109465 34233107 119. Jacob C Rani KA Holton PJ Boyce SR Weir NU Griffith CR Eynon CA . Malignant middle cerebral artery syndrome with thrombotic thrombocytopenia following vaccination against SARS-CoV-2. Pediatr Crit Care Med. 2021. doi: 10.1177/17511437211027496 120. Jackson SP Darbousset R Schoenwaelder SM . Thromboinflammation: challenges of therapeutically targeting coagulation and other host defense mechanisms. Blood. 2019;133 :906–918. doi: 10.1182/blood-2018-11-882993 30642917 121. Pomara C Sessa F Ciaccio M Dieli F Esposito M Garozzo SF Giarratano A Prati D Rappa F Salerno M . Post-mortem findings in vaccine-induced thrombotic thombocytopenia. Haematologica. 2021;106 :2291–2293. doi: 10.3324/haematol.2021.279075 34011138 122. Xu P Zhou Q Xu J . Mechanism of thrombocytopenia in COVID-19 patients. Ann Hematol. 2020;99 :1205–1208. doi: 10.1007/s00277-020-04019-0 32296910 123. Warkentin TE Kaatz S . COVID-19 versus HIT hypercoagulability. Thromb Res. 2020;196 :38–51. doi: 10.1016/j.thromres.2020.08.017 32841919 124. Koupenova M Corkrey HA Vitseva O Tanriverdi K Somasundaran M Liu P Soofi S Bhandari R Godwin M Parsi KM . SARS-CoV-2 initiates programmed cell death in platelets. Circ Res. 2021;129 :631–646. doi: 10.1161/CIRCRESAHA.121.319117 34293929 125. Iadecola C Anrather J Kamel H . Effects of COVID-19 on the nervous system. Cell. 2020;183 :16–27.e1. doi: 10.1016/j.cell.2020.08.028 32882182
PMC009xxxxxx/PMC9005110.txt	==== Front Clin Neurophysiol Clin Neurophysiol Clinical Neurophysiology 1388-2457 1872-8952 Published by Elsevier B.V. S1388-2457(22)00056-6 10.1016/j.clinph.2022.01.038 Article P 7 SARS-CoV2 infection causes a worsening of the modified ranking scale (mRS) in patients with neuromuscular diseases – first results of the German covid19-nme registry Worm A. a⁎ Aust F. a Hahn A. b Schänzer A. c Hasseli R. d Krämer-Best H.H. a a Justus-Liebig-Universität Gießen, Klinik für Neurologie – UKGM Gießen/ Marburg, Gießen, Germany b Justus-Liebig-Universität Gießen, Neuropädiatrie – UKGM Gießen/ Marburg, Gießen, Germany c Justus-Liebig-Universität Gießen, Neuropathologie – UKGM Gießen/ Marburg, Gießen, Germany d Justus-Liebig-Universität Gießen, Klinik f. Rheumatologie u. klinische Immunologie Campus Kerckhoff, Gießen, Germany ⁎ Corresponding author. 12 4 2022 5 2022 12 4 2022 137 e18e19 18 1 2022 Copyright © 2022 Published by Elsevier B.V. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background: Patients with neuromuscular diseases (NMD) are classified as risk groups for a potentially severe course of a SARS-CoV-2 infection. An online registry (www.covid19-nme.com) was developed to gather information about the severity of COVID19, a potential progression of NMD through the SARS-CoV-2 infection and the possible influence of medication on the course of the infection. Methods: Since February 2021, patients of all ages (children, adolescents and adults) with NMD and an infection with SARS-CoV-2 have been included in this register. In addition to demographic data, pre-existing diseases and therapies, information about the NMD, the course of the SARS-CoV-2 infection as well as the clinical findings before and after the infection are recorded. Results: So far 94 patients (37% female, age: median 60 years (1-94 years)) from Germany and Austria have been recorded. The diagnoses represent the entire spectrum of NMD: different forms of polyneuropathies (PN) including CIDP and hereditary PN, ICUAW, myasthenic syndromes, motor neuron diseases (SMA and ALS) as well as various muscle diseases such as dystrophinopathies and myotonic syndromes. The collected mRS (measure for description of neurological impairment) depicts a significant worsening after the SARS-CoV2 infection (p = 0.02; Wilcoxon), whereby the patients with ICUAW were excluded from the analysis. The duration of symptoms showed a positive correlation with age (r = 0.343; p = 0.005) and weight (r = 0.291; p = 0.030), but not with the type of NMD. In total, 13 patients deceased due to the SARS-CoV2 infection. The probability of a fatal outcome of COVID19 correlates with increasing age (r = 0.313; p = 0.004) but not the type of NMD. The ventilation situation did not change in NMD patients due to the infection with SARS-CoV2. Summary: The first results of the evaluation of the covid-19.nme registry indicate that the clinical symptoms of NMD progress due to an infection with SARS-CoV2. The underlying cause for this remains unclear. Autoimmunological processes and a possible neurotropy can be considered as pathophysiological mechanisms. ==== Body pmc
PMC009xxxxxx/PMC9005116.txt	==== Front Lancet Respir Med Lancet Respir Med The Lancet. Respiratory Medicine 2213-2600 2213-2619 Elsevier Ltd. S2213-2600(22)00136-9 10.1016/S2213-2600(22)00136-9 Spotlight Christine Jenkins—global advocate for asthma and COPD Kirby Tony 12 4 2022 12 4 2022 © 2022 Elsevier Ltd. All rights reserved. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmc Speaking to The Lancet Respiratory Medicine on her first overseas trip since COVID-19 arrived, Christine Jenkins reflects that “one of the few good things that COVID has done is to focus people's attention on how important lung health is for our overall health”. As Chair of the Lung Foundation Australia over the past 6 years, she has been actively raising the profile of lung diseases. For four decades, she has been a clinician-researcher aiming to improve outcomes and care for people with asthma and COPD, and currently heads the Respiratory Group at the George Institute, based at the University of New South Wales (UNSW), Sydney, Australia. Her professional life has been shaped by inspirational mentors, including her high school female science teacher, and her own experience of developing asthma while living on a farm outside Sydney as a teenager. “I was allergic to grass pollens and my GP gave me what was then quite advanced care—desensitising injections. It was my first ‘inside view' of medicine and general practice, and I found it fascinating.” For most of medical school at UNSW, Jenkins was determined to follow her family doctor into general practice, but working with exceptional specialist physicians made her reconsider. Before deciding, she worked for a year in remote Northern Nigeria, learning about various tropical diseases before returning to Australia. Then, during a respiratory medicine rotation at Sydney's Concord Hospital, she fell in love with the specialty and has never looked back. “It was the 1980s when asthma prevalence and mortality were increasing rapidly, especially in Australia and New Zealand. Better understanding of asthma as an inflammatory disease and availability of anti-inflammatory treatments made me see that many respiratory problems were treatable or preventable.“ Jenkins then completed her MD in airway hyper-responsiveness, and worked on trials to modify this and improve clinical outcomes. She was invited by the renowned Anne Woolcock to join her research group, now the Woolcock Institute of Medical Research, and continued this work, comparing the effects of inhaled and oral corticosteroids, and long-acting bronchodilators (LABAs) that were revolutionising asthma care. Jenkins reflects that while women face unique challenges when wanting careers as clinician-researchers, systems in place mean that, although never easy, “it is today more achievable than ever to balance your family and career”. Although she managed this balance with the help of her supportive husband, she is very excited by the number of women in respiratory medicine who have faced greater challenges and are making exceptional contributions. It was in the mid-1990s that COPD became an urgent priority, driven by historically high smoking rates and rising hospitalisations, and Jenkins shifted her interest in clinical practice guidelines from asthma to COPD. She then met another key mentor, Sonia Buist, and became part of the working group that produced the earliest GOLD guidelines for COPD management. She helped to establish the Burden of Obstructive Lung Disease (BOLD) study in Australia, and was a member of the GOLD executive for the next 15 years. Jenkins helped to lobby for asthma to become a National Health Priority in Australia, resulting in further improvements in asthma outcomes and a focus on primary care. She has dedicated many hours to teaching medical students and training GPs, who “are vital for diagnosis and management of many respiratory conditions.” Recent research has included studies of walking exercise for pulmonary rehabilitation in COPD, and a multicentre trial assessing theophylline and prednisone for COPD in China. In The Lancet Respiratory Medicine, Jenkins and colleagues publish a paper examining the wide range of risk factors for so-called non-smoking COPD, which could represent half of the global disease burden. “As with lung cancer, many people with COPD have no history of smoking”, Jenkins explains. “Non-smoking COPD is a huge challenge, especially in low- and middle-income countries, and poverty alleviation, education, and strong public health policies are needed, along with more research into causes and treatments.” Jenkins sees much more hope in COPD management today, but more effective therapies are desperately needed to blunt the relentless increase in this debilitating disease. Next on her “to do” list is building a clinical pathway to assess and manage breathlessness in primary care. She is also leading a multinational clinical trial addressing cardiac disease in COPD, and will continue her advocacy work, including helping to implement the National Strategic Action Plan for Lung Conditions that she and her colleagues nationwide developed, a plan that includes an urgent call for a national lung cancer screening plan and dedicated lung cancer nurses. “Christine is one of a handful of clinical researchers who not only contributes her own research, but gives attention to detail, contributes to any discussion in her kind and convincing manner, and brings an understanding of ‘the bigger picture'”, says Jørgen Vestbo, Professor of Respiratory Medicine at the NIHR Manchester Biomedical Research Centre, University of Manchester, UK. “She does not limit herself to isolated trial findings or health implications in only our wealthy part of the world, but uses her advocacy roles, insight, and abilities to promote lung health everywhere, including the low- and middle-income countries that need such help the most.” For the Origins of chronic obstructive pulmonary disease Series see www.thelancet.com/series/origins-of-COPD
PMC009xxxxxx/PMC9005117.txt	==== Front Lancet Respir Med Lancet Respir Med The Lancet. Respiratory Medicine 2213-2600 2213-2619 The Author(s). Published by Elsevier Ltd. S2213-2600(22)00139-4 10.1016/S2213-2600(22)00139-4 Corrections Correction to Lancet Respir Med 2022; published online Feb 25. https://doi.org/10.1016/S2213-2600(22)00045-5 12 4 2022 6 2022 12 4 2022 10 6 e60e60 © 2022 The Author(s). Published by Elsevier Ltd. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcMcKeigue PM, McAllister DA, Hutchinson SJ, Robertson C, Stockton D, Colhoun HM. Vaccine efficacy against severe COVID-19 in relation to delta variant (B.1.617.2) and time since second dose in patients in Scotland (REACT-SCOT): a case-control study. Lancet Respir Med 2022; published online Feb 25. https://doi.org/10.1016/S2213-2600(22)00045-5—We now offer open access for authors based at JISC-participating UK institutions for papers accepted after Jan 1, 2022. As a result, the copyright license for this Article has been updated to CC BY. This correction has been made to the online version as of April 12, 2022, and will be made to the printed version.
PMC009xxxxxx/PMC9005118.txt	==== Front Lancet Microbe Lancet Microbe The Lancet. Microbe 2666-5247 The Author(s). Published by Elsevier Ltd. S2666-5247(22)00098-2 10.1016/S2666-5247(22)00098-2 Corrections Correction to Lancet Microbe 2022; 3: e173–83 12 4 2022 5 2022 12 4 2022 3 5 e333e333 © 2022 The Author(s). Published by Elsevier Ltd. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcAhn JY, Lee J, Suh YS, et al. Safety and immunogenicity of two recombinant DNA COVID-19 vaccines containing the coding regions of the spike or spike and nucleocapsid proteins: an interim analysis of two open-label, non-randomised, phase 1 trials in healthy adults. Lancet Microbe 2022; 3: e173–83—In this Article Han Young Seo's name has been corrected. This correction has been made as of April 12, 2022.
PMC009xxxxxx/PMC9005119.txt	==== Front Lancet Infect Dis Lancet Infect Dis The Lancet. Infectious Diseases 1473-3099 1474-4457 Elsevier Ltd. S1473-3099(22)00224-9 10.1016/S1473-3099(22)00224-9 Correspondence Comparable neutralisation evasion of SARS-CoV-2 omicron subvariants BA.1, BA.2, and BA.3 Arora Prerna ab Zhang Lu ab Rocha Cheila ab Sidarovich Anzhalika ab Kempf Amy ab Schulz Sebastian c Cossmann Anne f Manger Bernhard d Baier Eva g Tampe Björn g Moerer Onnen h Dickel Steffen h Dopfer-Jablonka Alexandra ef Jäck Hans-Martin c Behrens Georg M N ef Winkler Martin S h Pöhlmann Stefan ab Hoffmann Markus ab a Infection Biology Unit, German Primate Center, Göttingen 37077, Germany b Faculty of Biology and Psychology, Georg-August-University Göttingen, Göttingen, Germany c Division of Molecular Immunology, Department of Internal Medicine 3, Friedrich-Alexander University of Erlangen-Nürnberg, Erlangen, Germany d Deutsches Zentrum für Immuntherapie, Department of Internal Medicine 3, Friedrich-Alexander University of Erlangen-Nürnberg, Erlangen, Germany e Department for Rheumatology and Immunology, Hannover Medical School, Hannover, Germany f German Centre for Infection Research, partner site Hannover-Braunschweig, Hannover, Germany g Department of Nephrology and Rheumatology, University Medical Center Göttingen, Göttingen, Germany h Department of Anaesthesiology, University Medical Center Göttingen, Göttingen, Germany 12 4 2022 12 4 2022 © 2022 Elsevier Ltd. All rights reserved. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcThe SARS-CoV-2 omicron (B.1.1.529) variant has rapidly become globally dominant, displacing the previously dominant delta (B1.617.2) variant. The viral spike (S) protein is the key target of the neutralising antibody response, and the omicron variant harbours more than 35 mutations in the S protein, which allow highly efficient evasion from neutralising antibodies.1 In keeping with these findings, the omicron variant efficiently spreads in populations with a high percentage of convalescent or vaccinated individuals.2, 3 The three main subvariants of the omicron variant are BA.1, BA.2, and BA.3. Initial data suggest that BA.2 might have a growth advantage over BA.1,4 posing a rapidly increasing threat to health systems. The omicron subvariants display remarkable differences regarding S protein mutations, particularly with respect to the N-terminal domain and the receptor-binding domain (appendix pp 2–3), which are known to harbour key epitopes of neutralising antibodies.5, 6 Here, we compared BA.1, BA.2, and BA.3 for sensitivity to neutralisation by antibodies induced by infection and vaccination, using pseudoviruses as a model system, which adequately mirrors SARS-CoV-2 neutralisation by antibodies.7 We analysed particles harbouring the S protein of B.1—which is identical to the wildtype strain apart from the D614G mutation—and S proteins of BA.1, BA.2, and BA.3. We first examined neutralisation by antibodies from convalescent patients, who were infected during the first (February to May, 2020) and second (December, 2020, to February, 2021) waves of COVID-19 in Germany (appendix pp 2–3, 4–6). Neutralisation of particles bearing the B.1 S protein (B.1pp) was robust, whereas neutralisation of BA.1pp and BA.3pp was at least 32-times less than B.1pp (BA.1 p=0·0020; BA.3 p=0·0020). Neutralisation of BA.2pp was also diminished, but the reduction was less pronounced than that measured for the other omicron subvariants (9·2-times less than B.1pp; p=0·0020). Analysis of neutralisation by antibodies induced by double vaccination with BNT162b2 (BNT) yielded similar results as neutralisation with antibodies from convalescent patients (appendix pp 2–3). Particles harbouring the S proteins of BA.1 and BA.3 showed 17-times lower neutralisation than B.1pp (BA.1 p=0·0020; BA.3 p=0·0020), whereas neutralisation of BA.2pp was 9-times reduced (p=0·0020). Triple BNT vaccination induced a more potent antibody response, and only modest evasion of neutralisation was seen for particles bearing omicron S proteins (BA.1 2·5-times, p=0·0039; BA.2 1·9-times, p=0·012; BA.3 2·4-times, p=0·0039; appendix pp 2–3). Finally, neutralisation by antibodies induced in fully vaccinated (three vaccine doses) individuals with breakthrough infection during the fourth wave in Germany (October, 2021, to January, 2022, dominated by the delta variant) was most potent and neutralisation of particles bearing omicron S protein was 9–12-times less efficient than B.1pp (BA.1 p=0·0020; BA.2 p=0·0039; BA.3 p=0·0039; appendix pp 2–3). However, no significant differences were observed between BA.1pp, BA.2pp, and BA.3pp (appendix pp 2–3). Our results show that all presently circulating omicron subvariants evade neutralisation by vaccine-induced antibodies with comparably high efficiency, suggesting that increased antibody evasion is not the reason for the current expansion of BA.2 in several countries.4, 8 Since currently available vaccines provided robust protection against early omicron isolates circulating in South Africa from Nov 15 to Dec 7, 2021,3 which was likely to be BA.1, our results suggest that this protection should extend to all omicron subvariants. SP acknowledges funding from Bundesministerium für Bildung und Forschung (BMBF; grant numbers 01KI2006D, 01KI20328A, 01KX2021), the Ministry for Science and Culture of Lower Saxony (grant numbers 14-76103-184, MWK HZI COVID-19), and the German Research Foundation (DFG; grant numbers PO 716/11-1, PO 716/14-1). MSW received unrestricted funding from Sartorius, Lung research. H-MJ received funding from BMBF (grant numbers 01KI2043, NaFoUniMedCovid19-COVIM 01KX2021), Bavarian State Ministry for Science and the Arts, and DFG through the research training groups RTG1660 and TRR130, the Bayerische Forschungsstiftung (Project CORAd), and the Kastner Foundation. GMNB acknowledges funding from the German Center for Infection Research (grant number 80018019238) and a European Regional Development Fund (Defeat Corona, grant number ZW7-8515131, together with AD-J). All other authors declare no competing interests. Supplementary Material Supplementary appendix ==== Refs References 1 Hoffmann M Krüger N Schulz S The omicron variant is highly resistant against antibody-mediated neutralization: implications for control of the COVID-19 pandemic Cell 185 2022 447 456 35026151 2 Altarawneh HN Chemaitelly H Hasan MR Protection against the omicron variant from previous SARS-CoV-2 infection N Engl J Med 2022 published online Feb 9. 10.1056/NEJMc2200133 3 Collie S Champion J Moultrie H Bekker LG Gray G Effectiveness of BNT162b2 vaccine against omicron variant in South Africa N Engl J Med 386 2022 494 496 34965358 4 Mahase E COVID-19: what do we know about omicron sublineages? BMJ 376 2022 o358 35149516 5 McCallum M De Marco A Lempp FA N-terminal domain antigenic mapping reveals a site of vulnerability for SARS-CoV-2 Cell 184 2021 2332 2347 33761326 6 Piccoli L Park YJ Tortorici MA Mapping neutralizing and immunodominant sites on the SARS-CoV-2 spike receptor-binding domain by structure-guided high-resolution serology Cell 183 2020 1024 1042 32991844 7 Schmidt F Weisblum Y Muecksch F Measuring SARS-CoV-2 neutralizing antibody activity using pseudotyped and chimeric viruses J Exp Med 217 2020 e20201181 8 Lyngse FP Kirkeby CT Denwood M Transmission of SARS-CoV-2 omicron VOC subvariants BA.1 and BA.2: evidence from Danish households medRxiv 2022 published online Jan 30. 10.1101/2022.01.28.22270044 (preprint).
PMC009xxxxxx/PMC9005162.txt	==== Front Environ Sci Pollut Res Int Environ Sci Pollut Res Int Environmental Science and Pollution Research International 0944-1344 1614-7499 Springer Berlin Heidelberg Berlin/Heidelberg 35414157 19979 10.1007/s11356-022-19979-1 Research Article Reverse vaccinology approach for multi-epitope centered vaccine design against delta variant of the SARS-CoV-2 Jalal Khurshid 1 Khan Kanwal 2 Basharat Zarrin 3 Abbas Muhammad Naseer 4 Uddin Reaz mriazuddin@iccs.edu 2 Ali Fawad 4 Khan Saeed Ahmad 4 http://orcid.org/0000-0001-5188-6540 Hassan Syed Shams ul 56 1 grid.471007.5 0000 0004 0640 1956 International Center for Chemical and Biological Sciences, HEJ Research Institute of Chemistry, University of Karachi, Karachi, Pakistan 2 grid.266518.e 0000 0001 0219 3705 Dr. Panjwani Center for Molecular Medicine and Drug Research, International Center for Chemical and Biological Sciences, University of Karachi, Karachi, Pakistan 3 grid.266518.e 0000 0001 0219 3705 Jamil‐ur‐Rahman Center for Genome Research, Dr. Panjwani Center for Molecular Medicine and Drug Research, ICCBS University of Karachi, Karachi, Pakistan 4 grid.411112.6 0000 0000 8755 7717 Department of Pharmacy, KUST, Khyber Pakhtunkhwa, Kohat, 26000 Pakistan 5 grid.16821.3c 0000 0004 0368 8293 Shanghai Key Laboratory for Molecular Engineering of Chiral Drugs, School of Pharmacy, Shanghai Jiao Tong University, Shanghai, 200240 People’s Republic of China 6 grid.16821.3c 0000 0004 0368 8293 Department of Natural Product Chemistry, School of Pharmacy, Shanghai Jiao Tong University, Shanghai, 200240 People’s Republic of China Responsible Editor: Lotfi Aleya 12 4 2022 2022 29 40 6003560053 7 1 2022 25 3 2022 © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. The ongoing COVID-19 outbreak, initially identified in Wuhan, China, has impacted people all over the globe and new variants of concern continue to threaten hundreds of thousands of people. The delta variant (first reported in India) is currently classified as one of the most contagious variants of SARS-CoV-2. It is estimated that the transmission rate of delta variant is 225% times faster than the alpha variant, and it is causing havoc worldwide (especially in the USA, UK, and South Asia). The mutations found in the spike protein of delta variant make it more infective than other variants in addition to ruining the global efficacy of available vaccines. In the current study, an in silico reverse vaccinology approach was applied for multi-epitope vaccine construction against the spike protein of delta variant, which could induce an immune response against COVID-19 infection. Non-toxic, highly conserved, non-allergenic and highly antigenic B-cell, HTL, and CTL epitopes were identified to minimize adverse effects and maximize the efficacy of chimeric vaccines that could be developed from these epitopes. Finally, V1 vaccine construct model was shortlisted and 3D modeling was performed by refinement, docking against HLAs and TLR4 protein, simulation and in silico expression. In silico evaluation showed that the designed chimeric vaccine could elicit an immune response (i.e., cell-mediated and humoral) identified through immune simulation. This study could add to the efforts of overcoming global burden of COVID-19 particularly the variants of concern. Supplementary Information The online version contains supplementary material available at 10.1007/s11356-022-19979-1. Keywords Reverse vaccinology Delta variant Chimeric vaccine Spike protein SARS-CoV-2 issue-copyright-statement© Springer-Verlag GmbH Germany, part of Springer Nature 2022 ==== Body pmcIntroduction Coronavirus disease 2019 (COVID-19) has appeared as one of the most life-threatening infectious diseases caused by the SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2). It was initially diagnosed in Wuhan, China in December 2019 (Ito et al. 2021). The COVID-19 cases spread with reproduction number (R0) (https://www.gov.uk/guidance/the-r-value-and-growth-rate) reaching 5.82 million deaths globally and leaving severe impact on infected survivors (Kar et al. 2020; Z. Yang et al. 2021). The WHO has confirmed over 410 million cases with 5.8 million reported deaths (as of 16 February 2022 https://www.who.int/emergencies/diseases/novel-coronavirus-2022/situation-reports). Currently, due to the mutations in SARS-CoV-2 genome, a number of variants such as alpha, beta, gamma, delta, kappa, and eta (Lazarevic et al. 2021) have been observed. These variants have higher transmissibility rates and immune escape mechanisms (Brown et al. 2021; Uddin et al. 2022) One of the recent and hazardous variants of concern (VOC) of SAR-CoV-2 is the delta variant (also called lineage B.1.617.2) derived from the alpha lineage B.1.617 (Ito et al. 2021). It was initially discovered in the Indian state of Maharashtra in the late 2020s (Salvatore et al. 2021), introduced by a traveler from the UK on 31st May 2021. It was designated the delta variant by the World Health Organization (WHO), as India witnessed a devastating second wave of it and it peaked during May 2021 (W. Yang and Shaman 2021). Lately, it spread to more than 93 countries (https://www.gisaid.org/hcov19-variants/) and became globally dominant variant, displacing the B.1.1.7 (alpha) variant and other pre-existing lineages (Campbell et al. 2021; Shu and McCauley 2017). It sustained an increase in the reproduction number (R0) to 1.2–1.4 (in the UK) (Riley et al. 2021). The B.1.617.2 carried the following mutations in the spike protein: G142D, T19R, R158G, L452R, T478K, D614G, P681R, D950N, T95I, E156del, and F157del (Kumar, Dwivedi, et al., 2021) at N-terminal domain. The N-terminal domain (NTD) mutations reduce the flexibility and revealed rigidity in spike protein of B.1.617.2 (Behl et al. 2022; Brown et al. 2021). These multiple mutations appear to give an advantage to Delta variant over other variants. The notable effect caused by these mutations is the increased transmission rate (within 10 s, 60% more transmissible than alpha variant) rendering it as most dangerous and dominant variant globally (https://nextstrain.org/ncov/gisaid/global?l=clock). In England, ~38,805 genotyped SARS-CoV-2 cases are linked with the delta variant with 2.61 times higher risk of hospitalization (https://nymag.com/intelligencer/2021/06/covid-b-1-617-2-delta-variant-what-we-know.html) and 1.67 higher risk of A&E care within 14 days of specimen as compared to the alpha variant (Angelo et al. 2017; Riley et al. 2021). Since the beginning of the COVID-19 pandemic, various variants (such as alpha, beta, gamma, delta, kappa, and currently omicron) have emerged as a consequence of mutations in the SARS-CoV-2 genome. These alterations have a greater influence on the rate of transmission and the immune system’s ability to escape (Negrut et al. 2021; Tao et al. 2021; Thakur et al. 2021). Lately, the WHO reported on November 24, 2021 that a novel mutant variant of SARS-CoV-2 was identified in Botswana, South Africa. On November 26, 2021, WHO designated the novel variant omicron as a variant of concern (VoC) (Poudel et al. 2022). The mutational profile of omicron is crucial for determining whether it shares or varies in clinical symptoms from other SARS-CoV-2 variants. As a result of the substantially modified form’s development, several nations have taken rigorous measures to reduce the transmission of the variants of SARS-CoV-2. Despite the availability of several vaccines, it is an immense logistical challenge to achieve universal coverage especially in populous nations (Mlcochova et al. 2021). The vaccination process has effectively controlled the hospitalization and deaths related to COVID-19 infection. However, the new viral variants with novel mutations and antigenic profiles are posing serious threats and diminishing the efficacy of the available vaccines (Angelo et al. 2017). It has been observed through sensitivity sera test performed by Davis et al. (2021) against SARS-CoV-2 variants that B.1.617.2 conferred reductions in neutralization of sera (for Pfizer and Oxford vaccine) 5.11-fold while ~ eightfold reduction in sensitivity observed through in vitro analysis in contrast to the wild type SARS-CoV-2 from Wuhan-1 (Krause et al. 2021). More recently, a study performed by the UK government found that only 33% protection was provided by single dose of COVID-19 vaccine against B.1.617.2 (compared with 51% protection against B.1.1.7) (Callaway 2021; W. Yang and Shaman 2021). Nevertheless, Israel on June 26, 2021 reported COVID-19 cases among which, about 90% of infections were likely caused by the delta variant, infecting ~50% of fully vaccinated adults (Papenfuss 2021). Therefore, new vaccine constructs are of dire necessity for efficacious operative long-term management of delta variant infections in current situation of pandemics (Kabir et al. 2020; Mlcochova et al. 2021). Traditionally, vaccine designing requires a pathogen growth (i.e., inactivation, isolation, and injection of the disease-causing virus) and assays (both in vivo and in vitro) that are time and money consuming approaches (Rappuoli et al. 2016). Traditional process requires more than a year for the production and availability of an efficacious vaccine, consequently contributing very little to control the current spread of infection (Ojha et al. 2020; Tagde et al. 2021). On the contrary, the development in computational biology and bioinformatics approaches has led the swift design of useful constructs, reducing the conventional laboratory-based experimentations (Sharma et al. 2021). One of the widely used computational approaches for the development of vaccine model is “Reverse Vaccinology” (D’Mello et al. 2019). It enables vaccine construction and designing based on the organism genome sequence information without the need to grow pathogens. It works on the construction of multiple fragments (i.e., epitopes) from viral proteins so that it can elicit cellular and humoral immune responses and reduces the adverse effects (Gheorghe et al. 2021; Shey et al. 2019). Reverse vaccinology approach has been successfully employed in prioritizing and designing vaccine targets against multiple pathogens (Jalal et al. 2022; Khan et al. 2022;; Nosrati et al. 2019; Srivastava et al. 2019; Tosta et al. 2021). Previous studies on SARS-CoV-2’s spike protein revealed that this protein is the key component that is enabling the entry of virus in human cells and playing a decisive role in infection (Kar et al. 2020). The emergence of multiple variants leads to the emergence of several waves of devastating pandemics around the world (Kumar et al. 2021a, b; Li et al. 2021). As a result, vaccines required a booster dose after 6 months due to high transmission of these SARS-CoV-2 variants. Therefore, current study applied the reverse vaccinology approach against the spike protein of delta variant to design multi-epitope-based vaccine model that can induce cellular (i.e., the activation of T helper cells, T cytotoxic cells), interferon-γ (i.e., IFN-γ), and humoral (i.e., B-cells activations and antibodies production) responses. Furthermore, the shortlisted vaccine molecules can be effectively expressed in E. coli vector model. We strongly believe that our findings may provide prolific information and better guidance for further vaccine development against the delta variant. Material and methods The current study’s pipeline is based on reverse vaccinology for the identification of novel vaccine constructs against the delta variant spike protein. Data retrieval The delta variant spike glycoprotein sequence was retrieved from Zhanglab’s database (Huang et al. 2020) and mutated with reported amino acid changes, i.e., T19R, G142D, E156del, F157del, R158G, T478K, L452R, D614G, P681R, D950N, and T95I (Quinonez et al. 2021) using PyMol software. The FASTA format was used for saving the variant spike protein for further analysis. Reverse vaccinology The recent advancements in vaccinomics, immunology, biochemistry, molecular biology, genomics, proteomics, and the conventional vaccinology have transformed into reverse vaccinology (RV). The RV is one of the novel and emerging computational approaches that have been tremendously used to optimize the vaccine target and vaccine model prediction, particularly those microbes which are difficult to cultivate in the laboratory (Rappuoli et al. 2016). It combines the immunology, molecular biology, biochemistry, genomics, and proteomics based in silico approaches to screen whole proteome of pathogens in order to determine novel vaccine candidates and assess its ability to induce host immune response. Antigenicity identification The antigenic analysis of delta spike protein was performed through VaxiJen v 2.0 (Doytchinova and Flower 2007) server using a threshold value of 0.4, in order to be recognized by the immune system. MHC-I T-cell epitope prediction Different T cell epitopes that can potentially activate human immune system and produce memory cells (immunomodulatory effects) were analyzed for delta spike protein through NetCTL server (Larsen et al. 2007). The predicted epitopes were chosen on the parameters of overall intrinsic peptide potential scores integrated with transporter-associated efficiency prediction, protease cleavage, prediction score for MHC I epitope affinity, along with combined score of predicted parameters with threshold value of 0.75. Additionally, the binding analysis of identified T-cell epitopes was studied by employing Immune Epitope Database and Analysis Resource (IEDB AR) server (Kim et al. 2012) through which T-cell recognized antigen represented by MHC-I. Default parameters of consensus method, i.e., ANN (Nielsen et al. 2003), SMM (Chen et al. 2009), CombLib (Sidney et al. 2008), and NetMHCpan (Lundegaard et al. 2008) and all HLA alleles were used for MHC-I prediction. The HLA alleles selected for the MHC-I analysis were HLA-A 0205, HLA_0201, HLA-A2, HLA-A 2.1, HLA-A3, HLA-B 5401, and HLA-B 5102. The threshold parameters based on IC50 < 100 uM and percentile rank (< 0.2) were set as cut-off values for the shortlisting of MHC-I epitopes (Solanki and Tiwari 2018). MHC-I immunogenicity prediction The predicted MHC-I epitopes should be immunogenic enough so that they can stimulate either CD4 or CD8 T-cells. Therefore, IEBD AR tool (Dhanda et al. 2019) was used for the prediction of MHC-I immunogenicity analysis. The score having positive value for MHC-I epitopes was selected for further study. Antigenicity, toxicity and conservancy assessments for MHC-I predicted epitope The shortlisted MHC-I epitopes having promiscuous immunogenic scores were further analyzed to predict their toxicity, conservancy, and antigenic properties. The conversancy analysis was performed through IEBD server (Angelo et al. 2017). The conversancy analysis is important to develop a broad-spectrum peptide-based vaccine against specific virus. Furthermore, antigenicity of conserve epitopes was predicted through VaxiJen server (Doytchinova and Flower 2007) with an accuracy of 70–80% and 0.5 probability threshold score. Finally, ToxinPred server was utilized for the prediction of relative toxicity levels for shortlisted antigenic MHC-I epitopes with cut-off value set as 0.5. T-cell MHC-II prediction Additionally, the MHC-II epitopes were also identified using IEBD server based on consensus method w. The cut-off value for MHC-II epitopes shortlisting was set as < 0.2 peptide rank and IC50 < 100 nM for top binders against the 95% HLA variability found in worldwide human population, i.e., DRB11301, DRB10101, DRB10301, DRB10401, DRB10701, DRB3∗01:01, DRB10801, DRB11101, HLA- HLA-DRB3∗02:02, HLA-DRB5∗01:01, HLA-DRB4∗01:01, and DRB11501 (Solanki and Tiwari 2018). In the current analysis, multiple epitopes with 9–14 residues length were shortlisted for further study. MHC restricted alleles cluster analysis The MHCcluster server (Thomsen et al. 2013) was used to cluster MHC restricted alleles with appropriate MHC epitopes for further validation of identified MHC-I/II epitopes. The tool results in clustering of MHC-I and II epitopes generating heat map and phylogenetic tree highlighting the functional relationship between HLAs and epitopes. B-cell epitope identification An ideal peptide vaccine must have the property of inducing long-lasting humoral immunity similar to the natural immune response generated by certain infections. B-cell epitopes are responsible for the stimulation of humoral immunity having the ability to eliminate the virus by producing antibodies against antigen exposed in human body through B lymphocytes. ABCpred, FBCpred, and BCpred (Saha and Raghava 2007) using sequence-based methodology with a cut-off score of > 0.51 and 75% specificity were employed for the identification of B-cell epitopes, respectively. Moreover, ElliPro server (Ponomarenko et al. 2008) was also used to characterize B-cell epitopes based on their hydrophobicity content (El‐Manzalawy et al. 2008), flexibility (Karplus and Schulz 1985), antigenicity (Emini et al. 1985), accessibility, beta-turn prediction through Chou and Fashman tool (Chou and Fasman 1978), Karplus and Schulz flexibility scale, and Parker hydrophilicity scale (Parker et al. 1986), respectively. Identified epitope mapping Consequently, the epitopes having ability to induce immune cells (B and T-cells) response are significant for the designing of epitope-based vaccine (Solanki and Tiwari 2018). Therefore, the shortlisted MHCI-II and B-cell epitopes of delta spike protein were mapped for the identification of binding affinity and similarity among them. The manual comparison was performed and overlapping epitopes characterized as probable peptides sequences were assembled and considered a final predicted epitopes for vaccine modeling. Vaccines construction and structure modeling In order to construct vaccine with reduced toxicity and allergenicity and increased immunogenicity, the sequential conjugations of shortlisted epitopes were performed with relevant adjuvant (beta-defensin, L7/L12 ribosomal protein, HBHA protein, and HBHA conserved sequence), PADRE (Pan HLA-DR reactive epitope) and linker (GGGS, HEYGAEALERAG and EAAAK) (Solanki and Tiwari 2018). The use of linkers enhances immunogenicity while the induction of CD4+ T-cells by PADRE peptide makes the vaccine efficacious and potent. HEYGAEALERAG and GGGS linkers were utilized to conjugate HTL, CTL, and B-cell epitopes, whereas EAAAK linkers were used to join adjuvant sequences at both N and C-terminus (Solanki and Tiwari 2018). Antigenicity, allergenicity, and solubility assessment for vaccines constructs Adverse allergic reactions may be associated with vaccine outcomes. In order to overcome the allergic features of vaccine model, AlgPred tool was used to examine the allergenicity of model vaccine sequences with cut-off score of −0.4 and 85% accuracy (N. Sharma et al. 2020). Predicted scores less than the threshold value were considered non-allergenic vaccines. To predict the antigenic nature of vaccine models, the VaxiJen and ANTIGENpro server (Magnan et al. 2010) were used with the threshold value > 0.5. Furthermore, to identify the solubility property of vaccine model upon over-expression in E. coli, SOLpro program was used having default parameters of 74% accuracy and corresponding probability (≥ 0.5) (Magnan et al. 2009). Physiochemical analysis of constructed vaccines The widely used Expasy ProtParam tool (Gasteiger et al. 2005) was utilized for the physiochemical analysis and functional characterization of vaccines based on pK values of different amino acids, GRAVY values, instability index, molecular weight, aliphatic index, approximate half-life, hydropathicity, and isoelectric pH parameters (Gasteiger et al. 2005). It is essential to evaluate physiochemical properties to determine the safety and efficacy of vaccine candidates. Comparative structure modeling Modeler Phyre2 tool, whereas PSIPRED (Buchan and Jones 2019), ProSA-web (Wiederstein and Sippl 2007), and PROCHECK (R. Laskowski et al. 1993) were applied for the model structure evaluation based on secondary structure analysis, error in 3D modeled identification, and tertiary structure stereochemistry analysis, respectively. The best-modeled vaccine construct was selected for further structure-based analysis. Molecular docking studies The interaction of vaccine construct was modeled with its receptors to generate the stable immune response of vaccine model to target cells. The molecular docking approach is an ideal method to study such interaction studies in terms of binding energies between epitopes and HLA proteins (Solanki and Tiwari 2018). The final potential vaccine constructs fulfilling all filters of the framework were docked into the binding cavity of six most common human alleles in human population HLA alleles, i.e., 3C5J (HLA-DRB302:02), 2Q6W (HLA-DR B301:01), 1H15 (HLA-DR B501:01), 2FSE (HLA-DR B101:01), 1A6A (HLA-DR B103:01), and 2SEB (HLA-DRB104:01) retrieved from Protein Data Bank. The PatchDock server (Schneidman-Duhovny et al. 2005) was used to estimate the HLA and vaccine interactions while the FireDock (fast interaction refinement in molecular docking) server was applied to further refine and re-score the docked complex obtained through PatchDock (Mashiach et al. 2008). Moreover, the docking step was validated by GRAMMX tool for vaccine and TLR4/MD complex. The TLR4 is implicated in viral protein recognition, leading to the production of inflammatory cytokines. According to several studies, TLR4 is critical for generating an efficient immune response against SARS-CoV-2 (Hu et al. 2012). Therefore, the molecular docking studies were performed for the vaccine candidate with TLR4. The UCSF Chimera (Pettersen et al. 2004) and PDBsum (Laskowski et al. 2018) tools were used for the binding interpretation based on hydrogen bond pattern and docked scores. Molecular dynamic simulation studies The molecular dynamic simulation (MDS) and energy minimization were performed through GROMACS (GROningen MAchine for Chemical Simulations) to determine the stability and flexibility of the vaccine construct. The MDS was performed to examine how vaccine model acts in biological system (Abraham et al. 2015). The topology files needed for energy minimization and equilibrium were generated using GROMACS MDS’s published approach. The solvation was executed with SPC216 water model with energy minimization using steepest algorithm and NVT and NPT ensembles for 50,000 steps (100 ps). Furthermore, the neutralization of vaccine construct was performed by adding charged ions to MDS system. Finally, the vaccine molecular dynamics simulation was carried out for 10 ns to find root mean square deviation (RMSD). The graphs obtained from MDS were visualized employing the Xmgrace plotting tool (Cowan and Grosdidier 2000). Furthermore, molecular dynamic simulation of docked complex (vaccine with TLR4) was performed using iMODS web-server (López-Blanco et al. 2014). It explains the collective movements of protein complex with vaccine construct through normal mode analysis (NMA). The intrinsic complex motion extent and direction were evaluated in terms of covariance, deformability, eigenvalue, and B-factors. Immune simulation of final vaccine construct The immunogenicity and immune response profile of a chimeric peptide vaccine were characterized using the C-ImmSim service. The C-ImmSim server is an in silico-based immune simulation method (Rapin et al. 2010). At three different intervals for 4 weeks, three injections of molded prophylactic delta variant vaccine at 1, 82, and 126 h time periods and 12,345 random seed were administered keeping all simulation parameters at default containing no LPS, volume and the stages of the simulation at 10, and 1000, respectively, with homozygous host haplotypes HLA-A0101, HLA-A0201, HLA-B0702, HLA-DRB10101, and HLA-DRB1*0401 (Rahman et al. 2020). Codon optimization and in-silico cloning The Java Codon Adaptation Tool (JCAT), based on codon adaptation approach, was used to reverse translate vaccine amino acid sequence in cDNA for the effective expression of vaccine construct in E. coli vector (Grote et al. 2005). Computed Codon Adaptation Index (CAI) values and GC content were used as a parameters for vaccine adaptation while avoiding the ribosome-binding prokaryote site, termination of Rho’s independent transcription, and the cleavage of restriction enzymes (Bibi et al. 2021). Finally, the insertion of adapted codon sequence was performed through SnapGene cloning module into the pET-28a (+) vector. Results Antigenicity prediction for delta spike protein The antigenicity analysis for delta variant spike protein predicted through VaxiJen v2.0 server was found to be 0.4703 using a cut-off value of 0.4 characterizing it as antigenic protein that can induce host-immune system response. Identification of MHC-I T-cell epitope The NetCTL server resulted in the prediction of 1262 T-cell epitopes using a threshold value of 0.75. MHC-I binding study was performed on these epitopes using IEBD tools. It identified ~34,075 epitopes for MHC-I. However, 284 epitopes were shortlisted that elicited high binding affinity using a cut-off parameter, i.e., IC50 < 100 and percentile rank ≤ 0.2 based on MHC-I and T-cell interaction. All these shortlisted epitopes were identified as optimal binders to T-cells and therefore assessed for further analysis. Immunogenicity prediction for shortlisted epitopes The efficacy of epitopes to induce T-cells response is based on their immunogenicity level. The shortlisted epitopes were analyzed for immunogenicity prediction. The greater the immunogenicity score, the higher will be the ability of epitopes to simulate naive T-cells and cellular immunity. From above 284 shortlisted MHC-I epitopes, 150 immunogenic epitopes with a cut-off value of the positive score were identified as having significant immunogenic values predicted through IEBD server. These immunogenic shortlisted epitopes were used for further study in vaccine designing. Antigenicity, conservancy, and toxicity analysis Furthermore, toxicity, antigenicity, and conservancy analysis were performed for the shortlisted 150 immunogenic epitopes. The ToxinPred and IEBD conservancy results revealed that all 150 sequences were non-toxic and 100% conserved within the spike protein of delta SARS-CoV-2. However, antigenicity analysis performed through VaxiJen server identified a total of 37 epitopes (Table 1) that are characterized as antigenic, i.e., scoring from 0.5 to 1.0 and selected for further evaluation while the remaining less/non-antigenic immunogenic epitopes were discarded.Table 1 Predicted antigenicity of MHC-I epitopes (all epitopes are non-toxic, and conservancy is 100% for each) S. no Epitopes Antigenicity 1 YIKWPWYIW 0.963 2 VTWFHAIHV 0.5426 3 QLTPTWRVY 1.2119 4 TLADAGFIK 0.5781 5 QYIKWPWYI 1.4177 6 IAIPTNFTI 0.7052 7 IPTNFTISV 0.882 8 SVYAWNRKR 0.765 9 KEIDRLNEV 0.53 10 WTAGAAAYY 0.6306 11 DIADTTDAV 1.0904 12 FNATRFASV 0.5609 13 YQPYRVVVL 0.5964 14 VVFLHVTYV 1.5122 15 LPFNDGVYF 0.5593 16 GVYFASTEK 0.7112 17 ASANLAATK 0.7014 18 NYNYRYRLF 0.9855 19 NATNVVIKV 0.6726 20 STQDLFLPF 0.6619 21 NASVVNIQK 0.8372 22 PFFSNVTWF 0.6638 23 YEQYIKWPW 0.869 24 IGAGICASY 0.6368 25 NLNESLIDL 0.6827 26 HWFVTQRNF 0.746 27 LPFFSNVTW 1.0808 28 FTISVTTEI 0.8535 29 PYRVVVLSF 1.0281 30 KVGGNYNYR 1.5212 31 VTYVPAQEK 0.8132 32 GYLQPRTFL 0.6082 33 GLTVLPPLL 0.6621 34 RLDKVEAEV 0.0765 35 AEIRASANL 0.7082 36 TLLALHRSY 0.8009 37 GQTGKIADY 1.4019 MHC-II epitopes identification and antigenicity, toxicity, and conservancy analysis Beside the MHC-I epitopes prediction, MHC-II epitopes were also identified using IEBD server. The epitopes having low percentile rank (> 0.2) and high binding affinity (IC50 < 100 nM) were analyzed. The server resulted in the prediction of 14 MHC-II epitopes. The ToxinPred and IEBD conservancy tools demonstrated that all 150 sequences inside the spike protein of delta SARS-CoV-2 were non-toxic and 100% conserved. However, using the VaxiJen server, antigenicity analysis revealed a total of 14 epitopes (Table 2) that are classified as antigenic, with scores ranging from 0.5 to 1.0, and hence chose for further examination. The remaining less/non-antigenic immunogenic epitopes were eliminated.Table 2 Predicted antigenicity of MHC-I epitopes (all epitopes are non-toxic, and conservancy is 100% for each) S. no Epitopes Antigenicity 1 QSLLIVNNATNVVIK 0.43 2 FGEVFNATRFASVYA 0.041 3 SLLIVNNATNVVIKV 0.47 4 TQSLLIVNNATNVVI 0.433 5 GEVFNATRFASVYAW −0.12 6 PFGEVFNATRFASVY 0.033 7 LLIVNNATNVVIKVC 0.099 8 EVFNATRFASVYAWN 0.08 9 KTQSLLIVNNATNVV 0.63 10 CPFGEVFNATRFASV 0.29 11 NCTFEYVSQPFLMDL 0.52 12 CTFEYVSQPFLMDLE 0.57 13 QQLIRAAEIRASANL 0.12 14 REGVFVSNGTHWFVT 0.44 MHC restriction cluster analysis of shortlisted epitopes MHCclusters tool used for the identification and evaluation of MHC-I/II epitopes with respect to MHC restricted allele and their appropriate peptides resulted in the confirmation of identified T-cells epitopes. The interactions between MHC-I/II and HLAs are displayed as heat map and phylogeny dynamic tree highlighting the stronger interaction in red color while yellow color represents weak interaction in terms of annotation (Fig. 1).Fig. 1 Clustering analysis for MHC I and II epitopes. The cluster analysis of MHC molecules and HLA alleles A MHCI clustering allele’s, B MHCII clustering alleles. Red color indicates strong interaction while yellow zone indicates the weaker interaction B-cell epitope prediction Both humoral and cellular immunity is needed simultaneously to successfully eliminate the virus from the body. Therefore, B-cell epitopes were also identified against delta spike protein using ABCPred, BCPred, and FBCPred. These tools are resulted in the identification of 36, 21, and 39 B-cell epitopes using a threshold value of 0.51, and 75% specificity (Table S1). Furthermore, resultant B-cells epitopes were further analyzed and shortlisted based on BepiPred linear epitope prediction, Kolaskar Tongaonkar antigenicity, Parker hydrophilicity prediction, Chou-Fasman beta-turn prediction, Karplus-Schulz flexibility prediction, and Emini surface accessibility prediction. The result of predicted B-cell epitopes for delta spike proteins is highlighted in Fig. 2.Fig. 2 B-cell epitopes analysis. A Bepipred linear epitope, B Chou & Fasman beta-turn prediction, C Emini surface accessibility prediction, D Karplus& Schulz flexibility prediction, E Kolaskar & Tongaonkar antigenicity, F Parker hydrophilicity prediction Epitope mapping and prioritization The shortlisted B-cell epitopes were used as a template and manually compared against MHC-I, and MHC-II epitopes to screen out the overlapping epitopes. The comparative analysis was resulted in the shortlisting of 4 epitopes having common MHC-I, MHC-II, and B-cell epitopes (Table 3), i.e., TQSLLIVNNATNVVIKVCEF, TRFASVYAWNRKRISNCV, LQELGKYEQYIKWPWYIWLG, and DQLTPTWRVYSTGSNVFQTR, respectively.Table 3 B cell, MHC-I, and MHC-II epitopes mapping information S. no Position B-cell epitopes MHC-I epitopes MHC-II epitopes Score 1 114–133 TQSLLIVNNATNVVIKVCEF NATNVVIKV SLLIVNNATNVVIKV 0.7 2 343–360 TRFASVYAWNRKRISNCV SVYAWNRKR FGEVFNATRFASVYA 0.8 3 1197–1216 LQELGKYEQYIKWPWYIWLG YIKWPWYIW 0.85 4 624–644 DQLTPTWRVYSTGSNVFQTR SVYAWNRKR 0.9 Vaccine model construction These four shortlisted epitopes from the above analysis were arranged in sequential manner along with the use of four adjuvants, PADRE sequences, GGGS, HEYGAEALERAG, and EAAAK linkers. This combination of different adjuvants and linkers helped to inflict strong immune response in body against viruses while PADRE sequences help to overcome the global polymorphism effect of HLA-DR molecules in various populations. Different vaccine models were constructed. The schematic diagram of vaccine construct is highlighted in Fig. 3 whereas supplementary Table S2 showed the detail of the vaccine constructs.Fig. 3 Schematic presentation of the final multi-epitope vaccine peptide. The 395 amino acid long peptide sequence containing adjuvant (brown) at both N and C terminal was linked with the multi-epitope sequence through an EAAAK linker (green). B cell epitopes and HTL epitopes are linked using GGGS linkers (red) while the CTL epitopes are linked with GGGS linkers (grey) Allergenicity, antigenicity, and solubility prediction The twelve constructed vaccine models were further analyzed to predict their antigenicity, solubility, and allergenicity. The AlgPred tool for allergenicity resulted in the identification of six vaccine constructs (V3, V4, V7, V8, V11, and V12) as highly allergenic in nature scoring from 0.2 to 0.3 and were excluded. However, the antigenicity predicted through ANTIGENpro server and solubility of remaining six vaccine constructs for expression in E. coli vector resulted in the prediction of highly antigenicity and solubility of remaining six vaccine constructs (V1, V2, V5, V6, V9, and V10) and scores were ranging from 0.7 to 0.9, and therefore selected for further study. Table 4 showed the allergenicity, solubility, and antigenicity for all twelve vaccine constructs.Table 4 Vaccine allergenicity, antigenicity, and solubility Vaccine Allergenicity Antigenicity Solubility V1 −0.50794438 0.844476 0.945605 V2 −0.41786815 0.792876 0.961907 V5 −0.50794438 0.824917 0.949964 V6 −0.41786815 0.789273 0.962925 V9 −0.50794438 0.951158 0.825384 V10 −0.41786815 0.963312 0.788187 Physicochemical properties analysis of shortlisted vaccine construct The ProtParam utilized for physicochemical characteristics’ prediction for all shortlisted six vaccines constructs resulted in estimated molecular weight of vaccines model as ~35KDa, pI score of ~5, while instability index score was in the range of 28–34 whereas high aliphatic score ranges from 83 to 86. The grand average of hydropathicity, on the other hand, was calculated to be in the range of 0.2 (Table 5).Table 5 Identified physicochemical properties of vaccine constructs Vaccine No of AA MW PI -ive AA +ive AA Instability index Aliphatic index GRAVY index V1 326 35435.83 5.76 40 43 29.31 83.47 −0.287 V2 317 34317.6 5.47 38 43 34.05 86.72 −0.25 V5 326 35435.83 5.76 43 40 28.83 83.47 −0.287 V6 317 34317.6 5.47 43 38 33.56 86.72 −0.25 V9 326 35435.83 5.76 43 40 29.42 83.47 −0.287 V10 317 34317.6 5.47 43 38 34.17 86.72 −0.25 Structure prediction and validation of selected vaccine construct Phyre2 tool was used for comparative 3D structure modeling of six shortlisted model vaccine constructs. Based on modeled structure and template sequence similarities, V1 vaccine construct was selected as final vaccine model. The template identified for V1 was PDB ID: 1EQ1A, Apolipophorin-II protein from Manduca sexta having sequence identity of 38% (Fig. 4).Fig. 4 Vaccine structure modeling and validation. A The 3D model of a multi-epitope vaccine was obtained by Swiss model B with template comparison and C and vaccine sequence Furthermore, the 3D structure evaluation through PROCHECK results in the stereochemical property identification of the final selected vaccine construct as, 80.9% residues in favorable region, 13.9% residue in additionally allowed region, and 0.9% residues in disallowed region, respectively (Fig. 5a). The ProSA tool predicted a Z-score value of −3.66, indicating the model is near to that of NMR/X-ray crystallography derived structures (Fig. 5b). Moreover, PSIPRED tool was used for the 2D (secondary structure) structure validation showed similar number of alpha helices, beta sheets, and beta turn as shown in Fig. 5c.Fig. 5 Structure evaluation through PSIPRED, ProSA, and PROCHECK. A Modeled structure validation through Ramachandran plot using PROCHECK showing 91% residues in favored region, B ProSA-web modeled structure evaluation indicating the z-score of −3.66 highlighting the protein in NMR and X-ray crystallographic region whereas, C shows structure confirmation for final vaccine construct generated through PSIPRED nearly same position of helices and beta-sheets as modeled structure Molecular docking of vaccine construct (V1) The interaction study of V1 model was performed with six HLAs and TLR4/MD2 complex (PDB 2Z65) to evaluate the enhancement in immune response. V1 contains the adjuvant HBHA conserved protein that is agonist to TLR4 protein which induced several immune responses. The PatchDock docking resulted in the −0.6 binding energy between V1 and TLR-4/MD2 complex (Table 6).Table 6 Docked score of HLA and vaccine model of delta variant Vaccine construct HLA alleles (PDB: ID) Score Area Hydrogen bond energy Global energy Ace V1 1A6A 15,698 2777.80 −1.06 9.52 360.52 3C5J 16,424 2206.70 −1.32 −1.81 405.23 1H15 18,840 3680.80 −0.30 −0.30 194.99 2FSE 17,542 2200.40 −2.90 −55.84 493.36 2Q6W 16,512 3323.10 −3.34 3.05 141.66 2SEB 17,332 2355.90 −2.83 −38.16 363.88 2Z65 18,750 2846.10 0.00 −0.06 −30.14 The PPI interactions of vaccine construct and TLR4/MD showed 179 non-bonded interactions and three hydrogen bond interactions between Ser265-Asp35, Lys224-Pro16, and Cys193-Ala17 of TLR4/MD and vaccine complex, respectively as shown in Fig. 6.Fig. 6 Docked vaccine construct with TLR4/MD. A Docked complex of vaccine (red) and TL4/MD (purple) (B), interaction occurs between the vaccine model and TLR4/MD protein. Interacting residues of vaccine represented in brown color, while protein interacting residues have been highlighted in orange-red color, C all interactions found between the docked complexes, i.e., blue lines represent hydrogen bonding, and orange lines represents non bounded interactions Molecular dynamics simulation study The GROMACS simulation tool was used to determine the movements of vaccine construct in biological environment. The molecular dynamics simulation was conducted for the best-docked model to validate the complex interactions. The simulations resulted in the stability of vaccine construct at 4 ns (Fig. 7).Fig. 7 Molecular dynamic simulation of construct V1. A Root mean square deviation (RMSD) of protein backbone, B plot of radius of gyration vs time during MDS However, the stability of vaccine and TLR4-vaccine complex was also determined through iMOD tool. Deformability graphs were produced that illustrates the normal mode analysis (NMA) for the mobility, flexibility, and stability of vaccine-protein in terms of peaks showing the deformability found in complex as shown in Fig. 8a. The visualization and co-variance of docked complex relationship are explained by the generation of B factor (Fig. 8b). The variance association plot generated the individual and cumulative variance represented through red and green colored highlighted in Fig. 8c. The eigenvalue of the complex was found to be 9.263151e−05 as shown in (Fig. 8d), and co-variance map representing the correlation, non-correlation, and anti-correlation motion found between a pair of residues represented by red, white, and blue color respectively (Fig. 8e).Fig. 8 The results of molecular dynamics simulation of vaccine construct and TLR4/MD docked complex. A Deformability, B B-factor, C variance (red color indicates individual variances and green color indicates cumulative variances), D eigenvalues, and E co-variance map (correlated (red), uncorrelated (white), or anti-correlated (blue) motions) Immune response simulation Furthermore, C-immune tool was employed for the prediction of human immune system response after the injection of vaccine at different time interval. It confirmed the consistency of immune response with real immune reactions, i.e., identification of B-cell, T-cytotoxic cells, T-helper cells, natural killer cells production, interleukins/interferons production, and antibodies production (Fig. 9). A raise in IgG1+IgG2, IgM, and IgG+IgM was observed after the induction of vaccine injection resulting in decrease antigen concentration (Fig. 9a and b). After the induction of vaccine construct injections, an increased production of Th (helper), Tc (cytotoxic) cells, and memory cells population (Fig. 9c–e) was observed. In addition, IFN-g production was also stimulated after immunization (Fig. 9f).Fig. 9 C-ImmSim presentation of an in silico immune simulation with the construct. A Immunoglobulin production in response to antigen injections (black vertical lines); specific subclasses are showed as colored peaks and B the evolution of B-cell populations after the three injections. C T-helper cell populations per state after the injections, D the evolution of T-cytotoxic, and E highlights the production of natural killer cells. The resting state represents cells not presented with the antigen while the anergic state characterizes tolerance of the T-cells to the antigen due to repeated exposures. F The main plot shows cytokine levels after the injections. The insert plot shows IL-2 level with the Simpson index, D shown by the dotted line. D is a measure of diversity. Increase in D over time indicates emergence of different epitope-specific dominant clones of T-cells. The smaller the D value, the lower the diversity Codon optimization and in silico cloning of V1 The shortlisted final vaccine construct V1 was back translated into cDNA in order to be expressed in E. coli (strain K12) and optimized and clone codon. The Codon Optimization Index (CAI) value for V1 was predicted to be 0.9603, whereas, GC content identified for the adapted sequence was 72% explicit high rate of expression. Finally, the recombinant plasmid with inserted adapted codon sequences was performed using SnapGene tool between 5369 and 6527 bp of pET28a(+) vector (Fig. 10).Fig. 10 Codon optimization and in silico cloning of vaccine model. In silico restriction cloning of the multi-epitope vaccine sequence into the pET30a (+) expression vector using SnapGene software, the red part represents the vaccine’s gene coding, and the black circle represents the vector backbone Discussion The COVID-19 is declared as pandemic and surged to cause increase in cases often leading to deaths due to various transmissible variants of the causative virus (UK (alpha/B.1.1.7), Brazil (gamma/P.1), India (delta/B.1.1.7.2), and South Africa (beta/B.1.351)) (World Health Organization: Interim recommendations for use of the Pfizer–BioNTech COVID-19 vaccine, BNT162b2, under emergency use listing: interim guidance, first issued 8 January 2021, updated 15 June 20212021). Till the end of April 2021, the delta (B.1.617.2) (https://www.who.int/en/activities/tracking-SARS-CoV-2-variants/) variant, first identified in India, has replaced the alpha (B.1.1.7) variant in usually gathered genomic data that resulted in growth of case numbers and hospitalizations (Brown et al. 2021). It currently makes up about 20% of newly diagnosed coronavirus cases in the USA alone (Farinholt et al. 2021), while 45% in Sydney declared the rise in delta cases as “worst health crisis in 120 years.” This variant of concern (VOC) (https://www.cdc.gov/coronavirus/2019-ncov/variants/variant-info.html#Consequence) is changing the current pandemic trajectory by having distinct epidemiological changes in transmission and antigenic escape, dominating in America, UK (Riley et al. 2021), Scotland (Angelo et al. 2017; Sheikh et al. 2021), Israel, Sydney, and South Asia (https://www.bbc.com/news/world-asia-53420537). A successful vaccination campaign in the start of 2021 has substantially increased the immunity level whereas the advent of the B.1.617.2 lineage of SARS-CoV-2 has put the delaying dosing tactics of immunization (first dose of vaccine followed by second dose given within 12 weeks) to a new challenge (Yang and Shaman 2021). The substitution and mutation in delta variant spike protein (the main immune dominant region) reduced the binding of several antibodies suggesting the role of these mutations in immune evasion (Planas et al. 2021). Recently, it was reported by Wall et al. (2021) that they identified the resistance of delta variant fourfold less potent compared to variant alpha (B.1.1.7) to neutralization of sera from convalescent patients whereas, sera from people who received a single dose of AstraZeneca or Pfizer vaccines hardly inhibited delta variant (Mlcochova et al. 2021; Planas et al. 2021). Nevertheless, some studies also suggested the use of multiple (three) dose regimes of vaccine to minimize delta variant effects and induce immunity (https://edition.cnn.com/2021/07/08/health/us-coronavirus-thursday/index.html). It has been reported that delta variant infected individuals on average having up to 1000 times more copies of the virus in their respiratory system compared to those infected by alpha strain (Mlcochova et al. 2021). Due to high transmission and immune evasion of ongoing delta variant, the development of new vaccines is urgently needed to overcome the spread of fourth wave of the pandemic. The traditional methods of vaccine designing utilize large proteins or the complete organisms that produced an unnecessary antigenic load and increased allergenic responses (Chauhan et al. 2019; Sette and Fikes 2003). On the other hand, the immunoinformatic approaches are cost effective and time saving that can address this problem through peptide-based vaccine construction and stimulates the strong but targeted immune response (He et al. 2018; Lu et al. 2017). Therefore, the current study employed the in silico approach-based reverse vaccinology to attain a multi-epitope vaccine against spike protein of delta variant that can induce the activation of immune cells (Abraham Peele et al. 2020). The design of multi-epitope-based vaccine is an emerging field that produces the vaccine models having not only protective immunity (Cao et al. 2017; Guo et al. 2014; Zhou et al. 2009) but also has been characterized in phase-I clinical trials (Jiang et al. 2017; Lennerz et al. 2014; Slingluff et al. 2013; Toledo et al. 2001). The present study resulted in the identification of probable immunogenic MHC-I, MHC-II, and B-cell epitopes to construct the multi-epitope vaccine using various filters such as (i) the epitopes must be non-toxic, antigenic, non-allergenic, highly conserved (Table 1) and (ii) have the ability to bind to MHC-I/II alleles, and should be overlapping to CTL, HTL, and B-cell epitopes (Table 2). Bazhan et al. applied the similar approach to design T-cell multi-epitope vaccine model against Ebola virus that was significantly immunogenic in mice (Bazhan et al. 2019). In the current study, twelve different vaccine constructs were modeled using four adjuvants, i.e., HBHA protein, HBHA conserved sequence, beta-defensin, and L7/L12 ribosomal protein along with GGGS, PADRE sequences, HEYGAEALERAG, and EAAAK linkers (Table S2). These twelve-vaccine models were further subjected to filtration steps, i.e., allergenicity, antigenicity, solubility, and physiochemical property analysis to shortlist a most promiscuous vaccine construct against the delta variant. The filtration resulted in the identification of V1 as the most potent vaccine constructs against delta variant as non-allergenic, most antigenic, highly soluble over expressing in E. coli and having suitable physicochemical properties (Tables 4 and 5). Similar in silico strategies were also applied by Foroutan et al. against Toxoplasma gondii to evaluate the allergenicity and physiochemical properties of their model vaccine and through laboratory validation. It was validated that this vaccine model design approach was able to trigger immune response in mice (Foroutan et al. 2020). Hence, the shortlisted V1 vaccine in this study was modeled through Phyre2 tool and validation of modeled structure was performed through PROCHECK. The Ramachandran plot analysis showed 80% of residues classified in favored region, validating the tertiary structure of the vaccine. Z-score assessment by ProSA web server, i.e., −3.66 indicated that the protein falls in experimentally approved structures of proteins solved by NMR and X-ray crystallographic methods. Furthermore, the spike protein of delta variant should interact with Toll-Like Receptor 4 (TLR4) expressed in immune cells to induce CTB (Boehme and Compton 2004; Carty and Bowie 2010; Xagorari & Chlichlia, 2008). It has been reported that the CTB lost ability to trigger inflammatory response in TLR4-deficient macrophages (Vaure and Liu 2014). It was demonstrated through ELISA-based assays that the direct binding of CTB with TLR4 inflicts the activation of NF-κB (Phongsisay et al. 2015). Hence, the interaction of modeled vaccine construct with human leukocyte antigen (HLA) and TLR4 to elucidate effective immune response was studied using molecular docking simulation studies. The docking study of TLR4 and the vaccine model resulted in the formation of three hydrogen bonds between Ser265-Asp35, Lys224-Pro16, and Cys193-Ala17. Several studies highlighted the importance of interaction of vaccine with TLR4 such as, Totura et al. demonstrated that the susceptibility of mice to SARS-CoV infection is relatively high in TLR4 deficient mice compared to wild type (Totura et al. 2015), similarly Hu et al. observed that upregulation in expression of TLR4 when exposed to SARS-CoV infection, suggesting the importance of TLR in immune response stimulation (Hu et al. 2012). Additionally, the vaccine model was simulated under the in vivo conditions to check its stability using GROMACS. The molecular dynamics simulation of the vaccine for 10 ns displayed the stability of vaccine model at 4 ns (Fig. 8). The codon optimization of V1 model was reverse translated to its cDNA to ensure a successful expression in E. coli pET-28a(+) expression vector. The GC and CAI values predicted for V1 were 72% and 0.9603, respectively resulting in the successful expression of vaccine (Fig. 9). Comparably, Foroutan et al. performed in silico codon optimization before expressing it in to mice (Foroutan et al. 2020). The immune simulation of vaccine models showed that the constructed vaccine model against delta variant significantly elicited immune response (Fig. 10). Correspondingly, the immune-simulation studies have been widely used for the construction of chimeric vaccine model against Klebsiella pneumoniae (Solanki et al. 2021), Mycobacterium tuberculosis (Bibi et al. 2021), Acinetobacter baumannii (Solanki and Tiwari 2018), Ebola virus (Ullah et al. 2020), as well as against cancerous antigens (Zhang 2018). Current study’s pipeline can be used further for the identification and designing a vaccine model against other SARS-CoV-2 variants too. Conclusion and future perspectives Briefly, the efforts to tackle the new mutated variants of SARS-CoV-2 are ongoing. However, low progress observed due to the immune evasion and high transmissibility of such variants. Although vaccines are being introduced, yet no effective results are reported. The elimination of SARS-CoV-2 and its variants will not be achieved without the innovative control strategies. Given that the spike proteins are the main source of viral infection, the peptide vaccine expressed from spike proteins might possibly provide therapeutic and prophylactic advantages. The chimeric vaccine models developed in this work could be utilized as a supplement to other approaches to eliminate the delta variant. However, the modeled V1 vaccine needs to be validated by in vitro as well as animal models and if successful then pre-clinical studies are required before administration. For top selected V1 construct, several ways to explore or validate in vivo cellular localization (e.g., immunofluorescence and western-blot tests), protein structure (e.g., crystallography), and protein–protein interactions (e.g., yeast two-hybrid) may be the future interests. Pilot vaccination trials are required to validate in vivo immunogenicity of peptides. For that various factors must be considered including antigen design/production (such as peptide, native protein, synthetic, polymers, type of host expression system, recombination with other promising antigens, and linkers), antigen administration (for example route/system, dose, adjuvant), host response (such as humoral and cellular immune response, physiological and clinical responses). Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 23 KB) Supplementary file2 (DOCX 13.9 KB) Acknowledgements The authors would like to acknowledge the Higher Education Commission of Pakistan for providing the financial support under National Research Program for Universities. Author contribution KJ and KK have done the experiments and written the original draft. MNA and SSH have rechecked the manuscript for any errors. ZB has analyzed the data. FA and SAK have done the computational studies. RU has designed and supervised an experiment. Declarations Ethics approval Not applicable. Consent to participate Not applicable. Consent for publication All the authors have approved the manuscript for publication. Competing interests The authors declare no competing interests. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Abraham MJ Murtola T Schulz R Páll S Smith JC Hess B Lindahl E GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers SoftwareX 2015 1 19 25 10.1016/j.softx.2015.06.001 Abraham Peele K Srihansa T Krupanidhi S Ayyagari VS Venkateswarulu T Dynamics Design of multi-epitope vaccine candidate against SARS-CoV-2: a in-silico study J Biomol Struct Dyn 2020 39 10 3793 3801 10.1080/07391102.2020.1770127 32419646 Angelo MA Grifoni A O'Rourke PH Sidney J Paul S Peters B de Silva AD Phillips E Mallal S Diehl SA Human CD4+ T cell responses to an attenuated tetravalent dengue vaccine parallel those induced by natural infection in magnitude, HLA restriction, and antigen specificity J Virol 2017 91 5 e02147 e12116 10.1128/JVI.02147-16 27974563 Bazhan SI Antonets DV Karpenko LI Oreshkova SF Kaplina ON Starostina EV Dudko SG Fedotova SA Ilyichev AA In silico designed ebola virus T-cell multi-epitope DNA vaccine constructions are immunogenic in mice Vaccines 2019 7 2 34 10.3390/vaccines7020034 Behl T, Kaur I, Aleya L, Sehgal A, Singh S, Sharma N, Bhatia S, Al-Harrasi A, Bungau S (2022) CD147-spike protein interaction in COVID-19: Get the ball rolling with a novel receptor and therapeutic target. Sci Total Environ 808:152072 Bibi S Ullah I Zhu B Adnan M Liaqat R Kong W-B Niu S In silico analysis of epitope-based vaccine candidate against tuberculosis using reverse vaccinology Sci Rep 2021 11 1 1 16 10.1038/s41598-020-79139-8 33414495 Boehme KW Compton T Innate sensing of viruses by toll-like receptors J Virol 2004 78 15 7867 7873 10.1128/JVI.78.15.7867-7873.2004 15254159 Brown KA, Gubbay J, Buchan SA, Daneman N, Mishra S, Patel S, Day T (2021) Inflection in prevalence of SARS-CoV-2 infections missing the N501Y mutation as a marker of rapid delta (B. 1.617. 2) lineage expansion in Ontario, Canada. medRxiv. 10.1101/2021.1106.1122.21259349 Buchan DW Jones DT The PSIPRED protein analysis workbench: 20 years on Nucleic Acids Res 2019 47 W1 W402 W407 10.1093/nar/gkz297 31251384 Callaway E Delta coronavirus variant: scientists brace for impact Nature 2021 595 17 18 10.1038/d41586-021-01696-3 34158664 Campbell F Archer B Laurenson-Schafer H Jinnai Y Konings F Batra N Pavlin B Vandemaele K Van Kerkhove MD Jombart T Increased transmissibility and global spread of SARS-CoV-2 variants of concern as at June 2021 Euro Surveill 2021 26 24 2100509 10.2807/1560-7917.ES.2021.26.24.2100509 Cao Y Li D Fu Y Bai Q Chen Y Bai X Jing Z Sun P Bao H Li P Rational design and efficacy of a multi-epitope recombinant protein vaccine against foot-and-mouth disease virus serotype A in pigs Antiviral Res 2017 140 133 141 10.1016/j.antiviral.2017.01.023 28161579 Carty M Bowie AG Recent insights into the role of toll-like receptors in viral infection Clin Exp Immunol 2010 161 3 397 406 10.1111/j.1365-2249.2010.04196.x 20560984 Chauhan V Rungta T Goyal K Singh MP Designing a multi-epitope based vaccine to combat Kaposi sarcoma utilizing immunoinformatics approach Sci Rep 2019 9 1 1 15 10.1038/s41598-018-37186-2 30626917 Chen H Wang Y Liu W Zhang J Dong B Fan X de Jong MD Farrar J Riley S Smith GJ Serologic survey of pandemic (H1N1) 2009 virus, Guangxi Province, China Emerging Infect Dis 2009 15 11 1849 10.3201/eid1511.090868 Chou P Fasman G Prediction of the secondary structure of proteins from their amino acid sequence Adv Enzymol 1978 47 1 45 148 364941 Cowan R, Grosdidier G (2000) Visualization tools for monitoring and evaluation of distributed computing systems. Paper presented at the Proc. of the International Conference on Computing in High Energy and Nuclear Physics, Padova, Italy D’Mello A Ahearn CP Murphy TF Tettelin H ReVac: a reverse vaccinology computational pipeline for prioritization of prokaryotic protein vaccine candidates BMC Genom 2019 20 1 1 21 10.1186/s12864-019-6195-y Davis C, Logan N, Tyson G, Orton R, Harvey W, Haughney J, Perkins J, Peacock T, Barclay WS, Cherepanov P (2021) Reduced neutralisation of the delta (B. 1.617. 2) SARS-CoV-2 variant of concern following vaccination. medRxiv. 10.1101/2021.1106.1123.21259327 Dhanda SK Mahajan S Paul S Yan Z Kim H Jespersen MC Jurtz V Andreatta M Greenbaum JA Marcatili P IEDB-AR: Immune epitope database—analysis resource in 2019 Nucleic Acids Res 2019 47 W1 W502 W506 10.1093/nar/gkz452 31114900 Doytchinova IA Flower DR VaxiJen: a server for prediction of protective antigens, tumour antigens and subunit vaccines BMC Bioinform 2007 8 1 1 7 10.1186/1471-2105-8-4 El-Manzalawy Y Dobbs D Honavar V Predicting linear B-cell epitopes using string kernels J Mol Recognit 2008 21 4 243 255 10.1002/jmr.893 18496882 Emini EA Hughes JV Perlow D Boger J Induction of hepatitis A virus-neutralizing antibody by a virus-specific synthetic peptide J Virol 1985 55 3 836 839 10.1128/jvi.55.3.836-839.1985 2991600 Farinholt T, Doddapaneni H, Qin X, Menon V, Meng Q, Metcalf G, Chao H, Gingras M-C, Farinholt P, Agrawal C (2021) Transmission event of SARS-CoV-2 Delta variant reveals multiple vaccine breakthrough infections. medRxiv. 10.1101/2021.1106.1128.21258780 Foroutan M, Ghaffarifar F, Sharifi Z, Dalimi A (2020) Vaccination with a novel multi-epitope ROP8 DNA vaccine against acute Toxoplasma gondii infection induces strong B and T cell responses in mice. Comp Immunol Microbiol Infect Dis 69:101413 Gasteiger E, Hoogland C, Gattiker A, Wilkins MR, Appel RD, Bairoch A (2005) Protein identification and analysis tools on the ExPASy server. Proteomics Protocols Handbook 571–607 Gheorghe G Ilie M Bungau S Stoian AMP Bacalbasa N Diaconu CC Is there a relationship between COVID-19 and hyponatremia? Medicina 2021 57 1 55 10.3390/medicina57010055 33435405 Grote A, Hiller K, Scheer M, Münch R, Nörtemann B, Hempel DC, Jahn D (2005) JCat: a novel tool to adapt codon usage of a target gene to its potential expression host. Nucleic Acids Res 33(suppl_2):W526–W531 Guo L, Yin R, Liu K, Lv X, Li Y, Duan X, Chu Y, Xi T, Xing Y (2014) Immunological features and efficacy of a multi-epitope vaccine CTB-UE against H. pylori in BALB/c mice model. Appl Microbiol Biotechnol 98(8):3495–3507 He R Yang X Liu C Chen X Wang L Xiao M Ye J Wu Y Ye L Efficient control of chronic LCMV infection by a CD4 T cell epitope-based heterologous prime-boost vaccination in a murine model Cell Mol Immunol 2018 15 9 815 826 10.1038/cmi.2017.3 28287115 Hu W Yen Y-T Singh S Kao C-L Wu-Hsieh BA SARS-CoV regulates immune function-related gene expression in human monocytic cells Viral Immunol 2012 25 4 277 288 10.1089/vim.2011.0099 22876772 Huang X Zhang C Pearce R Omenn GS Zhang Y Identifying the zoonotic origin of SARS-CoV-2 by modeling the binding affinity between the spike receptor-binding domain and host ACE2 J Proteome Res 2020 19 12 4844 4856 10.1021/acs.jproteome.0c00717 33175551 Ito K Piantham C Nishiura H Predicted dominance of variant delta of SARS-CoV-2 before Tokyo Olympic Games, Japan, July 2021 Euro Surveill 2021 26 27 2100570 10.2807/1560-7917.ES.2021.26.27.2100570 Jalal K Abu-Izneid T Khan K Abbas M Hayat A Bawazeer S Uddin R Identification of vaccine and drug targets in Shigella dysenteriae sd197 using reverse vaccinology approach Sci Rep 2022 12 1 1 19 10.1038/s41598-021-99269-x 34992227 Jiang P Cai Y Chen J Ye X Mao S Zhu S Xue X Chen S Zhang L Evaluation of tandem Chlamydia trachomatis MOMP multi-epitopes vaccine in BALB/c mice model Vaccine 2017 35 23 3096 3103 10.1016/j.vaccine.2017.04.031 28456528 Kabir M, Uddin M, Hossain M, Abdulhakim JA, Alam M, Ashraf GM, Bungau SG, Bin-Jumah MN, Abdel-Daim MM, Aleya L (2020) nCOVID-19 pandemic: from molecular pathogenesis to potential investigational therapeutics. Front Cell Dev Biol 616 Kar T Narsaria U Basak S Deb D Castiglione F Mueller DM Srivastava AP A candidate multi-epitope vaccine against SARS-CoV-2 Sci Rep 2020 10 1 1 24 10.1038/s41598-020-67749-1 31913322 Karplus P Schulz G Prediction of chain flexibility in proteins Naturwissenschaften 1985 72 4 212 213 10.1007/BF01195768 Khan K, Jalal K, Uddin R (2022) An integrated in silico based subtractive genomics and reverse vaccinology approach for the identification of novel vaccine candidate and chimeric vaccine against XDR Salmonella typhi H58. Genomics 110301 Kim Y Ponomarenko J Zhu Z Tamang D Wang P Greenbaum J Lundegaard C Sette A Lund O Bourne PE Immune epitope database analysis resource Nucleic Acids Res 2012 40 W1 W525 W530 10.1093/nar/gks438 22610854 Krause PR Fleming TR Longini IM Peto R Briand S Heymann DL Beral V Snape MD Rees H Ropero A-M SARS-CoV-2 variants and vaccines New Engl J Med 2021 385 179 186 10.1056/NEJMsr2105280 34161052 Kumar A Dwivedi P Kumar G Narayan RK Jha RK Parashar R Sahni C Pandey SN Second wave of COVID-19 in India could be predicted with genomic surveillance of SARS-CoV-2 variants coupled with epidemiological data: a tool for future medRxiv 2021 10.1101/2021.1106.1109.21258612 Kumar A Parashar R Kumar S Faiq MA Kumari C Kulandhasamy M Narayan RK Jha RK Singh HN Prasoon P Emerging SARS-CoV-2 variants can potentially break set epidemiological barriers in COVID-19 J Med Virol 2021 10.1002/jmv.27467 Larsen MV Lundegaard C Lamberth K Buus S Lund O Nielsen M Large-scale validation of methods for cytotoxic T-lymphocyte epitope prediction BMC Bioinform 2007 8 1 1 12 10.1186/1471-2105-8-1 Laskowski R MacArthur M Moss D Thornton J PROCHECK: a program to check the stereochemical quality of protein structures J Appl Crystallogr 1993 26 2 283 291 10.1107/S0021889892009944 Laskowski RA Jabłońska J Pravda L Vařeková RS Thornton JM PDBsum: Structural summaries of PDB entries Protein Sci 2018 27 1 129 134 10.1002/pro.3289 28875543 Lazarevic I Pravica V Miljanovic D Cupic M Immune evasion of SARS-CoV-2 emerging variants: what have we learnt so far? Viruses 2021 13 7 1192 10.3390/v13071192 34206453 Lennerz V, Gross S, Gallerani E, Sessa C, Mach N, Boehm S, Hess D, Von Boehmer L, Knuth A, Ochsenbein AF (2014) Immunologic response to the survivin-derived multi-epitope vaccine EMD640744 in patients with advanced solid tumors. Cancer Immunol Immunother 63(4):381–394 Li J, Lai S, Gao GF, Shi W (2021) The emergence, genomic diversity and global spread of SARS-CoV-2. Nature 1–11 López-Blanco JR Aliaga JI Quintana-Ortí ES Chacón P iMODS: internal coordinates normal mode analysis server Nucleic Acids Res 2014 42 W1 W271 W276 10.1093/nar/gku339 24771341 Lu C Meng S Jin Y Zhang W Li Z Wang F Wang-Johanning F Wei Y Liu H Tu H A novel multi-epitope vaccine from MMSA-1 and DKK 1 for multiple myeloma immunotherapy Br J Haematol 2017 178 3 413 426 10.1111/bjh.14686 28508448 Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M (2008) NetMHC-3.0: Accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8–11. Nucleic Acids Res 36(suppl_2):W509–W512 Magnan CN Randall A Baldi P SOLpro: Accurate sequence-based prediction of protein solubility Bioinformatics 2009 25 17 2200 2207 10.1093/bioinformatics/btp386 19549632 Magnan CN Zeller M Kayala MA Vigil A Randall A Felgner PL Baldi P High-throughput prediction of protein antigenicity using protein microarray data Bioinformatics 2010 26 23 2936 2943 10.1093/bioinformatics/btq551 20934990 Mashiach E, Schneidman-Duhovny D, Andrusier N, Nussinov R, Wolfson HJ (2008) FireDock: a web server for fast interaction refinement in molecular docking. Nucleic Acids Res 36(suppl_2):W229–W232 Mlcochova P, Kemp SA, Shanker Dhar M, Papa G, Meng B, Mishra S, Whittaker C, Mellan T, Ferreira I, Datir R, Collier D, Singh S, Pandey R, Ponnusamy K, Radhakrishnan VS, Sengupta S, Brown J, Marwal R, Ponnusamy K, Radhakrishnan VS, Goonawardne N, Abdullahi A, Devi P, Wattal C, Caputo D, Peacock T, Goel N, Vaishya R, Charles O, Chattopadhyay P, Agarwal M, Satwik A, Consortium I, Collaboration NB, Mavousian A, Brown J, Zhou J, Goonawardne N, Hyeon Lee J, Bassi J, Silacci-Fegni C, Saliba C, Pinto D, Irie T, Yoshida I, Hamilton WL, Sato K, James L, Corti D, Piccoli L, Bhatt S, Flaxman S, Barlcay W, Rakshit P, Agrawal A, Gupta RK (2021) SARS-CoV-2 B. 1.617. 2 Delta variant emergence and vaccine breakthrough. bioRxiv. 10.1101/2021.1105.1108.443253 Negrut N Codrean A Hodisan I Bungau S Tit DM Marin R Behl T Banica F Diaconu CC Nistor-Cseppento DC Efficiency of antiviral treatment in COVID-19 Exp Therapeutic Med 2021 21 6 1 7 10.3892/etm.2021.10080 Nielsen M Lundegaard C Worning P Lauemøller SL Lamberth K Buus S Brunak S Lund O Reliable prediction of T-cell epitopes using neural networks with novel sequence representations Protein Sci 2003 12 5 1007 1017 10.1110/ps.0239403 12717023 Nosrati M, Behbahani M, Mohabatkar H (2019) Towards the first multi-epitope recombinant vaccine against Crimean-Congo hemorrhagic fever virus: a computer-aided vaccine design approach. J Biomed Inf 93:103160 Ojha R, Gupta N, Naik B, Singh S, Verma VK, Prusty D, Prajapati VK (2020) High throughput and comprehensive approach to develop multiepitope vaccine against minacious COVID-19. Eur J Pharm Sci 151:105375 Papenfuss M (Jun. 25, 2021) Dangerous delta COVID-19 variant infecting vaccinated adults in Israel. Retrieved from https://www.huffpost.com/entry/covid-19-delta-infecting-vaccinated-in-israel_n_60d6661ce4b066ff5aba8faa Parker J Guo D Hodges R New hydrophilicity scale derived from high-performance liquid chromatography peptide retention data: correlation of predicted surface residues with antigenicity and X-ray-derived accessible sites Biochemistry 1986 25 19 5425 5432 10.1021/bi00367a013 2430611 Pettersen EF Goddard TD Huang CC Couch GS Greenblatt DM Meng EC Ferrin TE UCSF Chimera—a visualization system for exploratory research and analysis J Comput Chem 2004 25 13 1605 1612 10.1002/jcc.20084 15264254 Phongsisay V Iizasa EI Hara H Yoshida H Evidence for TLR4 and FcRγ–CARD9 activation by cholera toxin B subunit and its direct bindings to TREM2 and LMIR5 receptors Mol Immunol 2015 66 2 463 471 10.1016/j.molimm.2015.05.008 26021803 Planas D Veyer D Baidaliuk A Staropoli I Guivel-Benhassine F Rajah MM Planchais C Porrot F Robillard N Puech J Reduced sensitivity of SARS-CoV-2 variant delta to antibody neutralization Nature 2021 10.1038/s41586-41021-03777-41589 Ponomarenko J Bui H-H Li W Fusseder N Bourne PE Sette A Peters B ElliPro: a new structure-based tool for the prediction of antibody epitopes BMC Bioinform 2008 9 1 1 8 10.1186/1471-2105-9-514 Poudel S, Ishak A, Perez-Fernandez J, Garcia E, León-Figueroa DA, Romaní L, Bonilla-Aldana DK, Rodriguez-Morales AJ (2022) Highly mutated SARS-CoV-2 omicron variant sparks significant concern among global experts–what is known so far? Travel Med Infect Dis 45:102234 Quinonez E Vahed M Hashemi Shahraki A Mirsaeidi M Structural analysis of the novel variants of SARS-CoV-2 and forecasting in North America Viruses 2021 13 5 930 10.3390/v13050930 34067890 Rahman N Ali F Basharat Z Shehroz M Khan MK Jeandet P Nepovimova E Kuca K Khan H Vaccine design from the ensemble of surface glycoprotein epitopes of SARS-CoV-2: an immunoinformatics approach Vaccines 2020 8 3 423 10.3390/vaccines8030423 Rapin N, Lund O, Bernaschi M, Castiglione F (2010) Computational immunology meets bioinformatics: the use of prediction tools for molecular binding in the simulation of the immune system. PLoS One 5(4):e9862 Rappuoli R, Bottomley MJ, D’Oro U, Finco O, De Gregorio E (2016) Reverse vaccinology 2.0: human immunology instructs vaccine antigen design. J Exp Med 213(4):469–481 Riley S, Walters CE, Wang H, Eales O, Haw D, Ainslie KE, Atchinson C, Fronterre C, Diggle PJ, Page AJ (2021) REACT-1 round 12 report: resurgence of SARS-CoV-2 infections in England associated with increased frequency of the delta variant. medRxiv. 10.1101/2021.1106.1117.21259103 Saha S, Raghava GP (2007) Prediction methods for B-cell epitopes. Immunoinformatics 387–394 Salvatore M, Bhattacharyya R, Purkayastha S, Zimmermann L, Ray D, Hazra A, Kleinsasser M, Mellan TA, Whittaker C, Flaxman S (2021) Resurgence of SARS-CoV-2 in India: Potential role of the B. 1.617. 2 (delta) variant and delayed interventions. medRxiv. 10.1101/2021.1106.1123.21259405 Schneidman-Duhovny D, Inbar Y, Nussinov R, Wolfson HJ (2005) PatchDock and SymmDock: Servers for rigid and symmetric docking. Nucleic Acids Res 33(suppl_2):W363–W367 Sette A Fikes J Epitope-based vaccines: an update on epitope identification, vaccine design and delivery Curr Opin Immunol 2003 15 4 461 470 10.1016/S0952-7915(03)00083-9 12900280 Sharma N, Patiyal S, Dhall A, Pande A, Arora C, Raghava GP (2020) AlgPred 2.0: an improved method for predicting allergenic proteins and mapping of IgE epitopes. Brief Bioinform 22(4):bbaa294 Sharma R Rajput VS Jamal S Grover A Grover S An immunoinformatics approach to design a multi-epitope vaccine against Mycobacterium tuberculosis exploiting secreted exosome proteins Sci Rep 2021 11 1 1 12 10.1038/s41598-020-79139-8 33414495 Sheikh A McMenamin J Taylor B Robertson C SARS-CoV-2 Delta VOC in Scotland: Demographics, risk of hospital admission, and vaccine effectiveness Lancet 2021 397 10293 2461 2462 10.1016/S0140-6736(21)01358-1 34139198 Shey RA Ghogomu SM Esoh KK Nebangwa ND Shintouo CM Nongley NF Asa BF Ngale FN Vanhamme L Souopgui J In-silico design of a multi-epitope vaccine candidate against onchocerciasis and related filarial diseases Sci Rep 2019 9 1 1 18 10.1038/s41598-019-40833-x 30626917 Shu Y McCauley J GISAID: Global initiative on sharing all influenza data–from vision to reality Euro Surveill 2017 22 13 30494 10.2807/1560-7917.ES.2017.22.13.30494 28382917 Sidney J Assarsson E Moore C Ngo S Pinilla C Sette A Peters B Quantitative peptide binding motifs for 19 human and mouse MHC class I molecules derived using positional scanning combinatorial peptide libraries Immunome Res 2008 4 1 1 14 10.1186/1745-7580-4-2 18179690 Slingluff CL Lee S Zhao F Chianese-Bullock KA Olson WC Butterfield LH Whiteside TL Leming PD Kirkwood JM A randomized phase II trial of multiepitope vaccination with melanoma peptides for cytotoxic T cells and helper T cells for patients with metastatic melanoma (E1602) Clin Cancer Res 2013 19 15 4228 4238 10.1158/1078-0432.CCR-13-0002 23653149 Solanki V Sharma S Tiwari V Subtractive proteomics and reverse vaccinology strategies for designing a multiepitope vaccine targeting membrane proteins of Klebsiella pneumoniae Int J Pept Res Ther 2021 27 2 1177 1195 10.1007/s10989-021-10159-2 Solanki V Tiwari V Subtractive proteomics to identify novel drug targets and reverse vaccinology for the development of chimeric vaccine against Acinetobacter baumannii Sci Rep 2018 8 1 1 19 10.1038/s41598-018-26689-7 29311619 Srivastava S Kamthania M Kumar Pandey R Kumar Saxena A Saxena V Kumar Singh S Kumar Sharma R Sharma N Design of novel multi-epitope vaccines against severe acute respiratory syndrome validated through multistage molecular interaction and dynamics J Biomol Struct Dyn 2019 37 16 4345 4360 10.1080/07391102.2018.1548977 30457455 Tagde P, Tagde S, Tagde P, Bhattacharya T, Monzur SM, Rahman M, Otrisal P, Behl T, Abdel-Daim MM, Aleya LJB (2021) Nutraceuticals and herbs in reducing the risk and improving the treatment of COVID-19 by targeting SARS-CoV-2. 9(9):1266 Tao K Tzou PL Nouhin J Gupta RK de Oliveira T Kosakovsky Pond SL Fera D Shafer RW The biological and clinical significance of emerging SARS-CoV-2 variants Nat Rev Genet 2021 22 12 757 773 10.1038/s41576-021-00408-x 34535792 Thakur V, Bhola S, Thakur P, Patel SKS, Kulshrestha S, Ratho RK, Kumar P (2021) Waves and variants of SARS-CoV-2: Understanding the causes and effect of the COVID-19 catastrophe. Infection 1–16 Thomsen M Lundegaard C Buus S Lund O Nielsen M MHCcluster, a method for functional clustering of MHC molecules Immunogenetics 2013 65 9 655 665 10.1007/s00251-013-0714-9 23775223 Toledo H Baly A Castro O Resik S Laferté J Rolo F Navea L Lobaina L Cruz O Miguez J A phase I clinical trial of a multi-epitope polypeptide TAB9 combined with Montanide ISA 720 adjuvant in non-HIV-1 infected human volunteers Vaccine 2001 19 30 4328 4336 10.1016/S0264-410X(01)00111-6 11457560 Tosta SFDO Passos MS Kato R Salgado Á Xavier J Jaiswal AK Soares SC Azevedo V Giovanetti M Tiwari S Multi-epitope based vaccine against yellow fever virus applying immunoinformatics approaches J Biomol Struct Dyn 2021 39 1 219 235 10.1080/07391102.2019.1707120 31854239 Totura AL Whitmore A Agnihothram S Schäfer A Katze MG Heise MT Baric RS Toll-like receptor 3 signaling via TRIF contributes to a protective innate immune response to severe acute respiratory syndrome coronavirus infection Mbio 2015 6 3 e00638 e1615 10.1128/mBio.00638-15 26015500 Uddin R, Jalal K, Khan K (2022) Re-purposing of hepatitis C virus FDA approved direct acting antivirals as potential SARS-CoV-2 protease inhibitors. J Mol Struct 1250:131920 Ullah MA, Sarkar B, Islam SS (2020) Exploiting the reverse vaccinology approach to design novel subunit vaccines against Ebola virus. Immunobiology 225(3):151949 Vaure C Liu Y A comparative review of toll-like receptor 4 expression and functionality in different animal species Front Immunol 2014 5 316 10.3389/fimmu.2014.00316 25071777 Wall EC Wu M Harvey R Kelly G Warchal S Sawyer C Daniels R Adams L Hobson P Hatipoglu E AZD1222-induced neutralising antibody activity against SARS-CoV-2 delta VOC The Lancet 2021 398 10296 207 209 10.1016/S0140-6736(21)01462-8 Wiederstein M, Sippl MJ (2007) ProSA-web: Interactive web service for the recognition of errors in three-dimensional structures of proteins. Nucleic Acids Res 35(suppl_2):W407–W410 World Health Organization: Interim recommendations for use of the Pfizer–BioNTech COVID-19 vaccine, BNT162b2, under emergency use listing: interim guidance, first issued 8 January 2021, updated 15 June 2021. (2021). Retrieved from Geneva: WHO reference number: WHO/2019-nCoV/vaccines/SAGE_recommendation//BNT162b2/2021.2 Xagorari A Chlichlia K Toll-like receptors and viruses: Induction of innate antiviral immune responses Open Microbiol J 2008 2 49 10.2174/1874285800802010049 19088911 Yang W, Shaman J (2021) COVID-19 pandemic dynamics in India and impact of the SARS-CoV-2 Delta (B. 1.617. 2) variant. medRxiv. 10.1101/2021.1106.1121.21259268 Yang Z Bogdan P Nazarian S An in silico deep learning approach to multi-epitope vaccine design: a SARS-CoV-2 case study Sci Rep 2021 11 1 1 21 10.1038/s41598-020-79139-8 33414495 Zhang L Multi-epitope vaccines: a promising strategy against tumors and viral infections Cell Mol Immunol 2018 15 2 182 184 10.1038/cmi.2017.92 28890542 Zhou W-Y Shi Y Wu C Zhang W-J Mao X-H Guo G Li H-X Zou Q-M Therapeutic efficacy of a multi-epitope vaccine against Helicobacter pylori infection in BALB/c mice model Vaccine 2009 27 36 5013 5019 10.1016/j.vaccine.2009.05.009 19446591
PMC009xxxxxx/PMC9005163.txt	==== Front Ir J Med Sci Ir J Med Sci Irish Journal of Medical Science 0021-1265 1863-4362 Springer International Publishing Cham 35415774 3013 10.1007/s11845-022-03013-x Original Article The effect of N95 respirators on vital parameters, PETCO2, among healthcare providers at the pandemic clinics Karsli Emre 1 Yilmaz Atakan 2 Kemancı Aykut 3 Canacik Omer 4 Ozen Mert 2 Seyit Murat 2 Şahin Levent 5 Oskay Alten 2 http://orcid.org/0000-0003-4599-5833 Sabirli Ramazan ramazan_sabirli@hotmail.com 6 Turkcuer Ibrahim 2 1 Department of Emergency Medicine, Faculty of Medicine, Tınaztepe University, Izmir, Turkey 2 grid.411742.5 0000 0001 1498 3798 Department of Emergency Medicine, Faculty of Medicine, Pamukkale University, Denizli, Turkey 3 Department of Emergency Medicine, Tavsanli Doc Dr. Mustafa Kalemli State Hospital, Kutahya, Turkey 4 grid.490320.c Department of Emergency Medicine, Memorial Sisli Hospital, Istanbul, Turkey 5 grid.16487.3c 0000 0000 9216 0511 Department of Emergency Medicine, Kafkas University Faculty of Medicine, Kars, Turkey 6 Department of Emergency Medicine, Bakircay University Faculty of Medicine Cigli Training and Research Hospital, 35620 Izmir, Turkey 12 4 2022 2023 192 2 853860 8 3 2022 8 4 2022 © The Author(s), under exclusive licence to Royal Academy of Medicine in Ireland 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Background Wearing face shields and masks, which used to have very limited public use before the COVID-19 outbreak, has been highly recommended by organizations, such as CDC and WHO, during this pandemic period. Aims The aim of this prospective study is to scrutinize the dynamic changes in vital parameters, change in end tidal CO2 (PETCO2) levels, the relationship of these changes with taking a break, and the subjective complaints caused by respiratory protection, while healthcare providers are performing their duties with the N95 mask. Methods The prospective cohort included 54 healthcare workers (doctors, nurses, paramedics) who worked in the respiratory unit of the emergency department (ED) and performed their duties by wearing valved N95 masks and face shields. The vital parameters and PETCO2 levels were measured at 0–4th–5th and 9th hours of the work-shift. Results Only the decrease in diastolic BP between 0 and 9 h was statistically significant (p = 0.038). Besides, mean arterial pressure (MAP) values indicated a significant decrease between 0–9 h and 5–9 h (p = 0.024 and p = 0.049, respectively). In terms of the vital parameters of the subjects working with and without breaks, only PETCO2 levels of those working uninterruptedly increased significantly at the 4th hour in comparison to the beginning-of-shift baseline levels (p = 0.003). Conclusion Although the decrease in systolic blood pressure (SBP) and MAP values is assumed to be caused by increased fatigue due to workload and work pace as well as increase in muscle activity, the increase in PETCO2 levels in the ED healthcare staff working with no breaks between 0 and 4 h should be noted in terms of PPE-induced hypoventilation. Keywords Emergency department N95 respirators Pandemic clinics PETCO2 Vital parameters issue-copyright-statement© The Author(s), under exclusive licence to Royal Academy of Medicine in Ireland 2023 ==== Body pmcIntroduction After Sars-CoV-2 virus, a novel coronavirus that appeared in China at the end of 2019, spread all over the world, this outbreak was declared a global pandemic by WHO on March 11, 2020, and the health crisis induced by this virus was defined as COVID-19 disease [1, 2]. Sars-CoV-2 infection is transmitted from human to human by means of contact routes or respiratory droplets and leads to clinical conditions in a wide spectrum, ranging from asymptomatic infection to severe pneumonia and acute respiratory distress syndrome [3]. Personal protective equipment (PPE), such as face shields and masks, is considered critically important in minimizing the risk of disease transmission [4–6]. Wearing face shields and masks, which used to have very limited public use before the COVID-19 outbreak, has been highly recommended by organizations, such as CDC and WHO, during this pandemic period. Moreover, an increasing number of reports are being published in relation to these enhanced infection-prevention measures. Mask types can be primarily classified as full masks and half and quarter masks, and this mask classification is specified by the European Committee for Standardization (CEN). Surgical masks, namely filtering face piece (FFP) masks worn during the COVID-19 outbreak, are half-face masks [7]. FFP masks include mask types with varying filtering properties, such as FFP1, FFP2, and FFP3, with particle filtering at a rate of 80%, > 95%, and > 99%, respectively [7, 8]. In addition, respirator mask standards in the USA are specified as N95, N99, N100, R95, P95, P99, and P100 by the National Personal Institute for Occupational Safety and Health (NIOSH) [9]. N95 masks are FDA-approved mask models that provide filtering equivalent to FFP2 masks, because the former can filter > 95% of particles and droplets, while the latter have a protective effect at a rate of > 94% [7, 10]. Some models of these masks feature exhalation valves which reduce exhalation resistance and thus facilitate breathing out [11]. Studies documenting the efficacy of masks in suppressing the spread of viruses and recommending their widespread use were also conducted during the influenza virus pandemic [12]. In this COVID-19 pandemic, a recent study likewise revealed that using respiratory protection helped effectively to decrease the COVID-19 cases in Germany [13]. Furthermore, while WHO recommends surgical masks for general public based on local settings, resources, public preference, and culture, it encourages wearing N95-FFP2 or N99-FFP3 masks for healthcare professionals in care settings [6]. In accordance with the recommendations issued by WHO, healthcare workers in Turkey are obliged by law to wear at least N95/FFP2 medical masks during aerosol generating procedures (e.g., sampling, endotracheal intubation, mechanical ventilation, cardiopulmonary resuscitation, high flow oxygen therapy, respiratory secretion aspiration) in COVID-19 outpatient clinics [14]. Some recent clinical reports have addressed the adverse effects induced by N95 mask use both in various patient populations and in frontline healthcare providers. While N95 face masks reportedly impair cardiopulmonary exercise capacity in medical staff, they might also impose physiological stress on some parameters during dialysis, such as hypoxemia, reduced PaO2, increased respiratory distress, and rate as well as chest discomfort [15, 16]. Among the most frequent complaints by healthcare providers concerning respiratory protection equipment are headache, facial sensitivity, persistent erythema, and acne [17]. Against this background, the ultimate aim of this prospective study is to scrutinize the dynamic changes in vital parameters, change in end tidal CO2 levels, the relationship of these changes with taking a break, and the subjective complaints caused by respiratory protection, while healthcare providers are performing their duties with the N95 mask. Methods Study design and study population The ethical approval of this prospective cohort study was granted by the Ethics Committee of Pamukkale University (reference no E-6016787–020-11,772). The written informed consent forms were filled out and gathered from each subject prior to the study. The prospective cohort included the healthcare workers (doctors, nurses, paramedics) who worked in the respiratory unit of the emergency department (ED) between the dates of 09.01.2021 and 22.01.2021, performed their duties by wearing valved N95 masks (Fig. 1) and face shields, and overall in accordance with WHO guidelines [18] had no history or symptoms of any known disease, and were not on any drugs. In the emergency pandemic clinic, working hours were scheduled as shifts of 8–16 h. This study was carried out between 8 a.m. and 5 p.m. in the daytime shift. All measurements were made by the same person, who had no knowledge of the study, at the beginning of the working shift, before and after the lunch break, and at the end of the shift.Fig. 1 A valved N95 respirator The dataset of this report consisted of the information on the subjects’ age, gender, and smoking status, their vital parameters, and the total number of minutes when they took a break with the mask removed between the 0–4th and 5–9th hours. The primary outcome was the effect of using the N95 respirators on vital parameters and PETCO2, while the secondary outcome was the effect of wearing the N95 respirators on the comfort of the healthcare providers. Vital parameters and PETCO2 measurement Fever, heart rate, systolic and diastolic blood pressures (BP), fingertip oxygen saturations (sPO2), and PETCO2 levels were measured at the beginning of the shift (0th hour), before lunch (4th hour), at the return of lunch (5th hour), and at the end of the shift (9th hour). The shock index and MAP values were computed. Body temperature measurement As described in previous publications, fever of the subjects was measured by an infrared thermometer at a 0.5-cm distance from the mid-forehead [19]. Blood pressure measurement Blood pressure measurement was performed on the right arm with a manual sphygmomanometer using the auscultatory method after a 5-min rest [20]. Heart rate and sPO2 measurement The heart rate and sPO2 levels of the subjects were measured, waiting for 2 min after the device was attached to the fingertip, and the value displayed on the screen at the end of the 2nd min was recorded in the dataset. Shock index and mean arterial pressure calculation The measurement of both shock index and MAP values was performed in accordance with previous studies. The shock index was calculated using the heart rate/systolic blood pressure formula, while the MAP value was identified by the formula [21]. MAP=DP+1/3(SP-DP) [22]. PETCO2 measurement PETCO2 measurement was performed with a sidestream capnography device (GE Medical Systems, USA), and the PETCO2 level at the end of 2nd min was recorded in the dataset [23]. Statistical analysis The obtained data and information were evaluated for statistical analysis using the IBM SPSS 21.0 (Statistical Package for the Social Sciences) (SPSS Inc. Chicago, IL, USA) package data program. As clinical investigations with similar design focusing on the prolonged use of N95 face masks in ED were not available in the literature, a power analysis was run to obtain a hypothetical effect size. Assuming a hypothetical effect size of 0.5 at the standard 0.05 alpha error probability, a sample size of at least 54 people needed to be enrolled in the study to achieve 95% power. The normality of the original data was checked by the Kolmogrov-Smirnov test. The dependent variables with parametric distribution were expressed as mean ± standard deviation and analyzed by the paired t-test. For evaluation of variations between measurements, repeated measure anova (with Bonferroni correction) and Greenhouse Geisser tests were performed. For evaluation of variation between complaints of attendees and whether they take a break, chi square test was performed. Since the effect of the suitability of the breaks on vital parameters is important in the study, the effect power of the study was also evaluated by measuring the PETCO2 levels in the people who took a break in the first 4 h and those who did not. A p value of < 0.05 was set as the limit for statistical significance. Results Baseline data and break times 28 (51.9%) males and 26 (48.1%) females were enrolled in the study, and the average age of these subjects was 25.1 ± 3.48 years. 11 (20.4%) of the subjects were smokers. The participants were followed for 9th hours in their shifts. The average break time between 0–4 h and 5–9 h turned out to be 10.83 ± 8.5 min and 53.33 ± 27.47 min, respectively. 15 (27.7%) subjects continued working uninterruptedly between 0 and 4 h, whereas nobody preferred to work without a break between 5 and 9 h (Table 1).Table 1 Baseline characteristics of the study population Gender Male, n (%) 28 (51.9%) Female, n (%) 26 (48.1%) Age, year 25.1 ± 3.48 Smokers, n (%) 11 (20.4%) Break time (min) 0–4th hours 10.83 ± 8.5 5–9th hours 53.33 ± 27.47 Nonbreakers n ( %) 0–4th hours 15 (27.8%) 5–9th hours 0 BP blood pressure, PETCO2 partial end-tidal CO2 pressure, MAP mean arterial pressure Considering the 0–4th hour PETCO2 levels of those working with no break between 0 and 4 h, the effect size turned out to be high (f = 0.95) in the post-hoc power analysis, and 99.9% power was reached at the 95% confidence level. Regarding the dynamic changes of vital parameters in the 0–4 and 5–9 h, only the decrease in diastolic BP between 0 and 9 h was statistically significant (p = 0.038). Besides, MAP values indicated a significant decrease between 0–9 h and 5–9 h (p = 0.024 and p = 0.049, respectively) (Table 2).Table 2 Vital parameter measurements of the study population 0th hour 4th hours 5th hours 9th hours p values Heart rate (beat/min) 88.75 ± 13.85 89.24 ± 11.8 87.81 ± 11.1 88.27 ± 9.93 0.729* 0.772 0.697* 0.912 Body temperature (°C) 36.4 ± 0.16 36.42 ± 0.16 36.46 ± 0.18 36.4 ± 0.16 0.422* 1 0.104* 0.308 Systolic BP (mm/Hg) 129.09 ± 12.8 128.62 ± 12.64 128.07 ± 11.44 126.31 ± 12.69 0.737* 0.069 0.272* 0.512**** Diastolic BP (mm/Hg) 73.55 ± 11.1 72.59 ± 8.98 3.22 ± 9.57 70.35 ± 9.1 0.516* 0.038 0.056* 0.673**** sPO2 97.16 ± 1.29 97.37 ± 1.15 97.27 ± 1.12 97.42 ± 1.46 0.296* 0.284 0.459* 0.933**** PETCO2 (mm/Hg) 35.4 ± 4.35 35.81 ± 3.09 35.85 ± 2.98 35.61 ± 3.51 0.492* 0.721 0.513* 0.931**** Shock index 0.69 ± 0.13 0.7 ± 0.11 0.69 ± 0.11 0.7 ± 0.09 0.704* 0.589 0.369* 0.714**** MAP 92.06 ± 10.01 91.27 ± 8.6 91.5 ± 8.83 89 ± 9.18 0.488* 0.024 0.049* 0.483**** BP blood pressure, PETCO2 partial end-tidal CO2 p values are derived from paired sample t test and it refers to comparison between first measurement and second measurement; p values are derived from paired sample t test and it refers to comparison between first measurement and forth measurement; p values are derived from paired sample t test and it refers to comparison between third measurement and forth measurement; *p values are derived from repeated measure anova test (with Bonferroni correction) and Greenhouse Geisser test 15 (27.7%) individuals in the study group, all of whom were nonsmokers, continued performing their duties without a pause between 0 and 4 h. While baseline PETCO2 level was measured as 35.13 ± 2.64 mmHg in those working without breaks, this level increased to 36.66 ± 3.33 mmHg at the 4th hour. When it comes to those working by taking breaks, their baseline PETCO2 level, which was 34.92 ± 4.63 mmHg, rose to 36.07 ± 3.24 mmHg at the end of the 4th hour. In terms of the vital parameters of the subjects working with and without breaks, only PETCO2 levels of those working uninterruptedly increased significantly at the 4th hour in comparison to the beginning-of-shift baseline levels (p = 0.003) (Table 3). When the relation between state of taking break and PETCO2 levels was evaluated by measure anova (with Bonferroni correction) and greenhouse tests, it was observed that taking a break was effective in the measurements between 0 and 4th hours at PETCO2 level (p = 0.04).Table 3 Vital parameter measurements of breaker and nonbreaker subgroups 0th hour 4th hours p values Nonbreakers (N = 15) Heart rate (beat/min) 83.46 ± 10.57 81.8 ± 9.87 0.429 Fever (°C) 36.38 ± 0.18 36.43 ± 0.15 0.496* Systolic BP (mm/Hg) 129.13 ± 13.92 127.13 ± 13.75 0.275* Diastolic BP (mm/Hg) 72.46 ± 10.04 71.73 ± 8.72 0.794* sPO2 97.6 ± 1.21 97.4 ± 1.21 0.567* PETCO2 (mm/Hg) 35.13 ± 2.64 36.66 ± 3.33 0.003* 0.04** Shock index 0.65 ± 0.11 0.65 ± 0.12 0.876* MAP 91.35 ± 9.59 90.2 ± 8.82 0.978* Beakers (N = 39) Heart rate (beat/min) 90.79 ± 14.52 92.1 ± 11.31 0.457* Fever (°C) 36.41 ± 0.16 36.42 ± 0.17 0.628* Systolic BP (mm/Hg) 129.07 ± 12.53 129.2 ± 12.33 0.943* Diastolic BP (mm/Hg) 73.97 ± 11.58 72.92 ± 9.17 0.555* sPO2 97 ± 1.33 97.35 ± 1.18 0.128* PETCO2 (mm/Hg) 34.92 ± 4.63 36.07 ± 3.24 0.141* Shock index 0.71 ± 0.14 0.71 ± 0.1 0.676* MAP 92.34 ± 10.27 91.68 ± 8.59 0.641* BP blood pressure, PETCO2 partial end-tidal CO2, MAP mean arterial pressure p values are derived from paired samples t test, and it refers to comparison the parameters between 0th hour and 4th hours; p value is derived from repeated measure anova test (with Bonferroni correction) and Greenhouse Geisser test With respect to the subjective mask-driven complaints of those working with and without a pause between 0 and 4 h, 11 subjects (73.3%) suffered from shortness of breath, 11 (73.3%) individuals reported increased fatigue, 10 (66.7%) complained of headaches, and 15 (100%) came down with skin-bound complications, including persistent erythema and mask-induced scarring. The incidence of these perceived complaints among the subjects working with breaks remained significantly lower than their counterparts working without breaks (p = 0.005; p = 0.0001; p = 0.029, and p = 0.002, respectively) (Table 4).Table 4 Complaints of study population Complaints Nonbreakers (0–4th hours) (N = 15) Breakers (0–4th hours) (N = 39) p values Shortness of breath Yes 11 (73.3%) 11 (28.2%) 0.005 No 4 (26.7%) 28 (71.8%) Quick fatigue Yes 11 (73.3%) 4 (10.3%) 0.0001* No 4 (26.7%) 35 (89.7%) Headache Yes 10 (66.7%) 12 (30.8%) 0.029* No 5 (33.3%) 27 (69.2%) Skin problems Yes 15 (100%) 23 (59%) 0.002* No 0 16 (41%) Break time ≤ 30 min between 5 and 9th hours (N = 19) Break time > 30 min between 5 and 9th hours (N = 35) Shortness of breath Yes 12 (63.2%) 24 (58.6%) 0.043a No 7 (36.8%) 11 (31.4%) Quick fatigue Yes 11 (57.9%) 8 (22.9%) 0.017a No 8 (42.1%) 27 (77.1%) Headache Yes 16 (84.2%) 14 (40%) 0.004* No 3 (15.8%) 21 (60%) Skin problems Yes 19 (100%) 13 (37.1%) 0.0001* No 0 22 (62.9%) *p values are derived from Fisher exact test ap values are derived from chi square test The subjects, all of whom worked with a break between 5 and 9 h, were divided into two subgroups as those taking breaks of ≤ 30 min and > 30 min. Shortness of breath (n = 12, 63.2%), increased fatigue (n = 11, 57.9%), headaches (n = 16, 84.2%), and skin-bound complications (n = 19, 100%) accounted for the complaints of the subgroup with a break of ≤ 30 min. The incidence of these complaints in the > 30 min subgroup was significantly lower than that of the ≤ 30 min subgroup (p = 0.043, p = 0.017, p = 0.04, and p = 0.0001, respectively) (Table 4). Discussion This study reveals both the changes in vital signs of healthcare providers wearing N95 facemasks in the COVID-19 pandemic zones of ED and the subjective mask-induced complaints of these individuals. Accordingly, their diastolic BP and MAP values overall manifested a marked decrease between 0 and 9 h during the study, and the PETCO2 levels of those not taking a break in the first 4 h were observed to remain higher those working with a break. In addition, those taking breaks while performing their duties were less likely to suffer from shortness of breath, increased fatigue, headache, and skin-bound problems than those working uninterruptedly. Furthermore, the same complaints were expressed less frequently by the individuals with a break time of > 30 min than the ones taking shorter breaks. Cardiac output and peripheral vascular resistance are the major determinants of diastolic BP (DBP). During some exercise, such as running, cycling, and swimming, cardiac output increases in response to vasodilation of arterioles in exercising skeletal muscles, while peripheral vascular resistance decreases, thereby reducing DBP to some extent [24]. Moreover, Sainas et al. argued that MAP values tended to decrease following intense physical exertion [25]. Shahraki et al. likewise documented a decrease in DBP and MAP values of their subjects after 5 min of exercise [26]. Based on the measurements of vital signs and pCO2 levels of healthcare workers using N95 facemask after 1 h of treadmill exercise, Roberge et al. noted no significant difference in the physiological parameters between those wearing filtered and unfiltered masks [27]. Another trial in which surgical masks and N95 facemasks were tested on a total of 10 people found some clinical evidence for the impact of mask wearing on thermal stress and increased heart rate [28]. As identified by previous reports, while the overall patient population in ED showed a downward trend during the pandemic process, the burden of emergency services shifted dramatically to pandemic zones [29]. In the healthcare facility where the study data was collected, there was heavier workload and faster work pace during the period from the beginning of the morning shift to the noon time than in the afternoon period. On the other hand, our findings suggested that lower diastolic pressure and MAP values at the end of the shift than baseline measures might be attributed to the decreased peripheral vascular resistance and DBP as the individuals exerted physical effort. Though the systolic pressure and DBP did not drop significantly between 5 and 9 h, a slight decrease of both led to a significant reduction in the MAP values during this period. We also reckon that during the time interval of 5–9 h when all the healthcare providers took a break and relatively few patients visited the ED, the systolic and diastolic pressure levels did not manifest a decrease since they had the opportunity to rest longer. PETCO2, which refers to partial pressure of CO2 in the air exhaled during expiration, is one of the parameters considered in management of intubated patients in ED. Under normal circumstances, the value measured by the capnography at the end of expiration ranges between 35 and 45 mmHg. PETCO2, which is closely associated with pCO2 level in arterial blood, is also known to provide an indication for pCO2 levels [30]. In addition, previous literature reports established a direct correlation between PETCO2 levels and the decrease in cardiac output [31, 32]. However, physiological parameters were reported not to signify a marked change between wearing a mask with and without filter after 1 h of treadmill exercise amongst healthcare providers using N95 facemask [27]. A study conducted on children noted no marked difference in their PETCO2 levels measured while wearing N95 mask both at rest and during light exercise. Furthermore, hypoventilation developing after a prolonged use of PPE is considered to present major health concerns [33]. For instance, hypoventilation induced by long-term use of respiratory protection is likely to elevate PETCO2 levels [34]. Therefore, hypoventilation may account for increased PETCO2 levels relative to baseline in the healthcare providers working without rest in the first 4 h in our study. This assumption can be validated by the lack of change in PETCO2 levels among those taking breaks and between 5 and 9 h when everyone took a break. Though working without resting sounds inhumane, it may not always be possible to give a break during hectic hours in ED over the course of the pandemic. The healthcare providers who work at pandemic polyclinics should be employed in appropriate shifts, with appropriate breaks. PPE, especially protective masks, which has been re-introduced to the working life of medical staff through COVID-19 outbreak, affect working comfort, though they offer protection to healthcare workers against viral transmission. A substantial body of research in the literature draws attention to device-related discomfort of users while wearing N95 masks. For example, mask-induced complaints, such as shortness of breath, headache, and light-headedness, were reported to increase gradually among nurses wearing only an N95 or a surgical mask overlay with an N95 [35]. Besides, PPE-associated headache developed in 81% of healthcare providers based at pandemic outpatient clinics [36]. Another study likewise revealed that long-time wearing of N95 respirators was closely associated with headache complaints [37]. A recent study established that relatively long exposure time to N95 respirators (more than 6 h) also doubled (95%CI 1.35–3.01, and p < 0.01) the risk of developing skin damage among healthcare providers in addition to headache [38]. There was also some clinical evidence that healthcare staff reported increased fatigue and chest compression quality suffered when they performed cardiopulmonary resuscitation on manikin with an N95 respirator [39]. In line with the literature, complaints such as shortness of breath, increased fatigue, headache, and skin-related complications were expressed more frequently by those working without a break in the first 4-h period as well as those completing their shift with less than 30-min break. Even though these high rates declined substantially after giving a break, the fact that a total of 32 (59.25%) individuals reported persistent skin-related complaints is another aspect deserving attention in our study. We predict that the production materials of N95 respirators may be an underlying reason for such skin damage. The primary limitation of our study was the absence of arterial blood gas analysis of the subjects because we did not explore the effect of changes in vital parameters upon blood gas values. Another limitation was the lack of relevant information on their baseline effort capacity. Although this seems to have posed a drawback for between-group analyses, these subjects were assumed to be above a certain effort capacity due to practicing in the same facility and at a similar pace for a long time. Since their vital signs during the day were taken into consideration, the effect of absence of required information on their effort capacity must have been relatively minor. Conclusion Even though the wearing of N95 facemasks may cause physical discomfort, such as headache, shortness of breath, and increased fatigue among ED healthcare providers, such discomfort might not impose stress on vital signs if appropriate rest breaks are taken. However, lack of a significant effect of wearing respiratory protection on vital parameters does not necessarily entail the ignorance of comfort-related complaints. Hospital management and local authorities as well as policy makers should consider that mask-induced problems might impair the physical performance of healthcare providers who afflicted with various complications. Accordingly, appropriate rest periods should be provided to frontline health workers. Although the decrease in SBP and MAP values is assumed to be caused by increased fatigue due to workload and work pace as well as increase in muscle activity, the increase in PETCO2 levels in the ED healthcare staff working with no breaks between 0 and 4 h should be noted in terms of PPE-induced hypoventilation. Data availability All the data (other than patient names) are available to share. Declarations Conflict of interest The authors declare no competing interests. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. CDC (2019) Novel Coronavirus, Wuhan, China. CDC. Available at https://stacks.cdc.gov/view/cdc/84643/cdc_84643_DS1.pdf?. Accessed 27 Jan 2021 2. Gallegos A (2020) WHO declares public health emergency for novel coronavirus. Meds Med News. Available at: https://www.medscape.com/viewarticle/924596. Accessed 31 Jan 2021 3. Cascella M, Rajnik M, Cuomo A et al (2020) Features, evaluation, and treatment of coronavirus (COVID-19) [Updated 2020 Aug 10]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2020 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK554776/. Accessed 31 Jan 2021 4. Ortega R, Gonzalez M, Nozari A, Canelli R (2020) Personal protective equipment and Covid-19. N Engl J Med 25;382(26):e105. 10.1056/NEJMvcm2014809 5. Using PPE (2020). Summary of recent changes. Available from https://www.cdc.gov/coronavirus/2019-ncov/hcp/using-ppe.html. Accessed 24 Jan 2021 6. Advice on the use of masks in the context of COVID-19 Interim guidance WHO (2020) Accessed at: https://apps.who.int/iris/handle/10665/332293. Accessed 24 Jan 2021 7. Matuschek C Moll F Fangerau H Face masks: benefits and risks during the COVID-19 crisis Eur J Med Res 2020 25 1 32 10.1186/s40001-020-00430-5 32787926 8. Lepelletier D Grandbastien B Romano-Bertrand S What face mask for what use in the context of COVID-19 pandemic? The French guidelines J Hosp Infect 2020 105 3 414 418 10.1016/j.jhin.2020.04.036 32348833 9. Approven particulate filtering facepiece respirators (2021). Available from https://www.cdc.gov/niosh/npptl/topics/respirators/disp_part/default.html#:~:text=N95%20%E2%80%93%20Filters%20at%20least%2095,Not%20resistant%20to%20oil.&text=Surgical%20N95%20%E2%80%93%20. Accessed 24 Jan 2021 10. N95 respirators, surgical masks, and face masks (2021). Avaliable from: https://www.fda.gov/medical-devices/personal-protective-equipment-infectioncontrol/n95-respirators-surgical-masks-and-face-masks. Accessed 24 Jan 2021 11. Roberge RJ Are exhalation valves on N95 filtering facepiece respirators beneficial at low-moderate work rates: an overview J Occup Environ Hyg 2012 9 617 623 10.1080/15459624.2012.715066 22978255 12. Li Y Guo YP Wong KC Transmission of communicable respiratory infections and facemasks J Multidiscip Healthc 2008 1 17 27 10.2147/jmdh.s3019 21197329 13. Mitze T Kosfeld R Rode J Face masks considerably reduce COVID-19 cases in Germany Proc Natl Acad Sci U S A 2020 117 32293 32301 10.1073/pnas.2015954117 33273115 14. The Use of N95/FFP2 Mask. Turkish Ministery of Health COVID-19 brochures (2021) Available from https://covid19.saglik.gov.tr/Eklenti/37648/0/covid-19n95ffp2maskelerininkullanimi41x223kirimlibrosurpdf.pdf_tag1=E3799EB10E07CEC673D89257F0C3E63B512803D1.) Accessed 24 Jan 2021 15. Kao TW Huang KC Huang YL The physiological impact of wearing an N95 mask during hemodialysis as a precaution against SARS in patients with end-stage renal disease J Formos Med Assoc 2004 103 624 628 15340662 16. Fikenzer S Uhe T Lavall D (2020) Effects of surgical and FFP2/N95 face masks on cardiopulmonary exercise capacity Clin Res Cardiol 2020 109 12 1522 1530 10.1007/s00392-020-01704-y 32632523 17. Rosner E Adverse effects of prolonged mask use among healthcare professionals during COVID-19 J Infect Dis Epidemiol 2020 6 130 10.23937/2474-3658/1510130 18. COVID-19 table of PPE with description and related standard (simplified version) (2020). Available from: https://www.who.int/publications/m/item/from-DCP-v5-list-PPE-v8082020. Accessed 24 Jan 2021 19. Ataş Berksoy E, Bağ Ö, Yazici S et al (2018) Use of noncontact infrared thermography to measure temperature in children in a triage room. Medicine (Baltimore) 97(5):e9737. 10.1097/MD.0000000000009737 20. Muntner P Shimbo D Carey RM Measurement of blood pressure in humans: a scientific statement from the American Heart Association Hypertension 2019 73 e35 e66 10.1161/HYP.0000000000000087 30827125 21. Koch E Lovett S Nghiem T Shock index in the emergency department: utility and limitations Open Access Emerg Med 2019 11 179 199 10.2147/OAEM.S178358 31616192 22. DeMers D, Wachs D (2020) Physiology, mean arterial pressure. Accessed at:https://www.ncbi.nlm.nih.gov/books/NBK538226/. Accessed 25 Jan 2021 23. Pekdemir M Cinar O Yilmaz S Disparity between mainstream and sidestream end-tidal carbon dioxide values and arterial carbon dioxide levels Respir Care 2013 58 1152 1156 10.4187/respcare.02227 23322889 24. Farpour-Lambert NJ Aggoun Y Marchand LM (Physical activity reduces systemic blood pressure and improves early markers of atherosclerosis in pre-pubertal obese children J Am Coll Cardiol 2009 54 2396 2406 10.1016/j.jacc.2009.08.030 20082930 25. Sainas G Milia R Palazzolo G Mean blood pressure assessment during post-exercise: result from two different methods of calculation J Sports Sci Med 2016 15 424 433 27803621 26. Shahraki MR, Mirshekari H, Shahraki AR et al (2012) Arterial blood pressure in female students before, during and after exercise. ARYA Atheroscler 8(1):12–15. PMID:23056094 27. Roberge RJ Coca A Williams WJ Physiological impact of the N95 filtering facepiece respirator on healthcare workers Respir Care 2010 55 569 577 20420727 28. Li Y Tokura H Guo YP Effects of wearing N95 and surgical facemasks on heart rate, thermal stress and subjective sensations Int Arch Occup Environ Health 2005 78 501 509 10.1007/s00420-004-0584-4 15918037 29. Hartnett KP, Kite-Powell A, DeVies J et al (2020) Impact of the Covid-19 pandemic on emergency department visits—United States. Morbidity and Mortality Weekly Report (MMWR) 69(23):699–704. Accessed at: https://www.cdc.gov/mmwr/volumes/69/wr/mm6923e1.htm https://www.cdc.gov/mmwr/volumes/69/wr/mm6923e1.htm. (Accessed: 25 Jan 2021) 30. Pishbin E Ahmadi GD Sharifi MD The correlation between end-tidal carbon dioxide and arterial blood gas 2015 7 1095 1101 10.14661/2015.1095-1101 31. Shibutani K Muraoka M Shirasaki S Do changes in end-tidal PCO2 quantitatively reflect changes in cardiac output? Anesth Analg 1994 79 829 833 10.1213/00000539-199411000-00002 7978395 32. Senopathi TGA Wiryana M Sinardja K The End-Tidal CO2 correlation with a decreased cardiac output measured by ultrasonic cardiac output monitor in intubated ICU patients Bali Medical Journal 2017 6 12 16 10.15562/bmj.v6i1.372 33. Williams WJ Physiological response to alterations in [O2] and [CO2]: relevance to respiratory protective devices J Intl Soc Resp Protect 2010 27 27 51 10.1016/j.ajic.2013.02.017 34. https://www.physiocontrol.com/uploadedFiles/Physio85/Contents/Trade_Shows/Capno%20handout.pdf. Accessed 24 Jan 2021 35. Rebmann T Carrico R Wang J Physiologic and other effects and compliance with long-term respirator use among medical intensive care unit nurses Am J Infect Control 2013 41 1218 1223 10.1016/j.ajic.2013.02.017 23768438 36. Ong JJY Bharatendu C Goh Y Headaches associated with personal protective equipment — a cross-sectional study among frontline healthcare workers during COVID-19 Headache 2020 60 5 864 877 10.1111/head.13811 32232837 37. Lim EC Seet RC Lee KH Wilder-Smith EP Chuah BY Ong BK Headaches and the N95 face-mask amongst healthcare providers Acta Neurol Scand 2006 113 3 199 202 10.1111/j.1600-0404.2005.00560.x 16441251 38. Lan J Song Z Miao X Skin damage among health care workers managing coronavirus disease-2019 J Am Acad Dermatol 2020 82 5 1215 1216 10.1016/j.jaad.2020.03.014 32171808 39. Tian Y Tu X Zhou X Wearing a N95 mask increases rescuer’s fatigue and decreases chest compression quality in simulated cardiopulmonary resuscitation Am J Emerg Med 2021 44 434 438 10.1016/j.ajem.2020.05.065 33046304
PMC009xxxxxx/PMC9005164.txt	==== Front Ann Biomed Eng Ann Biomed Eng Annals of Biomedical Engineering 0090-6964 1573-9686 Springer International Publishing Cham 2958 10.1007/s10439-022-02958-5 Original Article Detection of COVID-19 from CT and Chest X-ray Images Using Deep Learning Models http://orcid.org/0000-0003-1047-1968 Zouch Wassim wzouch@kau.edu.sa 1 Sagga Dhouha 23 Echtioui Amira 2 Khemakhem Rafik 23 Ghorbel Mohamed 2 Mhiri Chokri 45 Hamida Ahmed Ben 2 1 grid.412125.1 0000 0001 0619 1117 King Abdulaziz University (KAU), Jeddah, Saudi Arabia 2 grid.412124.0 0000 0001 2323 5644 ATMS Lab, Advanced Technologies for Medicine and Signals, ENIS, Sfax University, Sfax, Tunisia 3 grid.442508.f 0000 0000 9443 8935 Higher Institute of Management of Gabes, Gabes University, Gabès, Tunisia 4 grid.413497.c Department of Neurology, Habib Bourguiba University Hospital, Sfax, Tunisia 5 grid.412124.0 0000 0001 2323 5644 Neuroscience Laboratory “LR-12-SP-19”, Faculty of Medicine, Sfax University, Sfax, Tunisia Associate Editor Stefan M. Duma oversaw the review of this article. 12 4 2022 111 2 8 2021 21 3 2022 © The Author(s) under exclusive licence to Biomedical Engineering Society 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Coronavirus 2019 (COVID-19) is a highly transmissible and pathogenic virus caused by severe respiratory syndrome coronavirus 2 (SARS-CoV-2), which first appeared in Wuhan, China, and has since spread in the whole world. This pathology has caused a major health crisis in the world. However, the early detection of this anomaly is a key task to minimize their spread. Artificial intelligence is one of the approaches commonly used by researchers to discover the problems it causes and provide solutions. These estimates would help enable health systems to take the necessary steps to diagnose and track cases of COVID. In this review, we intend to offer a novel method of automatic detection of COVID-19 using tomographic images (CT) and radiographic images (Chest X-ray). In order to improve the performance of the detection system for this outbreak, we used two deep learning models: the VGG and ResNet. The results of the experiments show that our proposed models achieved the best accuracy of 99.35 and 96.77% respectively for VGG19 and ResNet50 with all the chest X-ray images. Keywords COVID-19 Deep learning Chest X-ray CT Convolutional neural network http://dx.doi.org/10.13039/501100004054 King Abdulaziz University D-473-135-1435 Zouch Wassim ==== Body pmcIntroduction COVID-19 is a global pandemic caused by the (SARS-CoV-2). The epidemic first occurred in Wuhan, China, in early December 2019.27 The rapid outbreak of the disease in China, and the number of deaths and injuries rising sharply, is not only a local crisis but rather a crisis that worries the whole world. According to the latest statistics from the World Health Organization more than 185 million cases of COVID-19 have been reported in more than 188 countries and territories as of July 9, 2021, including more than 4010 million deaths and more than 1 million healings.25 Coronavirus infection can cause common symptoms such as sore throat, headache, cough with loss of smell, fever, and taste, and severe symptoms such as difficulty breathing, chills, feeling tired or weak, body aches.23 These symptoms differ from person to person depending on the strength of the immune system, the mutation of the virus, the possibility of reinfection, and the possible long-term health effects. Several means contribute to the spread of the epidemic, including aerosols, contaminants, and droplets.13 A rapid change in the spread of virus-laden bioaerosols through the respiratory tract to other different regions of the lungs cases a sudden deterioration few days after the infection.6 The lower airways of critically ill COVID-19 pneumonia patients are filled with a lot of exudate or very viscous mucus, resulting in airway obstruction. Estimating airway opening pressures and efficient mucus removal are therefore two issues of greatest concern to clinicians during mechanical ventilation. In Ref. [5], the authors retrospectively analyzed respiratory data from 24 patients with COVID-19 who received invasive mechanical ventilation. The results obtained show that the suction pressure could exceed 20 kPa and that the expected airway opening pressures could reach 40 to 50 cm H2O as the viscosity of the secretion simulators and the surface tension increased considerably, probably causing the closure of the distal airways. Clinicians need innovative ways to safely perform airway interventions on patients with COVID-19 with limited availability of personal protective equipment (PPE). In Ref. [18], the authors have described a new particle containment chamber for patients. The essential criteria of this chamber are: construction from inexpensive materials readily available, reduction in aerosol transmission of at least 90% measured by pragmatic tests, easy to clean and compatibility with common EMS stretchers. So, the more cases of COVID there are in a country, the more likely it is that new strains will emerge, as each infection alone gives the virus a chance to evolve, and the big concern is that new mutations could make them ineffective to vaccines. So coordinated efforts on a global scale are needed to prevent the spread of the virus. As a result, rapid and accurate identification of COVID-19 can save lives by reducing the spread of the disease and generating information for artificial intelligence (AI) models.15 In this regard, AI can make a valuable and useful contribution, especially in image-based medical diagnosis. Recently several researchers have used deep learning (DL) methods to solve health issues such as the prediction of musculoskeletal strength20 and the treatment of eye diseases.4 Based on the most recent assessment of using the AI to COVID-19 by researchers at the UN Global Pulse,11 the survey shows that, compared to traditional exams, AI is inherently as precise as humans and can save the radiologist time and effort. As a result, the diagnosis can be made more quickly and inexpensively.3 Radiography and computed tomography can be used.24 The use of AI and DL can aid in the discovery and identification of COVID-19. The main challenge is to provide access to specific and reasonably priced diagnostics for the diagnosis of COVID-19. The use of convolutional neural networks (CNNs) for extraction and learning has proven to be very useful, and researchers have adopted them widely.16 The objective of this study is to propose a method for the classification of COVID-19 and non-COVID-19 cases. We used two well-known CNN, VGG19 and ResNet50, based on the collected data. So far, only a small number of X-ray images for COVID-19 have been made available to the public. Therefore, we will not be able to train these models without making data augmentation. The main research contributions in this work may be summarized as follows:We proposed a method to classify radiographic and tomographic images of the lungs of patients as COVID-19 or not COVID-19. We applied the data augmentation to create a version of the converted COVID-19 image to enlarge the sample set. We have fine-tuned the last layer of our proposed models for using fewer category-labeled samples for training. The novelty is the development of an augmented dataset, which uses three augmentation strategies: random rotation, translation (width shift and height shift), and horizontal flip. Then, we have combined the last layer of two powerful algorithms that have already been used to enable efficient virus detection from images. It should be noted that VGGNet is a powerful and adaptable architecture for performing benchmarking on a specific task. Third, we have completed all modeling and training steps so that we can effectively detect the virus in CT and chest X-ray images. Specifically, our technique is designed to be used as a highly reliable tool to aid in clinical decision-making. The rest of the article is organized as follows. After the introduction, Sect. 2 presents related work. In Sect. 3, we explain how the dataset is put together and describe the proposed models. In Sect. 4, we present the experimental results and performances. Finally, a discussion is proposed in Sect. 5. Related Work Several existing studies focus on DL algorithms to learn the characteristics of radiographic and tomographic images of patients so that the model can detect pneumonia with greater precision. For example, in Ref. [9] the authors used the CNN to develop a predictive model to distinguish COVID-19 and influenza A, with a maximum precision of 86.7%. In Ref. [17], the authors developed an AI-based CT analysis tool for the detection and quantification of COVID-19. This system automatically extracts slices of opacities in the lungs to provide a quantitative measurement and a 3D representation of opacity volume. The developed system achieved a specificity of 92.2% and a sensitivity of 98.2%. In Ref. [21], the authors have developed a DL method based on support vector machine (SVM), using radiographic images. The deep features of the fully connected (FC) layers of the CNN have been extracted and integrated into SVM for classification. The highest level of accuracy achieved by SVM is 89.66%. In Ref. [12] the authors proposed to evaluate the performance of the most recent CNN dedicated architectures or classifications of medical images (MobileNet, Inception, VGG19, Xception, etc.) for the automatic detection of coronaviruses. To achieve these goals, the authors used transfer learning, which performs well in detecting various anomalies in small data sets. The results show that VGG19 and MobileNet-v2 are the best classifiers among the remaining CNNs, with an accuracy of 98.75 and 98.66% for VGG19 and MobileNet, respectively. In Ref. 4], the authors studied two datasets (COVIDx and COVIDNet) in order to detect COVID-19 using chest X-ray images. These datasets contain four classes of chest X-ray images broken down as follows: X-rays of uninfected cases, bacterial X-rays, viral X-rays for COVID-19, and non-positive X-rays for COVID-19 pneumonia. They obtained an overall accuracy of 83.5%. In Ref. [5], the authors developed a VB-net system to automatically segment all lungs and detect infection using a chest scanner. This process is time-consuming, so in order to speed it up, the authors proposed a strategy called “in the Loop (HITL)” whose goal is to generate a formation of samples in an iterative way. After three iterations of updating the model, the manual loop proposed a strategy that reduces the drawing time to 4 min. This method achieves an accuracy of 91.6%. In this work proposed a new method for COVID-19 detection based on CT and Chest X-ray images. Method In order to identify chest X-ray and CT images as normal or affected by COVID-19, we planned to use two DL models: ResNet50 and VGG19. The following diagram (Fig. 1) illustrates the overall design of the proposed system:Figure 1 Flow diagram of the proposed method. Dataset Since COVID-19 attacks the epithelial cells that line our airways, we will use X-ray image of the chest to analyze the health of a patient's lungs. In this work, we used two databases:Database of CT images: this database includes 408 NonCOVID-19 CT images and 349 COVID-19 CT images. These images were obtained from the open-source GitHub repository, shared by Dr. Jkooy.10 Database of chest X-ray images: this database includes 747 NonCOVID-19 chest X-ray images and 112 COVID-19 chest X-ray images. These images were obtained from the open-source GitHub in Ref. [11. Data Preprocessing The sizes of the input images for the two databases were different, then we changed the image sizes to 224 × 224 pixels. Data Augmentation DL models need data to be trained. The more data we have, the more performance the model gains since it will be able to capture more behaviors in the learning process. Since COVID-19 is still a relatively new disease, there is currently no publicly available dataset. In order to generate larger dataset, we used data augmentation techniques with three different augmentation strategies: random rotation with an angle ranging from -20 to 20 degrees, random noise, and horizontal flip. Detection of COVID-19 by Using VGG19 The design of the proposed VGG19 is illustrated in Fig. 2. VGG model was proposed by Karen Simonyan and Andrew Zisserman in their article “Very Deep Convolutional Networks for Large Scale Image Recognition”22 in 2014. VGG is considered to be very deep for image recognition. It has a basic and coherent system of 3 * 3 convolutional layers stacked on top of each other and with pooling applied between layers. In this article VGG consists of five blocks, the first two blocks contain two convolutional layers while the last three contain four convolutional layers commonality applied between the blocks. These blocks are followed by flatten, dropout, and dense layers.Figure 2 Proposed VGG19 architecture. Detection of COVID-19 by using ResNet50 Figure 3 shows the proposed ResNet50. This model is a very specific neural network introduced in 2015 by Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun in their publication “Residual Learning for Image Recognition”.14 The size of the model is actually much smaller due to the use of a global clustering method rather than a connected layer, which reduces the size of the ResNet model. The unique feature of ResNet is the recurring learning block. This means that each layer has to feed into the next layer, and the distance jumps directly into the layers of about 2 to 3 hops. This model consists of five blocks of convolution layer with max pooling and then a flatten, a dropout and finally a single FC layer.Figure 3 Proposed ResNet50 architecture. Performance Criteria We rated the model's performance using a variety of metrics including accuracy, recall, F1 score, and precision. Metrics are evaluated by various parameters in the confusion matrix, such as true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Figure 4 presents the general definition of the confusion.Figure 4 The confusion matrix definition. Definition of the Terms Metrics are defined as follows: TP: refers to situations where the prediction is correct and the actual value is also correct. TN: refers to situations where the prediction is negative and the true value is also negative. FP: the prediction is positive but the actual value is negative. FN: the prediction is negative but the actual value is positive. The following formulas show how to use a performance benchmark to assess the performance of several pre-workout models.1 Accuracy=TP+TNTP+TN+FP+FN, 2 Precision=TPTP+FP, 3 Recall=TPFN+TP, 4 F1score=2∗precision∗recallprecision+recall. Results In this section, we provide the results of binary classification of CT images and chest X-ray images using the two architectures VGG19 and ResNet50. In order to assess the performance and robustness of each model, we have optimized these models over a number of epochs equal to 50, a batch size of 32 and a learning rate of 0.0001 with the Adam optimizer. The data set has been divided into two parts 80% for training and 20% for testing. Based on the CT images, the two proposed models reach an average accuracy of 82.89%, 73.68% without the increase while with the increase they mark an overall precision of 84.87% and 76.32% respectively for VGG19 and ResNet50. Based on the chest X-ray images, the two proposed models reach an average accuracy of 98.06%, 95.48% without the increase while with the increase they mark an overall precision of 99.35% and 96.77% respectively for VGG19 and ResNet50. In terms of four performance criteria, it should be noted that the affine version of VGG19 outperforms the affine version of ResNet50. As shown in the four Tables 1, 2, 3 and 4, in computed tomography or radiography, the modified version of the VGG19 model produces better results than ResNet50, either before or after augmentation. As a result, we can conclude that VGG19 is the best option.Table 1 Classification report for ResNet50 and VGG19 with data augmentation using CT images Classifier Patient status Precision (%) Recall (%) F1 score (%) Accuracy (%) VGG19 Normal 80 90 85 84.87 COVID-19 90 80 85 ResNet50 Normal 98 56 68 76.32 COVID-19 71 94 81 Table 2 Classification report for ResNet50 and VGG19 without data augmentation using CT images Classifier Patient status Precision (%) Recall (%) F1 score (%) Accuracy (%) VGG19 Normal 79 86 82 82.89 COVID-19 87 80 84 ResNet50 Normal 88 50 64 73.68 COVID-19 69 94 79 Table 3 Classification report for ResNet50 and VGG19 with data augmentation using chest X-ray images Classifier Patient status Precision (%) Recall (%) F1 score (%) Accuracy (%) VGG16 Normal 99 100 100 99.35 COVID-19 100 96 98 ResNet50 Normal 98 98 98 96.77 COVID-19 88 92 90 Table 4 Classification report for ResNet50 and VGG19 without data augmentation using chest X-ray images Classifier Patient status Precision (%) Recall (%) F1 score (%) Accuracy (%) VGG16 Normal 98 100 99 98.06 COVID-19 100 88 94 ResNet50 Normal 96 99 97 95.48 COVID-19 95 76 84 Figures 5 and 6 represent the confusion matrix associated with the two models.Figure 5 The confusion matrix of our proposed VGG19 model with augmentation (a, c) and without augmentation (b, d) using CT images (a, b) and chest X-ray images (c, d). Figure 6 The confusion matrix of our proposed ResNet50 model with augmentation (a, c) and without augmentation (b, d) using CT images (a, b) and chest X-ray images (c, d). Figures 7, 8, 9, and 10 present the training results of the proposed models. The results show that for VGG19, the training accuracy rate reached 99.35 and 88.87% with a loss of the training reduced to 0.1, for the CT and chest X-ray images respectively. While for the ResNet50 models, the training precision attained 96.77 and 76.32% with a training loss varying between 0.1 and 0.2, for CT and chest X-ray images respectively.Figure 7 (a) Model accuracy; and (b) model loss using VGG19 with CT images. Figure 8 (a) Model accuracy; and (b) model loss using ResNet50 with CT images. Figure 9 (a) Model accuracy; (b) model loss using VGG19 with chest X-ray images. Figure 10 (a) Model accuracy; (b) model loss using ResNet50 with chest X-ray images. Discussion To conduct a detailed examination of COVID-19 patients, we planned to build a COVID-19 automated identification system to guide doctors. Our results show that the COVID-19 classification has a high recall rate with a low the number of false negatives. This is a desirable because the main objective of this study is to reduce the false negative cases of COVID-19. It should be noted that the model we presented has a good accuracy of around 99%. In Table 5, we presented a summary of the results of the automatic COVID-19 diagnosis based on chest X-ray and chest images, as well as a comparison between our proposed model and some state of the art models. The authors of the following reference have proposed a new method of diagnosing COVID-19 using DL methods.Table 5 Summary of the research on automatic diagnosis of COVID-19 based on CT and chest X-ray images References Proposed model Accuracy (%) 26 Location Attention + ResNet 86.70 19 DarkCovidNet 87.02 Our proposed methods VGG19 99.35 ResNet50 96.77 The authors of Ref. [26] have developed a new method for automatic detection of COVID-19 using a DL method. Their proposed method using an attention localization mechanism can accurately classify COVID-19 on chest X-rays, with an accuracy rate of around 86.7%. The authors of the Ref. [19] propose the DarkCovidNet model to automatically detect and classify cases of COVID-19. It is based on a top-to-bottom model that does not use any characterization extraction method. This model is capable of performing tasks with an accuracy of 87.02%. However, the majority of previous studies provide insufficient data to develop the model. Our model is less demanding in terms of computational work than other previously published models, and it has given promising results. Once training data is available, performance can be further improved. Despite promising results, our proposed method still requires research and clinical testing; nonetheless, due to its increased accuracy in detecting COVID-19, our model can be able to further assist radiologists and healthcare specialists. The disruption of supply chains across the world, caused by COVID-19, has resulted in a severe shortage of PPE for clinicians. Which produces collaborations between the BME community and local hospitals to address critical PPE shortages during the COVID-19 Pandemic.7 Early detection of patients with COVID-19 is key to preventing the disease from spreading to others. In this study, we proposed a DL-based method to detect COVID-19 pneumonia using chest X-ray and CT images. Our proposed classification models for COVID-19 detection can achieve greater than 99% accuracy. Based on our research, because of its strong overall performance, we believe it is in its nature to help physicians and healthcare professionals make clinical decisions. This research has a deep understanding of how to use DL in order to discover COVID-19 as soon as possible. This work was funded by the Deanship of Scientific Research (DSR), King Abdulaziz University, Jeddah, under Grant No. (D-473-135-1435). The authors, therefore, acknowledge with thanks DSR technical and financial support. Conflict of interest The authors declared no potential conflict of interest statements with respect to the research, authorship, and/or publication of this article. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Cabeza-Gil I Ríos-Ruiz I Calvo B Customised selection of the haptic design in C-loop intraocular lenses based on deep learning Ann. Biomed. Eng. 2020 48 2988 3002 10.1007/s10439-020-02636-4 33037510 2. Chen Z Zhong M Jiang L Effects of the lower airway secretions on airway opening pressures and suction pressures in critically ill COVID-19 patients: a computational simulation Ann. Biomed. Eng. 2020 48 3003 3013 10.1007/s10439-020-02648-0 33078367 3. Cohen, J. P., M. Paul, and D. Lan. COVID-19 image data collection, 2020. arXiv preprint. arXiv:2003.11597. 4. Farooq, M., and A. Hafeez. COVID-ResNet: a deep learning framework for screening of COVID19 from radiographs. J. Comput. Sci. Eng. 2020. 5. Fei, S., G. Yaozong, W. Jun, S. Weiya, S. Nannan, H. Miaofei, X. Zhong, S. Dinggang, and S. Yuxin. Lung infection quantification of COVID-19 in CT images with DL, 2020. arXiv preprint. arXiv:2003.04655. 6. Filipovic N Saveljic I Hamada K Abrupt deterioration of COVID-19 patients and spreading of SARS CoV-2 virions in the lungs Ann. Biomed. Eng. 2020 48 2705 2706 10.1007/s10439-020-02676-w 33140243 7. George MP Maier LA Kasperbauer S How to leverage collaborations between the BME community and local hospitals to address critical personal protective equipment shortages during the COVID-19 pandemic Ann. Biomed. Eng. 2020 48 2281 2284 10.1007/s10439-02002580-3 32710248 8. Globalpulse. Need for Greater Cooperation Between Practitioners and the AI Community. https://www.unglobalpulse.org/2020/05/need-for-greater-cooperation-between-practitionersand-the-ai-community/. 9. Gozes, O., M. Frid-Adar, H. Greenspan, et al. Rapid AI Development Cycle for the Coronavirus (COVID-19) Pandemic: Initial Results for Automated Detection and Patient Monitoring Using Deep Learning CT Image Analysis, 2020. arXivpreprint. arXiv:2003.05037. 10. https://github.com/UCSD-AI4H/COVID-CT/tree/master/Images-processed. 11. https://github.com/ieee8023/covid-chestxray-dataset. 12. Ioannis DA Tzani AM COVID-19: automatic detection from X-ray images utilizing transfer learning with convolutional neural networks Phys. Eng. Sci. Med. 2020 43 635 640 10.1007/s13246-020-00865-4 32524445 13. Jayaweera M Perera H Gunawardana B Transmission of COVID-19 virus by droplets and aerosols: a critical review on the unresolved dichotomy J. Environ. Res. 2020 188 109819 10.1016/j.envres.2020.109819 14. Kaiming, H., Z. Xiangyu, R. Shaoqing, and S. Jian. Deep residual learning for image recognition, 2016. arXiv:1512.03385v1. 15. Kallianos K Mongan J Antani S Henry T Taylor A Abuya J Kohli M How far have we come? Artificial intelligence for chest radiograph interpretation Clin. Radiol. 2019 74 338 345 10.1016/j.crad.2018.12.015 30704666 16. Krizhevsky A Ilya S Geoffrey EH Customised selection of the haptic design in C-loop intraocular lenses based on deep learning Ann. Biomed. Eng. 2012 48 2988 3002 10.1007/s10439-020-02636-4 17. Lin L Lixin Q Zeguo X Artificial intelligence distinguishes COVID-19 from community-acquired pneumonia on chest CT J. Radiol. 2020 10.1148/radiol.2020200905 18. Maloney LM Yang AH Princi RA A COVID-19 airway management innovation with pragmatic efficacy evaluation: the patient particle containment chamber Ann. Biomed. Eng. 2020 48 2371 2376 10.1007/s10439-020-02599-6 32856180 19. Ozturk T Talo M Yildirim EA Automated detection of COVID-19 cases using deep neural networks with X-ray images Comput. Biol. Med. 2020 121 103792 10.1016/j.compbiomed.2020.103792 32568675 20. Rane L Ding Z McGregor AH Deep learning for musculoskeletal force prediction Ann. Biomed. Eng. 2019 47 778 789 10.1007/s10439-018-02190-0 30599054 21. Sethy, P. K., and S. K. Behera. Detection of Coronavirus Disease (COVID-19) Based on Deep Features. Preprints 2020, 2020. 22. Simonyan, K., and Z. Andrew. Very deep convolutional networks for large-scale image recognition. In: ICLR, 2015. 23. Soufi GJ Hekmatnia A Nasrollahzadeh M SARS-CoV-2 (COVID-19): new discoveries and current challenges J. Appl. Sci. 2020 10 3641 10.3390/app10103641 24. Tao A Zhenlu Y Hongyan H Chenao Z Chong C Wenzhi L Qian T Ziyong S Liming X Correlation of chest CT and RT-PCR testing for coronavirus disease, (COVID-19) in China: a report of 1014 cases Radiology 2019 10.1148/radiol.2020200642 25. World Health Organization. Statement on the Second Meeting of the International Health Regulations (2005) Emergency Committee Regarding the Outbreak of Novel Coronavirus (2019-ncov). https://www.who.int/news-room/detail/30-01-2020-statement-on-the-secondmeeting-of-the-international-health-regulations-(2005)-emergency-committee-regardingthe-outbreak-of-novel-coronavirus-(2019-ncov). 26. Xiaowei, X., J. Xiangao, M. Chunlian, et al. Deep learning system to screen coronavirus disease 2019 pneumonia, 2020. arXiv. https://arxiv.org/abs/2002.09334. 27. Xu X Jiang X Ma C Du P Li X Lv S A deep learning system to screen novel coronavirus disease 2019 pneumonia Engineering 2020 6 10 1122 1129 10.1016/j.eng.2020.04.010 32837749
PMC009xxxxxx/PMC9005215.txt	==== Front Advances in Oral and Maxillofacial Surgery 2667-1476 2667-1476 Published by Elsevier Ltd on behalf of British Association of Oral and Maxillofacial Surgeons. S2667-1476(22)00039-5 10.1016/j.adoms.2022.100289 100289 Article Impact of COVID-19 lockdown on incidence of maxillofacial fractures: A retrospective analysis Boom L.J. a∗ Wolvius E.B. b Rozeboom A.V.J. b a Faculty of Medicine, Erasmus University Rotterdam and Erasmus University Medical Centre, Dr. Molewaterplein 40, 3015, GD, Rotterdam, the Netherlands b Department of Maxillofacial Surgery, Erasmus University Medical Centre, Dr. Molewaterplein 40, 3015, GD, Rotterdam, the Netherlands ∗ Corresponding author. 12 4 2022 12 4 2022 1002894 4 2022 8 4 2022 © 2022 Published by Elsevier Ltd on behalf of British Association of Oral and Maxillofacial Surgeons. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. COVID-19 lockdown restrictions greatly influenced people's behaviour and movements, and therefore patient presentation may differ in maxillofacial trauma surgery during lockdown. The aim of this study is to evaluate the impact of a lockdown on the incidence, types and mechanisms of injury of maxillofacial fractures. In this single-centre retrospective cohort study patients who visited the maxillofacial surgeon after traumatic injury between the 15th of March and the 1st of June in the years 2018, 2019, 2020 and 2021 were included. The primary outcome is the incidence of maxillofacial fractures during the lockdown in 2020 compared to the pre-lockdown and post-lockdown periods. Secondary outcomes are type of fracture and mechanism of injury. A total of 130 patients with maxillofacial fractures were identified. During the lockdown 0.51 (95% CI 0.32–0.84) times less maxillofacial fractures were reported. A significant association was found between mechanism of injury and lockdown compared to the post-lockdown period. No further associations were found between a lockdown and type of fracture or mechanism of injury. In conclusion, the incidence of maxillofacial fractures was significantly lower compared to equivalent time periods in other years, but recovered after lockdown. Keywords COVID-19 Maxillofacial injuries Quarantine Epidemiology ==== Body pmc1 Introduction After the first human cases of coronavirus disease (COVID-19) emerged in early December 2019, the virus was genetically identified as SARS-CoV-2 and globally shared on the 11th of January 2020. The virus has greatly stirred the world since, achieving the status of pandemic on the 11th of March 2020 [1]. Governments were forced to implement severe measures to prevent exponential transmission of the virus. In many countries the rapid spreading inevitably led to nationwide lockdowns, the Netherlands included [2]. The COVID-19 lockdown in the Netherlands was implemented on the 15th of March 2020. Additional measures were taken on 24th of March 2020 and an ‘intelligent’ lockdown was introduced [3]. The complete package of measures included social distancing, bans on festivals and gatherings, closure of schools, restaurants and sports facilities and the strong advice for non-essential occupations to work from home if possible. These restrictions instantly limited people's movements. Eventually most restrictions were lifted on the 1st of June 2020 as the end of the intelligent lockdown was announced [4]. The strict social and public health measures greatly influenced people's behaviour. In traumatology it is known that fracture epidemiology primarily depends on human behaviour and lifestyle [[5], [6], [7], [8]]. In the past, nationwide health emergencies such as natural disasters have influenced epidemiology of pathologies that normally occur in the population [[9], [10], [11]]. The COVID-19 pandemic may be considered as such a rare event. It is therefore a reasonable assumption that the pandemic and related lockdowns may impact patient presentation. Severe facial injuries involving mobility and bleeding of facial bones may promote life-threatening airway obstruction and hypovolemia [12]. Therefore traumatology is an important subspeciality in maxillofacial surgery that needs continuation of patient care, especially in a state of national health emergency. Differences in patient presentation may alter the demand for specific healthcare resources and therapies. Due to the pandemic, the presentation of patients, medical staffing levels and provision of resources are subject to change across almost all medical fields, with no exception for maxillofacial surgery [13]. Efficient reallocation is therefore necessary to ensure qualitative and sufficient patient care, especially for immediate and unscheduled maxillofacial trauma surgery. Epidemiological analysis of maxillofacial fractures during an unprecedented pandemic provides crucial insights to ensure adequate patient care. Eventually these new insights allow healthcare workers and policymakers to develop an efficient planning system for resource allocation to sustain future health crises. Furthermore the effects of a lockdown on behaviour can be monitored through fluctuations in incidence, type and aetiology of fractures. To evaluate if influences of a lockdown are either temporary or permanent, comparison to a post-lockdown time period is necessary. Therefore the aim of this study is to evaluate the impact of a COVID-19 lockdown on the incidence, types and mechanisms of injury of maxillofacial fractures compared to pre-lockdown and post-lockdown periods. 2 Methods 2.1 Design This study is a retrospective cohort study conducted at the Department of Oral & Maxillofacial Surgery of the Erasmus University Medical Centre in Rotterdam, the Netherlands. 2.2 Patient population The electronic patient record was searched for consultations concerning maxillofacial fractures between the 15th of March and the 1st of June in 2018, 2019, 2020 and 2021. Patients were included if they had either visited a maxillofacial surgeon or had undergone surgery for a maxillofacial fracture in predefined time periods. Maxillofacial fractures included zygomatic, nasal, orbital, sinus frontalis, maxillary and mandibular fractures. When patients had multiple fractures in different facial bones, each fracture was registered separately. Patients with maxillofacial fractures caused by other aetiology than trauma for example pathological fractures, were excluded. Patients with isolated dental trauma were excluded. Patients were divided in three different groups: a pre-lockdown group (2018 and 2019), a lockdown group (2020) and a post-lockdown group (2021). 2.3 Study outcomes The primary outcome of the study is the incidence of maxillofacial fractures. Secondary outcomes are mechanism of injury and type of maxillofacial fractures. Mechanism of injury was categorized as work-related, violence, domestic, fall, traffic, sports trauma and alcohol or drugs abuse. Trauma with bicycles with no other traffic involvement were considered as falls. 2.4 Variables Patient characteristics, as retrieved from the electronic patient record, were age at the time of injury and sex. In addition, time to presentation was calculated based on the date of injury and the date of presentation to a clinician. Secondly time to surgery and duration of hospitalization, if both applicable, were calculated based on the dates of surgery, hospitalization and hospital discharge. Lastly, the day of the week was calculated from the injury date and dichotomized to week or weekend. 2.5 Statistical analysis Descriptive statistics were used to generate baseline characteristics. Categorical data were described using absolute numbers and percentages. For comparison of categorical data between groups a Fisher's exact test or a Pearson's chi square test were used as appropriate. Poisson regression was performed to estimate the effect of a lockdown on the number of maxillofacial fractures. Independent Student's t-test were used for comparison of averages in each group. Data were analysed in IBM Statistical Package for Social Sciences version 25. A two-tailed p-value < 0.05 was considered statistically significant. 3 Results A total of 130 traumatic maxillofacial fractures were identified during the 2020 lockdown and equivalent periods in 2018, 2019 and 2021. General patient characteristics are summarized in Table 1 . The mean age of patients was 43.5 ± 21.5 years. The male to female ratio equals 1:2.6. The most reported mechanism of injury overall were traffic accidents.Table 1 Patient characteristics of all years combined. Table 1Patient characteristics No. % or mean ± SD Gender Male 94 72.3 Female 36 27.7 Age 130 43.5 ± 21.5 Type of fracture Frontal sinus 14 10.8 Nasal 26 12.0 Orbital 68 52.3 ZMC 49 37.7 Maxillary 37 28.5 Mandibular 43 33.1 Mechanism of injury Work-related 1 0.8 Violence 27 20.8 Fall 35 26.9 Traffic 60 46.2 Sports trauma 6 4.6 Unknown 1 0.8 Total 130 100.0 ZMC = zygomatic maxillary complex. Data separated by year are presented in Table 2 . Poisson regression showed that during a period of lockdown 0.51 (95% CI 0.32–0.84) times less maxillofacial fractures were reported, a statistically significant result (p = 0.007).Table 2 Patient characteristics separated per year. Table 2 2018 2019 2020 2021 p-values Pre Pre Lock Post Number of fractures 29 48 19 34 0.007* Type of fractures Sinus frontalis 5 3 4 2 Nasal 2 11 6 7 Orbital 17 22 13 16 ZMC 10 24 7 8 Maxillary 7 13 9 8 Mandibular 10 15 1 17 Mechanism of injury 0.959, 0.024ˆ Work-related 0 1 0 0 Violence 6 10 4 7 Fall 8 7 4 16 Traffic 14 28 11 7 Sports trauma 1 1 0 4 Unknown 0 1 0 0 Intoxications Alcohol 9 5 1 6 0.290, 0.400† Drugs 2 2 1 1 Age Mean ± SD 39.6 ± 18.7 43.6 ± 22.9 44.2 ± 21.6 46.3 ± 22.1 0.709, 0.732‡ Median 35.0 35.5 56.0 42.0 Q1-Q3 23.0–53.0 23.0–68.5 23.0–61.0 28.8–66.3 Sex 0.579, 0.749† Male (%) 20 (69.0) 35 (72.9) 15 (78.9) 24 (70.6) Female (%) 9 (31.0) 13 (27.1) 4 (21.1) 10 (29.4) Time variables Time to presentation 0 0.7 ± 1.8 0.2 ± 0.7 0.3 ± 1.1 0.478, 0.638‡ Time to surgery 3.4 ± 3.7 1.2 ± 1.8 4.6 ± 5.8 2.3 ± 5.2 0.325, 0.411‡ Duration of hospitalization 2.5 ± 3.2 4.7 ± 8.5 2.0 ± 0.8 1.4 ± 0.5 0.610, 0.227‡ Day of the week 0.073, 0.555† Weekend 17 22 5 12 Weekday 12 26 14 22 Pre, lock and post stand for pre-lockdown, lockdown and post-lockdown group respectively. The first P-value represents the comparison between pre-lockdown and lockdown group. The second P-value represents the comparison between lockdown and post-lockdown group. For Poisson regression data were put in by year, producing only one p-value. ZMC = zygomatic maxillary complex. * Poisson regression ˆ Pearson's chi square † Fisher's exact test ‡ Student's T test. Since patients may have had multiple maxillofacial fractures, no statistical analysis was performed on differences in type of fractures between the different time periods. Larger facial bones such as orbita, zygoma, maxilla and mandibula were mostly reported across all years. During the lockdown however only one mandibular fracture was reported. Furthermore, no significant association was found between the presence of a lockdown and mechanism of injury compared to the pre-lockdown period (p = 0.959). In comparison with the post-lockdown period, mechanism of injury was significantly associated with a lockdown (p = 0.024). In 2021 relatively more falls and less traffic accidents were reported compared to 2020. Overall traffic-related injuries were most reported across all years except for 2021. Alcohol-related maxillofacial injuries did not show a significant relation with the lockdown compared to both other periods (p = 0.290 and p = 0.400). No association was found for mean age and the presence of a lockdown (p = 0.709 and p = 0.732). In addition, no association was found for sex in relation to a lockdown (p = 0.579 and p = 0.749). Due to a low number of patients experiencing delays in time to presentation, time to surgery and duration of hospitalization, these time variables were unfit for Kaplan-Meier analysis. Significant differences were not found for time variables during lockdown with Student's t-test. Patients with maxillofacial fractures went to see a clinician within one day on average across all years. In the lockdown period relatively less injuries occurred during the weekend compared to the pre-lockdown period, an almost significant association (p = 0.073). No significant difference was found compared to the post-lockdown period (p = 0.555). 4 Discussion Following lockdown regulations we observed a significant decline in maxillofacial fractures up to 49% compared to the same period in 2018, 2019 and 2021. Other studies found a decline in maxillofacial fractures as well [[14], [15], [16]]. The overall reduction of maxillofacial fractures could be attributed to the decrease in traffic-related injuries, as numbers of other mechanisms of trauma remained approximately the same. Lockdown measures encouraged people to work from home and massively limited movements, thus decreasing overall traffic. Moreover the decrease in traffic-related injuries remained in 2021, leading to a significant association between mechanism of injury and a lockdown when compared to the post-lockdown period. One explanation could be that the Dutch government strongly advised to work from home and limit travel movements even after lockdown. Isolation and financial stressors during the COVID-19 pandemic may alter home-situations and raise tensions within families, possibly leading to greater numbers of violence [17]. Ludwig et al. reported a significant increase in assault-related facial trauma [18]. Though, Marchant et al. did not find an increase in interpersonal or domestic violence during lockdown [19]. Our study found no increase in cases of violence during lockdown as well. On one hand, our study was conducted at one of the eleven national trauma centres in the Netherlands, which treats more severe and complex trauma. Victims of violence may present themselves at lower-level trauma centres or other healthcare providers [20]. On the other hand, domestic violence is often underdiagnosed by clinicians at initial visits [21]. Therefore, the number of violence related injuries in this study may not be indicative for the prevalence of domestic violence in the Netherlands. No significant increase was found in time to presentation to a clinician. The short duration of hospitalization in 2020 persisted in 2021. A possible explanation may be the prevention of shortage of hospital beds. During the lockdown management of available beds was paramount. This may have carried on after the lockdown as healthcare workers remained cautious for a new surge in COVID-19 infections. This is important since maxillofacial fractures have the potential to be life-threatening and healthcare must remain easily accessible to patients especially in times of lockdown. Although not significant, a trend could be observed in the timing of injuries during the week. In the lockdown relatively less fractures were reported in the weekends. Restaurants, bars and nightlife were closed leading to less travel movements and less violence from person-to-person interactions. Thus, less maxillofacial fractures occurred. On the 28th of April 2021 bars and restaurants were gradually allowed to reopen, which might explain the light resurgence of trauma during weekends in 2021. Alcohol consumption might play a part in this. On one hand alcohol consumption might decrease when bars and restaurants were closed. But isolation and stress after abrupt alteration of everyday life may promote alcohol consumption on the other hand. Alcohol-related injuries however were not significantly associated with a lockdown in this study. Because average alcohol consumption remained the same during lockdowns worldwide [22], a lockdown might not provoke an increase in maxillofacial fractures involving alcohol. Several limitations in this study were observed. As with all single-centre studies the data do not capture all trauma in our region. As earlier mentioned, this study was performed at a national trauma centre, where severe and complex injuries are more common. Only patients who are directly admitted or transferred by peripheral hospitals are captured in this study. Patients that met our inclusion criteria, but with less severe trauma, are therefore out of scope. Secondly a single-centre study produces a lower sample size of injuries during lockdown, which sometimes makes it difficult to produce significant associations. When multiple centres are involved, higher numbers of injuries might be reported and several results may become significant. A strength of our study is the inclusion of an equivalent time period in 2021. Currently no studies reported on incidence of maxillofacial fractures after a lockdown. Observing the same time period after lockdown allows for evaluation of temporary and permanent changes made in the heat of the COVID-19 pandemic. Valuable lessons can be learned from observing these data. Future studies should therefore focus on mapping changes in the field of maxillofacial surgery during the COVID-19 pandemic. In summary, the incidence of maxillofacial fractures was significantly lower compared to equivalent time periods in other years, but recovered after lockdown. Lockdown measures seem to decrease the number of maxillofacial fractures, mainly attributed to the decrease in traffic-related injuries. Further associations between the lockdown period and changes in the presentation of maxillofacial fractures were not found. Conflict of interest The authors declare that there are no conflicts of interest regarding the publication of this article. Ethics statement/confirmation of patient permission This retrospective study was approved by the Ethics Committee of the Erasmus University Medical Centre. No personal details or identifying information of patients are included in this article. Declaration of interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ==== Refs References 1 World Health Organization Emergency Committee WHO Director-General's opening remarks at the media briefing on COVID-19 - 11 March 2020 Available from URL: https://www.who.int/dg/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19---11-march-2020 2020 2 Steffens I. A hundred days into the coronavirus disease (COVID-19) pandemic Euro Surveill 25 14 2020 Apr 2 3 Maart 2020 Maatregelen tegen verspreiding coronavirus, intelligente lockdown Available from URL: https://www-rijksoverheid-nl.eur.idm.oclc.org/onderwerpen/coronavirus-tijdlijn/maart-2020-maatregelen-tegen-verspreiding-coronavirus 2020 4 Juni 2020 Versoepeling coronamaatregelen en testen voor iedereen Available from URL: https://www-rijksoverheid-nl.eur.idm.oclc.org/onderwerpen/coronavirus-tijdlijn/juni-2020-versoepeling-coronamaatregelen-en-testen-voor-iedereen 2020 5 Porthouse J. Birks Y.F. Torgerson D.J. Risk factors for fracture in a UK population: a prospective cohort study QJM 97 9 2004 569 574 15317925 6 Bonavolontà P. Dell’aversana Orabona G. Abbate V. The epidemiological analysis of maxillofacial fractures in Italy: the experience of a single tertiary center with 1720 patients J Cranio-Maxillo-Fac Surg 45 8 2017 1319 1326 7 Sbordone C. Barca I. Petrocelli M. The influence of socioeconomic factors on the epidemiology of maxillofacial fractures in Southern Italy J Cranio-Maxillo-Fac Surg 29 8 2018 Nov 2119 2123 8 Qu X. Zhang X. Zhai Z. Association between physical activity and risk of fracture J Bone Miner Res 29 1 2014 Jan 202 211 23813682 9 Del Papa J. Vittorini P. D'Aloisio F. Retrospective analysis of injuries and hospitalizations of patients following the 2009 earthquake of L'Aquila city Int J Environ Res Publ Health 16 10 2019 May 14 1675 10 Deng Q. Lv Y. Xue C. Pattern and spectrum of tornado injury and its geographical information system distribution in Yancheng, China: a cross-sectional study BMJ 8 6 2018 Jun e021552 11 de Almeida M.M. van Loenhout J.A.F. Thapa S.S. Clinical and demographic profile of admitted victims in a tertiary hospital after the 2015 earthquake in Nepal PLoS One 14 7 2019 e0220016 12 Farber S.J. Kantar R.S. Rodriguez E.D. Facial trauma care in the austere environment J Spec Oper Med 18 3 2018 62 66 13 Maffia F. Fontanari M. Vellone V. Impact of COVID-19 on maxillofacial surgery practice: a worldwide survey Int J Oral Maxillofac Surg 49 6 2020 Jun 827 835 32414678 14 Lee D. Choi S. Kim J. The impact of COVID-19 on the injury pattern for maxillofacial fracture in Daegu city, South Korea Maxillofac Plast Reconstr Surg 43 1 2021 Sep 13 35 34515891 15 Yeung E. Brandsma D.S. Karst F.W. The influence of 2020 coronavirus lockdown on presentation of oral and maxillofacial trauma to a central London hospital Br J Oral Maxillofac Surg 59 1 2021 Jan 102 105 33208288 16 de Boutray M. Kün-Darbois J.D. Sigaux N. Impact of the COVID-19 lockdown on the epidemiology of maxillofacial trauma activity: a French multicentre comparative study Int J Oral Maxillofac Surg 50 6 2021 Jun 750 755 33172710 17 Mazza M. Marano G. Lai C. Danger in danger: interpersonal violence during COVID-19 quarantine Psychiatr Res 289 2020 Jul 113046 18 Ludwig D.C. Nelson J.L. Burke A.B. What is the effect of COVID-19-related social distancing on oral and maxillofacial trauma? J Oral Maxillofac Surg 79 5 2021 May 1091 1097 33421417 19 Marchant A.D. Gray S. Ludwig D.C. What is the effect of COVID-19 social distancing on oral and maxillofacial trauma related to domestic violence? J Oral Maxillofac Surg 79 11 2021 Nov 2319.e1 2319.e8 34454868 20 Nelms A.P. Gutmann M.E. Solomon E.S. What victims of domestic violence need from the dental profession J Dent Educ 73 4 2009 Apr 1 490 498 19339436 21 Chamberlain L. Perham-Hester K.A. The impact of perceived barriers on primary care physicians' screening practices for female partner abuse Women Health 35 2–3 2002 Jun 14 55 69 12201510 22 Grossman E.R. Benjamin-Neelon S.E. Sonnenschein S. Alcohol consumption during the COVID-19 pandemic: a cross-sectional survey of US adults Int J Environ Res Publ Health 17 24 2020 Dec 9 9189
PMC009xxxxxx/PMC9005216.txt	==== Front Adv Drug Deliv Rev Adv Drug Deliv Rev Advanced Drug Delivery Reviews 0169-409X 1872-8294 Published by Elsevier B.V. S0169-409X(22)00187-9 10.1016/j.addr.2022.114297 114297 Preface Current trends in delivery of non-viral nucleic acid-based therapeutics for improved efficacy Avci-Adali Meltem ⁎ Department of Thoracic and Cardiovascular Surgery, University Hospital Tuebingen, Calwerstr. 7/1, 72076 Tuebingen, Germany A. Santos Hélder ⁎ Head of Department, Department of Biomedical Engineering and W.J. Kolff Institute for Biomedical Engineering and Materials Science, University Medical Center Groningen/University of Groningen, Ant. Deusinglaan 1, 9713 AV Groningen, The Netherlands ⁎ Corresponding author. ⁎ Corresponding author. 12 4 2022 6 2022 12 4 2022 185 114297114297 © 2022 Published by Elsevier B.V. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Keywords Drug delivery Nanomedicine Nucleic acid RNA CRISPR Microfluidics ==== Body pmcIn recent years, the use of nucleic acid-based drugs to treat and prevent diseases has become increasingly important. Last year, two nanovaccines based on synthetic messenger RNAs (mRNAs) encoding viral proteins were approved by the FDA against COVID-19, successfully demonstrating the applicability of synthetic mRNAs for infectious disease prevention [1]. In addition to vaccination, synthetic mRNAs hold great potential for regenerative medicine and therapy by the expression of missing or defective proteins [2]. In contrast to synthetic mRNA, siRNA or shRNA can be applied to downregulate the expression of undesired proteins in cells [3]. miRNAs can be administered to restore metabolic health and re-establish a healthy environment in diseased tissue, such as inflamed or cancerous tissue [4]. In the future, CRISPR technology may offer new treatment options for patients to correct mutations or deletions [5]. Aptamers are short target-specific ssDNA or RNA molecules that can be used to functionalize vehicles loaded with drugs, such as siRNA or miRNA [6] to target the desired cell and tissue types for a specific and efficient treatment. Moreover, various types of vehicles, such as lipid nanoparticles and liposomes, and their fabrication using innovative technologies, such as microfluidics, are being investigated to enable the efficient production, delivery and uptake of nucleic acid molecules into the desired cells [7]. In addition, the introduction of new modifications in nucleic acids allows for increased stability, decreased activation potential of the innate immune system, and modulation of nucleic acid-based drugs [8], which increases the efficiency of nucleic acid-based therapy. This special issue collects 10 amazing contributions and focuses on novel and innovative strategies for the delivery of non-viral nucleic acid-based therapeutics. Worldwide contributions from experts in the fields of chemistry, biology, biotechnology, and pharmaceutics highlight the versatility of nucleic acid-based molecules for various applications in medicine and discuss translational challenges and required improvements for nucleic acid-based drugs. The review article of Prof. Avci-Adali and co-workers highlights novel and innovative strategies for the delivery of synthetic mRNA-based therapeutics for tissue regeneration. The versatility of synthetic mRNA molecules for various applications in the field of regenerative medicine are presented and translational challenges and required improvements for mRNA-based therapeutics are discussed to show the auspicious potential of the synthetic mRNA in the field of tissue regeneration. Prof. Remaut and colleagues describe the mechanisms behind the recognition of synthetic mRNA by the innate immune system and provide an extensive overview of strategies to control their innate immune-stimulating activity. These strategies range from modifications of the mRNA backbone to combination with inhibitors of the innate immune system to optimization of production and purification processes. In addition, the delicate balance of the self-adjuvant effect in mRNA vaccination strategies was discussed, which can be both beneficial and detrimental to therapeutic outcome. In their review, Prof. Santos and co-workers summarize the general principles for the development of RNAi vectors and discuss the practical aspects that should be considered in the production of these vectors, as well as the critical aspects of the vehicle that affect RNA encapsulation, targeting yield, and successful cytosolic release of RNA. Finally, recent advances in RNAi vectors and their further prospects for resolving therapeutic obstacles and promoting clinical translation are discussed. Prof. Prassl and co-workers described the key regulatory activities of miRNAs in the adipose tissue, discussed various miRNA replacement and inhibition strategies, summarized promising delivery systems for miRNAs, and reflected on the future of novel miRNA-based therapeutics to target adipose tissue with the ultimate goal to combat metabolic disorders. Dr. Hassan and colleagues discussed in their review article the opportunities offered by lipid nanoparticles (LNPs) for efficient and precise gene delivery. Furthermore, various synthesis strategies via microfluidics used for high-throughput fabrication of non-viral gene delivery vehicles are discussed. In addition, the application of these vehicles for the delivery of RNA and CRISPR editors for different diseases ranging from cancer to rare diseases are discussed. In their review article, Prof. Raemdonck and colleagues discuss both established and emerging techniques that can be used to evaluate the effects of different intracellular barriers on RNA transfection performance. They also demonstrate how various modulators, including small molecules but also genetic perturbation technologies, can promote RNA delivery by intervening at differing stages of the intracellular delivery process, such as cellular uptake, intracellular trafficking, endosomal escape, autophagy, and exocytosis. Gaining mechanistic insights into how RNA formulations are processed by cells is expected to fuel the rational design of the next generation of delivery carriers. Prof. Wang and colleagues from Melbourne, VIC, Australia outlined the advances in clinical and preclinical ultrasound technologies and the development of ultrasonic particles for ultrasound targeted gene delivery (UTGD). The use of UTGD in a variety of diseases was highlighted to readers. Furthermore, the advantages, future perspectives, and translational limitations of UTGD were discussed. Prof. Wang and colleagues briefly introduced the advantages and requirements of polymeric vectors for gene delivery into the skin. The incurable monogenic skin disease, recessive dystrophic epidermolysis bullosa (RDEB), and the main treatment methods and limitations were described. Then, the development of highly branched poly(β-amino ester)s (HPAEs) for in vitro evaluation and in vivo treatment of RDEB was summarized. In addition, the challenges, prospects, and favorable delivery routes of polymeric vectors for hereditary skin diseases were discussed. In the future, rapidly developing polymeric gene delivery systems could enable gene therapy for hereditary skin disorders. Prof. Cui and colleagues presented biomaterial-based strategies for delivering nucleic acids for various tissue regeneration approaches. First, the classes of nucleic acids and their mechanisms in tissue regeneration were introduced. Recent advances in the design of biomaterial-based nucleic acid delivery for tissue regeneration, including bone, cartilage, skin, nerve, and heart were then explained. In addition, the impact of gene delivery methods, therapeutic genes, and biomaterial selection on tissue regeneration was reviewed. Finally, the application perspectives of biomaterial-based nucleic acid delivery and the challenges of clinical implementation were highlighted. Prof. Zhang and his coworkers introduced the promising design of the nanocarriers and explained how they can enable the CRISPR/Cas9-based cancer therapy in vivo and how they can be used in the clinic in the future. First, the main features of the CRISPR/Cas9 gene-editing system were discussed and its applications in cancer therapy were summarized. Then, different types of nanocarriers for anticancer drug delivery were introduced. Finally, they focused on how to rationally design in vivo delivery systems for CRISPR/Cas9 to improve oncogene editing and cancer immunotherapy. Overall, this special issue is intended to provide important background and new insights into recent advances in the delivery of non-viral nucleic acid-based therapeutics to improve efficacy. We believe that advances in delivery systems will expand the non-viral nucleic acid-based therapeutics and provide promising options to address challenging clinical needs. Acknowledgments The guest editors of this special issue are very grateful to all authors who accepted our invitation and contributed with their amazing works to this exciting issue on the topic “Current trends in delivery of non-viral nucleic acid-based therapeutics for improved efficacy”. This special issue is dedicated to all scientists working in the field of nanomedicine and other related fields. Finally, our special thanks go also to the editors and staff of Advanced Drug Delivery Reviews for their continued support for this special issue and for making it possible. ==== Refs References 1 Le T.T. Cramer J.P. Chen R. Mayhew S. Evolution of the COVID-19 vaccine development landscape Nat. Rev. Drug Discov. 19 10 2020 667 668 32887942 2 Steinle H. Behring A. Schlensak C. Wendel H.P. Avci-Adali M. Concise review: application of in vitro transcribed messenger RNA for cellular engineering and reprogramming: progress and challenges Stem Cells 35 1 2017 68 79 27250673 3 Dong Y. Siegwart D.J. Anderson D.G. Strategies, design, and chemistry in siRNA delivery systems Adv. Drug Delivery Rev. 144 2019 133 147 4 Yang H. Qin X. Wang H. Zhao X. Liu Y. Wo H.-T. Liu C. Nishiga M. Chen H. Ge J. Sayed N. Abilez O.J. Ding D. Heilshorn S.C. Li K. An in Vivo miRNA Delivery System for Restoring Infarcted Myocardium ACS Nano 13 9 2019 9880 9894 31149806 5 You L. Tong R. Li M. Liu Y. Xue J. Lu Y. Advancements and obstacles of CRISPR-Cas9 technology in translational research Mol. Therapy-Methods Clin. Dev. 13 2019 359 370 6 Sivakumar P. Kim S. Kang H.C. Shim M.S. Targeted siRNA delivery using aptamer-siRNA chimeras and aptamer-conjugated nanoparticles Wiley Interdiscip. Rev.: Nanomed. Nanobiotechnol. 11 3 2019 e1543 30070426 7 Pardi N. Tuyishime S. Muramatsu H. Kariko K. Mui B.L. Tam Y.K. Madden T.D. Hope M.J. Weissman D. Expression kinetics of nucleoside-modified mRNA delivered in lipid nanoparticles to mice by various routes J. Controll. Release 217 2015 345 351 8 Zhao D. Yang G.e. Liu Q. Liu W. Weng Y. Zhao Y. Qu F. Li L. Huang Y. A photo-triggerable aptamer nanoswitch for spatiotemporal controllable siRNA delivery Nanoscale 12 20 2020 10939 10943 32207496
PMC009xxxxxx/PMC9005217.txt	==== Front Int J Cardiol Int J Cardiol International Journal of Cardiology 0167-5273 1874-1754 Published by Elsevier B.V. S0167-5273(22)00503-4 10.1016/j.ijcard.2022.04.023 Review 2021. The year in review. Structural heart interventions Marmagkiolis Konstantinos ad⁎ Iliescu Cezar A. b Grines Cindy L. c Matar Fadi d Cilingiroglu Mehmet e a Cardiology, University of Texas, MD Anderson Cancer Center, Houston, TX, United States of America b Medicine, Interventional Cardiology, Medical Director MD Anderson Cardiac Catheterization Laboratory, Houston, TX, United States of America c Northside Hospital, Atlanta, GA, United States of America d Cardiology, Univeristy of South Florida, Tampa, FL, United States of America e Cardiology, Tulane University, New Orleans, LA, United States of America ⁎ Corresponding author at: Cardiology, University of Texas, MD Anderson Cancer Center, Houston, Tampa Heart, 2727 W Dr Martin Luther King Jr Blvd #800, Tampa, FL 33607, United States of America. 12 4 2022 12 4 2022 10 1 2022 4 4 2022 8 4 2022 © 2022 Published by Elsevier B.V. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Since the beginning of 2020, the corona virus (COVID-19) pandemic redefined in many ways the practice of cardiology, research and cardiology conferences. Virtual conferences replaced most major in-person venues. The number of “elective” structural heart interventions declined and clinical research endured major setbacks in regards to academic and industry-sponsored clinical trials. In this review, we attempt to provide a broad overview of the field for general and interventional cardiologists with a specific interest in structural heart interventions. Keywords Structural TAVR MitraClip LAAO PFO 2021 Abbreviations ACC, American College of Cardiology SCAI, Society of Cardiovascular Angiography and Interventions TCT, Transcatheter Cardiovascular Therapeutics ESC, European Society of Cardiology CRT, Cardiovascular Research Technologies (CRT) AHA, American Heart Association EACTS, European Association for Cardio-Thoracic Surgery AS, aortic stenosis TAVR, transcatheter aortic valve replacement SAVR, Surgical aortic valve replacement RCT, randomized-controlled trial FDA, US Food and Drug Administration IDE, Investigational Device Exemption AMI, acute myocardial infarction HF, heart failure CAD, coronary artery disease PCI, percutaneous coronary intervention CT, computed tomography VKA, vitamin K antagonists KCCQ-OS, Kansas City Cardiomyopathy Questionnaire Overall Summary Score AKI, acute kidney injury CKD, Chronic kidney disease ==== Body pmc1 Introduction 2021 was a year of intense research on valvular heart disease and we have witnessed substantial progress in all aspects of structural heart interventions. Since the beginning of 2020, the corona virus (COVID-19) pandemic redefined in many ways the practice of cardiology, research and cardiology conferences. Virtual conferences replaced most major in-person venues. The number of “elective” structural heart interventions declined and clinical research endured major setbacks in regards to academic and industry-sponsored clinical trials. In this review, we attempt to provide a broad overview of the field for general and interventional cardiologists with a specific interest in structural heart interventions. We have included the major randomized controlled trials and late breaking studies presented at the ACC, SCAI, TCT, ESC, CRT and AHA conferences. In December 2021, ESC and EACTS published their 2021 ESC/EACTS Guidelines for the management of valvular heart disease [1], one year after the publication of ACC/AHA [2]. 1.1 Transcatheter aortic valve replacement (TAVR) Among low-risk AS patients who received the SAPIEN 3 valve (RCT,n:1000), the primary endpoint remained significantly lower at 2 years with TAVR versus surgery (SAVR) (11.5% vs. 17.4%; p: 0.007), but initial differences in death and stroke favoring TAVR were diminished; patients who underwent TAVR had increased rates of valve thrombosis (2.6% vs 0.7%; p:0.02) in the PARTNER-3 trial [3]. Valve thrombosis was defined according to Valve Academic Research Consortium (VARC) criteria: thrombus associated with an implanted valve that interferes with valve function or warrants treatment (anticoagulation or explantation). The 2-year data from the Evolut Low-Risk Trial (RCT, n: 1468) showed that the primary outcome of all-cause mortality and disabling stroke occurred in 4.3% of TAVR and 6.3% of SAVR patients at 2 years, a non-statistically significant difference. The need for permanent pacemaker was significantly higher in the TAVR arm (21.1% vs 7.9%) [4]. An analysis of the AMTRAC registry (n: 8626) evaluated TAVR in patients younger than 70 years old who were rejected for surgery. The outcomes were similar to those for older TAVR patients [5]. In the SURTAVI trial (RCT, n: 1660) of intermediate-risk patients the rate of all-cause mortality or disabling stroke was no different between the TAVR and SAVR arms at 5 years (31.3% vs 30.8%) [6]. Since mid-2020, FDA has removed the precaution from commercial labeling regarding TAVR in the patients with bicuspid aortic valve using SAPIEN-3 or Evolut-R/Pro [7]. For low-risk patients with bicuspid AS, TAVR appeared to be safe, with short length of hospital stay, zero mortality, and no disabling strokes at 30 days in the LRT trial [8]. Moreover, a meta-analysis of the FDA-approved IDE trials of low-risk patients with bicuspid AS undergoing TAVR demonstrated 30-day outcomes comparable to low-risk tricuspid AS patients, except for a trend toward higher stroke in bicuspid AS patients [9]. The AVATAR trial (RCT, n: 157) evaluated early SAVR in the treatment of asymptomatic severe AS. Patients randomized to early surgery had a significantly lower incidence of primary composite endpoint comprising all-cause death, AMI, stroke or unplanned HF hospitalization than those in the conservative arm (15.2% vs 34.7%, p = 0.02) [10]. Improved outcomes were mainly driven by a significant decrease in heart failure hospitalizations (4.01% vs 12.94%;HR:0.32, CI: 0.08–1.19). The results of the EARLY-TAVR (Evaluation of TAVR Compared to Surveillance for Patients With Asymptomatic Severe Aortic Stenosis) and PROGRESS (Management of Moderate Aortic Stenosis by Clinical Surveillance or TAVR) are expected in 2022. 1.1.1 Aortic Stenosis and PCI The ACTIVATION trial (RCT,n:235) compared PCI vs no-PCI prior to TAVR in patients with severe aortic stenosis and obstructive CAD. The observed rates of death (13.4% vs 12.1%) and rehospitalization at 1 year (34.5% vs 33.6%) were similar between PCI and no PCI prior to TAVR; however, the non-inferiority margin was not met, and PCI resulted in a higher incidence of bleeding (41.2% vs 21.7%) [11]. The majority of bleeding occurred in the first 30 days after TAVR. 1.1.2 Arrhythmias An analysis of the PARTNER-3 trial showed that patients with atrial fibrillation (AF) had a higher risk for the composite outcome of death, stroke or rehospitalization (HR 1.80, p = 0.0046) and rehospitalization alone (HR 1.8, p = 0.015), but not death or stroke [12]. In another analysis, early post-operative AF or flutter (POAF) was more frequent following SAVR compared with TAVR. Late POAF, but not early POAF, was significantly associated with worse outcomes at 2 years, irrespective of treatment modality [13]. A meta-analysis of 78 studies attempted to identify the predictors of permanent pacemaker implantation after TAVR. Male sex (OR, 1.16), baseline Mobitz type-1 s-degree atrioventricular block (OR, 3.13), left anterior hemiblock (OR, 1.43), bifascicular block (OR, 2.59), right bundle-branch block (OR, 2.48) and periprocedural atrioventricular block (OR, 4.17) were identified as potent predictors [14]. 1.1.3 Cerebral protection The REFLECT I trial (observational, n:375), which was stopped early, demonstrated that the TriGuard HDH cerebral embolic protection device during TAVR was safe in comparison with historical TAVR data but did not meet the predefined effectiveness endpoint compared with unprotected TAVR controls [15]. 1.1.4 Type of anesthesia The SOLVE-TAVI trial (RCT, n:447) in intermediate- to high-risk patients undergoing TAVR, showed that newer-generation self-expanding valves (SEV) and balloon-expandable valves (BEV) as well as conscious sedation (CS) and general anesthesia (GA) yield similar clinical outcomes at 1 year [16].The rates of all-cause mortality, stroke, moderate or severe paravalvular leakage, and permanent pacemaker implantation were similar between the BEV and SEV group (38.3% vs. 40.4%; p = 0.66) at 1 year. Regarding the anesthesia comparison, the combined endpoint of all-cause mortality, stroke, myocardial infarction, and acute kidney injury occurred with similar rates in the GA and CS groups (25.7% vs. 23.8%;p = 0.63). 1.1.5 Antithrombotic therapy In the POPTAVI trial (RCT, n: 665) the incidence of bleeding and the composite of bleeding or thromboembolic events at 1 year were significantly less frequent with aspirin than with aspirin plus clopidogrel administered for 3 months [17].Symptomatic clinical aortic valve thrombosis occurred in 3 patients (0.9%) in the aspirin-alone group and in 1 patient (0.3%) in the aspirin–clopidogrel group. In addition, an increased valve gradient (>10 mmHg) was observed in 10 patients (3.0%) and 11 patients (3.3%), respectively. In the POPular TAVI EU (RCT,n:213) patients undergoing TAVR who were receiving oral anticoagulation, the incidence of serious bleeding over a period of 1 month or 1 year was lower with oral anticoagulation alone than with oral anticoagulation plus clopidogrel (21.7% vs 34.6%; P = 0.01) [18]., The ENVISAGE-TAVI AF trial (RCT,n:1426) showed that in patients with AF who underwent TAVR, edoxaban was non-inferior to vitamin K antagonist [19]. The importance of subclinical leaflet thrombosis characterized by hypoattenuated leaflet thickening (HALT) and reduced leaflet motion by CT remains unclear [20]. HALT is more frequent in transcatheter compared with surgical valves at 30 days, but not at 1 year and it results in significantly increased aortic valve gradients [21]. In a substudy of the GALILEO-4D trial (RCT, n: 231) involving patients without an indication for long-term anticoagulation, rivaroxaban was more effective than an antiplatelet-based strategy in preventing subclinical leaflet-motion abnormalities. However, in the main trial, rivaroxaban was associated with a higher risk of death or thromboembolic complications and a higher risk of bleeding than the antiplatelet-based strategy [22]. The ATLANTIS trial (RCT,n:451) randomized TAVR patients in need for oral anticoagulation (OAC) to either apixaban 5 mg twice daily or vitamin K antagonist (VKA). In the intention-to-treat analysis, the primary endpoint—a composite of death, MI, stroke, systemic emboli, intra-cardiac or bioprosthesis thrombus, deep vein thrombosis or pulmonary embolism, and major bleeding over 1 year—was similar for the apixaban and VKA arms [23]. The importance of subclinical leaflet thrombosis and the optimal type and dose of anticoagulation to safely prevent it remain to be determined. 1.1.6 Vascular access In a STS/ACC TVT Registry analysis (n:4219), percutaneous trans-axillary access appeared to be safe and effective compared to surgical cut-down with similar rates of all-cause mortality (4.8% vs 4.1%), stroke (7.7% vs 6.5%), life-threatening bleeding (0.3% vs 0.1%; p = 0.31) but with a higher rate of major vascular complication (3.0% vs 1.5%, p = 0.02) [24]. A meta-analysis of five observational studies (2470 patients) comparing trans-carotid to transfemoral access for TAVR showed comparable procedural and clinical outcomes [25]. A single-center, retrospective analysis of 185 patients suggested that trans-caval access is a safe approach as compared to other alternative access techniques, with lower risk of kidney injury and shorter hospital stay [26]. The CHOICE-CLOSURE trial (RCT, n:516) compared a pure plug-based closure device (MANTA) with a primary suture-based technique (ProGlide) in TAVR. The MANTA was associated with a higher rate of access-site or access-related vascular complications (19.4% vs 12.0% p = 0.029) but a shorter time to hemostasis (80 vs. 240 s, p < 0.001) compared to ProGlide [27]. 1.1.7 Kidney disease In an analysis of the PARTNER 2A trial (RCT,n: 1045) intermediate-risk patients with severe AS and CKD, TAVR was associated with a similar risk at 5 years compared to SAVR for the primary endpoints [28]. The primary endpoint of the PARTNER 2A was a composite of death, stroke, rehospitalization, and new hemodialysis 5 years after SAVR or TAVR with the SAPIEN XT or SAPIEN 3. In the BRAVO 3 trial (RCT,n:802), acute kidney injury (AKI) occurred in 10.7% at 7 days and 17% at 30 days. AKI was associated with a significantly greater adjusted risk for 30-day death. Multivariate predictors of AKI at 30 days included baseline hemoglobin, body weight, and prior coronary artery disease, and predictors at 7 days included pre-existing vascular disease, CKD, transfusion, and valve post-dilation [29]. 1.1.8 Other Coronary obstruction during TAVR is a rare (0.7%) but disastrous complication with estimated 30-day mortality of 40–50%. The BASILICA technique entails intentional electrosurgical crossing and laceration of valve leaflets to prevent coronary obstruction during TAVR [30]. The international BASILICA registry (n: 214) demonstrated 86.9% procedural success and low rates of 30-day stroke (2.8%) and death (2.8%) [31].One-year outcomes from the BASILICA trial (observational, n:28) indicated no late stroke, myocardial infarction, or death related to BASILICA [32]. In the AMTRAC Valve Registry (n = 7303) 27.2% patients who underwent TAVR had a baseline MR grade ≥ moderate. MR regressed in 44.1%. 4-year mortality and CHF were higher for those with MR persistence, but not for those with MR regression after TAVR. In a propensity score-matched cohort with significant residual MR after TAVR, staged mitral intervention (repair or replacement) was associated with a better functional class [33]. A similar analysis (observational, n: 2964) showed a higher prevalence of baseline MR grade ≥ moderate (41.6%) which was also associated with increased mortality; however, the use of newer generation self expandable valves was associated with higher survival rate at 1 year irrespective of the degree of pre-procedural MR. [34] An analysis of the PARTNER-IIa trial showed that worsening tricuspid regurgitation (TR) occurred in 17.3% of TAVR and 27.0% of SAVR patients. Worsening TR is associated with female sex, AF, right ventricular enlargement, and SAVR. Regardless of mode of AVR, worsening TR was similarly associated with a poor prognosis [35]. Analysis of the PARTNER-3 trial demonstrated that predilation and direct TAVR were safe in patients with low surgical risk and favorable aortic valve anatomy. Direct TAVR decreased the procedure duration and did not predispose to more postdilation [36]. An analysis of the National Inpatient Sample and a meta-analysis demonstrated the safety of TAVR in cancer patients [37,38].Unlabelled TableKey points TAVR ➢ Sustained 2-year benefits in low-risk patients. ➢ Sustained 5-year benefits in intermediate-risk patients. ➢ Effective for bicuspid aortic valve even in low-risk patients. ➢ PCI as effective before or after TAVR. ➢ Late post-operative atrial fibrillation or flutter is associated with worse outcomes. ➢ The importance of subclinical leaflet thrombosis remains unclear. ➢ Single antiplatelet therapy without anticoagulation is probably the preferred anti-thrombotic regimen. ➢ Trans-axillary, trans-carotid and trans-caval alternative access are safe and effective in TAVR, although data are limited. ➢ Persistent at least moderate MR after TAVR is associated with worse outcomes. ➢ Worsening TR after TAVR is associated with poor prognosis. ➢ Direct TAVR safe is safe even in low risk patients with favorable anatomy. ➢ TAVR is safe in cancer patients. 1.2 Transcatheter edge-to-edge repair (TEER), transcatheter mitral valve replacement (TMVR) 1.2.1 Transcatheter Edge-to-Edge Repair (TEER) Transcatheter edge-to-edge repair (TEER) is now the standard of care for patients with symptomatic functional MR (FMR) despite guideline-directed medical therapy (GDMT) without an alternative indication for cardiac surgery. The 2020 ACC/AHA guidelines upgraded its use to a Class 2a recommendation for select primary MR (PMR), while the 2021 ESC/EACTS guidelines gave a IIb recommendation. A new class 2a recommendation for select FMR patients was given by both the ACC/AHA and the ESC/EACTS guidelines [1,2]. Over 33,000 patients have received TEER in the United States with continuously improving 30-day mortality (4.6%) and an average length of stay of one day [39]. 3-year outcomes of the COAPT trial (RCT, n: 614) that randomized patients with HF and moderate-to-severe or severe FMR who remained symptomatic despite GDMT, showed sustained 3-year improvements in MR severity, quality-of-life, and functional capacity with MitraClip compared to GDMT alone. The annualized rates of heart failure hospitalizations (HFHs) per patient-year were 35.5% vs 68.8% (p < 0.001) and mortality 42.8% vs 55.5% (p = 0.001).Moreover, patients assigned to GDMT alone who crossed over and were treated with TEER, the subsequent composite rate of mortality or HFH was reduced compared with those who continued on GDMT alone (p = 0.006) [40]. 2-year outcomes of CLASP study (observational, n: 124) demonstrated sustained favorable outcomes with the PASCAL device in FMR and PMR. Results showed high survival (72% FMR, 94% PMR) and freedom from HF rehospitalization rates (78% FMR, 97% PMR) with a significantly reduced annualized HFFs [41]. Although the COAPT trial has clearly defined the criteria for better TEER outcomes in FMR, up to half of real-world patients do not meet these highly selective criteria [42]. Several subgroup analyses have been performed in the COAPT trial. Baseline predictors of clinical super-responders were lower serum creatinine and KCCQ-OS score [43]. The impact of TEER in HFH was less pronounced in women compared with men beyond the first year after treatment [44] Diabetic and non-diabetic patients had consistent reductions in the 2-year rates of death and HFH and improvements in QOL and functional capacity following TEER [45]. Despite the worse prognosis of heart failure patients with a history of AF, MR reduction with the MitraClip still afforded substantial clinical benefits [46]. COPD was associated with attenuation of the survival benefit of TEER versus GDMT; however, the benefits of TEER on both HFH and health status were similar regardless of COPD [47]. The first report of CUTTING-EDGE registry (n: 332) reported that MV surgery after TEER carries high mortality (24.1% at 1 year and 31.7% at 3 years after MV surgery) and morbidity risks; moreover, only <10% of patients underwent MV repair [48]. The role of TEER in post-MI MR was evaluated in a retrospective international registry of 471 patients with at least moderate-to-severe MR following MI. The immediate procedural success did not differ between patients who underwent surgical MV repair or replacement (SMVR) and TEER (92% vs. 93%, P = 0.53). However, in-hospital and 1-year mortality rates were significantly higher in SMVR than in TEER (16% vs. 6%, P = 0.03 and 31% vs. 17%, P = 0.04) [49]. Two retrospective studies suggested that TEER can be safely performed with moderate conscious sedation and with same-day discharge [50]. 1.2.2 TMVR One-year outcomes were reported of the MITRAL trial (observational, n:30) evaluating transseptal mitral valve-in-valve (MViV) with the SAPIEN 3 in high-risk patients with failed surgical mitral bioprostheses. Transseptal MViV was associated with 100% technical success, low procedural complication rates, and very low mortality (3.4% in 1 month and 17.3% in 1 year) [51, 52]. The first single-arm prospective study evaluating transseptal mitral valve in ring (MViR) with the SAPIEN 3 in high-risk patients with failed surgical annuloplasty rings yielded a 30-day mortality rate 6.7% lower than predicted by the Society of Thoracic Surgeons score. At 1 year, transseptal MViR was associated with symptom improvement and stable valve performance [53]. 2-year outcomes after the implantation of the TENDYNE valve showed an all-cause mortality of 39.0% with the majority of deaths (43.6%) occurring during the first 90 days. 93.2% of surviving patients had no MR with decrease in heart failure hospitalizations. The improvement in symptoms at 1 year (88.5% NYHA functional class I or II) was sustained to 2 years (81.6% NYHA functional class I or II) [54].Unlabelled TableKey points Mitral valve interventions ➢ Transcatheter edge-to-edge repair (TEER) is now supported by ACC/AHA and ESC/EACTS guidelines for primary and functional MR. ➢ TEER mortality today is 4.6% at 30 days and the average length of stay is one day. ➢ Sustained 3-year TEER outcomes in patients with secondary MR. ➢ Sustained 2-year outcomes with the PASCAL system in patients with primary or functional MR. ➢ Creatinine, gender, diabetes, KCCQ-OS score, atrial fibrillation and COPD affect the outcomes of TEER in patients with functional MR. ➢ TEER in post-MI MR may be an alternative to surgery. ➢ TEER can be safely performed with moderate conscious sedation and with same-day discharge. ➢ Promising results of transseptal SAPIEN 3 implantation for mitral valve-in-valve or valve-in-ring in high-risk patients with failed bioprosthesis or surgical annuloplasty rings. ➢ Sustained 2-year outcomes with the TENDYNE device. 1.3 Tricuspid valve interventions Currently there are no FDA-approved transcatheter modalities for the management of tricuspid valve disease. In Europe PASCAL and TriClip are CE certified. They are both clip-based devices designed for right heart interventions. A single-center database analysis (n:80) compared the PASCAL versus MitraClip-XTR for the treatment of tricuspid regurgitation. Reduction in TR severity by at least one grade at 30 days was achieved in 91% and 96% respectively with similar 30-day mortality (5.0% vs 5.0%) [55]. 1-year outcomes of the TRILUMINATE trial (observational, n:85) found the TriClip to be safe and effective in patients with moderate or greater TR. TR was reduced to moderate or less in 71% of subjects while the overall major adverse event rate and all-cause mortality were both 7.1% at 1 year [56]. In the first 30-day report of the CLASP TR (observational, n:34) in the US, the PASCAL device performed as intended, with substantial TR reduction, low MAE rate (5.9%), no mortality or re-intervention, and significant improvements in functional status, exercise capacity, and quality of life [57]. 12-month outcomes from the multicenter compassionate-use experience with the PASCAL System (n:30) demonstrated survival of 93% and achievement of NYHA functional class I or II in 90% of the patients with improved 6-min walk distance. There was no stroke, endocarditis, or device embolization during the follow-up [58]. The Cardioband tricuspid system is designed to reduce functional TR through annular reduction. Via a steerable catheter the Cardioband implant is secured to the tricuspid annulus with stainless steel anchors. A size-adjustment tool enables controlled annular reduction to achieve optimal TR improvement. In the 30-day report of the TriBAND study (n:61), Cardioband demonstrated favorable outcomes at discharge and 30 days (all-cause mortality 1.6% and 19.7% at discharge and 30-days) in patients with symptomatic severe functional TR [59]. This first-in-human experience evaluating a percutaneous tricuspid valve (EVOQUE TTVR) in 25 patients demonstrated high technical success (92%), acceptable safety (30-day mortality 0%, 96% TR grade ≤ 2+, major bleeding 12% and 8% pacemaker implantation requirement) and significant clinical improvement [60].Unlabelled TableKey points Tricuspid valve interventions ➢ PASCAL and MitraClip-XTR showed reduction in TR severity by at least one grade at 30 days in 91% vs 96% respectively with similar 30-day mortality (5.0% vs 5.0%) ➢ TriClip reduced TR to moderate or less in 71% of patients with 1-year all-cause mortality 7.1%. ➢ First US experience with the PASCAL device: substantial TR reduction, low MAE rate (5.9%), no mortality or re-intervention, and significant improvements in functional status, exercise capacity, and quality of life. ➢ Cardioband demonstrated favorable outcomes at discharge and 30 days (all-cause mortality 1.6% and 19.7% at discharge and 30-days). ➢ First-in-human experience with the EVOQUE TTVR demonstrated high technical success (92%), acceptable safety (30-day mortality 0%, 96% TR grade ≤ 2+, major bleeding 12% and 8% pacemaker implantation requirement) and significant clinical improvement. 1.4 Percutaneous left atrial appendage occlusion 4-year outcomes from the PRAGUE-17 trial (RCT, n: 402) comparing left atrial appendage closure (LAAO) (Watchman or Amulet) with NOACs (95% apixaban) in non-valvular AF patients with a history of cardio-embolism, LAAO remains non-inferior to NOACs for preventing major cardiovascular, neurological or bleeding events [61]. A meta-analysis of 16 studies comprising 1428 patients suggested that LAAO combined with AF ablation is an effective and safe strategy. The long-term freedom rate from atrial arrhythmia was 66%, long-term successful rate sealing of LAAC 100%, and ischemic stroke/transient ischemic attack/systemic embolism during follow-up was 1%. Periprocedural adverse event rate (phrenic nerve palsy, intracoronary air embolus, device embolization, and periprocedural death) was 0%, procedure-related bleeding 3% and pericardial effusion 0% [62]. Another meta-analysis of 42 studies showed that intra-cardiac echocardiography (ICE) guided implantation is feasible and safe while it reduces exposure to general anesthesia and associated potential risks [63]. The next-generation Watchman FLX device approved by FDA in August 2020, is fully recapturable and repositionable with shorter device length and an atraumatic closed distal end. The PINNACLE FLX study (n:400) achieved primary effectiveness end point in 100%. Device-related thrombus was reported in 7 patients, no patients experienced pericardial effusion requiring open cardiac surgery, and there were no device embolizations [64]. A real-life analysis of the NCDR LAAO Registry (n: 49,357) suggested that women have a significantly higher risk of any in-hospital adverse events after LAAO (6.3% vs 3.9%, P < 0.001), major adverse event (4.1%vs 2.0%; P < 0.001) owing to pericardial effusion requiring drainage (1.2% vs 0.5%) or major bleeding (1.7% vs 0.8%). Women were also more likely than men to experience a hospital stay longer than 1 day (16.0% vs 11.6%; P < 0.001) or death (0.3% vs 0.1%; P < 0.001) [65]. In the Amulet IDE trial (RCT, n: 1878) Amulet was non-inferior to Watchman for the primary safety end point (14.5% versus 14.7%; P < 0.001 for non-inferiority). Major bleeding and all-cause death were similar between groups (10.6% versus 10.0% and 3.9% versus 5.1%, respectively). Procedure-related complications were higher for the Amulet occluder (4.5% versus 2.5%), largely related to more frequent pericardial effusion and device embolization. LAA occlusion was higher for the Amulet occluder than for the Watchman device (98.9% versus 96.8%; P < 0.001 for non-inferiority; P = 0.003 for superiority) [66]. Two retrospective studies suggested that LAAO can be safely performed with moderate conscious sedation and with same-day discharge [67,68].Unlabelled TableKey points Left atrial appendage occlusion ➢ LAAO is non-inferior to NOACs in patients with AF and history of cardioembolism. ➢ ICE-guided LAAO is safe. ➢ Combined LAAO and AF ablation is safe and effective. ➢ LAAO may carry higher risk in women. ➢ The next-generation Watchman FLX device demonstrates improved outcomes. ➢ Amulet in non-inferior to the Watchman device. ➢ LAAO can be safely performed with moderate conscious sedation and with same-day discharge. 2 Discussion Although the COVID-19 pandemic dominated public health headlines in 2021, important research progressed on the structural cardiology field. The most highlighted issues were the establishment of TAVR efficacy and safety in low risk, younger and cancer patients, and with the use of alternative access, the deeper understanding of the subclinical leaflet thrombosis, the advancement of TEER as a preferred therapy for selective patients, the newer data on TMVR and TTVR and the expansion of LAAO and PFO closure technologies. Declaration of Competing Interest Nothing to disclose. Appendix A Supplementary data Supplementary material 1 Image 1 Supplementary material 2 Image 2 Appendix A Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijcard.2022.04.023. ==== Refs References 1 Vahanian A. Beyersdorf F. Praz F. ESC/EACTS guidelines for the management of valvular heart disease EuroIntervention. Dec 22 2021 2021 2 Otto C.M. Nishimura R.A. Bonow R.O. 2020 ACC/AHA guideline for the management of patients with valvular heart disease: a report of the American College of Cardiology/American Heart Association joint committee on clinical practice guidelines Circulation 143 5 Feb 2 2021 e72 e227 33332150 3 Leon M.B. Mack M.J. Hahn R.T. Outcomes 2 years after Transcatheter aortic valve replacement in patients at low surgical risk J. Am. Coll. Cardiol. 77 9 Mar 9 2021 1149 1161 33663731 4 Forrest J. Complete 2-year follow-up from the Evolut low risk trial Presented EuroPCR 2021 2021. May 18, 2021 5 Witberg G. Landes U. Codner P. Clinical outcomes of transcatheter aortic valve implantation in patients younger than 70 years rejected for surgery: the AMTRAC registry EuroIntervention 17 16 Mar 18 2022 1289 1297 10.4244/EIJ-D-21-00613 34673502 6 NM VM 5-year clinician and echocardiographic outcomes from the randomized SURTAVI trial Presented at: TCT 2021. Orlando, FL November 5, 2021 7 Vincent F. Ternacle J. Denimal T. Transcatheter aortic valve replacement in bicuspid aortic valve stenosis Circulation 143 10 Mar 9 2021 1043 1061 33683945 8 Waksman R. Craig P.E. Torguson R. Transcatheter aortic valve replacement in low-risk patients with symptomatic severe bicuspid aortic valve stenosis JACC Cardiovasc. Interv. 13 9 May 11 2020 1019 1027 32381181 9 Medranda G.A. Rogers T. Doros G. Transcatheter aortic valve replacement in low-risk bicuspid and tricuspid patients: meta-analysis Cardiovasc. Revasc. Med. 33 Dec 2021 1 6 34253474 10 Banovic M. Putnik S. Penicka M. Aortic valve replacement versus conservative treatment in asymptomatic severe aortic stenosis: the AVATAR trial Circulation 145 2022 648 658 10.1161/CIRCULATIONAHA.121.057639 34779220 11 Patterson T. Clayton T. Dodd M. ACTIVATION (PercutAneous coronary inTervention prIor to transcatheter aortic VAlve implantaTION): a randomized clinical trial JACC Cardiovasc. Interv. 14 18 Sep 27 2021 1965 1974 34556269 12 Shahim B. Malaisrie S.C. George I. Atrial fibrillation and outcomes after Transcatheter or surgical aortic valve replacement (from the PARTNER 3 trial) Am. J. Cardiol. 148 Jun 1 2021 116 123 33691183 13 Shahim B. Malaisrie S.C. George I. Postoperative atrial fibrillation or flutter following Transcatheter or surgical aortic valve replacement: PARTNER 3 trial JACC Cardiovasc. Interv. 14 14 Jul 26 2021 1565 1574 34294398 14 Ullah W. Zahid S. Zaidi S.R. Predictors of permanent pacemaker implantation in patients undergoing transcatheter aortic valve replacement - A systematic review and meta-analysis J. Am. Heart Assoc. 10 14 Jul 20 2021 e020906 15 Lansky A.J. Makkar R. Nazif T. A randomized evaluation of the TriGuard HDH cerebral embolic protection device to Reduce the Impact of Cerebral Embolic LEsions after TransCatheter Aortic Valve ImplanTation: the REFLECT I trial Eur. Heart J. 42 27 Jul 15 2021 2670 2679 34000004 16 Feistritzer H.J. Kurz T. Stachel G. Impact of anesthesia strategy and valve type on clinical outcomes after transcatheter aortic valve replacement J. Am. Coll. Cardiol. 77 17 May 4 2021 2204 2215 33926657 17 Brouwer J. Nijenhuis V.J. Delewi R. Aspirin with or without Clopidogrel after Transcatheter aortic-valve implantation N. Engl. J. Med. 383 15 Oct 8 2020 1447 1457 32865376 18 Nijenhuis V.J. Brouwer J. Delewi R. Anticoagulation with or without Clopidogrel after Transcatheter aortic-valve implantation N. Engl. J. Med. 382 18 Apr 30 2020 1696 1707 32223116 19 Van Mieghem N.M. Unverdorben M. Hengstenberg C. Edoxaban versus vitamin K antagonist for atrial fibrillation after TAVR N. Engl. J. Med. 385 23 Dec 2 2021 2150 2160 34449183 20 Bogyi M. Schernthaner R.E. Loewe C. Subclinical leaflet thrombosis after Transcatheter aortic valve replacement: a meta-analysis JACC Cardiovasc. Interv. 14 24 Dec 27 2021 2643 2656 34949391 21 Makkar R.R. Blanke P. Leipsic J. Subclinical leaflet thrombosis in transcatheter and surgical bioprosthetic valves: PARTNER 3 cardiac computed tomography substudy J. Am. Coll. Cardiol. 75 24 Jun 23 2020 3003 3015 32553252 22 De Backer O. Dangas G.D. Jilaihawi H. Reduced leaflet motion after transcatheter aortic-valve replacement N. Engl. J. Med. 382 2 Jan 9 2020 130 139 31733182 23 Collet J.P. Oral anti-Xa anticoagulation after trans-aortic valve implantation for aortic stenosis: the randomized ATLANTIS trial Presented ACC 2021 2021 May 15, 2021 24 Chung C. Kaneko T. Tayal R. Dahle T.G. McCabe J.M. Percutaneous versus surgical transaxillary access for transcatheter aortic valve replacement: a propensity-matched analysis of the US experience EuroIntervention Nov 16 2021 10.4244/eij-d-21-00549 25 McGrath D.P. Kawabori M. Wessler B. Chen F.Y. Zhan Y. A meta-analysis of transcarotid versus transfemoral transcatheter aortic valve replacement Catheter. Cardiovasc. Interv. 98 4 Oct 2021 767 773 33979472 26 Barbash I.M. Segev A. Berkovitch A. Clinical outcome and safety of transcaval access for transcatheter aortic valve replacement as compared to other alternative approaches Front. Cardiovasc. Med. 8 2021 731639 27 Abdel-Wahab M. Hartung P. Dumpies O. Comparison of a pure plug-based versus a primary suture-based vascular closure device strategy for transfemoral transcatheter aortic valve replacement: the CHOICE-CLOSURE randomized clinical trial Circulation 145 2022 170 183 10.1161/CIRCULATIONAHA.121.057856 34738828 28 Garcia S. Cubeddu R.J. Hahn R.T. 5-year outcomes comparing surgical versus transcatheter aortic valve replacement in patients with chronic kidney disease JACC Cardiovasc. Interv. 14 18 Sep 27 2021 1995 2005 34556273 29 Chandrasekhar J. Sartori S. Mehran R. Incidence, predictors, and outcomes associated with acute kidney injury in patients undergoing transcatheter aortic valve replacement: from the BRAVO-3 randomized trial Clin. Res. Cardiol. 110 5 May 2021 649 657 33839885 30 Lederman R.J. Babaliaros V.C. Rogers T. Preventing coronary obstruction during transcatheter aortic valve replacement: from computed tomography to BASILICA JACC Cardiovasc. Interv. 12 13 Jul 8 2019 1197 1216 31272666 31 Khan J.M. Babaliaros V.C. Greenbaum A.B. Preventing coronary obstruction during transcatheter aortic valve replacement: results from the multicenter international BASILICA registry JACC Cardiovasc. Interv. 14 9 May 10 2021 941 948 33958168 32 Khan J.M. Greenbaum A.B. Babaliaros V.C. BASILICA trial: one-year outcomes of Transcatheter electrosurgical leaflet laceration to prevent TAVR coronary obstruction Circ. Cardiovasc. Interv. 14 5 May 2021 e010238 33 Witberg G. Codner P. Landes U. Effect of transcatheter aortic valve replacement on concomitant mitral regurgitation and its impact on mortality JACC Cardiovasc. Interv. 14 11 Jun 14 2021 1181 1192 33992550 34 Giannini C. Angelillis M. Fiorina C. Clinical impact and evolution of mitral regurgitation after TAVI using the new generation self-expandable valves Int. J. Cardiol. 335 Jul 15 2021 85 92 33811960 35 Cremer P.C. Wang T.K.M. Rodriguez L.L. Incidence and clinical significance of worsening tricuspid regurgitation following surgical or transcatheter aortic valve replacement: analysis from the PARTNER IIA trial Circ. Cardiovasc. Interv. 14 8 Aug 2021 e010437 36 Ternacle J. Al-Azizi K. Szerlip M. Impact of Predilation during Transcatheter aortic valve replacement: insights from the PARTNER 3 trial Circ. Cardiovasc. Interv. 14 7 Jul 2021 e010336 37 Monlezun D.J. Hostetter L. Balan P. TAVR and cancer: machine learning-augmented propensity score mortality and cost analysis in over 30 million patients Cardiooncology 7 1 Jun 28 2021 25 34183072 38 Marmagkiolis K. Monlezun D.J. Cilingiroglu M. TAVR in Cancer patients: comprehensive review, Meta-analysis, and Meta-regression Front. Cardiovasc. Med. 8 2021 641268 39 Mack M. Carroll J.D. Thourani V. Transcatheter mitral valve therapy in the United States: a report from the STS-ACC TVT registry J. Am. Coll. Cardiol. 78 23 Dec 7 2021 2326 2353 34711430 40 Mack M.J. Lindenfeld J. Abraham W.T. 3-year outcomes of Transcatheter mitral valve repair in patients with heart failure J. Am. Coll. Cardiol. 77 8 Mar 2 2021 1029 1040 33632476 41 Szerlip M. Spargias K.S. Makkar R. 2-year outcomes for Transcatheter repair in patients with mitral regurgitation from the CLASP study JACC Cardiovasc. Interv. 14 14 Jul 26 2021 1538 1548 34020928 42 Scotti A. Munafo A. Adamo M. Transcatheter edge-to-edge repair in COAPT-ineligible patients: incidence and predictors of 2-year good outcome Can. J. Cardiol. Dec 16 2021 10.1016/j.cjca.2021.12.003 43 Grayburn P.A. Sannino A. Cohen D.J. Predictors of clinical response to Transcatheter reduction of secondary mitral regurgitation: the COAPT trial J. Am. Coll. Cardiol. 76 9 Sep 1 2020 1007 1014 32854834 44 Kosmidou I. Lindenfeld J. Abraham W.T. Sex-specific outcomes of Transcatheter mitral-valve repair and medical therapy for mitral regurgitation in heart failure JACC Heart Fail. 9 9 Sep 2021 674 683 34391744 45 Shahim B. Ben-Yehuda O. Chen S. Impact of diabetes on outcomes after Transcatheter mitral valve repair in heart failure: COAPT trial JACC Heart Fail. 9 8 Aug 2021 559 567 34325886 46 Gertz Z.M. Herrmann H.C. Lim D.S. Implications of atrial fibrillation on the mechanisms of mitral regurgitation and response to MitraClip in the COAPT trial Circ. Cardiovasc. Interv. 14 4 Apr 2021 e010300 47 Saxon J.T. Cohen D.J. Chhatriwalla A.K. Impact of COPD on outcomes after MitraClip for secondary mitral regurgitation: the COAPT trial JACC Cardiovasc. Interv. 13 23 Dec 14 2020 2795 2803 33303119 48 Kaneko T. Hirji S. Zaid S. Mitral valve surgery after Transcatheter edge-to-edge repair: mid-term outcomes from the CUTTING-EDGE international registry JACC Cardiovasc. Interv. 14 18 Sep 27 2021 2010 2021 34556275 49 Haberman D. Estevez-Loureiro R. Benito-Gonzalez T. Conservative, surgical, and percutaneous treatment for mitral regurgitation shortly after acute myocardial infarction Eur. Heart J. 43 7 Feb 12 2022 641 650 10.1093/eurheartj/ehab496 34463727 50 Marmagkiolis K. Kilic I.D. Ates I. Kose G. Iliescu C. Cilingiroglu M. Feasibility of same-day discharge approach after transcatheter mitral valve repair procedures J. Invas. Cardiol. 33 2 Feb 2021 E123 E126
PMC009xxxxxx/PMC9005218.txt	==== Front J Infect J Infect The Journal of Infection 0163-4453 1532-2742 The British Infection Association. Published by Elsevier Ltd. S0163-4453(22)00202-X 10.1016/j.jinf.2022.04.021 Letter to the Editor Cohorting inpatients with Omicron and Delta variants of Sars-CoV-2 does not increase rates of mixed infection Davies Jessica a Gibani Malick M ab⁎ Portone Greta a McGregor Alastair a a Department of Infectious Diseases and Tropical Medicine, Northwick Park Hospital, London North West University Healthcare NHS Trust, London, United Kingdom b Department of Infectious Disease, Imperial College London, London, United Kingdom ⁎ Corresponding author at: Imperial College London, Department of Infectious Disease, London, United Kingdom. 12 4 2022 7 2022 12 4 2022 85 1 e18e20 6 4 2022 © 2022 The British Infection Association. Published by Elsevier Ltd. All rights reserved. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcMain text To the editor, We read with interest the letter by Zhang and colleagues1 which reported intra host gene variants occurring during COVID-19 infection. The co-circulation of novel variants raises the possibility of mixed infections, leading to a theoretical risk of generating interlineage recombinants within individual hosts. This of particular concern in the hospital setting, where multiple infectious patients are in close proximity. At the time of writing, the WHO has recognised five variants of concern (VOCs), demonstrating variable pathogenicity, transmissibility, and potential for immune evasion. The Omicron variant (B.1.1.529) is the most recently characterised, and was designated as a novel variant of concern on 26 November 2021.2 This variant harbors a large number of mutations in the genome including in the S gene, particularly focused around the region encoding the receptor binding motif.3 , 4 The omicron variant has demonstrated higher rates of reinfection when compared with other variants. This likely arises from partial immune evasion, particularly reduced serum antibody neutralization.5, 6, 7 In acute hospital settings in the UK, patients testing positive for SARS-CoV-2 are either isolated in a side-room or admitted to a cohort bay alongside other positive patients.8 Cohort arrangements are variant agnostic, as genotyping is either not-available, not-performed or the results are not available within a clinically meaningful timeframe. The cohorting of patients infected with different VOCs could be associated with a theoretical risk of re-infection or super-infection -- particularly when there is limited cross-immunity between variants. In addition there is a theoretical risk of generating interlineage recombinants when mixing patients infected with different VOC.9 We assessed the practicalities and safety of cohorting patients at a time of co-circulation of delta (B.1.617.2) and omicron variants. We retrospectively assessed a series of patients admitted to a designated COVID-19 ward over a three-week period. Patients admitted to hospital with a positive lateral flow test, a positive PCR test, or a positive test in the community within 14 days preceding admission and triaged on clinical grounds to one of the general medical wards were included. These patients underwent twice-weekly PCR testing for SARS-CoV-2. Positive samples were genotyped based on SNP typing. A total of 76 patients were treated as inpatients on the COVID ward over the 22 day period between 16th December 2021 and 7th January 2022. Of these 72/76 (95%) were tested on admission and 4/76 (5%) were moved to a COVID cohort ward pending a confirmatory PCR test having tested positive in the community. The median age was 69 years (range 18 to 101). The median length of stay was 8 days (range 1 to 42), contributing a total of 875 patient days. Of the 72 patients tested on admission, 66/72 (92%) tested positive, of which 58/72 (81%) were genotyped. Of the samples genotyped during the study period, 27/58 (47%) were classified as delta, and 27/58 (47%) were classified as omicron. Genotyping was indeterminate in 4/58 samples. Serial naso-pharyngeal swab sampling was performed twice weekly for the duration of their hospital stay. Paired samples were available for 36/72 (50%) patients, of which 27 (38%) were genotyped. Of these, 15 patients were infected with the Delta variant, 11 with Omicron and 2 were indeterminate. In all instances, the genotype identified on second swab was congruent with that identified on initial testing (Fig. 1 ). Three patients who had a positive result on initial swabbing were found to be negative on repeat PCR testing, whilst five did not undergo variant sequencing.Fig. 1 Results of SARS-COV-2 PCR and genotyping on serial nasopharyngeal swabs. Data points are ordered by length of hospital stay, normalised by day of admission and colour coded according to the results of genotyping. Patients with a single positive swab collected were typically discharged prior to serial swabbing. Paired data points are connected by horizontal lines. Fig. 1 Eight patients had a third serial swab genotyped. In all instances, the variant identified on repeat sampling corresponded to that initially identified (5/8 Delta infection, 3/8 Omicron - (Fig. 1). Two patients had a positive fourth swab genotyped, both of which were congruent with initial (2/2 Delta infection). Through repeated swabbing of inpatients on a COVID ward, we have attempted to evaluate the risk of co-infection and serial infection with different SARS-CoV-2 variants of concern inherent to current cohorting practices. Despite a sample size of 76 patients, we did not see any instances of co-infection with distinct variants of SARS-Cov-2, nor did we see any acquisition of alternative genotype during inpatient stay. These results suggest that current infection control practices – where bed management is guided by positive swab status rather than genotyping – are adequate for patient safety. This appears to apply during the acute stage of infection and convalescence despite uncertainty around the degree of antibody cross-neutralisation between different VOCs. We acknowledge the limitations of our study. In particular, the number of patients undergoing sequencing following serial sampling was limited, as this was an opportunistic study. In addition, SARS-CoV-2 variants were identified based on SNP typing as opposed to whole genome sequencing, limiting the resolution of our analysis. Future studies should include structured sampling timepoints, alongside sampling of healthcare workers and the environment and supported by whole genome sequencing. These findings are relevant to inform infection control procedures during times when different VOCs are co-circulating, such as during the delta and omicron wave. These findings give a degree of re-assurance that the cohorting of SARS-CoV-2 positive patients is safe at the point of admission in the absence of a genotype result. They are likely to be of increasing importance as novel variants of concern emerge where the degree of cross-immunity remains uncertain. We would advocate for continued active genomic surveillance of patients hospitalized with SARS-CoV-2 should continue, as there remains a risk of generating new interlineage SARS-CoV-2 recombinants when cohorting patients within healthcare settings. Declaration of Competing Interest No conflicts of interest exist. Appendix Supplementary materials Image, application 1 Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. M.M.G. is supported in part by the NIHR Imperial Biomedical Research Centre. Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jinf.2022.04.021. ==== Refs References 1 Zhang Y. Jiang N. Qi W. Li T. Zhang Y. Zhang H. Intra-host SARS-CoV-2 single-nucleotide variants emerged during the early stage of COVID-19 pandemic forecast population fixing mutations J Infect [Internet]. Jan 2022 [cited 2022 Apr 4]; Available from https://pubmed.ncbi.nlm.nih.gov/35041922/ 2 WHO. Classification of Omicron (B.1.1.529): sARS-CoV-2 Variant of Concern [Internet]. 2021 [cited 2022 Mar 28]. Available from: https://www.who.int/news/item/26-11-2021-classification-of-omicron-(b.1.1.529)-sars-cov-2-variant-of-concern 3 Mannar D. Saville J.W. Zhu X. Srivastava S.S. Berezuk A.M. Tuttle K.S. SARS-CoV-2 Omicron variant: antibody evasion and cryo-EM structure of spike protein–ACE2 complex Science 80-[Internet]. 2022 Feb 18 [cited 2022 Mar 28];375(6582):760–4. Available from https://www.science.org/doi/full/10.1126/science.abn7760 4 McCallum M. Czudnochowski N. Rosen L.E. Zepeda S.K. Bowen J.E. Walls A.C. Structural basis of SARS-CoV-2 Omicron immune evasion and receptor engagement Science [Internet]. 2022 Feb 25 [cited 2022 Mar 28];eabn8652. Available from http://www.ncbi.nlm.nih.gov/pubmed/35076256 5 Garcia-Beltran W.F. Denis-KJ St. Hoelzemer A. Lam E.C. Nitido A.D. Sheehan M.L. mRNA-based COVID-19 vaccine boosters induce neutralizing immunity against SARS-CoV-2 Omicron variant Cell [Internet]. 2022 Feb 3 [cited 2022 Mar 28];185(3):457–466.e4. Available from https://pubmed.ncbi.nlm.nih.gov/34995482/ 6 Dejnirattisai W. Shaw R.H. Supasa P. Liu C. Stuart A.S. Pollard A.J. Reduced neutralisation of SARS-CoV-2 omicron B.1.1.529 variant by post-immunisation serum Lancet [Internet]. 2022 Jan 15 [cited 2022 Mar 28];399(10321):234–6. Available from https://pubmed.ncbi.nlm.nih.gov/34942101/ 7 Rössler A. Riepler L. Bante D. von Laer D. Kimpel J. SARS-CoV-2 Omicron variant neutralization in serum from vaccinated and convalescent persons N Engl J Med [Internet]. 2022 Feb 17 [cited 2022 Mar 28];386(7):698–700. Available from https://www.nejm.org/doi/full/10.1056/NEJMc2119236 8 UKHSA. COVID-19: infection prevention and control (IPC) - GOV.UK [Internet]. [cited 2022 Mar 24]. Available from: https://www.gov.uk/government/publications/wuhan-novel-coronavirus-infection-prevention-and-control 9 Jackson B. Boni M.F. Bull M.J. Colleran A. Colquhoun R.M. Darby A.C. Generation and transmission of interlineage recombinants in the SARS-CoV-2 pandemic Cell 184 20 2021 Sep 30 5179 5188 .e8 34499854
PMC009xxxxxx/PMC9005219.txt	==== Front J Clin Virol J Clin Virol Journal of Clinical Virology 1386-6532 1873-5967 Published by Elsevier B.V. S1386-6532(22)00094-4 10.1016/j.jcv.2022.105161 105161 Article Immune response of booster doses of BBIBP-CORV vaccines against the variants of concern of SARS-CoV-2 Mahmoud Sally a⁎ Ganesan Subhashini bf Al kaabi Nawal ce Naik Shivaraj a Elavalli Santosh b Gopinath Prem a Ali Alaa Mousa a Bazzi Lara a Warren katherine b Zaher Walid Abbas befg Hosani Farida Al d a Biogenix labs, G42, Abu Dhabi, UAE b G42 Healthcare, Abu Dhabi, UAE c Sheikh Khalifa Medical City, SEHA, Abu Dhabi, UAE d Abu Dhabi Public Health Center, Abu Dhabi, UAE e College of Medicine and Health Sciences, Khalifa University, Abu Dhabi, UAE f IROS (Insights Research Organization & Solutions), Abu Dhabi, UAE g College of Medicine and Health Sciences, United Arab Emirates University, Al Ain, UAE ⁎ Corresponding author. 12 4 2022 6 2022 12 4 2022 150 105161105161 28 12 2021 5 4 2022 11 4 2022 © 2022 Published by Elsevier B.V. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background Booster doses for COVID-19 vaccinations are currently recommended and approved in many countries. However, we need more evidence on the immune response of individuals to booster doses of inactivated vaccines and the neutralizing effect against the variants of concerns of SARS-CoV-2. Objective To compare the fold reduction in antibody titers against the variants of concerns of SARS-CoV-2 between the primary doses and booster dose vaccine cohorts of inactivated BBIBP-CorV vaccine. Study design In this observational study Plaque Reduction Neutralization Test (PRNT) assay was done on pooled serum samples of the recipients of primary two doses of inactivated BBIBP-CorV and on the pooled serum samples of recipients of a booster dose of inactive BBIBP-CorV. The neutralizing antibody titers against the wild (Wuhan) strain and the variants of concern (alpha, beta and delta) were compared. Results The serum sample pool from the booster cohort had high neutralizing antibody titers against the SARS-CoV-2 variants compared to the pooled serum samples of the recipients of primary two doses of inactivated BBIBP-CorV and the difference was statistically significant. The observed fold reduction in antibody titers from the serum pool of recipients of two doses of BBIBP-CorV vaccine were 3.7-fold, 14.6-fold and 10.4-fold compared to 1.8 -fold, 6.5-fold and 3.8-fold reduction against the alpha, beta and delta lineages respectively in the serum pool of recipient of a booster dose (three doses of BBIBP-CorV). Conclusion Booster doses of inactive BBIBP-CORV offered better protection against the variants of concern of SARS-CoV-2. Keywords Neutralizing antibodies booster dose SARS-CoV-2 variants of concern COVID-19 ==== Body pmc1 Background The BBIBP-CORV is an inactivated vaccine prepared by multiplication of SARS-CoV-2 WIV04/HB02 strain in African green monkey kidney cells, the virus is further inactivated by beta-propiolactone and is mixed with an aluminum-based adjuvant. [1] Phase III trials reports of the inactivated vaccine (BBIBP-CORV) demonstrated a vaccine efficacy of 79% against symptomatic COVID-19 infection. [2] However we are yet to completely understand the humoral immunity offered by COVID-19 vaccination, studies have shown that there is a difference in humoral immunogenicity with regards to vaccine type and the vaccine-induced antibody response is less effective against the novel variants of SARS-CoV-2 than the wild strain. [3, 4] The newer, novel strains of SARS-CoV-2, particularly those harboring various mutations in the spike protein receptor binding domain (RBD) that increase viral affinity for ACE2 on target cells, have shown to mediate escape from vaccine-induced humoral immunity. Additionally, the population becomes more vulnerable to the novel strains due to a gradual decrease in the antibody titers over months post-vaccination. Both of these factors lead to the decreased effectiveness of the vaccines with less pronounced protection against the newer SARS-CoV-2 variants of concern. [5] In the current global situation where booster doses of the COVID-19 vaccines are widely recommended, we are challenged to answer the question of whether a booster dose is needed for inactivated vaccines and whether a booster dose is more effective in inducing a greater immune response and offer better protection against these emerging variants of SARS-CoV-2. [6,7] Neutralizing antibody titers against the SARS-CoV-2 virus have shown to be highly predictive of the host immune protection against infection. [8] Therefore in this study, neutralizing antibody titers in pooled serum samples from individuals immunized with primary 2 doses of inactivated BBIBP-CORV and from individuals who have received a booster dose of inactivated BBIBP-CORV were compared against the Wild (Wuhan) and the Alpha (B.1.1.7, UK Variant), Beta (B.1.351, South Africa Variant), Delta (B.1.617.2, Indian Variant) variants of concerns (VOCs) of SARS-CoV-2. 2 Method The study was approved by the medical research review board, Department of Health (DOH), Abu Dhabi, United Arab Emirates. Approval number: DOH/CVDC/2021/1424 2.1 Study design and study participants An observational study was conducted as a pilot study on a minimal number of samples to understand the reduction in antibody titers against the VOCs compared to the wild strain of SARS-Cov-2. The SARS-CoV-2 viral strains used in this study were obtained from nasopharyngeal samples collected for the purpose of PCR testing for COVID-19 infection. SARS-CoV-2-positive samples were sequenced to identify the viral strains, of which samples containing the Wild, Alpha, Beta, or Delta variants were used in our study. These three variants were chosen based on the World health organization (WHO) classification of the VOCs that was of greater concern during our study period from September to October, 2021 and the Delta variant was the most predominant lineage obtained from the COVID-19 nasopharyngeal test samples. [9] Therefore, in this study, 45 different isolates of the circulating Delta lineages, 15 different isolates of the Beta lineages, and 13 different isolates of Alpha lineages were used to study the effect of neutralising antibodies against each of the strains. Antibody titers were estimated from the pooled serum samples of post-vaccination individuals using a Plaque Reduction Neutralization Test (PRNT) assay. The serum samples pooled were chosen by convenience sampling based on the availability of the stored samples. PRNT was performed using the sera of vaccinated individuals who were categorized into two vaccine cohorts: 1) participants vaccinated with 2 doses of BBIBP-CORV vaccine (n= 35), this cohort included 15 (42.9%) females and 20 (57.1% male) participants and the mean age was 41.84 ± 10.45. The average interval post second dose of of BBIBP-CorV vaccine of the pooled serum sample was 73.35 ± 26.91 days. 2) participants vaccinated with 3 doses (2 doses plus one booster dose) of BBIBP-CORV vaccine (n= 20), this cohort included 8 (33.3%) females and 16 (66.1% male) participants and the mean age was 41.71 ± 9.86. The average interval post booster dose of of BBIBP-CorV vaccine of the pooled serum sample was 82 days. 2.2 Viral culture SARS-CoV-2 was isolated from the clinical sample received in laboratory. The virus was propagated at the biosafety level 3 by inoculating Vero E6 cells (ATCC CRL 1586), acquired from the ECACC. The infected cells were incubated for 2 – 3 days in Dulbecco's Modified Eagle Medium with F12 and Glutamax containing 5% heat-inactivated fetal bovine serum and 1% Antibiotic at 37°C with 5% carbon dioxide (CO2). The supernatant was clarified by centrifugation and 250uL aliquoted into cryovials. The resulting stock was quantified by plaque assay by serially diluting the viral stock and infecting Vero E6 cells in 12 well tissue culture plate with a 0.6% Agarose overlay and stained after 2 – 3 days with 0.5% crystal violet stain. Virus stocks were sequenced to identify the lineages. 2.3 The PRNT assay The PRNT was performed in duplicate using 12-well tissue culture plates. Serial dilutions of serum samples were incubated with 60–100 plaque-forming units of virus for 1h at 37°C. The virus–serum mixtures were added onto Vero E6 cell monolayers and incubated 2 h at 37°C in 5% CO2 incubator. Then the plates were overlaid with 0.6% agarose in cell culture medium and incubated for 2 - 3 days when the plates were fixed and stained with 0.5% crystal violet. Antibody titers were defined as the highest serum dilution that resulted in > 50% (PRNT50) reduction in the number of plaques compared to negative control. The neutralizing antibody titer against the Wild strain and against the variant strains were observed using the same serum sample pool and the average titers were recorded. The reduction in antibody titres against the variants of concern of SARS-CoV-2 in comparison to the Wild strain was estimated and compared among the vaccine cohorts. Independent sample t-test was performed to compare the antibody titers between the two vaccine groups. Analysis was carried out using R software version 4.0.4 and p values <0.05 were considered significant. 3 Results The antibody titers against the Wild strain were 1:320 in the two primary doses of BIBP-CORV vaccine cohort and 1:640 in the booster cohort of BBIBP-CORV. The mean titers of BBIBP-CorV two doses cohort against alpha, beta and delta lineages were 138.46 ± 94.68, 34.12 ± 21.52 and 45.78 ± 29.96 repectively, similarly the mean titers of BBIBP-CorV three doses cohort against Alpha, Beta and Delta lineages were 289.23 ±186.30, 103.53 ± 62.94 and 156.89 ± 104.44 respectively. The antibody titers were higher in the booster vaccine cohort compared to the primary two doses of the inactivated vaccine cohort. The average fold reduction in antibody titers from the serum pool of 2 doses of BBIBP-CORV vaccine were 3.7-fold, 14.6-fold and 10.4-fold against the Alpha, Beta and Delta lineages respectively when compared to the titers against the wild strain. Similarly average fold reduction in antibody titers from the serum pool of the booster dose (2 doses of BBIBP-CORV plus one booster dose after 6 months) were 1.8 -fold, 6.5-fold and 3.8-fold against the Alpha, Beta and Delta lineages respectively. In both the vaccination cohorts the fold reduction in antibody titers against the variants of concern were least pronounced against the Alpha variant and most pronounced against the Beta variants. Against all the variants of concern of SARS-COV-2 booster vaccine cohort showed statistically significant higher antibody titers compared to the primary vaccine cohorts. This shows that against the variants booster doses offered better protection than primary 2 doses of vaccination. [Fig.1 ]. The serum pool of recipients of booster dose of BBIBP-CORV vaccine neutralized all isolates of Alpha, Beta and Delta variants at titers of at least 1:20. [Fig. 2 ].Fig. 1 Neutralizing antibody titers against the variants of concern, Reciprocal titers of neutralizing antibodies against the Alpha, Beta and Delta lineages in the serum pool of Primary 2 doses of BIBP-CORV vaccine and in the serum pool after booster dose of the same vaccine. t-test was used test the difference in titers between the two vaccine cohorts. The mean titers of BBIBP-CorV two doses cohort against alpha, beta and delta lineages were 138.46 ± 94.68, 34.12 ± 21.52 and 45.78 ± 29.96 repectively, similarly the mean titers of BBIBP-CorV three doses cohort against alpha, beta and delta lineages were 289.23 ±186.30, 103.53 ± 62.94 and 156.89 ± 104.44 respectively Fig 1 Fig. 2 Paired comparison of the antibody titers against the same variant isolates from both the Primary and booster dose vaccine cohorts, Reciprocal titers of neutralizing antibodies against the Alpha, Beta and Delta lineages. Data are from 13 different isolates of Alpha; 15 different isolates of the Beta and 45 different isolates of the circulating Delta lineages and identical titer values may overlap. The reciprocal titer of 1:20 is marked to show titers below the cut-off value for PRNT assay. Fig 2 4 Discussion Our study results show that booster doses had better neutralizing effect against variants of concern thus emphasizing the need of booster doses for inactivated vaccines. The Delta variant is potentially a greater concern as it is less sensitive to antibodies in sera from individuals who had received only primary 2 doses of BBIBP-CORV, our findings were similar to the findings of another study on inactivated vaccines that showed that neutralizing antibodies titers against the Delta variant were several folds lower compared to the Wild and Alpha variants. [10] This could be because the Delta variant strains of SARS-CoV-2 with changing antigenicity of the spike protein can affect neutralization or escape post-vaccination antibodies. [11] However, newer evidence shows that reduction in vaccine effectiveness against Delta infections would probably be due to reducing antibody levels and immunity with time post-vaccination rather than the Delta variant more effectively escaping the antibodies. [12] This study showed that the fold reduction in antibody titers were more pronounced against the Beta variant in comparison to the Alpha variant. This might be due to the fact that the Beta variant was more severe and more resistant to immune response generated by vaccines. [13] Studies showed that compared to Alpha variant the odds of progressing to severe disease, requiring critical care admissions and death due to COVID-19 were higher with the Beta variant. [13,14] Supporting our study findings that booster doses for BBIBP-CORV vaccine increases antibody titers and showed lower fold reduction against the SARS-CoV-2 variants, studies on inactivated vaccine have also shown that antibody concentrations vigorously elevated along with spike-specific circulating follicular helper T cells after a booster dose of inactivated vaccine [15] The strength of our study is that our comparisons were standardized by comparing the neutralizing activity and the fold reductions between the two vaccine cohorts against the same viral isolate of all three variants of concern. This study has also some limitations as it did not take into consideration the influence of factors like age, gender, comorbidities or prior SARS-CoV-2 infections that could have influenced the antibody titers. However, the study used the same serum sample pool to understand the reduction in antibody titers against the variant strains compared to the wild strain. Therefore, these factors are unlikely to have made significant difference in the interpretation of the results. We conclude that while post-vaccination sera have reduced efficacy against variants of concern when compared to the wild strain, there is still highly neutralizing effect against the variants of concern. The Delta variant is potentially a greater concern as it is less sensitive to antibodies in sera from individuals who had received only primary 2 doses. However, sera of patients who have received booster dose have largely preserved protection against the Delta variant. This study suggest that booster doses can be recommended for vaccine recipients of inactivated vaccines for augmenting protection against the newer variants of concern. Further studies are needed to evaluate the impact of various other factors that can affect the antibody titers and the optimal timing of the booster doses. Authors contribution SM,SG - contributed to the study concept and design; SN, PG, AM, LB – contribute to data collection; SG, SE - contributed to the data analysis and data visualization; SM, SG- contributed to the original draft; NA, SN, KW, WZ, FA - revised the manuscript, SM, SG, NA, SN, SE, PG, AM, LB, KW, WZ, FA- critically reviewed and edited the final draft of the manuscript. All authors approved the final version of the manuscript. Data sharing Individual data reported in this study are not publicly shared. Request for accessing data should be submitted to the corresponding author (sally.mahmoud@g42.ai) and upon approval of the request de-identified data will be shared. Competing interest statement No funding was received for this study. The authors declare no competing financial or non-financial interests. Acknowledgment We thank Ms. Flavia Catarutti and all lab personnel at Biogenix lab for their support. ==== Refs References 1 Wang H Zhang Y Huang B Deng W Quan Y Wang W Development of an inactivated vaccine candidate, BBIBP-CorV, with potent protection against SARS-CoV-2 Cell 182 3 2020 Aug 6 713 721 e9 32778225 2 Al Kaabi N Zhang Y Xia S Effect of 2 Inactivated SARS-CoV-2 vaccines on symptomatic COVID-19 infection in adults: a randomized clinical trial JAMA 326 1 2021 35 45 34037666 3 Jalkanen P Kolehmainen P Häkkinen H Huttunen M Tähtinen P Lundberg R COVID-19 mRNA vaccine induced antibody responses against three SARS-CoV-2 variants Nat. Commun. 12 3991 2021 June 28 4 Planas D Veyer D Baidaliuk A Staropoli I Guivel-Benhassine F Rajah MM Planchais C Porrot F Robillard N Puech J Prot M. Reduced sensitivity of SARS-CoV-2 variant Delta to antibody neutralization Nature 596 7871 2021 Aug 276 280 34237773 5 Barros-Martins J Hammerschmidt SI Cossmann A Odak I Stankov MV Morillas Ramos G Dopfer-Jablonka A Heidemann A Ritter C Friedrichsen M Schultze-Florey C Immune responses against SARS-CoV-2 variants after heterologous and homologous ChAdOx1 nCoV-19/BNT162b2 vaccination Nat. Med. 2021 Jul 14 1 5 33442018 6 Callaway E. COVID vaccine boosters: the most important questions Nature 596 7871 2021 178 180 34354274 7 Schmidt T Klemis V Schub D Mihm J Hielscher F Marx S Abu-Omar A Ziegler L Guckelmus C Urschel R Schneitler S. Immunogenicity and reactogenicity of heterologous ChAdOx1 nCoV-19/mRNA vaccination Nat. Med. 27 9 2021 Sep 1530 1535 34312554 8 Khoury D.S. Cromer D. Reynaldi A. Neutralizing antibody levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection Nat. Med. 27 2021 1205 1211 34002089 9 World Health Organization (WHO) Tracking SARS-CoV-2 variants 20 August 2022 Available at: https://www.who.int/en/activities/tracking-SARS-CoV-2-variants/ Accessed on 10 Chen Y Shen H Huang R Tong X Wu C. Serum neutralizing activity against SARS-CoV-2 variants elicited by CoronaVac Lancet Infect. Dis. 21 8 2021 May 27 1071 1072 34051887 11 Harvey WT Carabelli AM Jackson B Gupta RK Thomson EC Harrison EM Ludden C Reeve R Rambaut A Peacock SJ Robertson DL. SARS-CoV-2 variants, spike mutations and immune escape Nat. Rev. Microbiol. 19 7 2021 Jul 409 424 34075212 12 Tartof SY Slezak JM Fischer H Hong V Ackerson BK Ranasinghe ON Frankland TB Ogun OA Zamparo JM Gray S Valluri SR. Effectiveness of mRNA BNT162b2 COVID-19 vaccine up to 6 months in a large integrated health system in the USA: a retrospective cohort study Lancet North Am. Ed. 2021 Oct 4 13 Callaway E. Remember Beta? New data reveal variant's deadly powers Nature 2021 Aug 9 Available at https://www.nature.com/articles/d41586-021-02177-3#ref-CR1 Accessed 18 March 2022 14 Abu-Raddad LJ Chemaitelly H Ayoub HH Yassine HM Benslimane FM HA Al Khatib Tang P Hasan MR Coyle P AlMukdad S Severity Al Kanaani Z. Criticality, and fatality of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) beta variant Clin. Infect. Dis. 2021 May 31 15 Liu Y Zeng Q Deng C Li M Li L Liu D Mei J Mo R Zhou Q Liu M Peng S. Robust induction of B cell and T cell responses by a third dose of inactivated SARS-CoV-2 vaccine medRxiv 2021 Jan 1
PMC009xxxxxx/PMC9005221.txt	==== Front Arch Bronconeumol Arch Bronconeumol Archivos De Bronconeumologia 0300-2896 1579-2129 Published by Elsevier España, S.L.U. on behalf of SEPAR. S0300-2896(22)00286-1 10.1016/j.arbres.2022.03.006 Review Article The Short- and Long-Term Clinical, Radiological and Functional Consequences of COVID-19 Consecuencias clínicas, radiológicas y funcionales a corto y largo plazo de la COVID-19Gao Yang a1 Liang Wei-quan b1 Li Yi-ran b1 He Jian-xing cd⁎ Guan Wei-jie bcd⁎⁎ a Department of Pulmonary and Critical Care Medicine, Beijing Anzhen Hospital, Capital Medical University, Beijing, China b Department of Respiratory and Critical Care Medicine, Foshan Second People's Hospital, Affiliated Foshan Hospital of Southern Medical University, Foshan, Guangdong, China c State Key Laboratory of Respiratory Disease, National Clinical Research Center for Respiratory Disease, Guangzhou Institute of Respiratory Health, First Affiliated Hospital of Guangzhou Medical University, Guangzhou Medical University, Guangzhou, China d Department of Thoracic Surgery, Guangzhou Institute for Respiratory Health, The First Affiliated Hospital of Guangzhou Medical University, China ⁎ Corresponding author. ⁎⁎ Corresponding author. 1 Yang Gao, Wei-quan Liang, and Yi-ran Li contributed equally to the work. Jian-xing He and Wei-jie Guan contributed equally to the work. 13 4 2022 4 2022 13 4 2022 58 3238 22 2 2022 4 3 2022 © 2022 Published by Elsevier España, S.L.U. on behalf of SEPAR. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. As with the rapid increase of the number of patients who have recovered from COVID-19 globally, there needs to be a major shift of the focus from rapid pathogen detection, treatment and prevention to the promotion of better recovery. Notwithstanding the scarcity of our understandings, recent studies have unraveled a plethora of pulmonary and systemic consequences which require medical attention. These consequences remained as of the end of follow-up which ranged from 1 month to 1 year. Here, we review the consequences of COVID-19 in terms of the residual symptoms, radiological and functional manifestations, and identify the potential risk factors that contribute to a prolonged recovery. We also summarize the benefits of clinical interventions (particularly the pulmonary rehabilitation program), and address several undetermined concerns and key future research directions. Como consecuencia del rápido aumento del número de pacientes que se han recuperado de la COVID-19 en todo el mundo, es necesario cambiar el enfoque de la detección rápida del patógeno, el tratamiento y la prevención para promover una mejor recuperación. A pesar de la escasez de nuestros conocimientos, estudios recientes han desvelado una plétora de consecuencias pulmonares y sistémicas que requieren atención médica. Estas consecuencias se mantienen al final del seguimiento, que oscila entre 1 mes y 1 año. Aquí se hace una revisión de las consecuencias de la COVID-19 en términos de síntomas residuales y manifestaciones radiológicas y funcionales y se identifican los posibles factores de riesgo que contribuyen a una recuperación demorada. También se resumen los beneficios de las intervenciones clínicas (en particular el programa de rehabilitación pulmonar) y se abordan varias preocupaciones no resueltas y direcciones clave de investigación futura. Keywords COVID-19 Long COVID Recovery Symptom Intervention Radiology Psychology Palabras clave COVID-19 COVID persistente Recuperación Síntomas Intervención Radiología Psicología ==== Body pmcIntroduction Coronavirus disease of 2019 (COVID-19) has strained the medical infrastructure, personnel and resources worldwide. The burden continues to increase amid the emergence of new variants and the progressively waning immunity after global vaccine roll-out. According to the latest statistics, the total number of laboratory-confirmed cases of COVID-19 has exceeded 0.4 billion globally,1 despite the dramatically improved rate of survival. Although a certain number of patients ultimately developed severe or critical illness, symptoms gradually resolved among most patients at convalescence.2, 3, 4 The number of patients who recover from COVID-19 will continue to increase considerably in the next foreseeable future. This constitutes a pressing need for caring a massive population which has not been fully characterized previously. Not all patients achieved complete remission of symptoms, radiologic findings and pulmonary function despite a prolonged convalescence phase.2, 3, 5, 6, 7, 8, 9, 10, 11 The consequences contributed to the ongoing suffering from physiologic dysfunction and the impaired quality-of-life which markedly dampened the overall well-being. Given the overwhelming tasks of disease prevention and control, global research efforts have characterized the clinical presentations at disease onset or during hospital admission, to identify the prognostic factors and disease phenotypes, to design and validate rapid diagnostic tests, and to verify the effects of therapeutic interventions and vaccines. Recent studies have documented the manifestations of consequences which ranged from respiratory to systemic (including psychological) disorders,12 with the follow-up duration ranging from 3 months to 1 year.2, 3, 4, 5, 10, 11, 13, 14, 15, 16, 17 In light of the variable study designs and research focus, there remains a considerable knowledge gap in the understanding of the magnitude and heterogeneity of the consequences of COVID-19, particularly in the era when the medical focus has been shifted from the diagnosis and treatment to the rehabilitation. Exploring the spectrum of consequences among patients recovered from COVID-19 might help characterize the residual disease burden and identify the major predictors of adverse outcomes, which have important implications to the early diagnosis and accelerate the development of novel interventions. Here, we review the major manifestations of short- and long-term consequences, highlighting the clinical characteristics and predictors of having any consequences. We also summarize the latest clinical interventions and elaborate the important future research directions. Clinical characteristics at acute phase Airborne droplets or virus-laden materials were the dominant sources of infection.18 After acquisition of virus, common symptoms which consisted of fever, cough, fatigue, dyspnea, sore throat and headache typically occurred after the median incubation period of 4 days.19 However, >50% of patients remained afebrile on hospital admissions and developed fever thereafter. Gastrointestinal symptoms, such as loss of appetite, nausea and vomiting, were less common as compared with those in patients with avian influenza.20 Neurological manifestations (e.g., myalgia, encephalopathy, dysgeusia and anosmia) were specific to aid in the diagnosis although appearing less common.21 Few patients developed autoimmune disorders, including haemolytic anemia,22 immune thrombocytopenia,23 Guillain-Barre syndrome,24 and Kawasaki disease-like syndrome in children.25 On laboratory examination, the most common manifestations consisted of lymphopenia, elevated C-reactive protein and elevated liver enzymes.19 These manifestations have important clinical implications for prognosis. Lymphopenia (seen in ∼85% of patients), has been associated with the risk of progression to critical illness and death.19 Several radiologic manifestations on chest X-ray or computed tomography (CT), such as ground-glass opacity, pulmonary infiltration, and consolidation usually developed at later stages of progression.26 The radiologic abnormality, albeit rapidly evolving during progression,26 could be readily detected at early stages among most patients and continued to persist.19 Lung function findings regarding airflow limitation and lung volumes were usually unremarkable.27 By contrast, diffusing capacity impairment and reduced exercise tolerance were more specific to reflect functional impairment of COVID-19.28 The path of disease progression Although approximately 85% of patients developed mild diseases,19 the risk of progression to critical illness (e.g., severe pneumonia, acute respiratory distress syndrome, sepsis, multisystem organ failure) or death among the remaining patients could not be underestimated. Older age, greater symptom burdens, severe hypoxemia,29 and comorbidities (e.g., hypertension, diabetes, heart disease, chronic obstructive pulmonary disease) have been associated with a worse outcome.30 Furthermore, certain chest radiologic31 and laboratory parameters (e.g., decreased CD3 + , CD4 + , and CD8 + T cell count, higher neutrophil-to-lymphocyte and neutrophil-to-CD8 + T cell ratio) might provide prognostic implications.32 Prevention of progression to critical illness or death has been the core mission of the clinical management, because the severity of illness correlated negatively with the likelihood of, and the time needed for, clinical recovery.33 Short- and long-term consequences Although the symptoms or signs of COVID-19 ultimately resolved among the majority of patients, some suffered from persistent or new-onset symptoms and conditions which have been described as “long COVID”.34 The definition is continuously evolving, although a new consensus definition for “post COVID-19 condition” has recently been proposed.35 A commonly accepted concept refers long COVID to the symptoms that continue or develop after acute COVID-19 infection which cannot be explained by an alternative diagnosis.36 However, this could be seen in many COVID-19 survivors >6 months after recovery.37 Thus, we summarize the short- (4 to 12 weeks after disease onset) and long-term (12 weeks to 1 year or longer) consequences of COVID-19 according to the ongoing symptomatic and post COVID-19 syndrome stages below, as proposed by the UK National Institute for Health and Care Excellence guideline.36 Short-term consequences Symptoms and signs: At 4-12 weeks after the initial symptom onset, a considerable proportion of patients reported residual symptoms such as cough (2%-40.3%), dyspnea (5.5%-91.5%), chest pain (0.2%-42%), sputum production (<10%), cognitive impairment (17%-28.3%), and fatigue (17%-84.8%), while fever might persist for a couple of weeks without secondary infection (Table 1 ).37, 38, 39, 40 Over 60% of patients reported amelioration of olfactory and/or gustatory dysfunction within the first 2 months (with the percentage decreasing to 40% and 20%, respectively), followed by a slower rate of recovery thereafter.41 Psychological disorders (e.g., post-traumatic stress disorder [PTSD]) occurred in 24.5% of survivors at 1 month after discharge and in 22.4% of patients at 3 month follow-up.42 Insomnia, anxiety and depression could still be identified at 3 months.42 Table 1 Proportion of patients with persistent symptoms at the short and long terms. Table 1Persistent symptoms 4-12 weeks 6 months 1 year or longer Patients with at least 1 symptom 41.0-94.0% 37.0-81.0% 45.0-60.9% Common symptoms Fatigue 17.0-84.8% 8.0-85.0% 10.0-60.9% Dyspnea 5.5-91.5% 11.9-42% 2.7-48% Chest pain 0.2-42% 0-21.0% 0-15.8% Cough 2.0-40.3% 2.1-24.0% 3.0-29.0% Hair loss 13.3-28.6% 2.5-26.3% 11.0-36.2% Anxiety or depression 4.3-45.1% 17.39-26.7% 3.3-42.0% Sleep disorder (e.g. insomnia) 3.6-53.6% 5.42-35% 10.7-43.3% Impaired memory or poor concentration 8.1-41.0% (up to several months) Less common symptoms Joint pain 7.6-27.3% 9.0-33.0% 12.8-32.5% Headache 1.6-61.0% 2.0-26.0% 2.3-17.6% Myalgia 4.5-26.0% 2.0-38.0% 4.0-27.0% Anosmia 1.7-10.0% 3.3-18.0% 1.3-26.3% Ageusia 2.5-10.0% 3.0-13.0% 1.4-30.2% Post-traumatic syndrome disorders 14.2-28.0% (up to several months) Sputum production, weight loss, sweating, skin rash, sore throat, edema of lower limbs, palpitation, diarrhea, constipation, nausea, decreased appetite, abdominal pain, dizziness, anorexia < 10% (up to several months) Fever < 5% (up to several weeks) Abnormalities reported within 4-12 weeks53, 75, 99, 100, 101, 102, 103, 104. Abnormalities reported at 6 months2, 3, 63, 70, 74, 83, 100, 105, 106, 107. Abnormalities reported at 12 months or longer 2, 46, 61, 62, 64, 65, 66, 67, 72, 108. Chest radiologic findings: Chest radiologic abnormalities could be identified in >75% of patients at initial presentation19 and could persist at convalescence. Ground-glass opacification (GGO) was highly prevalent at discharge (99.5%) but was only detected among 37.3% of patients at 3 months. However, consolidation and crazy-paving pattern might have resolved at hospital discharge. At 1 month after discharge, GGO (42.4%, 37.5%) and reticular pattern (22.0%, 47.5%) were most common in patients with mild and severe COVID-19. These figures decreased to 29.7% and 28.3% for GGO and 13.0% and 36.7% for reticular patterns at 3 months.43 By contrast, parenchymal bands, thickening of adjacent pleura and reticular lesions persisted in 34.4%, 13.9%, and 14.8% of patients at 3 months.44 Lung function abnormality: Although occurring in a small proportion of patients during acute phase, pulmonary function remained abnormal in the early post-acute phase especially in patients with severe COVID-19.45, 46 Reduced diffusing capacity, seen in 10-20% of patients at presentation, could substantially improve at 3 month in patients with severe COVID-19.46, 47, 48 Systemic complications: COVID-19 conferred notable adverse impacts on multiple organ systems, whose function was progressively normalized in the short term. Cardiovascular complications included myocarditis, myocardial infarction, heart failure, cardiogenic shock and arrhythmia.49 COVID-19 appeared to be associated with persistent cardiac injury after recovery, particularly subclinical myocardial injury in the short term and diastolic dysfunction in the long term.50 Despite the high incidence at acute phase,19 acute kidney injury could be generally resolved within 3 weeks without renal replacement therapy among patients with non-severe COVID-19;51, 52 new-onset chronic kidney disease was unusual but could relapse after discharge.3, 53 Acute liver injury or hepatitis occurred in few patients only and resolved during the acute phase among patients with non-critical illness, with minor impacts on long-term outcomes.30, 54, 55 Thyroid function might be impaired at acute phase in approximately 13.5% of patients and could be normalized within 2 months.56 Hypercoagulability (e.g., venous and arterial thromboses) was common at acute phase, especially in patients with severe or critical illness,57 but the incidence of venous thromboembolism fell to below 5% in the post-acute phase.38 Short-term mortality: Although most patients with COVID-19 had asymptomatic or mild-to-moderate illnesses, findings from the initial outbreak indicated the fatality rate of approximately 1.4% in mainland China.19 Findings from Spain have also lent support to these earlier observations. In a multicenter Spanish cohort of 12,126 patients, the mortality rate was 12.5% for non-hospitalized patients, 29.8% in hospitalized patients and 38.8% in patients admitted to the intensive care unit.58 The key predictors of short-term mortality consisted of the age over 50, obesity, cardiac diseases, fever, dyspnea, lung infiltration, lymphopenia, D-dimer levels above 1000 ng/mL, and requirement of mechanical ventilation.58 When taking into account the impact of the underlying diseases, in particular the chronic respiratory diseases, another multicenter Spanish cohort study of 5847 hospitalized patients with COVID-19 revealed a significantly higher 30-day mortality rate in patients with prior lung diseases than those without (29.5% vs. 17.9%).59 The study also revealed a more impressive 30-day mortality rate (approximately 40%) in patients with pre-existing COPD.59 Both studies, however, concurred that a greater age and higher D-dimer levels were associated with a significantly elevated risk of short-term death.58, 59 Patients aged greater than 60 years were at a particularly high risk of death at the short term, with the overall in-hospital mortality rate of 23.5% in a retrospective single-center study in Spain.60 Among these elderly patients, those with malignancy, chronic liver diseases, obesity and diabetes had an increased mortality risk despite the adjustment with the age.62 These studies have provided the scientific rationale for rapid identification of patients at risk of death and the implementation of more intensive therapeutic management. Medium- to long-term consequences Symptoms and signs: At week 12 and thereafter, symptoms or signs progressively waned but could still fluctuate over time. Some patients reported residual symptoms for up to 6 months or longer, especially those with severe COVID-19.3, 61 Cough lasted for 6 months or longer in 20% of patients (Table 1),62 which might be associated with pulmonary fibrotic changes.63 Breathlessness could persist for up to 1 year.2 Chest pain or tightness was reported in approximately 15.8% of patients over 1 year.64 Systemic symptoms such as fatigue persisted for 12 months among 10-60.8% of patients,2, 58, 64, 65, 66, 67, 68 which might be associated with the accentuated systemic inflammatory response and psychological disorders.68 Persistent taste and smell disorders have been reported at 6 months (7% and 11%, respectively),3 although recovered progressively thereafter.2 Albeit less commonly reported, arthralgia and myalgia could persist for over 6 months.2, 69, 70, 71 Hair loss was reported in > 20% of patients at 6 months while decreased to 11% at 1 year.2 Neurological disorders (e.g. sleep disorder, headache, dizziness) persisted with minor improvement for 1 year.2, 62, 71 Noticeably, 27% of patients suffered from sleep disorder at 6 months and 17% still had insomnia at 1 year.2 Cognitive impairment was reported to be persistent, the magnitude of which fluctuated from 4 months to 1 year, and was more prevalent in patients with severe COVID-19.62, 72, 73, 74, 75 COVID-19 imposed a considerable burden on psychology. More than 20% of patients suffered from PTSD at 1 year or longer.76 Anxiety and/or depression was reported in approximately 23% of patients at 6 months and 26% at 12 months.2, 62, 76 Chest radiologic findings: During recovery, GGO and interstitial fibrotic changes remained common.2, 44, 53 More than 60% of patients with mild or moderate COVID-19 ultimately had resolved radiologic abnormalities. However, GGO and fibrotic changes could persist while consolidation was rare on chest CT at 1 year follow-up.2, 77 Furthermore, the disease severity was associated with the magnitude of chest CT manifestations. At 3 months, pulmonary structural abnormalities were highly prevalent (up to 80.7%) in patients hospitalized in the intensive care unit due to ARDS.78 Radiologic abnormalities persisted in 20% of patients with severe COVID-19 who did not require mechanical ventilation at 12 months.79 Compared with patients who had completely resolved, a significantly higher proportion of patients diagnosed as having severe pneumonia and acute respiratory distress syndrome at the initial presentation had residual radiologic abnormality at 1 year.44 Lung function abnormality: Restrictive and obstructive patterns and impaired diffusing capacity were common in COVID-19 survivors,80 although improved considerably in non-critical survivors within 3 months.47 The improvement of diffusing capacity, however, seemed to slow down at 6 months. For up to 12 months, impaired diffusing capacity was identified in 20-30% of patients who had moderate COVID-19, whose prevalence remained constant since month 6. By contrast, impaired diffusing capacity could be found in approximately 54% of patients with critical illness at 1 year.2 In patients who had initially suffered from severe COVID-19 but did not require mechanical ventilation, impaired pulmonary function improved gradually over 1 year, with females having a higher risk of the ongoing impairment as compared with males.79 Impaired diffusing capacity could persist for a longer period, for instance, up to 18 months after discharge, among 57.92% of survivors with more severe disease.46, 81 Reduced exercise capacity was another important characteristic lung function change among survivors,37 which improved gradually since month 3 in patients with severe COVID-19.79 Systemic complications: As mentioned above, most systemic complications appeared to be transient. However, the estimated glomerular filtration rate decreased in 35% of patients after 6 months, 13% of which was categorized as the new-onset reduction of glomerular filtration rate despite the normal renal function at COVID-19 onset.3 At 6 months, hyperglycemia persisted in >30% of patients who experienced new-onset hyperglycemia during the hospitalization.82 A cohort study reported new-onset diabetes in 3.3% out of 1733 patients at 6 months.3 Functional abnormality: The magnitude of improvement depended considerably on the initial disease severity. Most patients with severe illness reported a failure of recovery in physical activity at 6 months.83 45% of hospitalized survivors exhibited limitations on daily living activities at 7 months (Table 1).84 The health care workers with severe COVID-19 showed improvement in functional fitness within 1 year, and rigorous intervention might be considered for accelerating the pulmonary rehabilitation (Fig. 1 ).85 Figure 1 Acute, short- and long-term effects of COVID-19 on multiple organ systems. Clinical manifestations a patient might experience at a specific stage of COVID-19. eGFR: estimated glomerular filtration rate; PTSD: post-traumatic stress disorder. Medium- and long-term mortality: The adverse impact of COVID-19 on the mortality risk extended to the long terms. In a cross-sectional study conducted in France, the mortality rate was 62.5% during the intensive care unit stay and increased to 72.1% to month 6 among patients aged greater than 80 years who were receiving mechanical ventilation.86 At 1 year, patients treated with corticosteroids had a markedly higher mortality rate as compared with those not (50.0% vs. 21.0%), although the data interpretation was limited by the small sample sizes.87 A study conducted in Italy has also demonstrated a markedly higher mortality rate among hospitalized patients with COVID-19 than those without (48.4% vs. 33.9%) at 1 year.88 The 1-year mortality rate was also higher in non-hospitalized emergency room visitors with COVID-19 than those without (18.0% vs. 8.7%).88 Future research perspective The continuously evolving global pandemic has elicited an unprecedented challenge for healthcare workers, policy makers and the public. Current research has outlined the clinical spectrum of the consequences, but this would surely benefit from the in-depth mechanistic insights which would unveil the biologic pathways that contribute to the ongoing disorders. Continuous profiling of the cytokine and chemokine expression levels within the lower airways (e.g., induced sputum) or peripheral blood will help understand the dynamics of residual inflammatory responses. The combination of chest imaging with high-resolution CT or functional magnetic resonance imaging with isotopes and lung function assessment (e.g., lung volume, diffusing capacity, exercise capacity) will provide clinicians an indispensable avenue for tracking the trend of changes in pulmonary functional abnormality.7, 8, 9, 17, 89 Further extension of the clinical observation is needed to reveal the long-term impact (e.g., 3-5 years) on symptom perception, lung function (particularly the exercise capacity), radiological characteristics, quality-of-life and psychology.10, 11, 16, 90 This will help reveal the duration of recovery of individual symptoms which would normally take, and determine the optimal timing for evaluating the magnitude of recovery and defining complete or partial recovery. Few intervention strategies have been proposed to accelerate recovery, with pulmonary rehabilitation demonstrating some promising findings.91, 92 Anti-inflammatory medications or biologics might have a role in antagonizing the residual inflammation, especially when taking into account the pulmonary (e.g., cough, sputum production) and systemic manifestations (e.g., fatigue). There also lacks a crystal-clear indication for the population which would benefit most from therapeutic interventions after discharge. Whether non-pharmaceutical interventions are superior to drug therapy for improving the functional well-being (e.g., exercise capacity) merits further investigations from the personalized interventions. It would be important to address the optimal duration of such interventions should they be proven effective. Previous studies have indicated a strong immune protective effect against acquiring re-infections among the COVID-19 rehabilitees.93 The natural immunity has constituted the front line which prevents from being susceptible. However, approximately 4% of patients were susceptible to re-infections with the precise mechanisms not fully explored.93, 94, 95 Despite the formation of pan-sarbecovirus antibodies,96 it remains unclear whether COVID-19 survivors would be immune to other coronavirus. Continuous follow-up might help delineate whether re-infections were caused by the waning immunity alone, the underlying immunodeficiency, or emerging variants evading the established immunity. A minority of patients were tested continuously RNA-positive or re-positive after discharge which might not necessarily reflect prolonged viral shedding.97, 98 A work-up with the government, epidemiologists and clinicians will help ascertain the true infectivity of COVID-19 survivors by gathering the information of the dynamic nucleic acid assays monitoring, enforcing symptom monitoring of patients and close contacts, and integrating multiple laboratory tests (e.g., viral culture, high-throughput sequencing). This will be important for redefining the community-based prevention measures, since no existing study has disclosed the infectivity of COVID-19 survivors who were discharged after having at least two consecutive RNA-negative laboratory test findings. Conclusions The world has come across a tough period when COVID-19 resulted in an overwhelming burden. Thanks to the efforts of the medical and scientific communities, the rapid deployment of effective vaccines and accelerated development of repurposed and novel medications, most patients have either partially or completely recovered from COVID-19. Although many questions remain unanswered, we keep optimistic that the ongoing efforts will better characterize the trajectory of clinical recovery and unveil new interventions which would promote recovery. The goal of enabling a better recovery would ultimately be realized for most COVID-19 rehabilitees. Support statement This work was supported by Guangzhou Institute for Respiratory Health Open Project (funded by China Evergrande Group) - Project No. 2020GIRHHMS09 and 2020GIRHHMS19 (Prof. Guan) and the Open Project of the State Key Laboratory of Respiratory Disease (SKLRD-OP-201909) (Dr. Gao). Author contributions Yang Gao, Wei-quan Liang, Yi-ran Li, Jian-xing He, and Wei-jie Guan drafted the manuscript; Wei-jie Guan, and Jian-xing He critically revised the manuscript. All authors have approved the final submission. Conflicts of interest The authors declared no conflict of interest with any financial organization regarding the material discussed in the manuscript. Acknowledgment We thank Prof. Jiang Xie (Department of Pulmonary and Critical Care Medicine, Beijing Anzhen Hospital, Capital Medical University) for his helpful comments on the writing of the review. ==== Refs References 1 WHO Coronavirus (COVID-19) Dashboard. 2021. Accessed Feb 21st, 2022. 2 Huang L. Yao Q. Gu X. Wang Q. Ren L. Wang Y. 1-year outcomes in hospital survivors with COVID-19: a longitudinal cohort study Lancet. 398 2021 747 758 34454673 3 Huang C. Huang L. Wang Y. Li X. Ren L. Gu X. 6-month consequences of COVID-19 in patients discharged from hospital: a cohort study Lancet. 397 2021 220 232 33428867 4 Zhang J. Xu J. Zhou S. Wang C. Wang X. Zhang W. The characteristics of 527 discharged COVID-19 patients undergoing long-term follow-up in China Int J Infect Dis. 104 2021 685 692 33540129 5 Martinez-Garcia M.A. Aksamit T.R. Aliberti S. Bronchiectasis as a long-term consequence of SARS-COVID-19 pneumonia: future studies are needed Arch Bronconeumol. 57 2021 739 740 6 Vaira L.A. Salzano G. Le Bon S.D. Maglio A. Petrocelli M. Steffens Y. Prevalence of persistent olfactory disorders in patients with COVID-19: a psychophysical case-control study with 1-year follow-up Otolaryngol Head Neck Surg. 2021 1945998211061511 7 Han X. Fan Y. Alwalid O. Li N. Jia X. Yuan M. Six-month follow-up chest CT findings after severe COVID-19 pneumonia Radiology. 299 2021 e177 e186 33497317 8 Clavario P. De Marzo V. Lotti R. Barbara C. Porcile A. Russo C. Cardiopulmonary exercise testing in COVID-19 patients at 3 months follow-up Int J Cardiol. 340 2021 113 118 34311011 9 Faverio P. Luppi F. Rebora P. Busnelli S. Stainer A. Catalano M. Six-month pulmonary impairment after severe COVID-19: a prospective, multicentre follow-up study Respiration. 100 2021 1078 1087 34515212 10 Yin X. Xi X. Min X. Feng Z. Li B. Cai W. Long-term chest CT follow-up in COVID-19 survivors: 102-361 days after onset Ann Transl Med. 9 2021 1231 34532368 11 Zhao Y. Wang D. Mei N. Yin B. Li X. Zheng Y. Longitudinal radiological findings in patients with COVID-19 with different severities: from onset to long-term follow-up after discharge Front Med. 8 2021 711435 12 Bonazza F. Borghi L. di San Marco E.C. Piscopo K. Bai F. Monforte A.D. Psychological outcomes after hospitalization for COVID-19: data from a multidisciplinary follow-up screening program for recovered patients Res Psychother. 23 2020 491 33585298 13 Lombardo M.D.M. Foppiani A. Peretti G.M. Mangiavini L. Battezzati A. Bertoli S. Long-term coronavirus disease 2019 complications in inpatients and outpatients: a one-year follow-up cohort study Open Forum Infect Dis. 8 2021 ofab384 14 Huang S. Zhou Z. Yang D. Zhao W. Zeng M. Xie X. Persistent white matter changes in recovered COVID-19 patients at the 1-year follow-up Brain. 2021 awab435 15 Du H.W. Fang S.F. Wu S.R. Chen X.L. Chen J.N. Zhang Y.X. Six-month follow-up of functional status in discharged patients with coronavirus disease 2019 BMC Infect Dis. 21 2021 1271 34930161 16 Fogante M. Cavagna E. Rinaldi G. COVID-19 follow-up: Chest X-ray findings with clinical and radiological relationship three months after recovery Radiography. S1078–8174 2021 00172 173 17 Wu Q. Zhong L. Li H. Guo J. Li Y. Hou X. A follow-up study of lung function and chest computed tomography at 6 months after discharge in patients with coronavirus disease 2019 Can Respir J. 2021 2021 6692409 33628349 18 Stadnytskyi V. Bax C.E. Bax A. Anfinrud P. The airborne lifetime of small speech droplets and their potential importance in SARS-CoV-2 transmission Proc Natl Acad Sci U S A. 117 2020 11875 11877 32404416 19 Guan W.J. Ni Z.Y. Hu Y. Liang W.H. Ou C.Q. He J.X. Clinical characteristics of coronavirus disease 2019 in China N Engl J Med. 382 2020 1708 1720 32109013 20 Redd W.D. Zhou J.C. Hathorn K.E. McCarty T.R. Bazarbashi A.N. Thompson C.C. Prevalence and characteristics of gastrointestinal symptoms in patients with severe acute respiratory syndrome coronavirus 2 infection in the United States: a multicenter cohort study Gastroenterology. 159 2020 765 767 e762 32333911 21 Romero-Sánchez C.M. Díaz-Maroto I. Fernández-Díaz E. Sánchez-Larsen A. Layos-Romero A. García-García J. Neurologic manifestations in hospitalized patients with COVID-19: The ALBACOVID registry Neurology. 95 2020 e1060 e1070 32482845 22 Lazarian G. Quinquenel A. Bellal M. Siavellis J. Jacquy C. Re D. Autoimmune haemolytic anaemia associated with COVID-19 infection Br J Haematol. 190 2020 29 31 32374906 23 Zulfiqar A.A. Lorenzo-Villalba N. Hassler P. Andr C. Immune thrombocytopenic purpura in a patient with COVID-19 N Engl J Med. 382 2020 e43 32294340 24 Toscano G. Palmerini F. Ravaglia S. Ruiz L. Invernizzi P. Cuzzoni M.G. Guillain-Barré) syndrome associated with SARS-CoV-2 N Engl J Med. 382 2020 2574 2576 32302082 25 Feldstein L.R. Rose E.B. Horwitz S.M. Collins J.P. Newhams M.M. Son M.B.F. Multisystem inflammatory syndrome in U.S. children and adolescents N Engl J Med. 383 2020 334 346 32598831 26 Shi H. Han X. Jiang N. Cao Y. Alwalid O. Gu J. Radiological findings from 81 patients with COVID-19 pneumonia in Wuhan China: a descriptive study. Lancet Infect Dis. 20 2020 425 434 27 Mo X. Jian W. Su Z. Chen M. Peng H. Peng P. Abnormal pulmonary function in COVID-19 patients at time of hospital discharge Eur Respir J. 55 2020 2001217 32381497 28 Gao Y. Chen R. Geng Q. Mo X. Zhan C. Jian W. Cardiopulmonary exercise testing might be helpful for interpretation of impaired pulmonary function in recovered COVID-19 patients Eur Respir J. 57 2021 2004265 33361097 29 Xie J. Covassin N. Fan Z. Singh P. Gao W. Li G. Association between hypoxemia and mortality in patients with COVID-19 Mayo Clin Proc. 95 2020 1138 1147 32376101 30 Mao R. Qiu Y. He J.S. Tan J.Y. Li X.H. Liang J. Manifestations and prognosis of gastrointestinal and liver involvement in patients with COVID-19: a systematic review and meta-analysis Lancet Gastroenterol Hepatol. 5 2020 667 678 32405603 31 Feng Z. Yu Q. Yao S. Luo L. Zhou W. Mao X. Early prediction of disease progression in COVID-19 pneumonia patients with chest CT and clinical characteristics Nat Commun. 11 2020 4968 33009413 32 Huang W. Li M. Luo G. Wu X. Su B. Zhao L. The inflammatory factors associated with disease severity to predict COVID-19 progression J Immunol. 206 2021 1597 1608 33579725 33 Truffaut L. Demey L. Bruyneel A.V. Roman A. Alard S. De Vos N. Post-discharge critical COVID-19 lung function related to severity of radiologic lung involvement at admission Respir Res. 22 2021 29 33478527 34 Mahase E. COVID-19: What do we know about “long COVID”? BMJ. 370 2020 m2815 32665317 35 Soriano J.B. Murthy S. Marshall J.C. Relan P. Diaz J.V. A clinical case definition of post-COVID-19 condition by a Delphi consensus Lancet Infect Dis. 22 2022 e102 e107 34951953 36 Guidelines NIfHaCEC. COVID-19 rapid guideline: managing the long-term effects of COVID-19. London: National Institute for Health and Care Excellence (UK); 2020. 37 Groff D. Sun A. Ssentongo A.E. Ba D.M. Parsons N. Poudel G.R. Short-term and long-term rates of postacute sequelae of SARS-CoV-2 infection: a systematic review JAMA Netw Open. 4 2021 e2128568 34643720 38 Nalbandian A. Sehgal K. Gupta A. Madhavan M.V. McGroder C. Stevens J.S. Post-acute COVID-19 syndrome Nat Med. 27 2021 601 615 33753937 39 Nasserie T. Hittle M. Goodman S.N. Assessment of the frequency and variety of persistent symptoms among patients with COVID-19: a systematic review JAMA Netw Open. 4 2021 e2111417 34037731 40 Fernández-de-Las-Peñas C. Palacios-Ceña D. Gómez-Mayordomo V. Florencio L.L. Cuadrado M.L. Plaza-Manzano G. Prevalence of post-COVID-19 symptoms in hospitalized and non-hospitalized COVID-19 survivors: A systematic review and meta-analysis Eur J Intern Med. 92 2021 55 70 34167876 41 Petrocelli M. Cutrupi S. Salzano G. Maglitto F. Salzano F.A. Lechien J.R. Six-month smell and taste recovery rates in coronavirus disease 2019 patients: a prospective psychophysical study J Laryngol Otol. 135 2021 436 441 33888166 42 de Lorenzo R. Cinel E. Cilla M. Compagnone N. Ferrante M. Falbo E. Physical and psychological sequelae at three months after acute illness in COVID-19 survivors Panminerva Med. 2021 Jun 1 10.23736/S0031-0808.21.04399-8 Online ahead of print 43 Chen Y. Ding C. Yu L. Guo W. Feng X. Yu L. One-year follow-up of chest CT findings in patients after SARS-CoV-2 infection BMC Med. 19 2021 191 34365975 44 Pan F. Yang L. Liang B. Ye T. Li L. Li L. Chest CT patterns from diagnosis to 1 year of follow-up in COVID-19 Radiology. 2021 211199 45 Daher A. Balfanz P. Cornelissen C. Müller A. Bergs I. Marx N. Follow up of patients with severe coronavirus disease 2019 (COVID-19): Pulmonary and extrapulmonary disease sequelae Respir Med. 174 2020 106197 33120193 46 Steinbeis F. Thibeault C. Doellinger F. Ring R.M. Mittermaier M. Ruwwe-Glösenkamp C. Severity of respiratory failure and computed chest tomography in acute COVID-19 correlates with pulmonary function and respiratory symptoms after infection with SARS-CoV-2: An observational longitudinal study over 12 months Respir Med. 191 2022 106709 34871947 47 Chen M. Liu J. Peng P. Jian W. Gao Y. Fang L. Dynamic changes of pulmonary diffusion capacity in survivors of non-critical COVID-19 during the first six months EClinicalMedicine. 43 2022 101255 35018338 48 Sonnweber T. Sahanic S. Pizzini A. Luger A. Schwabl C. Sonnweber B. Cardiopulmonary recovery after COVID-19: an observational prospective multicentre trial Eur Respir J. 2021 57 49 Desai A.D. Lavelle M. Boursiquot B.C. Wan E.Y. Long-term complications of COVID-19 Am J Physiol Cell Physiol. 322 2022 c1 c11 34817268 50 Ramadan M.S. Bertolino L. Zampino R. Durante-Mangoni E. Cardiac sequelae after coronavirus disease 2019 recovery: a systematic review Clin Microbiol Infect. 27 2021 1250 1261 34171458 51 Chan L. Chaudhary K. Saha A. Chauhan K. Vaid A. Zhao S. AKI in hospitalized patients with COVID-19 J Am Soc Nephrol. 32 2021 151 160 32883700 52 Heung M. Steffick D.E. Zivin K. Gillespie B.W. Banerjee T. Hsu C.Y. Acute kidney injury recovery pattern and subsequent risk of CKD: an analysis of veterans health administration data Am J Kid Dis. 67 2016 742 752 26690912 53 Morin L. Savale L. Pham T. Colle R. Figueiredo S. Harrois A. Four-month clinical status of a cohort of patients after hospitalization for COVID-19 JAMA. 325 2021 1525 1534 33729425 54 Bangash M.N. Patel J. Parekh D. COVID-19 and the liver: little cause for concern Lancet Gastroenterol Hepatol. 5 2020 529 530 32203680 55 Guo H. Zhang Z. Zhang Y. Liu Y. Wang J. Qian Z. Analysis of liver injury factors in 332 patients with COVID-19 in Shanghai China. Aging. 12 2020 18844 18852 33001040 56 Khoo B. Tan T. Clarke S.A. Mills E.G. Patel B. Modi M. Thyroid function before, during, and after COVID-19 J Clin Endocrinol Metabol. 106 2021 e803 e811 57 Higgins V. Sohaei D. Diamandis E.P. Prassas I. COVID-19: from an acute to chronic disease? Potential long-term health consequences. Crit Rev Clin Lab Sci. 58 2021 297 310 33347790 58 Muñoz-Rodríguez J.R. Gómez-Romero F.J. Pérez-Ortiz J.M. López-Juárez P. Santiago J.L. Serrano-Oviedo L. Characteristics and risk factors associated with mortality in a multicenter spanish cohort of patients with COVID-19 pneumonia Arch Bronconeumol. 57 2021 34 41 34629641 59 Signes-Costa J. Núñez-Gil I.J. Soriano J.B. Arroyo-Espliguero R. Eid C.M. Romero R. Prevalence and 30-day mortality in hospitalized patients with COVID-19 and prior lung diseases Arch Bronconeumol. 57 2021 13 20 60 Posso M. Comas M. Román M. Domingo L. Louro J. González C. Comorbidities and mortality in patients with COVID-19 aged 60 years and older in a university hospital in Spain Arch Bronconeumol (Engl Ed). 56 2020 756 758 61 Tessitore E. Handgraaf S. Poncet A. Achard M. Höfer S. Carballo S. Symptoms and quality of life at 1-year follow up of patients discharged after an acute COVID-19 episode Swiss Med Week. 151 2021 w30093 62 Seeßle J. Waterboer T. Hippchen T. Simon J. Kirchner M. Lim A. Persistent symptoms in adult patients one year after COVID-19: a prospective cohort study Clin Infect Dis. ciab611 2021 Jul 5 10.1093/cid/ciab611 Online ahead of print 63 Caruso D. Guido G. Zerunian M. Polidori T. Lucertini E. Pucciarelli F. Post-acute sequelae of COVID-19 pneumonia: six-month chest CT follow-up Radiology. 301 2021 E396 E405 34313468 64 Fang X. Ming C. Cen Y. Lin H. Zhan K. Yang S. Post-sequelae one year after hospital discharge among older COVID-19 patients: A multi-center prospective cohort study J Infect. 84 2022 179 186 34902448 65 Fumagalli C. Zocchi C. Tassetti L. Silverii M.V. Amato C. Livi L. Factors associated with persistence of symptoms 1 year after COVID-19: A longitudinal, prospective phone-based interview follow-up cohort study Eur J Intern Med 97 2022 36 41 34903448 66 Heightman M. Prashar J. Hillman T.E. Marks M. Livingston R. Ridsdale H.A. Post-COVID-19 assessment in a specialist clinical service: a 12-month, single-centre, prospective study in 1325 individuals BMJ Open Respir Res. 2021 8 67 Zhang X. Wang F. Shen Y. Zhang X. Cen Y. Wang B. Symptoms and health outcomes among survivors of COVID-19 infection 1 year after discharge from hospitals in Wuhan China. JAMA Netw Open. 4 2021 e2127403 34586367 68 Crook H. Raza S. Nowell J. Young M. Edison P. Long covid-mechanisms, risk factors, and management BMJ. 374 2021 n1648 34312178 69 Karaarslan F. Güneri F.D. Kardeş S. Long COVID: rheumatologic/musculoskeletal symptoms in hospitalized COVID-19 survivors at 3 and 6 months Clin Rheumatol. 41 2022 289 296 34713356 70 Taquet M. Dercon Q. Luciano S. Geddes J.R. Husain M. Harrison P.J. Incidence, co-occurrence, and evolution of long-COVID features: A 6-month retrospective cohort study of 273,618 survivors of COVID-19 PLoS Med. 18 2021 e1003773 34582441 71 Fernández-de-Las-Peñas C. Rodríguez-Jiménez J. Fuensalida-Novo S. Palacios-Ceña M. Gómez-Mayordomo V. Florencio L.L. Myalgia as a symptom at hospital admission by severe acute respiratory syndrome coronavirus 2 infection is associated with persistent musculoskeletal pain as long-term post-COVID sequelae: a case-control study Pain. 162 2021 2832 2840 33863864 72 Maestre-Muñiz M.M. Arias Á. Mata-Vázquez E. Martín-Toledano M. López-Larramona G. Ruiz-Chicote A.M. Long-term outcomes of patients with coronavirus disease 2019 at one year after hospital discharge J Clin Med. 2021 10 35011750 73 Mattioli F. Piva S. Stampatori C. Righetti F. Mega I. Peli E. Neurologic and cognitive sequelae after SARS-CoV2 infection: Different impairment for ICU patients J Neurol Sci. 432 2022 120061 34894422 74 Pilotto A. Cristillo V. Cotti Piccinelli S. Zoppi N. Bonzi G. Sattin D. Long-term neurological manifestations of COVID-19: prevalence and predictive factors Neurol Sci. 42 2021 4903 4907 34523082 75 Tleyjeh I.M. Saddik B. AlSwaidan N. AlAnazi A. Ramakrishnan R.K. Alhazmi D. Prevalence and predictors of post-acute COVID-19 syndrome (PACS) after hospital discharge: A cohort study with 4 months median follow-up PloS One. 16 2021 e0260568 34874962 76 Mazza M.G. Palladini M. De Lorenzo R. Bravi B. Poletti S. Furlan R. One-year mental health outcomes in a cohort of COVID-19 survivors J Psych Res. 145 2021 118 124 77 Zhou F. Tao M. Shang L. Liu Y. Pan G. Jin Y. Assessment of sequelae of COVID-19 nearly 1 year after diagnosis Front Med. 8 2021 717194 78 González J. Benítez I.D. Carmona P. Santisteve S. Monge A. Moncusí-Moix A. Pulmonary function and radiologic features in survivors of critical COVID-19: a 3-month prospective cohort Chest. 160 2021 187 198 33676998 79 Wu X. Liu X. Zhou Y. Yu H. Li R. Zhan Q. 3-month, 6-month, 9-month, and 12-month respiratory outcomes in patients following COVID-19-related hospitalisation: a prospective study Lancet Respir Med. 9 2021 747 754 33964245 80 Sanchez-Ramirez D.C. Normand K. Zhaoyun Y. Torres-Castro R. Long-term impact of COVID-19: a systematic review of the literature and meta-analysis Biomedicines. 2021 9 35052688 81 Xu B. Ma F.Q. He C. Wu Z.Q. Fan C.Y. Mao H.R. Incidence and affecting factors of pulmonary diffusing capacity impairment with COVID-19 survivors 18 months after discharge in Wuhan China. J Infect 84 2022 e16 e18 82 Montefusco L. Ben Nasr M. D’Addio F. Loretelli C. Rossi A. Pastore I. Acute and long-term disruption of glycometabolic control after SARS-CoV-2 infection Nat Metab. 3 2021 774 785 34035524 83 Horwitz L.I. Garry K. Prete A.M. Sharma S. Mendoza F. Kahan T. Six-month outcomes in patients hospitalized with severe COVID-19 J Gen Intern Med. 36 2021 3772 3777 34355349 84 Fernández-de-Las-Peñas C. Palacios-Ceña D. Gómez-Mayordomo V. Palacios-Ceña M. Rodríguez-Jiménez J. de-la-Llave-Rincón A.I. Fatigue and dyspnoea as main persistent post-COVID-19 symptoms in previously hospitalized patients: related functional limitations and disability Respiration. 101 2022 132 141 34569550 85 Xiong L. Li Q. Cao X. Xiong H. Huang M. Yang F. Dynamic changes of functional fitness, antibodies to SARS-CoV-2 and immunological indicators within 1 year after discharge in Chinese health care workers with severe COVID-19: a cohort study BMC Med. 19 2021 163 34256745 86 Guillon A. Laurent E. Godillon L. Kimmoun A. Grammatico-Guillon L. Long-term mortality of elderly patients after intensive care unit admission for COVID-19 Intensive Care Med. 47 2021 710 712 33844045 87 Munch M.W. Granholm A. Nørregaard K.M. Aksnes T.S. Sølling C.G. Christensen S. Perner A. Long-term mortality and health-related quality of life in the COVID STEROID trial Acta Anaesthesiol Scand. 2022 Jan 24 10.1111/aas.14029 88 Di Bari M. Tonarelli F. Balzi D. Giordano A. Ungar A. Baldasseroni S. COVID-19, vulnerability, and long-term mortality in hospitalized and nonhospitalized older persons J Am Med Dir Assoc. 23 2022 414 420 34990587 89 Qin W. Chen S. Zhang Y. Dong F. Zhang Z. Hu B. Diffusion capacity abnormalities for carbon monoxide in patients with COVID-19 at 3-month follow-up Eur Respir J. 2021 58 90 Cicco S. Vacca A. Cittadini A. Marra A.M. Long-term follow-up may be useful in coronavirus disease 2019 survivors to prevent chronic complications Infect Chemother. 52 2020 407 409 32757498 91 Sun J. Liu J. Li H. Shang C. Li T. Ji W. Pulmonary rehabilitation focusing on the regulation of respiratory movement can improve prognosis of severe patients with COVID-19 Ann Pall Med. 10 2021 4262 4272 92 Gloeckl R. Leitl D. Jarosch I. Schneeberger T. Nell C. Stenzel N. Pulmonary rehabilitation in long COVID: more than just natural recovery!? ERJ Open Res. 2021 7 93 He Z. Ren L. Yang J. Guo L. Feng L. Ma C. Seroprevalence and humoral immune durability of anti-SARS-CoV-2 antibodies in Wuhan China: a longitudinal, population-level, cross-sectional study. Lancet. 397 2021 1075 1084 94 To K.K. Hung I.F. Ip J.D. Chu A.W. Chan W.M. Tam A.R. Coronavirus Disease 2019 (COVID-19) re-infection by a phylogenetically distinct severe acute respiratory syndrome coronavirus 2 strain confirmed by whole genome sequencing Clin Infect Dis. 73 2021 e2946 e2951 32840608 95 Maier H.E. Kuan G. Saborio S. Bustos Carrillo F.A. Plazaola M. Clinical spectrum of SARS-CoV-2 infection and protection from symptomatic re-infection Clin Infect Dis. ciab717 2021 Aug 19 10.1093/cid/ciab717 Online ahead of print 96 Tan C.W. Chia W.N. Young B.E. Zhu F. Lim B.L. Sia W.R. Pan-sarbecovirus neutralizing antibodies in BNT162b2-immunized SARS-CoV-1 survivors N Engl J Med. 385 2021 1401 1406 34407341 97 Killerby M.E. Ata Ur Rasheed M. Tamin A. Harcourt J.L. Abedi G.R. Shedding of culturable virus, seroconversion, and 6-month follow-up antibody responses in the first 14 confirmed cases of coronavirus disease 2019 in the United States J Infect Dis. 224 2021 771 776 33693830 98 Wu X. Wang Z. He Z. Li Y. Wu Y. Wang H. A follow-up study shows that recovered patients with re-positive PCR test in Wuhan may not be infectious BMC Med. 19 2021 77 33715626 99 Buttery S. Philip K.E.J. Williams P. Fallas A. West B. Cumella A. Patient symptoms and experience following COVID-19: results from a UK-wide survey BMJ Open Respir Res. 2021 8 100 Eloy P. Tardivon C. Martin-Blondel G. Isnard M. Turnier P.L. Marechal M.L. Severity of self-reported symptoms and psychological burden 6-months after hospital admission for COVID-19: a prospective cohort study Int J Infect Dis. 112 2021 247 253 34517049 101 Hossain M.A. Hossain K.M.A. Saunders K. Uddin Z. Walton L.M. Raigangar V. Prevalence of long COVID symptoms in Bangladesh: a prospective Inception Cohort Study of COVID-19 survivors BMJ Glob Health. 2021 6 102 Nesan G. Keerthana D. Yamini R. Jain T. Kumar D. Eashwer A. 3-month symptom-based ambidirectional follow-up study among recovered COVID-19 patients from a tertiary care hospital using telehealth in Chennai India. Inquiry. 58 2021 469580211060165 103 Xiong Q. Xu M. Li J. Liu Y. Zhang J. Xu Y. Clinical sequelae of COVID-19 survivors in Wuhan China: a single-centre longitudinal study Clin Microbiol Infect. 27 2021 89 95 32979574 104 Bottemanne H. Gouraud C. Hulot J.S. Blanchard A. Ranque B. Lahlou-Laforêt K. Do anxiety and depression predict persistent physical symptoms after a severe COVID-19 episode?. A prospective study Front Psych. 12 2021 757685 105 Augustin M. Schommers P. Stecher M. Dewald F. Gieselmann L. Gruell H. Post-COVID syndrome in non-hospitalised patients with COVID-19: a longitudinal prospective cohort study Lancet Reg Health Eur. 6 2021 100122 34027514 106 Fernández-de-Las-Peñas C. Palacios-Ceña D. Gómez-Mayordomo V. Rodríguez-Jiménez J. Palacios-Ceña M. Velasco-Arribas M. Long-term post-COVID symptoms and associated risk factors in previously hospitalized patients: A multicenter study J Infect. 83 2021 237 279 107 Chen X. Li Y. Shao T.R. Yang L.L. Li S.J. Wang X.J. Some characteristics of clinical sequelae of COVID-19 survivors from Wuhan China: A multi-center longitudinal study. Influ Other Respir Virus. 19 2021 Nov 10.1111/irv.12943. doi: 10.1111/irv.12943. Online ahead of print. 108 Willi S. Lüthold R. Hunt A. Hänggi N.V. Sejdiu D. Scaff C. COVID-19 sequelae in adults aged less than 50 years: A systematic review Travel Med Infect Dis. 40 2021 101995 33631340
PMC009xxxxxx/PMC9005222.txt	==== Front J Clin Virol J Clin Virol Journal of Clinical Virology 1386-6532 1873-5967 Elsevier B.V. S1386-6532(22)00091-9 10.1016/j.jcv.2022.105158 105158 Article Comparison between an in-house SARS-CoV-2 ELISpot and the T-Spot® Discovery SARS-CoV-2 for the assessment of T cell responses in prior SARS-CoV-2-infected individuals Mak Willem A. a⁎ Koeleman Johannes G.M. a Ong David S.Y. ab a Department of Medical Microbiology and Infection Control, Franciscus Gasthuis en Vlietland, Rotterdam, the Netherlands b Department of Epidemiology, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, the Netherlands ⁎ Corresponding author. 12 4 2022 6 2022 12 4 2022 150 105158105158 30 1 2022 28 3 2022 © 2022 Elsevier B.V. All rights reserved. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcAdaptive immune responses induced by SARS-CoV-2 vaccination or infection decrease over time and, especially neutralizing antibodies, are less effective against new SARS-CoV-2 variants of concern. [1], [2], [3], [4], [5] As immune protection is also determined by the cellular immunity, it is crucial to have T cell assays that can adequately assess SARS-CoV-2-specific T cell responses. In our previous work, we assessed SARS-CoV-2 antigen-specific T cell responses after SARS-CoV-2 infection and vaccination using the T-Spot® Discovery SARS-CoV-2 kit from Oxford Immunotec. [6] Although the T-Spot® Discovery is a commercial kit that has been extensively used in clinical studies, this kit might not be economically feasible for all laboratories and adaptations to the assay are not possible, such as using different antigen peptide pools containing virus mutations. Therefore, we also developed an in-house SARS-CoV-2 enzyme-linked immunospot (ELISpot) assay. [7] In the current analysis, we compared the performances of our in-house SARS-CoV-2 ELISpot to the T-Spot® Discovery SARS-CoV-2. Both assays detect SARS-CoV-2 antigen-specific interferon-gamma (IFN-γ)-secreting T cells, which include predominantly CD4+ T helper type 1 (Th1) cells and CD8+ cytotoxic T cells that are crucial for an anti-virus immune response. [8], [9], [10] Also, peripheral blood mononuclear cells (PBMCs) are stimulated with overlapping peptide pools of SARS-CoV-2 spike subunit 1 (S1), nucleocapsid protein (N), and membrane protein (M) in these assays. The T-Spot® Discovery excludes S1, N, and M peptides that are homologous to endemic coronaviruses. [6] In contrast, our in-house ELISpot does not exclude sequences homologous to endemic coronaviruses. [7] We determined T cell responses of 90 blood samples collected from 55 healthcare workers who tested SARS-CoV-2 positive 12 months before the first blood collection (Fig. 1 ). As described in our previous study, blood was collected either before (n = 32) or after (n = 58) receiving the first and second COVID-19 vaccination. [6] T cell responses were not statistically different between both assays after S1 and N stimulation but was significantly higher in the in-house ELISpot after M stimulation (P = 0.04). Furthermore, we found a strong association between S1 responses (r = 0.85) but only moderate associations of N and M responses between both assays (r = 0.43 and r = 0.56, respectively). [11] Fig. 1 Comparison between the in-house SARS-CoV-2 ELISpot and T-Spot® Discovery SARS-CoV-2. A total of 90 samples were included from our study cohort (n = 55) that tested SARS-CoV-2 positive 12 months before the first blood collection. The healthcare workers provided a blood sample once (n = 30), twice (n = 15), or thrice (n = 10), of which 32 samples were collected before and 58 samples were collected median 18 (IQR 14–69) days after receiving the first or second COVID-19 vaccination. PBMCs were stimulated for 16–20 h with SARS-CoV-2 antigens in both assays. (A) Total magnitude of IFN-γ responses to tested antigens of in-house ELISpot (red) and T-Spot® Discovery (blue). Datasets were compared with a Mann-Whitney U test. (B) Associations between antigen-specific responses assessed by Spearman's rank correlation coefficient (r). Fig. 1 SARS-CoV-2 S1 is considered the most prominent target for achieving protective immunity and is thus solely integrated in most SARS-CoV-2 vaccines. [12, 13] Therefore, evaluating the T cell response against S1 is most valuable. S1 responses between both assays were strongly associated. Unlike S1, the N and M proteins of SARS-CoV-2 are highly homologous to proteins in endemic coronaviruses. [14, 15] The observed larger inter-assay differences in N and M responses might be attributable to considerable homologous sequences being removed from the T-Spot® Discovery N and M peptide pools, whereas for the in-house assay no homologous sequences were removed. In conclusion, we showed that our in-house ELISpot assay was highly correlated with the commercially available T-Spot® Discovery for the assessment of T cell responses against SARS-CoV-2 S1. A great advantage of using an in-house ELISpot is the possibility to easily adapt the S1 peptide pools to more accurately assess specific T cell responses against current circulating viruses (e.g., Omicron (B.1.1.529) variant) and future SARS-CoV-2 variants of concern. Funding Franciscus Gasthuis and Vlietland, Rotterdam, the Netherlands ==== Refs References 1 Patricia Almendro-Vázquez Rocio Laguna-Goya Maria Ruiz-Ruigomez Alberto Utrero-Rico Antonio Lalueza de la Calle Guillermo Maestro Longitudinal dynamics of SARS-CoV-2-specific cellular and humoral immunity after natural infection or BNT162b2 vaccination PLoS Pathog. 17 12 2021 1 23 10.1371/journal.ppat.1010211 2 Sandile Cele Laurelle Jackson S Khoury David Khadija Khan David Khoury Thandeka Moyo-Gwete SARS-CoV-2 Omicron has extensive but incomplete escape of Pfizer BNT162b2 elicited neutralization and requires ACE2 for infection MedRxiv 2021 3 Vivek Naranbhai Anusha Nathan Clarety Kaseke Cristhian Berrios Shawn Choi Getz Matthew A. T cell reactivity to the SARS-CoV-2 Omicron variant is preserved in most but not all prior infected and vaccinated individuals MedRxiv 2022 4 Jordan Stanley C. Bong-Ha Shin Gadsden Terry-Ann M. Maggie Chu Anna Petrosyan Le Catherine N. T cell immune responses to SARS-CoV-2 and variants of concern (Alpha and Delta) in infected and vaccinated individuals Cell Mol. Immunol. 2021 1 3 10.1038/s41423-021-00767-9 5 Delphine Planas David Veyer Artem Baidaliuk Isabelle Staropoli Florence Guivel-Benhassine Michael Rajah Maaran Reduced sensitivity of SARS-CoV-2 variant Delta to antibody neutralization Nature 596 2021 276 280 10.1038/s41586-021-03777-9 34237773 6 Mak Willem A. Koeleman Johannes G.M. van der Vliet Marijke Keuren Frans Ong David S.Y. SARS-CoV-2 antibody and T cell responses one year after COVID-19 and the booster effect of vaccination: a prospective cohort study J. Infect. 84 2022 171 178 10.1016/j.jinf.2021.12.003 34896516 7 Mak Willem A. Koeleman Johannes G.M. Ong David S.Y. Development of an in-house SARS-CoV-2 interferon-gamma ELISpot and plate reader-free spot detection method J. Virol. Methods 300 114398 2022 1 6 10.1016/j.jviromet.2021.114398 8 Ong David S.Y. Fragkou Paraskevi C. Schweitzer Valentijn A. Chemaly Roy F. Moschopoulos Charalampos D. Chrysanthi Skevaki How to interpret and use COVID-19 serology and immunology tests Clin. Microbiol. Infect. 27 7 2021 981 986 10.1016/j.cmi.2021.05.001 33975005 9 Calarota Sandra A. Fausto Baldanti Enumeration and characterization of human memory t cells by enzyme-linked immunospot assays Clin. Dev. Immunol. 2013 1 8 10.1155/2013/637649 10 Alba Grifoni Daniela Weiskopf Ramirez Sydney I. Jose Mateus Dan Jennifer M. Rydyznski Moderbacher Carolyn Targets of T Cell Responses to SARS-CoV-2 Coronavirus in Humans with COVID-19 Disease and Unexposed Individuals Cell 181 7 2020 1489 1501 10.1016/j.cell.2020.05.015 e15 32473127 11 Patrick Schober Boer Christa Schwarte Lothar A. Correlation coefficients: appropriate use and interpretation Anesth. Analg. 126 5 2018 1763 1768 10.1213/ANE.0000000000002864 29481436 12 Thibault Fiolet Yousra Kherabi Conor-James MacDonald Jade Ghosn Nathan Peiffer-Smadja Comparing COVID-19 vaccines for their characteristics, efficacy and effectiveness against SARS-CoV-2 and variants of concern: a narrative review Clin. Microbiol. Infect. 28 2 2021 202 221 10.1016/j.cmi.2021.10.005 34715347 13 Antonio Bertoletti T Tan Anthony Le Bert Nina The T-cell response to SARS-CoV-2: kinetic and quantitative aspects and the case for their protective role Oxford Open Immunol. 2 1 2021 1 9 10.1093/oxfimm/iqab006 14 Mei Yang Suhua He Xiaoxue Chen Zhaoxia Huang Ziliang Zhou Zhechong Zhou Structural Insight Into the SARS-CoV-2 Nucleocapsid Protein C-Terminal Domain Reveals a Novel Recognition Mechanism for Viral Transcriptional Regulatory Sequences Front. Chem. 8 624765 2021 1 12 10.3389/fchem.2020.624765 15 Rhia Kundu Sam Narean Janakan Lulu Wang Joseph Fenn Timesh Pillay Nieves Derqui Fernandez Cross-reactive memory T cells associate with protection against SARS-CoV-2 infection in COVID-19 contacts Nat. Commun. 13 1 2022 1 8 10.1038/s41467-021-27674-x 34983933
PMC009xxxxxx/PMC9005223.txt	==== Front Inform Med Unlocked Inform Med Unlocked Informatics in Medicine Unlocked 2352-9148 The Authors. Published by Elsevier Ltd. S2352-9148(22)00092-2 10.1016/j.imu.2022.100945 100945 Article Challenges of deep learning methods for COVID-19 detection using public datasets Hasan Md. Kamrul a⁎ Alam Md. Ashraful a Dahal Lavsen b Roy Shidhartho a Wahid Sifat Redwan a Elahi Md. Toufick E. a Martí Robert c Khanal Bishesh b a Department of Electrical and Electronic Engineering (EEE), Khulna University of Engineering & Technology (KUET), Khulna 9203, Bangladesh b Nepal Applied Mathematics and Informatics Institute for Research (NAAMII), Nepal c Computer Vision and Robotics Institute, University of Girona, Spain ⁎ Correspondence to: Department of EEE, KUET, Khulna 9203, Bangladesh. 12 4 2022 2022 12 4 2022 30 100945100945 12 2 2022 3 4 2022 4 4 2022 © 2022 The Authors 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Since the COVID-19 pandemic, several research studies have proposed Deep Learning (DL)-based automated COVID-19 detection, reporting high cross-validation accuracy when classifying COVID-19 patients from normal or other common Pneumonia. Although the reported outcomes are very high in most cases, these results were obtained without an independent test set from a separate data source(s). DL models are likely to overfit training data distribution when independent test sets are not utilized or are prone to learn dataset-specific artifacts rather than the actual disease characteristics and underlying pathology. This study aims to assess the promise of such DL methods and datasets by investigating the key challenges and issues by examining the compositions of the available public image datasets and designing different experimental setups. A convolutional neural network-based network, called CVR-Net (COVID-19 Recognition Network), has been proposed for conducting comprehensive experiments to validate our hypothesis. The presented end-to-end CVR-Net is a multi-scale-multi-encoder ensemble model that aggregates the outputs from two different encoders and their different scales to convey the final prediction probability. Three different classification tasks, such as 2-, 3-, 4-classes, are designed where the train–test datasets are from the single, multiple, and independent sources. The obtained binary classification accuracy is 99.8% for a single train–test data source, where the accuracies fall to 98.4% and 88.7% when multiple and independent train–test data sources are utilized. Similar outcomes are noticed in multi-class categorization tasks for single, multiple, and independent data sources, highlighting the challenges in developing DL models with the existing public datasets without an independent test set from a separate dataset. Such a result concludes a requirement for a better-designed dataset for developing DL tools applicable in actual clinical settings. The dataset should have an independent test set; for a single machine or hospital source, have a more balanced set of images for all the prediction classes; and have a balanced dataset from several hospitals and demography. Our source codes and model are publicly available1 for the research community for further improvements. Keywords COVID-19 disease Chest computed tomography and X-ray Convolutional neural networks Ensemble classifier ==== Body pmc1 Introduction Pneumonia of unknown cause detected in Wuhan, China, was reported to the World Health Organization (WHO) office in China on 31st December 2019. This was subsequently named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) on 11th February 2020, as the virus causing the disease is genetically related to the coronavirus responsible for the SARS outbreak of 2003. The new disease was referred to as “COVID-19” by WHO on 11th February 2020 [1]. As of August 2020, the outbreak of 2019 in Wuhan (China), has extended worldwide with 386,548,962 confirmed COVID-19 cases including 5,705,754 deaths in last 2 years (5 February 2022) [2], as presented in Fig. 1. The clinical attributes of severe COVID-19 epidemic are bronchopneumonia that causes cough, fever, dyspnea, and subtle respiratory anxiety ailment [3], [4], [5]. The clinical screening test for COVID-19 is Reverse Transcription Polymerase Chain Reaction (RT-PCR) using respiratory specimens. However, this test is a manual, complicated, tedious, and time-consuming procedure with an estimated true-positive rate of 63.0% [6]. There is a significant lack of inventory of RT-PCR kits, leading to a delay in efforts to prevent and cure coronavirus disease [7]. Furthermore, the RT-PCR kit is estimated to cost around 120∼130 USD and requires a specially designed biosafety laboratory to house the PCR unit, each of which can cost 15,000∼90,000 USD [8]. Nevertheless, the utilization of a costly screening device with delayed test results makes it more challenging to suppress the spread of the disease. However, it is observed that most of the COVID-19 cases have common characteristics on radiographic images, such as Computed Tomography (CT) and Chest X-ray (CXR), including bilateral, multi-focal, ground-glass opacities with a peripheral or posterior distribution, mainly in the lower lobes and early- and late-stage pulmonary consolidation [10], [11], [12], [13]. Those features can be utilized to develop a sensitive Computer-aided Diagnosis (CAD) tool to detect COVID-19 Pneumonia and be considered as a screening tool [14]. Currently, deep Convolutional Neural Networks (CNNs) allow for building an end-to-end model, without the need for manual feature extraction [15], [16], which have demonstrated tremendous success in many domains of medical imaging, such as arrhythmia detection [17], [18], [19], skin lesion segmentation and classification [20], [21], [22], [23], [24], breast cancer detection [25], [26], [27], brain disease classification [28], pneumonia detection from CXR images [29], fundus image segmentation [30], [31], minimally invasive surgery [32] and lung segmentation [33]. Several deep CNN-based methods have been published to detect COVID-19 from CXR and CT images. Though the results obtained are promising, they exhibit limited scope as a CAD tool. Most of the works, especially on CXR images, have been based on data from different sources for two different classes (COVID vs. Normal). This brings inherent bias on the algorithms as the model tends to learn the distribution and artifacts of the data source for binary classification problems. Therefore, these models perform very poorly when used in practical settings where the model has to adapt to data from different domains. To accelerate the development of DL tools that could be utilized in realistic clinical settings, the scientific community needs to emphasize more on making publicly systematically-designed and documented datasets that have information, such as inclusion and exclusion criteria, symptomatic vs. asymptomatic cases, and the disease severity stage at which these images were taken. In this work, we design various experiments with a proposed CNN-based COVID-19 detection method to justify this proposition.Fig. 1 A world heat map of the corona pandemic per capita [9] [Accessed on 25 December 2021]. The rest of the paper is structured as follows: Section 2 reviews the earlier published literature for COVID-19 detection, and Section 3 highlights the significant contributions to this article. We explain the proposed framework for the recognition of COVID-19 and datasets in Section 4. The results and different experiments are reported in Section 5. We interpret the obtained results from the proposed CVR-Net in Section 6. Finally, Section 7 concludes the article with future working directions. 2 Review of literature Different CNN architectures have already been proposed for COVID-19 detection as a binary (COVID vs. No-Findings) or multi-class (COVID vs. No-Findings vs. Pneumonia) problem [34], [35], [36]. Ghoshal and Tucker [37] investigated uncertainty of the COVID-19 classification report, using a drop-weights-based Bayesian CNN, as the availability of uncertainty-aware DL can ensure more extensive adoption of DL in clinical applications. Abbas et al. [38] proposed a framework by adopting a deep CNN, called Decompose, Transfer, and Compose (DeTraC) [39] for the classification of COVID-19 CXR images, where the authors implemented the DeTraC in two phases. Firstly, using gradient descent optimization, they trained the backbone pre-trained CNN model of DeTraC to extract deep local features from each image. Secondly, they used the class-composition layer of DeTraC to refine the final classification of the images. Zhao et al. [40] developed diagnosis methods based on multi-task learning and self-supervised learning, where the authors proposed an open-source COVID-19 dataset of CT images with a binary class (COVID and Non-COVID). For the classification task, they trained DenseNet-169 and ResNet-50, via a pre-trained model on ImageNet [16] weights, with their newly proposed dataset. Afshar et al. [41] proposed a CNN model named COVID-CAPS, which was based on the Capsule Networks (CapsNets) for handling the small datasets of COVID-19. CapsNets are alternative models of CNN, which are capable of capturing spatial information using routing by agreement. Capsules try to reach a mutual agreement on the existence of the objects. Their proposed COVID-CAPS model had 4 convolutional layers and 3 capsule layers, where batch normalization [42] followed the former layers. The authors fine-tuned all the capsule layers, while the conventional layers were frozen with pre-trained weights of ImageNet. He et al. [43] built a COVID-19 CT dataset, called China Consortium of Chest CT Image Investigation (CC-CCII), with three classes: novel coronavirus Pneumonia, common Pneumonia, and healthy controls. The authors trained 3D DenseNet3D-121 on their proposed CC-CCII dataset, and they experimentally validated that 3D CNNs outperform 2D CNNs in general. Singh et al. [13] implemented a CNN-based model named multi-objective differential evolution-based CNN for the classification of COVID-19. They fine-tuned the parameters of the CNN model using a multi-objective fitness function. The differential evolution algorithm was used to optimize the multi-objective fitness function. The model was optimized iteratively using mutation, crossover, and selection operation to determine the best available solution in differential evolution. Farooq and Hafeez [44] employed ResNet-50 using transfer learning with progressively resizing [45] the input images to 128 × 128 × 3, 224 × 224 × 3, and 229 × 229 × 3 pixels, where the authors also fine-tuned the network at each stage. Ozkaya et al. [46] extracted deep features using VGG-16, GoogleNet [47], and ResNet-50 models, which were classified by Support Vector Machine (SVM) [48] with linear kernel function. They also applied the modified T-test [49], a feature ranking algorithm, to select the features [50] for avoiding overfitting. Rajaraman et al. [51] evaluated ImageNet pre-trained CNN models such as VGG-16, VGG-19, InceptionV3, Xception, Inception-ResNetV2, MobileNetV2, DenseNet-201, and NasNet-mobile [52]. Then, they optimized the hyperparameters of the CNNs using a randomized grid search method [53]. In the end, the authors proposed an ensemble of those CNN models for the final COVID-19 recognition. Toğaçar et al. [54] restructured the data classes using a fuzzy color technique, where they stacked a structured image with the original images. The authors trained MobileNetV2 and SqueezeNet to extract the deep features, which were then processed using the social mimic optimization method [55]. After that, selected features were combined and classified using the SVM to recognize COVID-19. Khan et al. [56] developed a 15-layered CNN architecture for extracting deep features from two different layers like global average pool and fully connected layers, which were then merged employing the max-layer detail approach. The most discriminant features from the pool of features were selected using a Correntropy technique, and a one-class kernel extreme learning machine classifier was applied for the classification. CNN-based models like ResNet50, ResNet101, ResNet152, InceptionV3, and Inception-ResNetV2 were proposed and implemented by Narin et al. [57] for the detection of COVID-19-infected patient using CXR radiographs. Sedik et al. [58] classified CT and CXR images of COVID-19 vs. normal using CNN and convolutional long short-term memory (ConvLSTM) based models. Sanida et al. [59] employed lightweight modified MobileNetV2 to classify the COVID-19, normal, viral Pneumonia, and lung opacity images for the real-time operations in a low-power embedded system. Authors in [60] proposed a COVIDetectioNet using a pre-trained AlexNet to extract the deep features. The useful features were selected using the Relief algorithm from all layers of the architecture were then classified using the SVM approach. An efficient Grayscale Spatial Exploitation Net (GSEN) is designed by employing web pages crawling across cloud computing environments in [61], utilizing the accuracy rates improvement in a positive relationship to the cardinality of crawled CXR dataset. Their model consists of four convolutional blocks where each is composed of a single convolutional, batch normalization, ReLU activation function, and max-pooling layer. Monday et al. [62] proposed a neurowavelet capsule network. Firstly, they presented a multi-resolution analysis of a discrete wavelet transform to filter noisy and incompatible information from the CXR data to enhance the feature extraction robustness of the network. Secondly, the discrete wavelet transform of the multi-resolution analysis was also conducted a sub-sampling procedure to minimize the loss of spatial details, thereby improving the overall classification performance. Sakthivel et al. [63] proposed an ensemble-based CNN model where five DL models like ResNet, FitNet, IRCNN, MobileNet, and EfficientNet are ensembled and fine-tuned to classify the CXR images. An application-specific hardware architecture had been incorporated by carefully exploiting the data flow and resource availability. 3 Our contributions Many DL-based Artificial Intelligence (AI) algorithms have been proposed in the past year to automatically classify COVID-19 cases from normal and other Pneumonia cases. These published works reported high COVID-19 binary classification accuracy using either CT scans or CXRs [13], [34], [35], [45], [54], [64], [65], [66], [67], [68], [69], [70], [71]. Although the reported metrics, such as sensitivity and/or specificity, are very high in most cases, these results are obtained on cross-validation studies without an independent test set coming from a separate dataset having biases, such as the two classes predicted from two unique datasets. AI models are likely to overfit training data distribution when independent test sets are not used or are prone to learn dataset-specific artifacts rather than the actual disease characteristics. Additionally, the publicly available datasets for COVID-19 classification used in the recent studies have class and dataset source biases, resulting in AI models learning dataset-specific distributions rather than the underlying pathology. Many recent studies proposing COVID-19 classification based on DL using imaging data do not emphasize the importance of avoiding overfitting and having an independent test set with images from a separate dataset than the training and validation dataset. However, the critical contributions in this article are pointed out as follows: • Proposing an end-to-end and multi-scale-multi-encoder CVR-Net, aggregating the outputs from two different encoders and their different scales to obtain the final prediction probability. • Designing various experiments to investigate the issues of overfitting and biasing; exploring the limitations of existing large public datasets that have been widely used for developing and evaluating COVID-19 detection algorithms in the past years. • Validating multi-class classification models to distinguish various Pneumonia types, including COVID-19, requires a balanced set of images for all the prediction classes coming from a single site and demography and having several balanced sets coming from separate scanners or hospitals and demography. • Comparing the proposed architecture to other state-of-the-art methods using an independent test set for evaluation, where some of the identified bias and overfitting issues are minimized. 4 Materials and methods This section presents the materials and methods for conducting this research. Section 4.1 briefly describes utilized datasets. The designing of the proposed network (CVR-Net) is explained in Section 4.2. Finally, Section 4.3 describes the training protocol of our network and the evaluation metrics. 4.1 Datasets This section illustrates the experimental setup for various classification tasks utilizing chest CT scans or CXRs from several publicly available datasets. The classes applied for different experimentations are taken from the following set: • NOR: Normal; no Pneumonia and COVID-19 negative • CVP: COVID-19 positive Pneumonia • OVP: Other Viral Pneumonia; Viral Pneumonia but not COVID-19 • OBP: Other Bacterial Pneumonia; Bacteria induced Non-COVID Pneumonia • NCP: Non-COVID Pneumonia; OBP + OVP • NCV: Non-COVID; NOR + NCP Table 1 demonstrates the details of the experimental setup with various tasks and how various datasets are combined for these tasks. Three different types of classification tasks are designed: NCV vs. CVP (2-classes, CL2); NOR vs. NCP vs. CVP (3-classes, CL3); and NOR vs. OBP vs. OVP vs. CVP (4-classes, CL4). Several different combinations of the publicly available datasets are utilized for chest CT scans (labeled CT) and for chest X-rays (labeled CXR) [40], [71], [72], [73], [74], [75], [76], [77], [78], [79]. For each binary (CL2) or multi-class (CL3/CL4) classification task, we design experiments to study the impact of having single separate vs. multiple mixed sources of data for individual classes during training, labeled Single and Multiple, respectively. The setup where the test set contains images from an independent source whose images are never used during training and validation is labeled as Independent. For adding diversity in each class of CXR-Multiple-CL2, we include images from more datasets: CXR images from ChestX-ray8 [74] to NCV, and from CCXRI [75] and PadChest [76] to CVP. To evaluate the ability to distinguish various Pneumonia types, we design CXR-Multiple-CL3 and CXR-Multiple-CL4 having the same number of images as CXR-Multiple-CL2, but the NCV class split into individual Pneumonia types.Table 1 Various classification tasks utilizing CT scans or CXRs in different combinations from publicly available datasets. Different studiesa Class categories # of images Data source references Modality Utilization CXR-Single-CL2 NCV 5,856 CXRI [72] X-ray [34], [35], [65] CVP 500 CIDC [73] CXR-Multiple-CL2 NCV 7,864 CXRIL [72], ChestX-ray8 [74] X-ray Proposed CVP 4,015 CCXRIL [75], CIDC [73], PadChest [76] CXR-Independent-CL2 NCV (Train/Test) 6,958/1,227 CheXpert [77]＋ CXRI [72]/ChestX-ray8 [74] X-ray Proposed CVP (Train/Test) 3,515/500 CCXRI [75]＋ PadChest [76]/CIDC [73] CT-Single-CL2 NCV 1,227 SCoV [78] CT Proposed CVP 1,252 SCoV [78] CT-Multiple-CL2 NCV 7,864 SCoVL [78], CCII [71], MGC [40] CVP 4,015 SCoVL [78], CCII [71], MGC [40] CT Proposed CT-Independent-CL2 NCV (Train/Test) 16,616/1,227 MGC [40]＋ CCII [71]＋ iCTCF [79]/SCoV [78] CT Proposed CVP (Train/Test) 6,472/1,252 MGC[40]＋ CCII [71]＋ iCTCF [79]/SCoV [78] CXR-Single-CL3 NOR 1,583 CXRI [72] X-ray [65], [80] [34], [35] NCP 4,273 CXRI [72] CVP 500 CIDC [73] CXR-Multiple-CL3 NOR 3,591 CXRIL [72], ChestX-ray8 [74] X-ray Proposed NCP 4,595 CXRIL [72], ChestX-ray8 [74] CVP 4,015 CCXRIL [75], CIDC [73], PadChest [76] CXR-Multiple-CL4 NOR 3,591 CXRIL [72], ChestX-ray8 [74] X-ray Proposed OBP 2,780 CXRI [72] OVP 1,493 CXRI [72] CVP 4,015 CCXRIL [75], CIDC [73], PadChest [76] a X-Y-CL#: X is CXR or CT; Y denotes the way images from different sources are combined for each class during training or evaluation; CL# is the number of classes. Similar to CXR, publicly available CT scan datasets are also utilized, where most of these datasets contained manually selected 2D slices instead of complete 3D volumes. Hence, all of the CT images referred to in this paper are 2D slices of CT scans. CT-Single-CL2 utilizes NCV and CVP samples from SCoV [78], while we have multiple sources to each class in CT-Multiple-CL2 adding NCV and CVP samples from MGC [40], SCoV [78], and CCII [71]. Due to a lack of publicly available images, some of the designs were not possible, for example, CT-Multiple-CL3 and CT-Multiple-CL4. To evaluate the network’s performance on an independent test set from a separate dataset source whose images are never used during the network’s training, we design CXR-Independent-CL2 and CT-Independent-CL2, utilizing train data from a large study in Spain and test data from the other sources. Table 1 details the train/test split for these two setups. Fig. 2 shows example images from these datasets. In the setup where an independent test dataset is not available, 5-fold cross-validation is applied to evaluate the performance of the proposed CVR-Net (see in Section 4.2). Fig. 2 Samples of chest radiography images from the utilized datasets (a) Normal (X-ray), (b) Normal (CT), (c) Pneumonia viral (X-ray), (d) Pneumonia bacterial (X-ray), (e) COVID-19 (X-ray), and (f) COVID-19 (CT). 4.2 Proposed CVR-Net Architecture We propose a CNN-based end-to-end multi-tasking network, where we apply multi-encoder and multi-scale ensembling, as depicted in Fig. 3. The proposed CVR-Net consists of two encoders, for the same input image, where each of the encoders has five blocks, namely E1n and E2n, n=1,2,…,5, for encoder-1 and encoder-2, respectively. The encoder-1 consists of the residual and convolutional blocks [81], as presented in Fig. 4, well-known as ResNet [81]. The residual connections, also known as skip connections, allow gradients to flow through a network directly, without passing through non-linear activation functions and thus avoiding the problem of vanishing gradients [81]. In residual connections, the output of a weight layer series is added to the original input and then passed through the non-linear activation function, as shown in Fig. 4. However, in encoder-1, 7 × 7 input convolution, followed by max-pooling with the stride of 2, and pool size of 3 × 3, is used before identity and convolutional blocks. By stacking these blocks on top of each other (see Fig. 3), an encoder-1 is formed to get the feature map, where the notation (n×) under the identity block denotes the number of repetitions (n times). The different blocks of encoder-1 (E1n and n=1,2,…,5) downsample the input image resolutions in half of the input resolutions, while the resolution inside the blocks is kept constant. The outputs of those blocks generate the feature maps with different scales. Within the encoder-2 (Xception), three components of information flow blocks are used, which were initially proposed by Chollet [82], such as entry flow, middle flow, and exit flow, as depicted in Fig. 3. The batch of input images first passes through the input flow, then the central flow, eight times (8×) repeated, and finally through the exit flow. All flows, as in the proposed network (see in Fig. 3), have Depth-wise Separable Convolution (DwSC) [82] and residual connections. As in the case of encoder-1, the resolution after each block is downsampled by the factor of two, and the exact resolution is maintained at each block for encoder-2. After the two encoder blocks, the two different 2D feature maps are concatenated channel-wise to enhance the depth information of the feature map. We use differently scaled feature maps to build the proposed CVR-Net, where each feature map is passed through the Fully Connected Layer (FCL) block. A Global Average Pooling (GAP) [83] layer and four fully connected layers are used in our FCL block, where the GAP layer performs an extreme dimensionality reduction to avoid overfitting. In GAP, an height×weight×depth dimensional tensor is reduced to a 1×1×depth vector by transferring height×width feature map to a single number contributes to the lightweight design of the proposed CVR-Net. Table 2 presents the implementational details of the proposed CVR-Net. We utilize the feature maps E13∼E15 from encoder-1 and E23∼E25 from encoder-2, where we concatenate E15 and E25 to increase the depth of the feature information. The final prediction, in CVR-Net, is the average of different probabilities, such as P1, P2, P3, P4, and P5 respectively for E13, E14, [E15++E25], E23, and E24, which was trained end-to-end fashion. However, designing of such a multi-encoder and multi-scale network, as CVR-Net, has several benefits, especially for the small datasets, such as: if one encoder fails to generate responsible features, another encoder can compensate it and vice-versa; if the feature quality is reduced in the deeper blocks (lower resolution), the prior blocks (higher resolution) can also compensate it and vice-versa; if one or more P predicts wrong class, other P can overcome it, as the final result is average of all P’s. Another positive prospect of the CVR-Net is that during the training, it can be anticipated that if the gradient of one or more branches vanishes, other branches can recover it as the final gradient is the average of all the individual gradients. Fig. 3 The proposed network, called CVR-Net, for the automatic COVID-19 recognition from radiography images, where we ensemble the multi-encoder and multi-scale of the network, via fully connected blocks, obtain final recognition probability. Fig. 4 The convolutional (left) and residual (right) blocks [81] of the proposed CVR-Net, where the output map is the summation of the input map and the generated map from the process (convolutions). Table 2 Details of the proposed CVR-Net have used feature maps, shapes, and the number of parameters, where the input resolution is M×N pixels. Feature block Shape of features Prediction Parameters E13 M8×N8×512 P1=FCL(E13) 1,796,867 E14 M16×N16×1024 P2=FCL(E14) 9,181,827 [E15++E25] M32×N32×4096 P3=FCL([E15++E25]) 46,620,971 E23 M8×N8×512 P4=FCL(E23) 1,371,131 E24 M16×N16×1024 P5=FCL(E24) 15,954,283 Proposed CVR-Net P=Avg(P1∼P5) 48,596,087 4.3 Training protocol and evaluation Since most images in all the datasets have a 1:1 aspect ratio, we resize the images to 224 × 224 pixels using nearest-neighbor interpolation. We apply the following stochastic augmentation on the resized images with: rotation (with a probability of 0.45), height & width shift (with a probability of 0.20), and vertical & horizontal flipping around the X- and Y-axis (with a probability of 1.0), respectively. We employ categorical cross-entropy as a loss function [84], penalizing the majority class by giving higher weight in the loss function to the samples from the minority class. Each class’s weights are computed as wj=Nj/N, where wj and Nj are the weights, and the total number of samples for the jth class and N is the total sample numbers. The network weights are initialized using transfer learning [85] where we use the ImageNet pre-trained weights of ResNet-50 and Xception to initialize the weights of the two respective branches. We use Adam optimizer to optimize the training network with initial learning rate (LR), exponential decay rates (β1,β2) as LR=0.0001, β1=0.9, and β2=0.999, respectively, without AMSGrad variant [86]. The initial learning rate is reduced after 12 epochs by 10.0% if validation loss stops improving. The training is terminated after 25 epochs if the validation performance stops improving. The models were implemented using the Python programming language and Keras framework [87] and the experiments were carried out on a machine running Windows-10 operating system with the following hardware configuration: Intel® CoreTM i7−7700 HQ CPU @ 3.60GHz processor with Install memory (RAM): 32.0GB and GeForce GTX 1080 GPU with 8GB memory. When comparing against other state-of-the-art methods (see in Table 5), the same above-described protocol was operated for all the networks. We use different metrics, such as recall, precision, F1-score, and accuracy, to evaluate our multi-tasking CVR-Net for COVID-19 recognition, which is mathematically defined [88] as follows: Recall=TPTP+FN Precision=TPTP+FP F1−score=2×TP2×TP+FN+FP Accuracy=TP+TNTP+FN+FP+TN where the TP, FN, FP, and TN respectively denote true positive (patient with coronavirus symptoms recognized as the positive patient), false negative (patient with coronavirus symptoms recognized as the negative patient), false positive (patient without coronavirus symptoms recognized as the positive patient), and true negative (patient without coronavirus symptoms recognized as the negative patient). The recall quantifies the type-II error (the patient, with the positive syndromes, inappropriately fails to be nullified), and precision quantifies the positive predictive values (percentage of truly positive recognition among all the positive recognition). The F1-score indicates the harmonic mean of recall and precision, which shows the tradeoff between them. Accuracy quantifies the fraction of correct predictions (both positive and negative). 5 Experimental results This section initially presents the results of binary and multi-class classification tasks for various setups described in Section 4.1 using the architecture proposed in Section 4.2. Finally, we compare the proposed network’s performance with state-of-the-art classification networks by training them on the same training set and evaluating an independent test set whose images are not used during training. 5.1 Binary classification: COVID vs. Non-COVID Table 3 presents the quantitative results of the proposed CVR-Net on the binary task: COVID-19 (CVP) vs. Non-COVID (NCV). The 5-fold cross-validation results are conveyed with average and standard deviation. In contrast, a single value is reported when a separate test set from an independent data source is used to evaluate the results. Table 3 demonstrates very high precision and recall in both the cases of CXR-Single-CL2 and CXR-Multiple-CL2. A slight reduction in accuracy for CXR-Multiple-CL2 compared to CXR-Single-CL2 may be because of relatively more minor overfitting to the distribution of the single particular dataset from which the individual classes were coming from in CXR-Single-CL2. As expected, the results for CXR-Independent-CL2 show reduced precision and recall, with accuracy dropping from 98−99% in the cross-validation results to around 88%, when using an independent test set. The observations in the experiments with CXR are consistent in CT as well. Table 3 shows the same pattern with CT-Single-CL2 and CT-Multiple-CL2 having very high accuracy compared to CT-Independent-CL2. The cross-validation results reflect the large DL models’ overfitting nature on a relatively small dataset with limited variability of the real-world scenarios. The accuracy in CT-Independent-CL2 drops from 87−96% in the cross-validation results to around 79% when using the independent test set. We also notice that the accuracy with CT is lower than CXR.Table 3 COVID-19 recognition results from different studies of binary classification applying the proposed network on two different modalities of chest radiography images, wherein for single and multiple sources, we employ 5-fold cross-validation. Different studiesa Dataset distribution Metrics (Train/Val/Test) Recall Precision Accuracy NCV: 3,514/1,171/1,171 CXR-Single-CL2 CVP: 300/100/100 0.997±0.001 0.997±0.001 0.998±0.001 NCV: 4,719/1,573/1,572 CXR-Multiple-CL2 CVP: 2,409/803/803 0.984±0.001 0.984±0.002 0.984±0.001 NCV: 5,567/1,391/1,227 CXR-Independent-CL2 CVP: 2,812/703/500 0.887 0.885 0.887 NCV: 737/245/245 CT-Single-CL2 CVP: 752/250/250 0.976±0.003 0.976±0.003 0.976±0.003 NCV: 4,719/1,573/1,572 CT-Multiple-CL2 CVP: 2,409/803/803 0.969±0.003 0.970±0.003 0.969±0.003 NCV: 13,293/3,323/1,227 CT-Independent-CL2 CVP: 5,178/1,294/1,252 0.799 0.821 0.799 a X-Y-CL#: X is CXR or CT; Y denotes the way images from different sources are combined for each class during training or evaluation; CL# is the number of classes. Details in Table 1. 5.2 Multi-class classifications: Normal, COVID, other bacterial, and viral pneumonia Table 4 and Fig. 5 present quantitative results of the proposed CVR-Net on two different multi-class tasks: (i) 3-class problem for NOR vs. NCP vs. CVP (ii) 4-class problem for NOR vs. OBP vs. OVP vs. CVP. Similar to the binary classification, cross-validation results are reported with average and standard deviation. Fig. 5 shows that in CXR-Single-CL3, NOR and NCP rarely get predicted as CVP while a small number of CVP gets predicted as NCP and NOR. Compared to CVP, a higher fraction of NOR gets predicted as NCP. This is perhaps because the NOR and NCP classes come from the same dataset source, while CVP images are from separate sources. We see that in CXR-Multiple-CL3, fractions of NOR and CVP getting predicted as NCP are much closer. It is worth noting that NOR and NCP in CXR-Multiple-CL3 have images coming from two different datasets, but these sources still do not have the CVP images coming from separate sources. It can also be observed that adding multiple data sources in NOR and NCP has substantially increased the fraction of NCP being predicted as NOR in CXR-Multiple-CL3. From Table 3, Table 4, we see that inter-fold variation is increasing with the decreased performance metrics when a new class is added with the same number of total samples when comparing CXR-Single-CL2 vs. CXR-Single-CL3 and CXR-Multiple-CL2 vs. CXR-Multiple-CL3. In CXR-Multiple-CL4, NCP is further split into other bacterial and viral Pneumonia: OBP and OVP. As seen in Fig. 5, the network confuses much more between OBP and OVP, both coming from the same dataset CXRI. Following the pattern of CXR-Single-CL3, we can also observe that nearly 14% of OBP and OVP still gets classified as NOR. CVP has relatively high precision and recall, but it is noteworthy that the source of the CVP images and the rest of the three classes do not intersect. These results further reinforce the observation in the binary classification task that seemingly high accuracy could be due to the network learning bias in the dataset design and peculiarities of individual data sources rather than the actual underlying pathology. Unlike binary classification problems, we could not evaluate with an independent test set and perform the experiments with CT scans due to the lack of publicly available datasets for these multiple classes.Table 4 COVID-19 recognition results from different experiments of multi-class classification (see in Table 1) applying the proposed network on CXR images employing 5-fold cross-validation. Different studiesa Dataset distribution Metrics (Train/Val/Test) Recall Precision Accuracy NOR: 951/316/316 0.925±0.011 0.940±0.009 0.925±0.012 NCP: 2,565/854/854 0.978±0.003 0.969±0.006 0.977±0.003 CVP: 300/100/100 0.944±0.041 0.976±0.010 0.946±0.041 CXR-Single-CL3 Weighted Average 0.964±0.005 0.963±0.004 0.964±0.005 NOR: 2,155/718/718 0.970±0.018 0.844±0.029 0.970±0.018 NCP: 2,757/919/919 0.863±0.029 0.990±0.004 0.863±0.029 CVP: 2,409/803/803 0.980±0.008 0.968±0.019 0.980±0.008 CXR-Multiple-CL3 Weighted Average 0.933±0.013 0.940±0.011 0.933±0.013 NOR: 2,155/718/718 0.962±0.023 0.902±0.026 0.962±0.023 OBP: 1,668/556/556 0.741±0.021 0.874±0.023 0.741±0.021 OVP: 897/298/298 0.705±0.050 0.646±0.032 0.705±0.051 CVP: 2,409/803/803 0.975±0.007 0.968±0.011 0.975±0.007 CXR-Multiple-CL4 Weighted Average 0.882±0.003 0.886±0.004 0.882±0.003 a X-Y-CL#: X is CXR or CT; Y denotes the way images from different sources are combined for each class during training or evaluation; CL# is the number of classes. Details in Table 1. 5.3 Comparison to the state-of-the-art Several recent studies report the DL models’ performance using datasets that are not publicly available [89], [90], [91]. However, we compare these methods utilizing publicly available data using the experimental setup CXR-Independent-CL2 and CT-Independent-CL2, i.e., the setup, where test set images coming from an independent dataset whose images are never used during training of the models. Table 5 manifests the performance of the proposed CVR-Net along with other widely used and state-of-the-art classification networks and COVID-19 detection networks. The hyperparameters for all the networks used in Table 5, such as learning rate, regularizations, number of epochs, optimization algorithm, etc., are described in Section 4.3 at the end. The proposed CVR-Net performs the best concerning the precision, recall, and overall accuracy in CXR and CT images. The second best is Inception-v3 for CXR and VGG-19 for CT scan. Fig. 6 visualizes the regions in the input image where the neural network is activating most of its signal from when predicting COVID-19 positive class. The activation maps are shown using GradCAM with a threshold 0.6 (maximum 1) [92]. In the figure, the input images are the top three true positive images for CXR and CT, having the highest softmax prediction output for the COVID-19 class from CVR-Net. The activation map for CVR-Net as a whole is smooth and focused within the lung region, while the two branches of CVR-Net having ResNet and Xception architecture have more dispersed activation maps outside the lung region as well. This reveals that combining the two branches make the activation map more focused on the lung region. However, it is remarkably noticed that the focused region we see in the figure in the activation maps of CVR-Net does not always align with the pathology of COVID-19 seen in the CXR and CT images. For Inception, the activation maps are dispersed and smooth, but it is important to note that the images were chosen based on the highest confidence in predicting COVID-19 for CVR-Net and not for Inception.Fig. 5 Confusion matrix for CXR-Single-CL3, CXR-Multiple-CL3, and CXR-Multiple-CL4 employing our CVR-Net. Table 5 Comparison of various methods, including the proposed network (CVR-Net), where the methods are trained on the same dataset and evaluated using an independent test set, not used during training. The top three performing metrics are denoted by bold-font, underline, and double-underline. Methods Parameters CXR-Independent-CL2 CT-Independent-CL2 Recall Precision Accuracy Recall Precision Accuracy VGG-19 46M 0.833 0.846 0.833 0.785 0.816 0.785 Xception 124M 0.869 0.881 0.869 0.718 0.788 0.718 EfficientNet-b1 7M 0.832 0.850 0.832 0.716 0.803 0.716 DenseNet-169 96M 0.850 0.865 0.850 0.718 0.794 0.718 ResNet-152 84M 0.829 0.866 0.829 0.705 0.784 0.705 Inception-v3 74M 0.871 0.884 0.871 0.737 0.782 0.737 DarkNeta[34] 1.94M 0.712 0.699 0.712 0.495 0.245 0.495 CoroNeta[35] 124M 0.869 0.877 0.869 0.689 0.776 0.689 Proposed CVR-Net 48M 0.887 0.885 0.887 0.799 0.821 0.799 a We have implemented those models in our experimental settings for ablation studies. Fig. 6 GradCAM visualizations example, showing activation map on input query CXR and CT images of COVID-19 positive class for proposed CVR-Net, encoder-1 (ResNet), encoder-2 (Xception), and Inception. 6 Discussion and observations We have studied the issues and challenges of DL methods on publicly available datasets for COVID-19 detection using CXRs and CT scans in this work. The results show that many current DL-based methods for COVID-19 classification over-estimate their performance. In particular, we observed two significant issues leading to such high accuracy that is likely not to translate to real-world settings: (i) the prediction classes training data come from separate individual dataset sources. This can result in the network learning the peculiarities of the dataset from which the particular class comes rather than the underlying pathology’s characteristics or features. (ii) the cross-validation results without an independent test set whose images are never used during training can overestimate the network’s performance. It is important to note that both the mentioned issues are common knowledge in machine learning but seems to have been overlooked or not emphasized enough in many recent works involving DL and COVID-19 detection [13], [34], [35], [36], [65], [80], [93], [94]. To reduce such bias and overfitting problems to some extent, we have designed an experiment where the training set contains images in each class from various dataset sources, and an independent test set is used to evaluate the deep neural networks. The results show that, as expected, the performance of the DL model reduces in this scenario. In this more realistic setting, CVR-Net performed the best when compared against other state-of-the-art classification networks. CVR-Net (architecture detailed in Section 4.2) uses multiple branches and aggregates information from different scales, creating a form of ensembling within a single network that seems to be more robust than other DL models, such as VGG, Xception, ResNet, Inception, DenseNet, and EfficientNet, as seen in Table 5. While some of the hyperparameters, such as learning rate and epochs, are adapted for each model dynamically during training, we did not exhaustively optimize the hyperparameters, regularization methods, and training protocol for each of the models separately (details in Section 4.2). For a more detailed comparison, these networks require extensive experimentation with each model to separately tune the hyperparameters and select the best regularization methods outside the scope and objective of the current paper. 4-class classification task showed the model’s difficulty distinguishing bacterial Pneumonia from other viral Pneumonia. Although the results in Fig. 5 for CXR-Multiple-CL4 suggest that COVID-19 Pneumonia is well distinguished from other Pneumonia, the underlying reason is likely that these two classes come from separate data sources. To evaluate the model’s ability to distinguish different classes properly, we suggest that it is essential to have images for each class coming from the same settings, such as the same imaging protocol, machines, demography, etc. Images from multiple settings should also be included when the objective is to assess the algorithm’s ability to work in diverse settings. However, it is essential to include images from all these settings in each class in this case. Table 5 shows higher accuracy when using CXR images compared to CT. We utilized 2D slices rather than the whole CT volume, which was not publicly available for most experimental setups. CT volume may capture details of 3D spatial information, potentially missed in these 2D slices manually selected. Thus, we cannot conclude from the results that CXR is more sensitive than CT for COVID-19 diagnosis. Moreover, the publicly available datasets come from many different sources where it is challenging to track inclusion and exclusion criteria, symptomatic vs. asymptomatic cases, and the disease severity stage at which these images were taken. Building a dataset containing these details may help identify the sensitivity of CXR vs. CT at different stages and symptom severity. This might facilitate a more informed decision for deciding between CT and CXR, which has several tradeoffs, such as patient conditions and the availability of the resource [95], [96]. 7 Conclusion This paper has explored the insights of the COVID-19 detection using the DL framework and publicly available datasets. An end-to-end DL-based model, called CVR-Net, recognizes the COVID-19 from chest radiography images with fewer false negatives. The multi-scale-multi-encoder design of the CVR-Net ensures robustness in recognition, as the final prediction probability is the aggregation of multiple scales and encoders. The experimental results show that many DL-based methods overestimate their interpretation as the data come from different individual dataset references and the cross-validation results without an independent test set. The training set from diverse sources and an independent test set can ameliorate such bias and overfitting troubles to some extent. It is also observed and suggested that it is necessary to have images for each class from identical settings like imaging protocol, machines, and demography. The results also reveal that the CXRs exhibit higher accuracy when compared to CT. We utilized 2D slices rather than the whole CT volume, unavailable for most experimental setups. CT volume may capture 3D spatial information, potentially missed in these manually selected 2D slices. However, the CXRs images can be a good choice for COVID-19 recognition as it has better performance in our experimentation, especially where CT is unavailable to collect. It can be remarked and concluded from the experiments that to accelerate the development of practical clinical DL tools, the scientific community needs to emphasize more on making publicly systematically-designed and documented datasets that have information, such as inclusion and exclusion criteria, symptomatic vs. asymptomatic cases, and the disease severity stage at which these images were taken. Future work will improve the performance by segmenting the lung and adding more distinctive training samples to all the classes. We also intend to deploy our trained CVR-Net to a web application for clinical utilization. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data and code availability Source code and trained model available at https://github.com/kamruleee51/CVR-Net. Acknowledgment We thank the people who contributed to making the COVID-19 related radiography images public. We also thank radiologist Dr. Ram Kumar Ghimire, for feedback on the characteristics of COVID-19 seen in specific CXR and CT images. 1 https://github.com/kamruleee51/CVR-Net. ==== Refs References 1 World Health Organization Naming the coronavirus disease (COVID-19) 2020 https://tinyurl.com/25r7muwv [Accessed: 16 December 2021] 2 World Health Organization WHO Coronavirus disease (COVID-19) dashboard 2022 https://covid19.who.int/ [Accessed: 12 December 2021] 3 Wang D. Hu B. Hu C. Zhu F. Liu X. Zhang J. Clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus–infected pneumonia in Wuhan, China JAMA 323 11 2020 1061 1069 32031570 4 Chen N. Zhou M. Dong X. Qu J. Gong F. Han Y. Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study Lancet 395 10223 2020 507 513 32007143 5 Li Q. Guan X. Wu P. Wang X. Zhou L. Tong Y. Early transmission dynamics in Wuhan, China, of novel coronavirus–infected pneumonia N Engl J Med 2020 6 Wang W. Xu Y. Gao R. Lu R. Han K. Wu G. Detection of SARS-CoV-2 in different types of clinical specimens JAMA 323 18 2020 1843 1844 32159775 7 Yang T. Wang Y.-C. Shen C.-F. Cheng C.-M. Point-of-care RNA-based diagnostic device for COVID-19 2020 Multidisciplinary Digital Publishing Institute 8 NEWS A.J. Bangladesh Scientists create $3 kit. Can it help detect COVID-19? 2020 https://bit.ly/aj2020corona [Accessed: 19 December 2021] 9 COVID C. Global cases by the center for systems science and engineering (CSSE) at johns hopkins university (JHU) ArcGIS. johns hopkins CSSE. retrieved april, Vol. 8 2020 19 10 Huang C. Wang Y. Li X. Ren L. Zhao J. Hu Y. Clinical features of patients infected with 2019 novel coronavirus in wuhan, China Lancet 395 10223 2020 497 506 31986264 11 Corman V.M. Landt O. Kaiser M. Molenkamp R. Meijer A. Chu D.K. Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR Eurosurveillance 25 3 2020 2000045 12 Xu X. Jiang X. Ma C. Du P. Li X. Lv S. A deep learning system to screen novel coronavirus disease 2019 pneumonia Engineering 2020 13 Singh D. Kumar V. Kaur M. Classification of COVID-19 patients from chest CT images using multi-objective differential evolution–based convolutional neural networks Eur J Clin Microbiol Infect Diseases 2020 1 11 14 Lee E.Y. Ng M.-Y. Khong P.-L. COVID-19 pneumonia: what has CT taught us? Lancet Infect Diseases 20 4 2020 384 385 32105641 15 LeCun Y. Bengio Y. Hinton G. Deep learning Nature 521 7553 2015 436 444 26017442 16 Krizhevsky A. Sutskever I. Hinton G.E. Imagenet classification with deep convolutional neural networks Advances in neural information processing systems 2012 1097 1105 17 Yıldırım O. Pławiak P. Tan R.-S. Acharya U.R. Arrhythmia detection using deep convolutional neural network with long duration ECG signals Comput Biol Med 102 2018 411 420 30245122 18 Hannun A.Y. Rajpurkar P. Haghpanahi M. Tison G.H. Bourn C. Turakhia M.P. Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network Nat Med 25 1 2019 65 30617320 19 Acharya U.R. Oh S.L. Hagiwara Y. Tan J.H. Adam M. Gertych A. A deep convolutional neural network model to classify heartbeats Comput Biol Med 89 2017 389 396 28869899 20 Hasan M.K. Dahal L. Samarakoon P.N. Tushar F.I. Martí R. DSNet: Automatic dermoscopic skin lesion segmentation Comput Biol Med 120 2020 103738 21 Esteva A. Kuprel B. Novoa R.A. Ko J. Swetter S.M. Blau H.M. Dermatologist-level classification of skin cancer with deep neural networks Nature 542 7639 2017 115 118 28117445 22 Codella N.C. Nguyen Q.-B. Pankanti S. Gutman D.A. Helba B. Halpern A.C. Smith J.R. Deep learning ensembles for melanoma recognition in dermoscopy images IBM J Res Dev 61 4/5 2017 5–1 23 Hasan M.K. Elahi M.T.E. Alam M.A. Jawad M.T. Martí R. DermoExpert: Skin lesion classification using a hybrid convolutional neural network through segmentation, transfer learning, and augmentation Inform Med Unlock 2022 100819 24 Hasan M.K. Roy S. Mondal C. Alam M.A. Elahi M.T.E. Dutta A. Dermo-DOCTOR: A framework for concurrent skin lesion detection and recognition using a deep convolutional neural network with end-to-end dual encoders Biomed Signal Process Control 68 2021 102661 25 Celik Y. Talo M. Yildirim O. Karabatak M. Acharya U.R. Automated invasive ductal carcinoma detection based using deep transfer learning with whole-slide images Pattern Recognit Lett 2020 26 Cruz-Roa A. Basavanhally A. González F. Gilmore H. Feldman M. Ganesan S. Automatic detection of invasive ductal carcinoma in whole slide images with convolutional neural networks Medical imaging 2014: digital pathology, Vol. 9041 2014 International Society for Optics and Photonics 904103 27 Hasan M.K. Aleef T.A. Roy S. Automatic mass classification in breast using transfer learning of deep convolutional neural network and support vector machine 2020 IEEE region 10 symposium 2020 IEEE 110 113 28 Talo M. Yildirim O. Baloglu U.B. Aydin G. Acharya U.R. Convolutional neural networks for multi-class brain disease detection using MRI images Comput Med Imaging Graph 78 2019 101673 29 Rajpurkar P. Irvin J. Zhu K. Yang B. Mehta H. Duan T. Chexnet: Radiologist-level pneumonia detection on chest x-rays with deep learning 2017 ArXiv:1711.05225 30 Tan J.H. Fujita H. Sivaprasad S. Bhandary S.V. Rao A.K. Chua K.C. Automated segmentation of exudates, haemorrhages, microaneurysms using single convolutional neural network Inform Sci 420 2017 66 76 31 Hasan M.K. Alam M.A. Elahi M.T.E. Roy S. Martí R. DRNet: Segmentation and localization of optic disc and fovea from diabetic retinopathy image Artif Intell Med 111 2021 102001 32 Hasan M.K. Calvet L. Rabbani N. Bartoli A. Detection, segmentation, and 3D pose estimation of surgical tools using convolutional neural networks and algebraic geometry Med Image Anal 70 2021 101994 33 Gaál G. Maga B. Lukács A. Attention u-net based adversarial architectures for chest x-ray lung segmentation 2020 ArXiv:2003.10304 34 Ozturk T. Talo M. Yildirim E.A. Baloglu U.B. Yildirim O. Acharya U.R. Automated detection of COVID-19 cases using deep neural networks with X-ray images Comput Biol Med 2020 103792 35 Khan A.I. Shah J.L. Bhat M.M. Coronet: A deep neural network for detection and diagnosis of COVID-19 from chest x-ray images Comput Methods Programs Biomed 2020 105581 36 Narin A. Kaya C. Pamuk Z. Automatic detection of coronavirus disease (covid-19) using x-ray images and deep convolutional neural networks 2020 ArXiv:2003.10849 37 Ghoshal B. Tucker A. Estimating uncertainty and interpretability in deep learning for coronavirus (COVID-19) detection 2020 ArXiv:2003.10769 38 Abbas A. Abdelsamea M.M. Gaber M.M. Classification of COVID-19 in chest X-ray images using DeTraC deep convolutional neural network 2020 ArXiv:2003.13815 39 Abbas A. Abdelsamea M.M. Gaber M.M. Detrac: Transfer learning of class decomposed medical images in convolutional neural networks IEEE Access 8 2020 74901 74913 40 Zhao J. Zhang Y. He X. Xie P. COVID-CT-dataset: a CT scan dataset about COVID-19 2020 ArXiv:2003.13865 41 Afshar P. Heidarian S. Naderkhani F. Oikonomou A. Plataniotis K.N. Mohammadi A. Covid-caps: A capsule network-based framework for identification of covid-19 cases from x-ray images 2020 ArXiv:2004.02696 42 Ioffe S. Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift 2015 ArXiv:1502.03167 43 He X. Wang S. Shi S. Chu X. Tang J. Liu X. Benchmarking deep learning models and automated model design for COVID-19 detection with chest CT scans 2020 Cold Spring Harbor Laboratory Press MedRxiv 44 Farooq M. Hafeez A. Covid-resnet: A deep learning framework for screening of covid19 from radiographs 2020 ArXiv:2003.14395 45 Hasan M.K. Jawad M.T. Hasan K.N.I. Partha S.B. Al Masba M.M. Saha S. COVID-19 identification from volumetric chest CT scans using a progressively resized 3D-CNN incorporating segmentation, augmentation, and class-rebalancing Inform Med Unlock 26 2021 100709 46 Ozkaya U. Ozturk S. Barstugan M. Coronavirus (COVID-19) classification using deep features fusion and ranking technique 2020 ArXiv:2004.03698 47 Szegedy C, Liu W, Jia Y, Sermanet P, Reed S, Anguelov D, et al. Going deeper with convolutions. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2015, p. 1–9. 48 Furey T.S. Cristianini N. Duffy N. Bednarski D.W. Schummer M. Haussler D. Support vector machine classification and validation of cancer tissue samples using microarray expression data Bioinformatics 16 10 2000 906 914 11120680 49 Zhou N. Wang L. A modified T-test feature selection method and its application on the HapMap genotype data Genom Proteom Bioinform 5 3–4 2007 242 249 50 Hasan M.K. Jawad M.T. Dutta A. Awal M.A. Islam M.A. Masud M. Associating measles vaccine uptake classification and its underlying factors using an ensemble of machine learning models IEEE Access 9 2021 119613 119628 51 Rajaraman S. Siegelman J. Alderson P.O. Folio L.S. Folio L.R. Antani S.K. Iteratively pruned deep learning ensembles for COVID-19 detection in chest X-rays 2020 ArXiv:2004.08379 52 Pham H. Guan M.Y. Zoph B. Le Q.V. Dean J. Efficient neural architecture search via parameter sharing 2018 ArXiv:1802.03268 53 Bergstra J. Bengio Y. Random search for hyper-parameter optimization J Mach Learn Res 13 1 2012 281 305 54 Toğaçar M. Ergen B. Cömert Z. COVID-19 detection using deep learning models to exploit social mimic optimization and structured chest X-ray images using fuzzy color and stacking approaches Comput Biol Med 2020 103805 55 Balochian S. Baloochian H. Social mimic optimization algorithm and engineering applications Expert Syst Appl 134 2019 178 191 56 Khan M.A. Kadry S. Zhang Y.-D. Akram T. Sharif M. Rehman A. Prediction of COVID-19-pneumonia based on selected deep features and one class kernel extreme learning machine Comput Electr Eng 90 2021 106960 57 Narin A. Kaya C. Pamuk Z. Automatic detection of coronavirus disease (covid-19) using x-ray images and deep convolutional neural networks Pattern Anal Appl 24 3 2021 1207 1220 33994847 58 Sedik A. Hammad M. El-Samie A. Fathi E. Gupta B.B. El-Latif A. Efficient deep learning approach for augmented detection of coronavirus disease Neural Comput Appl 2021 1 18 59 Sanida T. Sideris A. Tsiktsiris D. Dasygenis M. Lightweight neural network for COVID-19 detection from chest X-ray images implemented on an embedded system Technologies 10 2 2022 37 60 Turkoglu M. COVIDetectioNet: COVID-19 diagnosis system based on X-ray images using features selected from pre-learned deep features ensemble Appl Intell 51 3 2021 1213 1226 61 ElAraby M.E. Elzeki O.M. Shams M.Y. Mahmoud A. Salem H. A novel gray-scale spatial exploitation learning net for COVID-19 by crawling internet resources Biomed Signal Process Control 73 2022 103441 62 Monday H.N. Li J. Nneji G.U. Nahar S. Hossin M.A. Jackson J. COVID-19 pneumonia classification based on NeuroWavelet capsule network Healthcare, Vol. 10 2022 MDPI 422 63 Sakthivel R. Thaseen I.S. Vanitha M. Deepa M. Angulakshmi M. Mangayarkarasi R. An efficient hardware architecture based on an ensemble of deep learning models for COVID-19 prediction Sustainable Cities Soc 2022 103713 64 Apostolopoulos I.D. Aznaouridis S.I. Tzani M.A. Extracting possibly representative COVID-19 biomarkers from X-Ray images with deep learning approach and image data related to pulmonary diseases J Med Biol Eng 2020 1 65 Apostolopoulos I.D. Mpesiana T.A. Covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks Phys Eng Sci Med 2020 1 66 Hall L.O. Paul R. Goldgof D.B. Goldgof G.M. Finding covid-19 from chest x-rays using deep learning on a small dataset 2020 ArXiv:2004.02060 67 Huang L. Han R. Ai T. Yu P. Kang H. Tao Q. Serial quantitative chest ct assessment of covid-19: Deep-learning approach Radiol: Cardiothoracic Imaging 2 2 2020 e200075 68 Mahmud T. Rahman M.A. Fattah S.A. CovXNet: A multi-dilation convolutional neural network for automatic COVID-19 and other pneumonia detection from chest X-ray images with transferable multi-receptive feature optimization Comput Biol Med 122 2020 103869 69 Minaee S. Kafieh R. Sonka M. Yazdani S. Jamalipour Soufi G. Deep-COVID: Predicting COVID-19 from chest X-ray images using deep transfer learning Med Image Anal 1361-8415 65 2020 101794 10.1016/j.media.2020.101794 70 Oh Y. Park S. Ye J.C. Deep learning covid-19 features on cxr using limited training data sets IEEE Trans Med Imaging 2020 71 Zhang K. Liu X. Shen J. Li Z. Sang Y. Wu X. Clinically applicable AI system for accurate diagnosis, quantitative measurements, and prognosis of covid-19 pneumonia using computed tomography Cell 2020 72 Mooney P. Chest X-Ray images (pneumonia) 2018 https://tinyurl.com/33sjpfz7 [Accessed: 22 December 2021] 73 Cohen J.P. Morrison P. Dao L. Roth K. Duong T.Q. Ghassemi M. COVID-19 image data collection: Prospective predictions are the future 2020 ArXiv:2006.11988 74 Wang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM. Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2017, p. 2097–106. 75 Chowdhury M.E. Rahman T. Khandakar A. Mazhar R. Kadir M.A. Mahbub Z.B. Can AI help in screening viral and COVID-19 pneumonia? 2020 arXiv preprint arXiv:2003.13145 76 Bustos A. Pertusa A. Salinas J.-M. de la Iglesia-Vayá M. Padchest: A large chest x-ray image dataset with multi-label annotated reports Med Image Anal 66 2020 101797 77 Irvin J. Rajpurkar P. Ko M. Yu Y. Ciurea-Ilcus S. Chute C. Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison Proceedings of the AAAI conference on artificial intelligence, Vol. 33 2019 590 597 78 Angelov P. Almeida E. Explainable-by-design approach for covid-19 classification via ct-scan 2020 MedRxiv 79 Ning W. Lei S. Yang J. Cao Y. Jiang P. Yang Q. iCTCF: an integrative resource of chest computed tomography images and clinical features of patients with COVID-19 pneumonia 2020 80 Wang L. Wong A. COVID-Net: A tailored deep convolutional neural network design for detection of COVID-19 cases from chest X-Ray images 2020 arXiv preprint arXiv:2003.09871 81 He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2016, p. 770–8. 82 Chollet F. Xception: Deep learning with depthwise separable convolutions. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2017, p. 1251–8. 83 Lin M. Chen Q. Yan S. Network in network 2013 ArXiv:1312.4400 84 Zhang Z. Sabuncu M. Generalized cross entropy loss for training deep neural networks with noisy labels Advances in neural information processing systems 2018 8778 8788 85 Shin H.-C. Roth H.R. Gao M. Lu L. Xu Z. Nogues I. Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning IEEE Trans Med Imaging 35 5 2016 1285 1298 26886976 86 Kingma D.P. Ba J. Adam: A method for stochastic optimization 2014 ArXiv:1412.6980 87 Chollet F. Keras 2015 GitHub Repository,GitHub, https://github.com/fchollet/keras 88 Hasan M.K. Alam M.A. Roy S. Dutta A. Jawad M.T. Das S. Missing value imputation affects the performance of machine learning: A review and analysis of the literature (2010–2021) Inform Med Unlock 27 2021 100799 89 Harmon S.A. Sanford T.H. Xu S. Turkbey E.B. Roth H. Xu Z. Artificial intelligence for the detection of COVID-19 pneumonia on chest CT using multinational datasets Nature Commun 11 1 2020 1 7 31911652 90 Wang Z. Xiao Y. Li Y. Zhang J. Lu F. Hou M. Automatically discriminating and localizing COVID-19 from community-acquired pneumonia on chest X-rays Pattern Recognit 2020 107613 91 Song Y. Zheng S. Li L. Zhang X. Zhang X. Huang Z. Deep learning enables accurate diagnosis of novel coronavirus (COVID-19) with CT images 2020 Cold Spring Harbor Laboratory Press MedRxiv 92 Selvaraju RR, Cogswell M, Das A, Vedantam R, Parikh D, Batra D. Grad-cam: Visual explanations from deep networks via gradient-based localization. In: Proceedings of the IEEE international conference on computer vision. 2017, p. 618–26. 93 Hemdan E.E.-D. Shouman M.A. Karar M.E. Covidx-net: A framework of deep learning classifiers to diagnose covid-19 in x-ray images 2020 ArXiv:2003.11055 94 Sethy P.K. Behera S.K. Detection of coronavirus disease (covid-19) based on deep features, Vol. 2020030300 2020 Preprints 2020 95 Cleverley J. Piper J. Jones M.M. The role of chest radiography in confirming covid-19 pneumonia Bmj 370 2020 96 Rubin G.D. Ryerson C.J. Haramati L.B. Sverzellati N. Kanne J.P. Raoof S. The role of chest imaging in patient management during the COVID-19 pandemic: a multinational consensus statement from the fleischner society Chest 2020
PMC009xxxxxx/PMC9005224.txt	==== Front Comput Biol Med Comput Biol Med Computers in Biology and Medicine 0010-4825 1879-0534 The Authors. Published by Elsevier Ltd. S0010-4825(22)00305-5 10.1016/j.compbiomed.2022.105513 105513 Article Towards a multi-scale computer modeling workflow for simulation of pulmonary ventilation in advanced COVID-19 Middleton Shea a1 Dimbath Elizabeth a1 Pant Anup a George Stephanie M. a Maddipati Veeranna c Peach M. Sean b Yang Kaida b Ju Andrew W. b Vahdati Ali a∗ a Department of Engineering, College of Engineering and Technology, East Carolina University, Greenville, NC, USA b Department of Radiation Oncology, Brody School of Medicine, East Carolina University, Greenville, NC, USA c Division of Pulmonary and Critical Medicine, Brody School of Medicine, East Carolina University, Greenville, NC, USA ∗ Corresponding author. Ross Hall, East Carolina University, Greenville, NC, USA. 1 The first and second author contributed equally to this work. 12 4 2022 6 2022 12 4 2022 145 105513105513 17 1 2022 10 3 2022 8 4 2022 © 2022 The Authors 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Physics-based multi-scale in silico models offer an excellent opportunity to study the effects of heterogeneous tissue damage on airflow and pressure distributions in COVID-19-afflicted lungs. The main objective of this study is to develop a computational modeling workflow, coupling airflow and tissue mechanics as the first step towards a virtual hypothesis-testing platform for studying injury mechanics of COVID-19-afflicted lungs. We developed a CT-based modeling approach to simulate the regional changes in lung dynamics associated with heterogeneous subject-specific COVID-19-induced damage patterns in the parenchyma. Furthermore, we investigated the effect of various levels of inflammation in a meso-scale acinar mechanics model on global lung dynamics. Our simulation results showed that as the severity of damage in the patient's right lower, left lower, and to some extent in the right upper lobe increased, ventilation was redistributed to the least injured right middle and left upper lobes. Furthermore, our multi-scale model reasonably simulated a decrease in overall tidal volume as the level of tissue injury and surfactant loss in the meso-scale acinar mechanics model was increased. This study presents a major step towards multi-scale computational modeling workflows capable of simulating the effect of subject-specific heterogenous COVID-19-induced lung damage on ventilation dynamics. Keywords Pulmonary mechanics COVID-19 Pulmonary ventilation Computer modeling Acute respiratory distress syndrome SARS-CoV-2 Lung mechanics ==== Body pmc1 Introduction Coronavirus Disease 2019 (COVID-19) infection due to the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) virus can cause extensive damage to many tissues and organs, including the lungs. Millions of lives have been lost to the disease, while others face long-term effects of this viral infection [1,2]. Though computer models may assist with managing and tracking the spread of COVID-19, yet new variants of the SARS-CoV-2 virus may spread more easily and can pose an increased risk of severe disease and other complications linked to COVID-19 [3,4]. Alveolar damage is seen in imaging and histopathological studies of COVID-19 infected lungs [5], and heterogeneous damage throughout the acinar regions of the lung is observed in computed tomography (CT) images [5]. In particular, CT imaging can provide useful spatial information on patterns of lung injury, including ground-glass opacities (GGO) and areas of consolidation [6]. Also, some patients suffering from COVID-19 acute respiratory distress syndrome (CARDS) exhibit increased hypoxemia compared to typical acute respiratory distress syndrome (ARDS) [7]. However, there is currently little quantitative information on how acinar level patho-mechanics contribute to whole lung function in COVID-19 patients. A better understanding of how COVID-19 impairs regional ventilation may give insight into overall lung dynamics specific to CARDS. To provide a four-dimensional view of airflow patterns in the lung, CT images can be used to develop geometric models of the lung, which then can be combined with fluid flow and tissue mechanics physics-based models. Such physics-based computer models of the lung present an opportunity for gaining valuable insights into pulmonary ventilation dynamics [8,9]. For instance, computational modeling of lung dynamics across multiple spatial scales may allow a deeper understanding of how mechanical changes at the alveolar and acinar levels affect lobar and whole-lung dynamics in CARDS. Previous physics-based computer models of the lung have provided a detailed four-dimensional view of pulmonary ventilation in both healthy and disease states. In the aforementioned computer models, CT images were used to determine lobar volumes and airway branching patterns of large airways with small peripheral airways constructed utilizing volume-filling tree-generating algorithms [10,11] and the coupling of airway trees to compliant acinar regions provided realistic flow distribution among the lung lobes [12]. Along with realistic geometries for airways and compliant acinar regions, in silico models have also incorporated other factors such as tissue deformation, gravity, acinar-level interdependence, and surfactant all of which contribute to distribution of ventilation in the healthy lung [[13], [14], [15]]. Disease can lead to remodeling of airways and parenchymal tissue thus inducing changes in airflow patterns and tissue mechanics that can be implemented in in silico models [16]. Insight into the effects of disease on ventilation and pressure distribution of the lung has been gained through modification of model geometry and mechanics in previous studies [17,18]. Additionally, information on distribution and manifestation of damage throughout the lung is important for realistic representation of disease states [18]. Registration of high resolution and 4D CT images has been used to identify damaged areas of parenchyma and changes in airway geometry and regional ventilation in disease states [16,19]. Utilization of imaging techniques also opened the door for patient-specific modeling of airflow and pressure distribution [10,18]. Multi-scale in silico lung models with geometries resolved from CT imaging have proven useful in understanding various pulmonary diseases like cystic fibrosis, chronic obstructive pulmonary disease (COPD) and asthma [18,20,21]. However, there is a need for computer modeling studies on pulmonary ventilation dynamics in COVID-19 patients.. The main objective of this study is to develop a physics-based in silico modeling workflow for studying the pulmonary ventilation of COVID-19-infected lungs by bringing existing methodologies for coupling airflow and lung tissue mechanics together [10,11,22]. The presented in silico modeling approach is the foundation and first step towards a virtual hypothesis testing platform for a better understanding of COVID-19 pulmonary dynamics utilizing 4D CT data from COVID-19-infected lungs and accoutning for patient-specific lung geometry and disease distribution with varying levels of damage. In addition, we aimed to develop an in silico multi-scale approach to simulate and compare the regional changes in lung dynamics associated with heterogeneous subject-specific COVID-19-induced damage in the parenchyma of the lung. To this end, we present an investigation of the effect of various levels of inflammation in a meso-scale acinar mechanics model on global lung dynamics. 2 Methods 2.1 Imaging This study utilized the 4D CT scan of a male patient recently hospitalized in Vidant Medical Center (Greenville, North Carolina, USA) for an advanced case of COVID-19. The 4D CT scan was obtained during tidal breathing using an Optima CT580 RT scanner (GE Healthcare, Waukesha, WI). The methodology used in this paper was approved by the East Carolina University and Medical Center Institutional Review Board (UMCIRB) with study ID 20–001447. Informed consent was obtained from the patient. Sorting of the CT images into phases of the breathing cycle was accomplished using the Varian 4DCT Real-time Position Management (RPM) system (Varian Medical Systems, Palo Alto, CA). When a series of CT images was attained over a time comparable to that of a normal breathing cycle, the Varian RPM camera captured the patient's real-time external chest motion amplitude, and Advantage 4D (GE Healthcare, Waukesha, WI) software retrospectively sorted the CT data into corresponding phases of the respiratory cycle from 0% to 90%, with 0% corresponding to end-inspiration and 50% corresponding to end-expiration [23]. The images were taken at 120 kVp, 300 mAs, and 20 mm collimation and were reconstructed through a 512 × 512 matrix with a 2.5 mm slice thickness and reconstructed retrospectively to a slice thickness of 1.25 mm. 2.2 Segmentation Following imaging, the geometry of the major airways visible in CT and lungs lobes was segmented. The major conducting airway geometry was segmented from the end-inspiratory phase using a combination of dynamic region growing and manual editing in Materialise Mimics 23.0 (Materialise NV, Belgium) (Fig. 1 a). Centerline detection and extraction as described in Bordas et al.‘s work [10] for the major airways were also achieved using Mimics 23.0. Five different lung lobe (right upper, right middle, right lower, left upper, and left lower) geometries were segmented at the end-inspiratory phase and end-expiratory phase using the Chest Imaging Platform [24] available in 3D Slicer [25,26]. Total segmented lobe volume was validated against total segmented lung volume. 3D Slicer was used to segment the COVID-19 affected GGO and consolidated regions of the lungs at end-inspiratory and end-expiratory phases using the LungCTAnalyzer [27] extension of the Chest Imaging Platform (Fig. 1b). This segmentation was based on Hounsfield unit (HU) values in the CT images. Inflated lung Hounsfield unit thresholds were determined based on the study by Kassin et al. [28] with a range of −1000 to −650 HU. The difference between ground glass opacities and consolidated regions were determined based on the 3D Slicer Chest Imaging Platform [24] and Lung CT Analyzer [27] (https://github.com/rbumm/SlicerLungCTAnalyzer) extensions preset values for COVID-19 lung analysis.Fig. 1 CT image as segmented in 3D Slicer and Mimics software. (a) Purple regions show major airways at end-inspiration. (b) Areas affected by COVID-19 damage, such as GGO (orange) and consolidated (blue) are highlighted based on HU values used for segmentation. The remaining dark grey areas are aerated lung tissue; (c) Geometrical representation of the entire generated tree, viewed from the front, including the airways generated using the space-filling algorithm. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Fig. 1 2.3 Geometry Segmented major airway geometry was limited to the first four to six generations. Further airway segmentation was constrained by CT image resolution. Subsequently, a space-filling airway generation algorithm with random heterogeneity based on the work of Tawhai et al. [11,29] was used to create up to 16 generations of conducting airways. Airway diameter, length, and branching angles were based on the segmented geometry and lung lobe models (Fig. 1c) following the methods described in Refs. [[29], [30], [31]]. The number of acini per lung was approximately 15,000 [32]. The lumen diameter for each one-dimensional line segment was assigned based on the Horsfield number [10,11]:(1) logD(x)=(x−Max)logRdH+logDMax where x, D , Max, DMax represented the current Horsfield order, the airway diameter, the maximum Horsfield order and the maximum diameter, respectively. RdH represented the anti-log slope of airway diameter plotted against Horsfield order and was assigned to be 1.15. 2.4 Airflow and acinar mechanics model Here, our multi-scale computational models of the lung are developed through the C++ simulation package CHASTE (Cardiac, Heart, and Soft Tissue Environment) [10,33]. Airflow in the airway tree geometry was described by a reduced dimensional airway model implemented in CHASTE, which was coupled to the tissue mechanics acinar models [10,34]. The flow was presented as a modified Poiseuille flow following the approach developed by Swan et al. [13] and Ismail et al. [12] and by assuming isotropic expansion of acini. Corrections to the dynamic resistance based on work by Pedley et al. [10,35] were applied as shown in Equation (2):(2) R=γ(ReDawlaw)1/2Rp where R is the dynamic resistance, R p is the Poiseuille resistance, D aw and l aw are the diameter and length of an airway, respectively, Re is the Reynold's number, and γ was set to be generation-dependent based on the work of van Ertbruggen et al. and Ismail et al. [12,36]. Atmospheric pressure was assigned at the trachea, hence airflow into the lungs was driven by variations in the volume of acini as a function of changes in the transpulmonary pressure during breathing. All nodal pressures and edge fluxes were solved for simultaneously using multifrontal lower–upper factorization solver UMFPACK. Flow from the airway model was fed into each acinar model to calculate the change in acinar volume using the stretch ratio during each time step taken by the solver. A sigmoidal acinar mechanics model based on the work of Fujioka et al. and Venegas et al. [37,38] was coupled to the generated airway tree in the simulations (Fig. 2 ). In this model, as shown in Equation (3), V a is the acinar volume, Pa is the transpulmonary pressure for each acinus (defined as the difference between pressure in the acinus and the pleural pressure), and A, B, C, and D are constants that vary based on surfactant level and consequently tissue compliance:(3) Va=A+B1+e−(Pa−C)/D Fig. 2 The sigmoidal model derived from Refs. [22,39] used in a simulation for healthy lung and COVID-19 affected lungs. The pressure – volume curve for a healthy lung is shown in blue, with pressure being equal to the transpulmonary pressure. As damage progresses in the lungs, there is a progressive reduction of surfactant amount and compliance of the acinar units. The decrease in surfactant shifts the pressure-volume curves to the right, as seen in the 20% reduced surfactant (orange), 40% reduced surfactant (grey), and 60% reduced surfactant (yellow) cases. The dashed grey lines indicate the pressure range used for the simulation. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Fig. 2 Time-derivative of equation (3) was solved at the end of each terminal bronchiole, where an acinus was connected to an airway, thus coupling the pressure in the airway tree and the acinar model. 2.5 Boundary conditions and simulation settings 2.5.1 Boundary and loading conditions Pulmonary ventilation dynamics during tidal breathing was simulated in accordance with the aforementioned coupled airway-acinar model. To simulate tidal breathing, a varying pleural pressure was applied at the acini while a constant atmospheric pressure boundary condition was applied at the trachea. The varying pleural pressure was assumed to be [12,40]:(4) Ppl=Pplmax+ΔPpl2(1−cos(2πtT+Φ)+π) where Pplmax=−5cmH2O (-490 Pa), ΔPpl=−3.2cmH2O (-314 Pa), and the phase shift Φ=π/11. 2.5.2 Simulation of COVID-19-afflicted lungs The coupled airway-acinar model was used to simulate COVID-19 lung damage based on the segmented CT images and the associated region-specific levels of damage, corresponding to GGO and consolidated regions. Since the CT images were obtained from a patient recovering from advanced COVID-19, hypothetical healthy lung simulations were also performed to provide a basis for comparison of the results. To simulate the changes between healthy and COVID-19 afflicted lung function, different mechanical behavior of the sigmoidal acinar model (Fig. 2) was implemented based on the amount of surfactant and compliance of the acinar units. Simulations were run for the healthy case considering the normal amount of surfactant. The COVID-19 affected lung simulations considered the reduced amount of surfactant and decreased compliance in both GGOs and consolidation regions [[41], [42], [43]]. Previously segmented GGO and consolidated regions in lung CT images were used to identify acinar regions affected by COVID-19. We assumed GGOs to represent partial filling of air spaces while consolidated regions may signify more severe damage where a large proportion of airspace is filled with liquid and inflammatory infiltrates [5,6]. While the airspace becomes compromised and inflamed, as seen in CT images of GGO and consolidation regions, the amount of surfactant and compliance of the acini can be altered [38,44,45]. Different acinar properties were applied to the GGO and consolidated regions to simulate these varying levels of tissue damage. Two simulations were run to visualize the effects of varying levels of COVID-19 severity: for the first simulation, a 20% reduction in the surfactant amount was used for the GGO regions, and 40% reduction in the surfactant for the consolidation regions was considered. For the second simulation, a 20% reduction of surfactant in the GGO regions and a 60% reduction in the consolidated regions was considered (Fig. 2). Coefficients for the healthy sigmoidal model were fit to physiologically relevant coefficients based on patient-specific lung volumes [39]. Lung volumes were estimated based on the gender, height, and age of the COVID-19 positive patient used for CT scans in this study: male, 167.6 cm, 51 years, and 88.5 kg, respectively [46]. The following equations from Boren et al. [47] were used to estimate lung capacities and volumes for a healthy adult male:(5) TLC=0.078H−7.30 (6) FRC=0.032H−2.94 (7) RV=0.019H+0.0115Ag−2.24 (8) VC=0.052H−0.022Ag−3.60 where TLC is the total lung capacity in liters, H is the subject's height in centimeters, FRC is functional residual capacity in liters, RV is residual volume in liters, Ag is the subject's age in years, and VC is vital capacity in liters. Based on these equations for the patient in this study, we found that TLC = 5.77 L, FRC = 2.42 L, RV = 1.53 L, and VC = 3.99 L. Estimated lung volumes were then used to define coefficients A and B, as A is approximated by the residual volume and B is approximated by the vital capacity [39]. The values of C and D in the sigmoidal model correspond to the inflection point of the curve and the pressure range in which the volume change takes place, respectively [39]. As such, C and D are not as easily determined by available physiological data and were estimated such that in the healthy case, they were estimated based on the calculated FRC and the tidal volume pressure range. Then, to define the regions of GGO and consolidation, C and D coefficients signifying reduced surfactant as reported by Fujioka et al. [38], were scaled to fit our patient data. Fujioka et al.‘s [38] coefficients were limited to normal surfactant, 20% reduced, and 40% reduced surfactant amounts. Coefficients for 60% reduced surfactant was not directly available. Hence, the coefficients for 60% reduced surfactant were extrapolated. Assumptions for extrapolation were based on work by Fujioka et al. [38] that only coefficient C varied markedly when surfactant level was changed (Table 1 ).Table 1 Coefficients for use in the sigmoidal model of equation (4), defining the acinar unit's pressure-volume relationship. Table 1 A (L) B (L) C (Pa) D (Pa) Healthy 1.53 4.24 1078.73 449.14 20% reduced surfactant 1.53 4.24 1420.00 451.11 40% reduced surfactant 1.53 4.24 1818.15 356.96 60% reduced surfactant 1.53 4.24 2359.48 356.96 Three different simulations were executed for this study:1. A hypothetical control study where the lung was assumed to be healthy; GGO and consolidated regions are considered to be normal inflated regions with normal surfactant levels. 2. COVID-19-afflicted lung with 20% reduced surfactant for the GGO region and 40% reduced surfactant for the consolidated region 3. COVID-19-afflicted lung with 20% reduced surfactant for the GGO region and 60% reduced surfactant for the consolidated region. The total simulation time was 12 s, with each breathing cycle assumed to be 4 s between consecutive end-inspirations. The time step used by the solver was 0.001 s, and data were saved every 100 time steps. A smaller 0.0001 s time step was also tested for the healthy case and showed no significant difference in results other than an increase in simulation time. Three breathing cycles were executed, with the first two cycles discarded so that only steady-state conditions were analyzed. Following the onset of steady-state conditions, the data from that breathing cycle was used to calculate the flow rate at the trachea, flow rate into the individual lobes, total tidal volume of the lungs at different time points, the flow rate through individual lobes. Flowcharts summarizing the entire model development and simulation process are shown in Fig. 3, Fig. 4 , respectively.Fig. 3 Flowchart for generating the airway tree geometric model from CT images. The CT image was segmented to determine major conducting airways, lungs, and lobes as well as COVID-19-affected regions. CHASTE [33] was then used to generate the complete airway tree down to terminal bronchioles. Fig. 3 Fig. 4 Flowchart for running the tidal breathing simulation for healthy and diseased lungs. The airway tree model was coupled with the sigmoidal acinar model in CHASTE [33] to simulate tidal breathing. Different levels of surfactant reduction were applied to the acinar model to simulate lung function in disease states. Fig. 4 3 Results In total, three simulations were performed consisting of a hypothetical healthy lung with normal surfactant levels as well as two COVID-19-afflicted simulations created through reduction of surfactant levels based on the degree of damage present in the CT images. The first simulation will be referred to as the hypothetical “healthy” case, while the second and third simulations will be referred to as “diseased 20–40” and “diseased 20–60” cases, respectively. Airflow during the simulations was generated due to the negative pressure as a result of the acinar expansion. Fig. 5 shows the flow rate of air at the trachea and into each lobe after reaching steady-state over an entire breathing cycle. Different colored lines represent each simulation scenario, with the highest flow rate in the healthy simulation and the lowest in the diseased 20–60 simulation. The lobar flow rate values were determined by calculating the flow rate at the first airway branch that enters each lobe.Fig. 5 Flow through the trachea (a) and each lung lobe (b–f) during tidal breathing over one breathing cycle, plotted by tracking the flow every 100 time steps. Fig. 5 The flow rate at various points in the lung can be integrated to determine the total and lobar tidal volumes (Fig. 6 ). The maximum values of tidal volumes for the whole lung and each lobe are also reported in Table 2 , in addition to the percent change in tidal volumes in diseased cases versus the hypothetical healthy case. From Table 2, it can be seen that the healthy, diseased 20–40, and diseased 20–60 simulation tidal volumes were 0.592 L, 0.392 L, and 0.248 L, respectively. Thus, the results in Fig. 5 and Table 2 demonstrate that the tidal volume of the whole lung decreases as the severity of COVID-19 in the affected consolidated regions increases. It can also be seen in Table 2 that the right middle lobe shows the least difference in its tidal volume between the healthy and diseased simulations.Fig. 6 Time-dependent volume change of the entire lung, determined at the trachea (a), and individual lobes (b–f). Values for volume change obtained by integration of flow rates. Fig. 6 Table 2 Tidal volume of the different lobes during one breathing cycle based on the integration of simulation results. Table 2Lobes Tidal Volume under different conditions (L) Healthy Diseased 20-40 % change between healthy and 20-40 Diseased 20-60 % change between healthy and 20-60 Whole lung 0.592 0.392 −33.8% 0.248 −58.1% Right upper 0.106 0.070 −33.8% 0.040 −61.9% Right middle 0.079 0.071 −9.8% 0.060 −23.3% Right lower 0.111 0.062 −44.2% 0.030 −72.6% Left upper 0.196 0.137 −29.9% 0.091 −53.7% Left lower 0.101 0.052 −48.8% 0.026 −74.0% The volumes of GGO and consolidated regions calculated from CT images at end-inspiration were also quantified and are presented in Table 3 . As previously described, these volumes were determined directly by thresholding the CT images. Note that actual COVID-19 affected volumes are likely to be slightly larger than presented, as very small unconnected “islands” of damage had to be excluded from analysis for volume meshing purposes. In Table 3, “COVID-Afflicted %” is the combined portion of consolidated and GGO regions. It can be seen that at end-inspiration, the right middle lobe shows the lowest percentage of damage by COVID-19, followed by the left upper lobe by a large margin. The two lobes with the smallest volume of air and the highest amount of damage in both states are the left and right lower lobes (Table 3).Table 3 Total volume of air, GGO, and consolidated regions at end-inspiration obtained from segmenting the CT image. Table 3Volume (L) Right Upper Lobe Right Middle Lobe Right Lower Lobe Left Upper Lobe Left Lower Lobe TOTAL Air 0.252 0.324 0.136 0.518 0.118 1.348 GGO 0.189 0.079 0.263 0.187 0.126 0.844 Consolidated 0.169 0.057 0.225 0.239 0.228 0.918 Total 0.61 0.46 0.624 0.944 0.472 3.110 Air % 41.3% 70.4% 21.8% 54.9% 25.0% 43.3% GGO% 31.0% 17.2% 42.1% 19.8% 26.7% 27.1% Consolidated % 27.7% 12.4% 36.1% 25.3% 48.3% 29.5% COVID-Afflicted % 58.7% 29.6% 78.2% 45.1% 75.0% 56.7% Table 4 compares the share of ventilation that goes into each lobe during tidal breathing for each of the three simulated scenarios versus the values calculated from the CT using images at end-inspiration and end-expiration. Based on these results, it can be seen that the right middle lobe which shows the lowest COVID-19 infiltration of 29.6% at end-inspiration, demonstrates a 10% increase in tidal volume percent share as the simulations progress from healthy to the diseased 20–40 and 20–60 states. The left upper lobe which has a relatively low 45.1% infiltration at end-inspiration, shows an increase in tidal volume percent share with COVID-19 progression, though not as drastically as the right middle lobe, with only a 3.4% increase (Table 4). At end-inspiration, the right upper, right lower, and left lower lobes demonstrated remarkable COVID-19-induced damage with 58.7%, 78.2%, and 75.0% affliction, respectively, and each lobe showed a decrease in tidal volume percent share with disease progression. The less-afflicted right upper lobe showed a far more minor change in tidal volume (−1.6%) than the right lower lobe (−6.5%) or the left lower lobe (−6.5%).Table 4 Difference between the lobar share of tidal volume obtained from 4D CT (4th column), simulations of healthy and different disease states (columns 1 to 3), and Jahani et al. [48] study on lobar distribution of ventilation in healthy subjects (5th column). Table 4 Lobar Share of Tidal Volume of Healthy simulation Lobar Share of Tidal Volume of Diseased 20–40 simulation Lobar Share of Tidal Volume of Diseased 20–60 simulation Lobar Share of Tidal Volume from CT Mean Lobar Air Volume from Jahani et al. Right upper 17.9% 17.9% 16.3% 23.5% 20.9% Right middle 13.3% 18.1% 24.3% 8.6% 10.7% Right lower 18.7% 15.8% 12.2% 13.2% 22.8% Left upper 33.1% 35.0% 36.5% 38.8% 25.8% Left lower 17.1% 13.2% 10.6% 13.7% 20.6% Additionally, Table 4 contains a column showing lobar air volume fraction, averaged over inspiration and expiration and converted to a percentage, as defined in a Jahani et al. [48] study investigating regional healthy lung deformation and ventilation with 4D-CT. This column was added for validation purposes to show a general agreement with our healthy model and CT-based estimations of lobar ventilation distribution. Note that lungs differ from person to person morphologically, but general trends in shape persist in most lungs; in this case, our healthy simulation successfully predicts that the left upper lobe is the largest, followed by the right lower, right upper, left lower, and right middle lobes, which mirrors Jahani et al.‘s results [48]. The pressure distributions of the lung in the three different simulation scenarios are visualized in Fig. 7 . All pressure distribution results are shown at both maximum inspiration and maximum expiration. Qualitatively, it can be discerned from Fig. 7 that the subject had the highest magnitude of pressure during both inhalation and exhalation in the right middle lobe and left upper lobe. These results correlate with the percentage of healthy (air-filled) acini determined from the CT images (Table 3). As these two lobes had the lowest percentage of COVID-19 affliction, they were expected to have the most unrestricted airflow. It is also notable that the three more COVID-19 affected lobes, the right upper and right and left lower lobes, show very different pressure distributions and magnitudes when comparing the healthy case to the diseased simulations. Furthermore, the diseased simulation scenarios show much more heterogeneity in air pressure distribution throughout the lung. Table 5 shows the average pressure at the terminal bronchioles in the entire lung for each simulation. From these values, it can be seen that the healthy lung simulation produced the highest average pressure magnitude and the least heterogeneity, while progressive disease states caused an increase in standard deviation and a substantial decrease in mean pressure.Fig. 7 Lung pressure distribution as viewed from the front in a healthy simulation (a, b), diseased 20–40 simulation (c, d), and diseased 20–60 simulation (e, f), at maximum inhalation (a,c,e) and maximum exhalation (b,d,f). Note that the right lung appears on the left in this image and vice versa. Fig. 7 Table 5 Mean and standard deviation of lung pressure at the terminal bronchioles in each simulation. Table 5Average Pressure (Pa) Maximum Inspiration Maximum Expiration Healthy −16.9 ± 1.46 17.5 ± 1.58 Diseased 20-40 −9.63 ± 2.63 8.99 ± 2.99 Diseased 20-60 −5.06 ± 3.03 4.85 ± 3.02 4 Discussion Few computer modeling studies have been performed with the aim of better understanding ventilation dynamics changes in advanced cases of COVID-19. Three purely mathematical yet elegant and informative models of COVID-19 effects on pulmonary ventilation and perfusion were presented by Voutouri et al. [49], Busana et al. [50] and Herrmann et al. [51]. Another mathematical modeling study by Weaver et al. [52] simulated the effect of increased respiratory effort of patients with COVID-19 acute hypoxemic respiratory failure during spontaneous breathing on parameters associated with lung injury such as tidal swings in pleural pressure. Additionally, a computational fluid dynamics model of airflow in the upper airways of COVID-19 patients was also recently presented by Pan et al. [53]. In our simulation-based study, a virtual physics-based hypothesis-testing platform was developed and presented to study lobar ventilation dynamics of COVID-19-infected lungs. In particular, the mechanical changes in severely COVID-19-afflicted lungs compared to theoretically healthy lungs were modeled and examined in a multi-scale modeling framework. To the authors' knowledge, this is the first in silico modeling approach to use 4D CT images of COVID-19-induced lung damage to simulate regional airflow and pressure distribution in the COVID-19-infected lung. We present this study as the first step towards patient-specific physics-based models of COVID-19-afflicted lung dynamics and to lay the foundation for more detailed and individualized in silico models currently in development by our group. The multi-scale approach presented here uses patient-specific lung geometry and injury patterns obtained from CT images of a patient with advanced COVID-19 and accounts for changes in flow rate into individual lobes as a consequence of COVID-19-induced lung injury and decreased tissue compliance. Heterogeneous damage in the parenchyma of the lung due to COVID-19 was represented through changing the in surfactant levels and tissue compliance at the acinar level. In addition, our computational models enabled the visualization of acinar-level impacts of the infection on global lung dynamics. Images obtained through 4D CT were processed for efficient image segmentation and conversion to physics-based computer models. This methodology provides a solid foundation for future investigations of other potential mechanisms of COVID-19 damage to the lungs and the ensuing effects on global lung function through a computationally efficient approach. Our disease scenario simulation results showing differences in the lobar distribution of tidal volume are in agreement with previous studies demonstrating that lung damage in advanced ARDS, which resembles Type H COVID-19 pneumonia as categorized by Gattinoni et al. [54], can decrease the compliance of injured regions to a large extent, hence decreasing airflow [55]. As expected, the least damaged lobe (right middle, 29.6% damage at end-inhalation) showed a large increase (83%) in tidal volume share when comparing the healthy and the most severe simulated disease case, and the most damaged lobe (right lower, 78.2% damage measured at end-inhalation) experienced a significant decrease in ventilation share (−35%) in the same comparison. Meanwhile, the full lung (56.7% damage at end-inhalation) also showed a notable overall reduction in tidal volume when comparing the healthy and the most severe simulated disease state (−58%). In this study, analysis of damaged regions from the segmented CT images showed that lower lobes contained higher amounts of damage (combined GGO and consolidated regions) than other lobes (Table 2, Table 3), in this patient. These findings correspond to those from previous studies of CT imaging of COVID-19 lungs [56]. In a study using ventilation/perfusion single-photon emission computed tomography combined with computed tomography (V/Q SPECT/CT) performed in COVID-19 patients, large ventilation defects in the subpleural areas were observed [57]. The researchers observed that ventilation defects were present in the GGO areas while perfusion was largely preserved. However, in more severely damaged areas of the lung where complete alveolar filling and parenchymal lesions and fibrosis were present, perfusion defects were additionally observed which can be a sign of involvement of capillary walls [57]. Furthermore, the researchers proposed a potential adaptive mechanism where redistribution of ventilation towards the healthy parenchyma occurs [57]. This is indeed what our multi-scale model demonstrates: as the severity of damage in the right lower, left lower, and to some extent in the right upper lobe was increased, ventilation was redistributed to the least injured right middle and left upper lobes (Table 4). While our computer model in its current version only includes ventilation, we are actively developing a perfusion model which will be coupled to our ventilation dynamics model and will enable us to simulate microangiopathy induced by COVID-19. Our simulations reasonably predicted a decrease in overall tidal volume as the level of lung damage and surfactant loss was increased. Although preserved compliance has been reported in early CARDS, lungs in more advanced COVID-19, such as those of our patient, are reported to have decreased compliance, in line with typical ARDS [54]. The decrease in tidal volume represents this reduction in compliance, as stiffer acini in the pressure range of tidal breathing would have a smaller change in volume compared to healthy acini with normal stiffness. Overall, the data in this study exhibit trends of decreased airflow to areas most affected by COVID-19 damage in an advanced case of the infection. The lower lobes showed a greater change in percent share of tidal volume in disease state simulations, while the upper lobes showed a less overall shift in percent share of tidal volumes and were less affected by COVID-19 (Table 3, Table 4). Our hypothetical simulation of healthy lung ventilation distribution among the different lobes was in qualitative agreement with the results of Jahani et al.‘s [48] 4D CT analysis of healthy lungs (Table 4). However, the ventilation distribution was notably altered in the COVID-19-infected lung simulations, with less damaged lobes receiving higher portions of tidal volume and more damaged lobes receiving a lower share of tidal volume (Table 4). Thus, regional differences seen in tidal volume distribution and subsequent changes in lobar share of tidal volume due to damage hint at how heterogenous damage in acinar regions may affect global lung function in a region-specific manner. However, discrepancies in the lobar share of lung volumes between our CT analysis and simulations exist and are most evident when considering the right middle lobe (Table 4). The difference in values found in our study compared to 4D CT measured values could be the result of simplifications and assumptions of our model, specifically not accounting for gravity with respect to the position of the patient during imaging and the interdependence of gravity and tissue mechanics [58] or not incorporating the interdependence of acini [59] and collateral ventilation. We would like to emphasize here that the presented work is the first step toward patient-specific modeling of COVID-19 lung dynamics and hence is not intended to make quantitative predictions about ventilation dynamics, nor is it meant to serve as a clinical decision support system for clinicians at this stage. Rather our aim in this study is to lay the foundation for and take the first steps toward developing a patient-specific modeling workflow for the investigation of COVID-19-afflicted lung dynamics. We acknowledge this study has limitations that we intend to address in future studies. For example, a uniform pleural pressure was applied to all acini in our models. However, gravity has been shown to affect the spatial distribution of pleural pressure, tissue compliance, and acinar volumes [58]. Swan et al. [60] showed that even in the healthy lung, distribution of tissue compliance is spatially non-uniform. We plan on adding the effect of gravity on pleural pressure and tissue compliance distribution in future studies. Similarly, collateral ventilation and inter-acinar interactions can affect the uniformity of pressure distributions and thus may improve the fidelity of human lung digital twins [59]. Airway deformation can also affect airflow and pressure distributions but was not included in our model; a fluid-structure interaction modeling approach would be able to capture the detailed interaction of airflow and airway wall mechanics but can be computationally more expensive compared to a reduced-order model like the one presented in our study [53,61]. Moreover, gas exchange and ventilation-perfusion coupling are important considerations for comprehensive pulmonary dynamics modeling and need to be accounted for in future studies of COVID-19-afflicted lungs [62,63]. Likewise, the lack of spirometry data of the patient and not accounting for lung motion via image registration did not allow for the implementation of patient-specific boundary conditions and direct validation of the model against 4D CT data. We are developing image registration approaches based on the work of other researchers [64,65] to account for lung motion to make the model's predictions more credible and accurate. While surfactant dysfunction accounted for acinar level damage in this study, other forms of damage, such as diffuse alveolar damage, microangiopathy, edema, and fibrosis, lead to altered lung mechanics in the COVID-19 affected lung [36]. Including more detailed alveolar mechanics models and accounting for different types of damage in the simulations, could further elucidate the impact of the disease on global lung function. Furthermore, we modeled surfactant loss at two levels of damage based on ground-glass opacities and consolidations. Damage distribution in the lung will be more accurately represented if a continuous range of damage based on CT-derived Hounsfield values is incorporated in the model. Furthermore, we acknowledge that population-based in silico studies are bringing researchers a step closer to making virtual clinical trials a reality [66]. While the modeling workflow was demonstrated for one patient here, we aim to perform the modeling and simulation for a larger cohort of patients to study intersubject variabilities of airflow as the imaging data is collected and analyzed and our modeling workflow becomes more automated and streamlined. In conclusion, this study presents a major step towards a modeling workflow capable of simulating the effect of heterogenous COVID-19-induced lung damage on ventilation dynamics in a patient-specific manner. The in silico model reasonably predicted redistribution of ventilation from severely damaged lung lobes to the lobes the least affected by viral insult in advanced COVID-19. This modeling study lays the foundation for patient-specific investigations of pulmonary ventilation in COVID-19 patients and individualized treatment strategies. Conflicts of interest The authors declare no competing interests. Data availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Author contributions A.V., E.D., S.M., and A.P., performed the method development and simulations and wrote the manuscript. M.S.P and A.J. and K.Y. collected the CT data. S.M.G. and V.M.contributed to study conceptualization and protocol development. All authors reviewed the manuscript and contributed to the scientific discussion. Declaration of competing interest None Declared. Acknowledgments This material is based on the work supported by the 10.13039/100000001 National Science Foundation under 2034964. Furthermore, this material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. 2037785 (Shea Middleton). Special thanks to nurses and respiratory therapists at Vidant Medical Center for their help with this study, and to Vidant Radiation Oncology (Greenville, North Carolina, USA) for the use of CT scanner in the facility. Special thanks to Rafel Bordas for insightful email correspondence on lung model generation. ==== Refs References 1 WHO coronavirus (COVID-19) dashboard \| WHO coronavirus (COVID-19) dashboard with vaccination data. https://covid19.who.int/. 2 Lopez-Leon, S. et al. More than 50 Long-term effects of COVID-19: a systematic review and meta-analysis. Sci. Rep. . 3 WHO COVID-19 Weekly Epidemiological Update 2021 4 Youssef H.M. Alghamdi N.A. Ezzat M.A. El-Bary A.A. Shawky A.M. A modified SEIR model applied to the data of COVID-19 spread in Saudi Arabia AIP Adv. 10 2020 125210 33304643 5 Barisione E. Fibrotic progression and radiologic correlation in matched lung samples from COVID-19 post-mortems Virchows Arch. 2020 10.1007/s00428-020-02934-1 6 Bayraktaroğlu S. Çinkooğlu A. Ceylan N. Savaş R. The novel coronavirus pneumonia (COVID-19): a pictorial review of chest CT features Diagn. Interventional Radiol. 27 2021 188 194 7 Gattinoni L. COVID-19 does not lead to a “typical” acute respiratory distress syndrome Am. J. Respir. Crit. Care Med. 201 2020 1299 1300 32228035 8 Burrowes K.S. Iravani A. Kang W. Integrated lung tissue mechanics one piece at a time: computational modeling across the scales of biology Clin. BioMech. 66 2019 20 31 9 Burrowes K.S. De Backer J. Kumar H. Image-based computational fluid dynamics in the lung: virtual reality or new clinical practice? Wiley Interdiscip. Rev.: Syst. Biol. Med. 9 2017 e1392 10 Bordas R. Development and analysis of patient-based complete conducting airways models PLoS One 10 2015 e0144105 11 Tawhai M.H. CT-based geometry analysis and finite element models of the human and ovine bronchial tree J. Appl. Physiol. 97 2004 2310 2321 15322064 12 Ismail M. Comerford A. Wall W.A. Coupled and reduced dimensional modeling of respiratory mechanics during spontaneous breathing Int. J. Numer. Methods Biomed. Eng. 29 2013 1285 1305 13 Swan A.J. Clark A.R. Tawhai M.H. A computational model of the topographic distribution of ventilation in healthy human lungs J. Theor. Biol. 300 2012 222 231 22326472 14 Ma H. Fujioka H. Halpern D. Gaver D.P. Surfactant-mediated airway and acinar interactions in a multi-scale model of a healthy lung Front. Physiol. 11 2020 941 32922307 15 Roth C.J. Ismail M. Yoshihara L. Wall W.A. A comprehensive computational human lung model incorporating inter-acinar dependencies: application to spontaneous breathing and mechanical ventilation Int. J. Numer. Methods Biomed. Eng. 33 2017 e02787 16 Berger L. A poroelastic model coupled to a fluid network with applications in lung modelling Int. J. Numer. Methods Biomed. Eng. 32 2016 1 17 17 Yoon S. An integrated 1D breathing lung simulation with relative hysteresis of airway structure and regional pressure for healthy and asthmatic human lungs J. Appl. Physiol. 129 2020 732 747 32758040 18 Burrowes K.S. A combined image-modelling approach assessing the impact of hyperinflation due to emphysema on regional ventilation–perfusion matching Comput. Methods Biomech. Biomed. Eng. Imag. Vis. 5 2015 110 126 10.1080/21681163.2015.1023358 19 Yin Y. Choi J. Hoffman E.A. Tawhai M.H. Lin C.L. Simulation of pulmonary air flow with a subject-specific boundary condition J. Biomech. 43 2010 2159 2163 20483412 20 Choi S. 1D network simulations for evaluating regional flow and pressure distributions in healthy and asthmatic human lungs J. Appl. Physiol. 127 2019 122 133 31095459 21 Hasler D. A multi-scale model of gas transport in the lung to study heterogeneous lung ventilation during the multiple-breath washout test PLoS Comput. Biol. 15 2019 e1007079 22 Fujioka H. Halpern D. Gaver D.P. A model of surfactant-induced surface tension effects on the parenchymal tethering of pulmonary airways J. Biomech. 46 2013 319 328 23235110 23 O'Connell B.F. Irvine D.M. Cole A.J. Hanna G.G. McGarry C.K. Optimizing geometric accuracy of four-dimensional CT scans acquired using the wall- and couch-mounted Varian® Real-time Position ManagementTM camera systems Br. J. Radiol. 88 2015 20140624 24 Krishnan K. Ibanez L. Turner W.D. Jomier J. Avila R.S. An open-source toolkit for the volumetric measurement of CT lung lesions Opt Express 18 2010 15256 25 3D Slicer Image Computing Platform \| 3D Slicer.. 26 Fedorov A. 3D slicer as an image computing platform for the quantitative imaging network Magn. Reson. Imag. 30 2012 1323 1341 27 Bumm R. First results of spatial reconstruction and quantification of COVID-19 chest CT infiltrates using lung CT analyzer and 3D slicer Br. J. Surg. 108 2021 28 Kassin M.T. Generalized chest CT and lab curves throughout the course of COVID-19 Sci. Rep. 11 1 11 2021 1 13 2021 33414495 29 Tawhai M.H. Pullan A.J. Hunter P.J. Generation of an Anatomically Based Three-Dimensional Model of the Conducting Airways 2000 30 Tawhai M.H. CT-based geometry analysis and finite element models of the human and ovine bronchial tree J. Appl. Physiol. 97 2004 2310 2321 15322064 31 Bordas R. Development and analysis of patient-based complete conducting airways models PLoS One 10 2015 e0144105 32 McDonough J.E. Regional differences in alveolar density in the human lung are related to lung height J. Appl. Physiol. 118 2015 1429 1434 25882386 33 Cooper F. Chaste: cancer, Heart and Soft tissue environment J. Open Source Softw. 5 2020 1848 34 Cooper F. Chaste: cancer, Heart and Soft tissue environment J. Open Source Softw. 5 2020 1848 35 Pedley T.J. Schroter R.C. Sudlow M.F. The prediction of pressure drop and variation of resistance within the human bronchial airways Respir. Physiol. 9 1970 387 405 5425201 36 Van Ertbruggen C. Hirsch C. Paiva M. Anatomically based three-dimensional model of airways to simulate flow and particle transport using computational fluid dynamics J. Appl. Physiol. 98 2005 970 980 Bethesda, Md. : 1985 15501925 37 Venegas J.G. Harris R.S. Simon B.A. A comprehensive equation for the pulmonary pressure-volume curve J. Appl. Physiol. 84 1998 389 395 9451661 38 Fujioka H. Halpern D. Gaver D.P. A model of surfactant-induced surface tension effects on the parenchymal tethering of pulmonary airways J. Biomech. 46 2013 319 328 23235110 39 Venegas J.G. Harris R.S. Simon B.A. A comprehensive equation for the pulmonary pressure-volume curve J. Appl. Physiol. 84 1998 389 395 9451661 40 Swan A.J. Clark A.R. Tawhai M.H. A computational model of the topographic distribution of ventilation in healthy human lungs J. Theor. Biol. 300 2012 222 231 22326472 41 Duarte-Neto A.N. Pulmonary and systemic involvement in COVID-19 patients assessed with ultrasound-guided minimally invasive autopsy Histopathology 77 2020 186 197 32443177 42 Polak S.B. Van Gool I.C. Cohen D. von der Thüsen J.H. van Paassen J. A systematic review of pathological findings in COVID-19: a pathophysiological timeline and possible mechanisms of disease progression Mod. Pathol. 2020 1 11 10.1038/s41379-020-0603-3 43 Mason R.J. Thoughts on the alveolar phase of COVID-19 Am. J. Physiol. Lung Cell Mol. Physiol. 319 2020 L115 L120 32493030 44 Dimbath E. Implications of microscale lung damage for COVID-19 pulmonary ventilation dynamics: a narrative review Life Sci. 274 2021 119341 45 Barisione E. Fibrotic progression and radiologic correlation in matched lung samples from COVID-19 post-mortems Virchows Arch. 2020 10.1007/s00428-020-02934-1 46 Boren H.G. Kory R.C. Syner J.C. The veterans administration-army cooperative study of pulmonary function. II. The lung volume and its subdivisions in normal men Am. J. Med. 41 1966 96 114 47 Boren H.G. Kory R.C. Syner J.C. The veterans administration-army cooperative study of pulmonary function. II. The lung volume and its subdivisions in normal men Am. J. Med. 41 1966 96 114 48 Jahani N. Assessment of regional ventilation and deformation using 4D-CT imaging for healthy human lungs during tidal breathing J. Appl. Physiol. 119 2015 1064 1074 26316512 49 Voutouri C. In silico dynamics of COVID-19 phenotypes for optimizing clinical management Proc. Natl. Acad. Sci. U. S. A 118 2021 50 Busana M. The impact of ventilation–perfusion inequality in COVID-19: a computational model J. Appl. Physiol. 130 2021 865 876 33439790 51 Herrmann J. Mori V. Bates J.H.T. Suki B. Modeling lung perfusion abnormalities to explain early COVID-19 hypoxemia Nat. Commun. 11 2020 52 Weaver L. High risk of patient self-inflicted lung injury in COVID-19 with frequently encountered spontaneous breathing patterns: a computational modelling study Ann. Intensive Care 11 2021 109 34255207 53 Pan S. yu Ding M. Huang J. Cai Y. Huang Y. zi Airway resistance variation correlates with prognosis of critically ill COVID-19 patients: a computational fluid dynamics study Comput. Methods Progr. Biomed. 208 2021 106257 54 Gattinoni L. COVID-19 pneumonia: different respiratory treatments for different phenotypes? Intensive Care Med. 2020 46 1099 1102 31690968 55 Perlman C.E. The contribution of surface tension-dependent alveolar septal stress concentrations to ventilation-induced lung injury in the acute respiratory distress syndrome Front. Physiol. 11 2020 56 Barisione E. Fibrotic progression and radiologic correlation in matched lung samples from COVID-19 post-mortems Virchows Arch. 478 2021 471 485 32989525 57 Cobes N. Ventilation/perfusion SPECT/CT findings in different lung lesions associated with COVID-19: a case series Eur. J. Nucl. Med. Mol. Imag. 47 2020 2453 2460 58 Kang W. Clark A.R. Tawhai M.H. Gravity outweighs the contribution of structure to passive ventilationperfusion matching in the supine adult human lung J. Appl. Physiol. 124 2018 23 33 29051337 59 Roth C.J. Ismail M. Yoshihara L. Wall W.A. A comprehensive computational human lung model incorporating inter-acinar dependencies: application to spontaneous breathing and mechanical ventilation Int. J. Numer. Methods Biomed. Eng. 33 2017 e02787 60 Swan A.J. Clark A.R. Tawhai M.H. A computational model of the topographic distribution of ventilation in healthy human lungs J. Theor. Biol. 300 2012 222 231 22326472 61 Ismail M. Comerford A. Wall W.A. Coupled and reduced dimensional modeling of respiratory mechanics during spontaneous breathing Int. J. Numer. Methods Biomed. Eng. 29 2013 1285 1305 62 Matuszak J. Tabuchi A. Kuebler W.M. Ventilation and perfusion at the alveolar level: insights from lung intravital microscopy Front. Physiol. 11 2020 63 Clark A.R. Tawhai M.H. Temporal and spatial heterogeneity in pulmonary perfusion: a mathematical model to predict interactions between macro-and micro-vessels in health and disease ANZIAM J. 59 2018 562 580 64 Jahani N. Yin Y. Hoffman E.A. Lin C.L. Assessment of regional non-linear tissue deformation and air volume change of human lungs via image registration J. Biomech. 47 2014 1626 1633 24685127 65 Yin Y. Hoffman E.A. Lin C.-L. Local Tissue-Weight-Based Nonrigid Registration of Lung Images with Application to Regional Ventilation Spiromics View Project Cardiopulmonary Mechanisms Affecting Cognition in COPD View Project Local Tissue-Weight-Based Nonrigid Registration of Lung Images with Application to Regional Ventilation 2009 10.1117/12.811715 66 Tawhai M.H. Lin C.-L. Image-based modeling of lung structure and function J. Magn. Reson. Imag. 32 2010 1421 1431
PMC009xxxxxx/PMC9005225.txt	==== Front Infect Genet Evol Infect Genet Evol Infection, Genetics and Evolution 1567-1348 1567-7257 The Author(s). Published by Elsevier B.V. S1567-1348(22)00079-X 10.1016/j.meegid.2022.105282 105282 Article Comparative genomics, evolutionary epidemiology, and RBD-hACE2 receptor binding pattern in B.1.1.7 (alpha) and B.1.617.2 (delta) related to their pandemic response in UK and India Chakraborty Chiranjib a⁎1 Sharma Ashish Ranjan b1 Bhattacharya Manojit c1 Mallik Bidyut d Nandi Shyam Sundar e Lee Sang-Soo b⁎ a Department of Biotechnology, School of Life Science and Biotechnology, Adamas University, Barasat-Barrackpore Rd, Jagannathpur, Kolkata, West Bengal 700126, India b Institute for Skeletal Aging &Orthopedic Surgery, Hallym University-Chuncheon Sacred Heart Hospital, Chuncheon, Gangwon-Do 24252, Republic of Korea c Department of Zoology, Fakir Mohan University, Vyasa Vihar, Balasore 756020, Odisha, India d Department of Applied Science, Galgotias College of Engineering and Technology, Knowledge Park-II, Greater Noida, Uttar Pradesh 201306, India e ICMR-National Institute of Virology, (Mumbai unit), Indian Council of Medical Research, Haffkine Institute Compound, A. D. Marg, Parel, Mumbai, 400012, India ⁎ Corresponding author. 1 Authors contributed equally to this work. 13 4 2022 13 4 2022 10528216 9 2021 4 4 2022 8 4 2022 © 2022 The Author(s). Published by Elsevier B.V. 2021 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background The massive increase in COVID-19 infection had generated a second wave in India during May–June 2021 with a critical pandemic situation. The Delta variant (B.1.617.2) was a significant factor during the second wave. Conversely, the UK had passed through the crucial phase of the pandemic from November to December 2020 due to B.1.1.7. The study tried to comprehend the pandemic response in the UK and India to the spread of the B.1.1.7 (Alpha, UK) variant and B.1.617.2 (Delta, India) variant. Methods This study was performed in three directions to understand the pandemic response of the two emerging variants. First, we served comparative genomics, such as genome sequence submission patterns, mutational landscapes, and structural landscapes of significant mutations (N501Y, D614G, L452R, E484Q, and P681R). Second, we performed evolutionary epidemiology using molecular phylogenetics, scatter plots of the cluster evaluation, country-wise transmission pattern, and frequency pattern. Third, the receptor binding pattern was analyzed using the Wuhan reference strain and the other two variants. Results The study analyzed the country-wise and region-wise genome sequences and their submission pattern, molecular phylogenetics, scatter plot of the cluster evaluation, country-wise geographical distribution and transmission pattern, frequency pattern, entropy diversity, and mutational landscape of the two variants. The structural pattern was analyzed in the N501Y, D614G L452R, E484Q, and P681R mutations. The study found increased molecular interactivity between hACE2-RBD binding of B.1.1.7 and B.1.617.2 compared to the Wuhan reference strain. Our receptor binding analysis showed a similar indication pattern for hACE2-RBD of these two variants. However, B.1.617.2 offers slightly better stability in the hACE2-RBD binding pattern through MD simulation than B.1.1.7. Conclusion The increased hACE2-RBD binding pattern of B.1.1.7 and B.1.617.2 might help to increase the infectivity compared to the Wuhan reference strain. Graphical abstract Unlabelled Image Keywords Comparative genomics Evolutionary epidemiology Receptor binding Variance of UK and India Abbreviations BFE, Binding free energy CDC, Centre for Disease Prevention and Control ECDC, European Centre for Disease Prevention and Control GISAID, Global Initiative on Sharing All Influenza Data hACE2, Human angiotensin-converting enzyme 2 ICU, Intensive Care Unit MDS, Molecular dynamics simulation MM/GBSA, Molecular mechanics/generalized born surface area NAMD, Nanoscale molecular dynamics PDB, Protein Data Bank RBD, Receptor-binding domain RMSF, Root mean square fluctuation RMSD, Root mean square deviation S-glycoprotein, Spike glycoprotein VMD, Visual molecular dynamics VOC, Variant of concern VOI, Variant of interest WHO, World Health Organization nAb, Neutralizing antibody ==== Body pmc1 Introduction India has passed through a massive pandemic. A severe increase in the number of infected cases has been observed since March 2021, which has created the current COVID-19 surge in India. New COVID-19 patients were approximately 273,810 (as of April 19, 2021) in a single day, the maximum number of daily cases recorded from daily data worldwide. This incident set a new record during the pandemic (Mallapaty, 2021). A sudden increase in infection and death cases resulted in the second wave and a massive public health crisis in India. At present, this is an international concern (Rubin et al., 2021; Thiagarajan, 2021). At first, a sudden rise in infection cases was observed in Maharashtra (India)in March 2021. During that period, Indian scientists attempted to determine the biological cause of the spread of the infection. They found that one variant might be responsible for the sudden burst in the cases. A distinct B.1.617.2 variant was identified from India with three significant mutations in spike proteins from the genome sequence (Chakraborty et al., 2021a). Now the variant is circulating throughout the globe, and the dominance of the B.1.617.2 variant is noted in several countries (Bhattacharya et al., 2021a). The identified three important mutations were L452R, E484Q, and P681R (Cherian et al., 2021). Among these three mutations, L452R and E484Q are located in the RBD region. The B.1.617.2 variant was primarily identified in December 2020 in India (WHO, 2020). This B.1.617.2 variant is a significant variant in India and is currently circulating. It is claimed that this variant is associated with more infections and is linked with India's current COVID-19 surge, which is regarded as the second wave of COVID-19 (Vaidyanathan, 2021). B.1.617.2 variant has spread to approximately 50 nations, including Singapore, the UK, the USA, Germany, and Australia. Due to its high spreading capacity and increased infectivity, the ECDC declared B.1.617.2 as a variant of concern (VOC)(ECDC, 2021a). However, the WHO affirmed B.1.617.2 variant as a variant of interest (VOI) (WHO, 2020). The CDC, USA, described the B.1.617.2 variant as VOI (Control and Prevention, 2021). Several B.1.617.2 variant sequences were uploaded to the GISAID database from different countries. Most of the sequences were gathered from India and other countries such as the UK, Singapore, and the USA (Rambaut et al., 2020). Three distinct sub-lineages were observed within B.1.617. The first sub-lineage is B.1.617.1, the second sub-lineage is B.1.617.2, and the third sub-lineage is B.1.617.3. Some significant mutations were associated with spike protein alteration within B.1.617.2. The major mutations observed were L452R, E484Q, and P681R (ECDC, 2021a, b). L452R is related to a reduction in therapeutic antibodies and is associated with neutralization using convalescent plasma. This mutation is also associated with increased transmissibility (Deng et al., 2021). While, E484Q has been reported to decrease neutralization in convalescent sera (Jangra et al., 2021). Furthermore, the P681R mutation is located outside the RBD, near the furin cleavage site of the S-glycoprotein. This mutation may be associated with cell entrance and infectivity (Afrin et al., 2021; Cherian et al., 2021). Another significant mutation, D614G, was reported by researchers and was found outside the RBD (Zhang et al., 2020). However, more scientific data are needed regarding variant B.1.617.2, such as the transmissibility and reproduction number (R0) variant, the possibility for hospitalization and ICU admission, more structural detail in the mutation site, and receptor binding pattern. Another more infective and rapidly spreading variant of SARS-CoV-2 is B.1.1.7 (Graham et al., 2021). This variant was first observed in different places in the UK and spread to 122 countries (Andrew Banchich, 2021; Volz et al., 2021). Volz et al. collected whole-genome SARS-CoV-2 sequences from the UK. They identified 9134 VOC sequences from London, 4413 VOC sequences from East England, and 5,609,413 VOC sequences from southeast England. In addition, they identified 52,795 non-VOC genomes. They observed that VOCs had spread widely across England. This variant's reproduction number (R0) is higher than other variants. The study reported that the R0 of this variant was 50%–100% higher (Volz et al., 2021). In addition, another study showed a high proportion of infections (>80%) among patients due to B.1.1.7 (Graham et al., 2021). Grint et al. observed that B.1.1.7 may be responsible for the augmented death rate among patients (Grint et al., 2021). Therefore, due to the high risk, the variant was also declared as VOC by the CDC, USA (Control and Prevention, 2021), WHO (WHO, 2020), and ECDC (ECDC, 2021c). The ECDC demonstrated that this variant might cause a higher possibility of hospitalization and ICU admission (ECDC, 2021c). The variant has significant mutations. One primary mutation is N501Y, which is located in the RBD. Ostrov also confirmed this significant mutation associated with B.1.1.7 (N501Y), which increased the binding ability between the S-glycoprotein and ACE2 (Angiotensin-Converting Enzyme 2) receptor (Ostrov, 2021). Widera et al. found a high viral load in patients infected with B.1.1.7 variant having the same mutation (Widera et al., 2021).A decreased affinity was observed in B.1.1.7 variant S-glycoprotein and neutralizing mAbs (monoclonal antibodies) due to the mutation (N501Y substitution) (Cheng et al., 2021). The variant has another mutation, P681H, located outside the RBD and near the furin cleavage site (S1/S2 furin cleavage) (Lubinski et al., 2021). The D614G mutation, also observed in S-glycoprotein, was reported in this variant and was related to increased infectivity and virion density (Zhang et al., 2020). Several single-molecule drugs and several combination therapies were assessed to comprehend the treatment strategies against COVID-19. In this direction, numerous small to major clinical trials were performed, and the significant clinical trials are ACTT-1, ACTT-2, ACTT-3, ACTT-4 study group RECOVERY trials, etc. Several single drug molecules were assessed like Ritonavir/Lopinavir, Hydroxychloroquine, Remdesivir, Favipiravir, Dexamethasone, Ivermectin, and Heparin. At the same time, several immunotherapeutic drugs were also evaluated, such as interferons, baricitinib, mavrilimumab, and tocilizumab. Some clinicians also used convalescent plasma therapy to treat COVID-19 patients. It has been noted that single dosages therapy or combination therapy of Remdesivir, Dexamethasone, Baricitinib, Tofacitinib, Tocilizumab, and Sarilumab might provide better results against hospitalized COVID-19 patients (Chakraborty et al., 2021b; Alam et al., 2021). Simultaneously, several vaccines have been approved for the immunization against the SARS CoV-2 virus. Among the vaccines, DNA, mRNA, protein, and attenuated vaccines are important platforms for vaccine development. The significant vaccines are Johnson and Johnson, AstraZeneca/Oxford vaccine, Moderna, Pfizer/BioNTech, Sinopharm, Sinovac, COVAXIN, and Covovax (Chakraborty et al., 2021c; Chakraborty et al., 2021d; Chakraborty et al., 2021e). Due to the continuous generation of variants, some major issues are different escape or resistance abilities such as immune escape, nAb escape, partial vaccine escape, and therapeutic resistance. Developing a new drug or vaccine that can protect against these significant variants and their mutations is necessary. Therefore, to understand the significant variants and their major mutations, comparative genomics epidemiological and evolutionary epidemiology properties are necessary for new therapeutic and vaccine development to combat the variants. This study was aimed in three directions to comprehend the biological and epidemiological properties of two emerging variants, namely B.1.1.7 (Alpha, UK origin) and B.1.617.2 (Delta, India origin).The first aim was to comprehend the comparative genomics such as country-wise and region-wise genome sequences and their submission patterns, the mutational landscape of two variants, and the structural landscape of significant mutations such as N501Y, D614G, L452R, E484Q, and P681R of these two variants. The second aim was to evaluate the evolutionary epidemiology, such as molecular phylogenetics, the molecular clock of the evolution, country-wise geographical distribution, and transmission pattern. The third aim was to analyze the receptor binding interaction, such as molecular docking analysis, calculation of binding free energy, and molecular dynamics simulation to understand the hACE2 receptor-RBD interactions. 2 Material and methods 2.1 Data collection Data was collected for the Wuhan strain, B.1.1.7 and B.1.617.2, from different sources such as NCBI, CDC(USA),(Control and Prevention, 2021), WHO (WHO, 2020), and ECDC (ECDC, 2021c). Relevant keywords were searched in several significant databases such as PubMed (Liu, 2020; Zuo et al., 2021), Web of Science (Farooq et al., 2021), and Google Scholar. Different keywords for database search such as “B.1.1.7,” “B.1.617.2,” “variants of interest (VOI),” “variants of concern (VOC),” “variants of consequence” were used in this study. The data was also collected using several databases and servers such as Github and Pango lineages (O'Toole and McCrone, 2020), Pango lineages (O'Toole et al., 2020), SARS-CoV-2 resources GISAID (Velazquez et al., 2020), and Nextstrain server (Hadfield et al., 2018; Nextstrain, 2020). The NextStrain server fetches the records from the GISAID server. All the data were obtained from the different servers during the first week of July 2021.The statistical calculation and verification were performed using different advanced statistical tools and techniques wherever it is needed, such as MATLAB, and PAST 4.03 software (Hammer et al., 2001; MathWorks, 1992). 2.2 Comparative genomics, evolutionary epidemiology analysis 2.2.1 Country-wise and region-wise genome sequences analysis of B.1.617.2 and B.1.1.7 and their submission pattern For analysis of genome sequences and their submission pattern, for these two variants, we used servers such as GitHub and Pango lineages (O'Toole and McCrone, 2020), Pango lineages (O'Toole et al., 2020), and GISAID (Velazquez et al., 2020). 2.2.2 Molecular phylogenetics, the molecular clock of evolution, country-wise transmission prototype and geographical positioning B.1.617.2 and B.1.1.7 Here, we used servers and databases like Nextstrain (Hadfield et al., 2018; Rambaut et al., 2020) and GISAID (Velazquez et al., 2020) to analyze the molecular phylogenetics, the molecular clock of evolution, country-wise geographical distribution, and transmission pattern, and entropy diversity of these two variants. The Nextstrain server is used to evaluate a pathogen's real-time evolution and is a very efficient tool for phylodynamics analysis (Hadfield et al., 2018). At the same time, GISAID is one of the largest genome sequences servers. This is one of the significant resources for analyzing SARS CoV-2 variants during pandemics, especially for VOC and VOI. Several scientists have used the GISAID to analyze the millions of sequences in different variants such as B.1.1.7 (Alpha; originated from the UK), B.1.617.2 (Delta; originated from India), P1 (Gamma; originated from Brazil), B.1.351 (Beta; originated from South Africa) (Kalia et al., 2021; Zelenova et al., 2021; Khare et al., 2021). The classification methodology as proposed by Rambaut et al. (2020))was followed. 2.3 Analysis of the structural landscape of significant mutations such as N501Y, D614G L452R, E484Q, and P681R, which are found in these two variants The COVID-3D server was used for the structural analysis of significant mutations (N501Y, D614G L452R, E484Q, and P681R) in two emerging variants (Portelli et al., 2020). Subsequently, the GISAID database (Velazquez et al., 2020) was used for significant mutational landscape analysis. COVID-3D server is used to analyze and visualize the structural landscape of genetic variation (Portelli et al., 2020). 2.4 Molecular docking analysis In this analysis, the binding interaction between human ACE2 (hACE2) and RBD of SARS-CoV-2 in the Wuhan strain (PDB ID: 6VW1) and more infective variants such as B.1.1.7 (PDB ID: 7BWJ) and B.1.617.2were predicted. The Wuhan strain was used for molecular docking experiments, and the infective variants were used for comparative analysis. Using UCSF chimera software, the PDB files were pre-processed and downloaded from the RCSB protein data bank (Kouranov et al., 2006). The S-glycoprotein 3D structure of B.1.617.2was not found in the RCSB PDB. Therefore, the generated RDD of the B.1.617.2 variant by introducing two significant mutations in control RBD, 452R and E484Q. The molecular docking was performed using the HDOCK server and visualized and edited the PDB structure using PyMOL software. The HDOCK server predicts the binding complexes between two molecules, such as protein-protein docking, using a unique hybrid algorithm. The hybrid algorithm was designed using a template-based method and an ab initio free docking modeling server (Yan et al., 2020; De Vries et al., 2010; Yan et al., 2017). 2.5 Binding free energy (BFE) calculation In this study, the HawkDock online server was employed to evaluate the BFE of our docked complexes using MM/GBSA approaches (Weng et al., 2019; Chen et al., 2016;). For MM/GBSA analysis, the binding free energy (ΔGbind) of the system can be defined as a change in free energy, which is expressed as(1) ΔGbind=ΔEMM+ΔGsolv–TΔS In the above, ΔEMM depicts the changes in the MM energy of the system, ΔGsolv indicates changes in solvation free energy, and TΔS denotes the total system binding entropy. Further equations calculate how the ΔEMM energy and ΔGsolv solvation free energy are obtained from the system. The equation is(2) ΔEMM=ΔEinternal+ΔEelectrostatic+ΔEvdw WhereΔEinternal is the change in internal MM energy, ΔEelectrostatic is the change in electrostatic energy, and ΔEvdw is the change in van der Waals energy.(3) ΔGsolv=ΔGGB+ΔGSA The equation represents deviations of solvation free energy combined with non-electrostatic solvation (ΔGSA) and energy electrostatic solvation (ΔGGB). 2.6 Molecular dynamics simulation (MDS) MDS is an efficient method for identifying and predicting biological or chemical processes at the atomic level (Cao et al., 2017). To generate the MDS analysis, the NAMD 2.14b2_win64-multicore software package was used for three docked complexes (Kalé et al., 1999; Phillips et al., 2005; Phillips et al., 2020: Melo et al., 2018) and optimized the parameters using CHARMM22 and CHARM36 all-atom force field (MacKerell Jr et al., 2000). Protein topologies were prepared using the psf module, water solvation was prepared around the complex molecules, and equilibrated TIP3P (transferable intermolecular potential 3P) water was added as solvation. After that, the system was placed at least 1.0 nm from the box edge, and 56 sodium ions and 30 Cl ions were included in the solvation molecule, which was used to neutralize the solvated protein structure. The energy minimization of the system uses the NAMD minimizer of 1000 steps using an algorithm called the steepest descent algorithm. The system equilibration was then achieved in two-step: NVT ensembles and NPT ensembles for 1000 ps. MD simulations of the equilibrated system were 20 ns (Timestep) at 2 fs speed. The simulation trajectories were analyzed with VMD 1.9.3 graphic visualization software and evaluated with some other types of secondary analysis such as bond angle, dihedral, etc. (Humphrey et al., 1996). The entire study methodology is represented through a flowchart to understand the overall study process (Fig. 1A). A schematic diagram of two variants along with the Wuhan strain is illustrated in Fig. 1B. Another schematic diagram of these two variants and their pandemic response is depicted in Fig. 1C. Fig. 1 The flowchart of the study methodology and schematic diagram of the S-protein of three strains (one control and two variants). (A) Flowchart of our study methodology that describes the workflow in three directions which are comparative genomics, evolutionary epidemiology, and receptor binding. (B) The schematic diagram shows the features of RBD of one control (Wuhan strain) and two variants (B.1.617.2 and B.1.1.7). The diagram shows significant mutations in B.1.1.7 (N501Y, D614G, and P681R) and B.1.617.2 (N501Y, D614G, L452R, E484Q, and P681R) (C).Two variants (B.1.1.7 and B.1.617.2) and their pandemic response in their country of origin.B.1.1.7 (Alpha) originated in the UK, and it is responsible for the second wave generation in the UK in the second half of 2020. Similarly, B.1.617.2 (Delta) is originated in India, and the variant is responsible for the generation second wave in India during the first half of 2021. Fig. 1 3 Results 3.1 Country-wise and region-wise submitted genome sequences number and pattern of these two variants The PANGO lineage data shows these two variants' region-wise submitted genome sequence pattern. Data indicates that more sequences were submitted for the B.1.1.7 (Fig. 2A) than the B.1.617.2 variant (Fig. 2B). The country-wise submission patterns of these variants were also mapped. Furthermore, genome sequences for the B.1.1.7 (Fig. 2C) compared to the B.1.617.2 variant were also found (Fig. 2D). We observed the highest number of genome sequences for B.1.1.7, submitted from various places in the UK (247,549 no), and the second-highest number of genome sequences for this variant was introduced from the USA (175,511 no). The genome summation pattern of the B.1.1.7 variant indicates that this variant was transmitted to different parts of the world from the UK.Fig. 2 Country-wise and region-wise submitted genome sequences pattern of two variants. (A) Region-wise submitted genome sequences pattern of B.1.1.7. (B) Region-wise submitted genome sequences pattern of B.1.617.2. (C) Country-wise submitted genome sequences number of B.1.1.7. It shows the maximum sequences were deposited from the UK. The Scatter plot of the cluster shows the dots of B.1.617.2, located in the upper position of the regression line. (D) Country-wise submitted genome sequences number of B.1.617.2. It also shows the maximum sequences were deposited Form India. Fig. 2 For the B.1.617.2 variant, 6732 genome sequences were found from India, 63,629 genome sequences from the UK, and 4271 genome sequences from the USA. The study shows that the B.1.1.7 variant is widely spread and well-studied. However, this B.1.617.2 variant has just started to spread from India to other countries. Probably, this is the time to alert other countries for the spared of the B.1.617.2 variant. 3.2 Molecular phylogenetics of these two variants Molecular phylogenetic analysis was performed using the genome sequences of these two variants. A radial phylogenetic tree of the B.1.1.7 variant was developed using 890 sequences between September 2020 and May 2021 (Fig. 3A). A radial phylogenetic tree of the B.1.617.2 variant used 66 sequences between January 2021 and May 2021 (Fig. 3B).Fig. 3 Evaluation of molecular phylogenetics and scatter plot of a cluster of two variants. (A) Radial phylogenetic tree of the B.1.1.7 using 890 sequences. (B) Radial phylogenetic tree of the B.1.617.2 using 66 sequences. (C) Scatter plot of a cluster of the B.1.1.7 using 890 sequences. The Scatter plot of the cluster shows the dots of B.1.1.7, located in the upper position of the regression line. (d) Scatter plot of a cluster of the B.1.617.2 using 66 sequences. The Scatter plot of the cluster shows the dots of B.1.617.2, located in the upper position of the regression line. Fig. 3 This study used the Nextstrain server and GISAID to access the molecular phylogenetics, the molecular clock of evolution, country-wise geographical distribution, transmission pattern, etc., two variants. Researchers have frequently applied these two servers to understand the emergence and evolution of SARS-CoV-2 variants (Islam et al., 2021; Alai et al., 2021). 3.3 Molecular clock of the evolution of these two variants The molecular clock estimation was performed and depicted as scatter plots, showing the cluster genome samples of the sampling date and estimated mutation with regression lines. The advanced tools estimated the substitution rate (23.948 substitutions per year). The molecular clock of the B.1.1.7 variant using 1132 sequences (Fig. 3C) and the B.1.617.2 variant using 235 sequences was developed (Fig. 3D). The molecular clock was generated using a linear regression model. This model shows the sample values scatter as plots that are spread around the regression line. The plot illustrates the general patterns of the molecular clock and the mutations of these two variants through the plotted points. 3.4 Speard and country-wise transmission prototype of these two variants The geographical distribution (divisional) and country-wise transmission prototype of B.1.1.7 were evaluated. The geographical spread and country-wise transmission prototype highlighting B.1.1.7 in Europe are represented in Fig. 4A and B, respectively. This lineage emerged in the UK. From the transmission model, it appears that the variant is transmitted all over the countries of the European Union, different parts of the USA, and Canada. It is also widespread in several parts of South Africa, Latin America, and Asia. It also got transmitted to India, Sri Lanka, Australia, and the Philippines. The geographical distribution (divisional) and country-wise transmission pattern of the B.1.617.2 variant is also depicted. The geographical spread and country-wise transmission pattern highlighting B.1.617.2 in Asia are represented in Fig. 4C and D, respectively. The lineage was originated in India and transmitted to several regions of the UK, the USA, and some parts of South Africa. This variant also got transmitted to Australia, Malaysia, and Thailand.Fig. 4 Geographical distribution and country-wise transmission pattern of the two variants. (A) Geographical distribution of the B.1.1.7. (B) Country-wise transmission pattern of the B.1.1.7.(C) Geographical distribution of the B.1.617.2. (D) Country-wise transmission pattern of the B.1.617.2. Fig. 4 3.5 Entropy diversity of each nucleic acid position throughout the genome, S-glycoprotein, and RBD region of these two variants The entropy diversity of each nucleic acid position throughout the genome of the B.1.1.7 variant in a frame was evaluated (Fig. 5A ). The entropy diversity of each nucleic acid position all over the genome of the Delta variant in a frame was also recorded (Fig. 5B). Fig. 5 Frequencies of two variants(A) Entropy diversity of each nucleic acid position throughout the genome of the B.1.1.7.(B) Entropy diversity of each nucleic acid position throughout the genome of the B.1.617.2.(C)The entropy diversity of each nucleic acid position of S-glycoprotein of B.1.1.7.(D) The entropy diversity of each nucleic acid position of S-glycoprotein of B.1.617.2.(E) The entropy diversity of each nucleic acid position of the RBD region of the B.1.1.7. (F) The entropy diversity of each nucleic acid position of the RBD region of the B.1.617.2. Fig. 5 Furthermore, the entropy diversity of each nucleic acid position of the S-glycoprotein of the B.1.1.7 variant was developed, shown in Fig. 5C. The entropy diversity of each nucleic acid position of the S-glycoprotein of the B.1.617.2 variant was also evaluated (Fig. 5D). Finally, we evaluated the entropy diversity of each nucleic acid position in the RBD region of the B.1.1.7 variant (Fig. 5E). We also evaluated the entropy diversity of each nucleic acid position of the RBD region of the B.1.617.2 variant (Fig. 5F). We found the maximum entropy of each nucleic acid position throughout the genome, S-glycoprotein, and RBD region of B.1.617.2, compared to B.1.1.7. 3.6 Mutational landscape of S-glycoprotein of these two variants We evaluated the mutational landscape of all mutations in these two significant variants. Based on the results, we found several mutations in a different positions in the S-glycoprotein of the B.1.1.7 variant (Fig. 6A). We found nineteen mutations in codon 70, six in codon 501, and two in codon 614 in this variant. Several mutations at different positions in the S-glycoprotein of the B.1.617.2 variant were also observed (Fig. 6B). We also observed fifteen mutations in codon 142, one mutation in codon 452, one mutation in codon 680, and five mutations in codon 950 in this variant.Fig. 6 Mutational landscape of S-glycoprotein of two variants.(A) Mutations in a different particular position in the S-glycoprotein of the B.1.1.7.(B) Mutations in a different position in the S-glycoprotein of the B.1.617.2. Fig. 6 3.7 The structural landscape of significant mutations (N501Y, D614G L452R, E484Q, and P681R) found in these two variants The structural landscape of mutation analysis revealed significant mutations, such as N501Y, D614G L452R, E484Q, and P681R, which are frequently found in these two variants. We evaluated the N501Y mutation in the UK variant (Alpha). Within the N501Y structure, the amino acid was altered from Asn501→Tyr. The structural analysis of N501Y showed different forms of interactions with other residues such as Q506 etc. In this variant mutation, the interactivity among the other residues in the wild form (N501) is shown in Fig. 7A . Again, we have shown a schematic interaction demonstrating the interaction of the wild residue via different bond formations with the nearby residues. The schematic interaction pattern among the residues in the wild form (N501) is shown in Fig. 7B and displays diverse types of interaction with different types of bonds with other nearby residues. Likewise, we have shown the interaction of the mutant-type residue and its molecular interaction with the other residues. The mutant-type residue (Y501) and its molecular interaction among the other nearby residues are shown in Fig. 7C. The schematic diagram shows the mutant-type residue (Y501) and its interaction among the other nearby residues (Fig. 7D). It extensively shows the different types of interaction with different types of bond formation among other nearby residues.Fig. 7 The structural landscape of significant mutations of the B.1.1.7.(A) Molecular association between the residues in the wild-type N501Y mutation. (B) The schematic diagram shows the molecular association between the residues in the wild-type N501Y mutation. (C) Molecular association between the residues in the mutant type N501Y mutation. (D) The schematic diagram shows the molecular association between the residues in the mutant type N501Y mutation. (E) Molecular association between the residues in the wild-type D614G mutation. (F) The schematic diagram shows the molecular association between the residues of wild type D614G mutation. (G) Molecular association between the residues in the mutant type D614G mutation. (H) The schematic diagram shows the molecular association between the residues of mutant type D614G mutation. Fig. 7 Again, like the previous point AA mutation, we also evaluated the D614G mutation, which was identified as the B.1.1.7 variant. In the D614G structure, the amino acid changed from Asp614→Gly. The structural evaluation of the D614G mutation was performed, showing different forms of interactions. In this variant mutation, the molecular association of the wild type (D614) and the surrounding residues is shown in Fig. 7E. At the same time, a schematic interrelation diagram has been shown to depict the different bond formation and interaction between the wild type residue (D614) and the adjacent residues (Fig. 7F). Similarly, the mutant type residue (G614) and its interactions between the neighboring residues are shown in Fig. 7G. An interrelation diagram has also been depicted to show the interaction and bond formation between the mutant residue (G614) and adjacent residues (Fig. 7H). Furthermore, the L452R mutation was evaluated, which was identified in B.1.617.2 variant. In the L452R structure, the amino acid was altered from Leu452→Arg. Structural analysis of the L452R mutation showed different forms of interactions. In this variant, the interrelation between the wild-type residue (L452) and adjacent residues is shown in Fig. 8A . Similarly, we have drawn a schematic interaction diagram between the wild-type residue (L452) and nearby residues, and a schematic diagram for interrelation is shown in Fig. 8B. Again, we have shown the molecular association between the mutant type residue (R452) and surrounding residues (Fig. 8C). To make the interaction clearer, we depicted a schematic interaction diagram and a schematic interrelation between the mutant type residue (R452) and adjacent residues (Fig. 8D). Fig. 8 The structural landscape of significant mutations of the B.1.617.2.(A) Molecular association between the residues in the wild-type L452R mutation. (B) The schematic diagram shows the molecular association between the residues in the wild-type L452R mutation. (C) Molecular association between the residues in the mutant type L452R mutation. (D) The schematic diagram shows the molecular association between the residues of mutant type L452R mutation. (E) Molecular association between the residues in the wild-type E484Q mutation. (F) The schematic diagram shows the molecular association between the residues in the wild-type E484Q mutation. (G) Molecular association between the residues in the mutant type E484Q mutation. (H) The schematic diagram shows the molecular association between the residues of mutant type E484Q mutation. (I) Molecular association between the residues in the wild-type P681R mutation. (J) The schematic diagram shows the molecular association between the residues in the wild-type P681R mutation. (K) Molecular association between the residues in the mutant type P681R mutation. (L) The schematic diagram shows the molecular association between the residues of mutant type P681R mutation. Fig. 8 We next evaluated the E484Q mutation, which was identified in B.1.617.2 variant. In the E484Q structure, the amino acid changed from Glu484→Gln. Structural analysis of the E484Q mutation showed different interactions between the residues. In this AA variant (E484Q), the interaction between the wild type (E484) is shown in Fig. 8E. We have drawn a schematic representation to show the interaction and the bond formation between the wild type (E484) and nearby residues (Fig. 8F). Simultaneously, we have also shown the interrelation between mutant type residue (Q484) and adjacent residues in Fig. 8G. A schematic representation was developed to show the molecular association among the mutant type residue (Q484) and adjacent residues, and a schematic interaction figure is shown in Fig. 8H. Finally, we evaluated the P681R mutation identified in the B.1.617.2 variant. In the P681R structure, the amino acid was altered from Pro681→Arg. The structural analysis of P681R showed different forms of interactions between the residues. In variant P681R, the wild-type residue (P681) and its interrelation with other residues are shown in Fig. 8I. Furthermore, a schematic diagram illustrates the molecular association among the wild-type residue (P681) and neighboring residues (Fig. 8J). The interactivity between the residue of the mutant type (R681) and others is shown in Fig. 8K. Simultaneously, a schematic diagram was drawn to visualize the graphical interaction between the mutant type residue (R681) and other residues (Fig. 8L). COVID-3D server is noteworthy severer to analyze the mutational landscape. Presently, many researchers are using the COVID-3D server to understand the mutational landscape of SARS-CoV-2 variants. Jacob et al. have used the COVID-3D server in their mutational assessment of the variants and apprehend the destabilizing or stabilizing properties (Jacob et al., 2020). We have also used previously to understand the mutational residue characteristics of the SARS-CoV-2 variants using the COVID-3D server (Chakraborty et al., 2021f). 3.8 Molecular docking analysis The results of the molecular docking analysis of the Wuhan strain are illustrated in Fig. 9A and B. Fig. 9A represents the top view (zoom) of the interaction between hACE2 and RBD residues of the Wuhan strain. Fig. 9B depicts the bottom view (zoom) of the interaction between hACE2 and RBD residues of the Wuhan strain.Fig. 9 The diagram shows the protein-protein docking complexes of hACE2 and RBD. The hACE2 is colored by cyan, and its interacting RBD domains are colored by purple. Comparison of the binding interaction of residues of RBD and hACE2 of Wuhan strain and other two variants. (A) Top zoom view of the interaction between RBD and hACE2 residues of the Wuhan strain (B) Bottom zoom view of the interaction between RBD and hACE2 residues of the Wuhan strain. (C) Top zoom view interaction between RBD and hACE2 residues of B.1.1.7 variant. (D) The bottom zoom view of the interaction between RBD and hACE2 residues of B.1.1.7 variant and yellow color stick represent the mutation of N501Y. (E) The top zoom view of the interaction between RBD and hACE2 residues of B.1.617.2 variant and yellow color stick represent the mutation of L452R and E484Q. (F) Bottom zoom view of the interaction between RBD and hACE2 residues of B.1.617 variant. Fig. 9 The molecular docking analysis of the B.1.1.7 variant is shown in Fig. 9C and D. Fig. 9C shows the top view (zoom) of the interaction between hACE2 residues and RBD residues of the B.1.1.7 variant. Similarly, Fig. 9D illustrates the bottom view (zoom) of the interaction between hACE2 and RBD residues of B.1.1.7. The results of the molecular docking evaluation of the B.1.617.2 variant are shown in Fig. 9E and F. Fig. 9E illustrates the top view (zoom) of the interaction between hACE2 residues and RBD residues of the B.1.617.2 variant. Simultaneously, Fig. 9F shows the bottom view (zoom) of the interaction between RBD and hACE2residuesof the Delta variant. The molecular docking study shows the comparative hydrogen bonds formed during the interaction of hACE2 and RBD residues of the three variants (Wuhan strain, Alpha variant, and Delta variant) and is presented in Table 1 . The study showed that the molecular interactivity of the residues of hACE2 and RBD of B.1.1.7 formed more hydrogen bonds during the interaction (12 numbers) compared to the B.1.617.2 variant (11 numbers).Table 1 Comparative analysis of hydrogen bonds formed during the interaction of hACE2 and RBD residues of one control and two variants. Table 1Sl. no. Variant Number of hydrogen bonds (H-bonds) Residues involve in hydrogen bond (ACE2_Spike protein) Hydrogen bond length 1 Wuhan strain 9 TYR83-ASN487 ASP30-LYS417 ASP38-TYR449 GLN42-TYR449 GLN42-GLy446 TYR41-THR500 LYS353-GLY496 LYS353-GLY502 ASP355-THR500 2.302 Å 2.423 Å 2.369 Å 2.582 Å 2.885 Å 2.553 Å 2.734 Å 2.703 Å 3.529 Å 2 B.1.1.7 (UK) variant 11 GLN24-ASN487 TYR83-ASN487 ASP30-LYS417 GLU35-GLN493 ASP38-TYR449 GLN42-TYR449 GLN42-GLy446 TYR41-THR500 LYS353-GLY496 LYS353-GLY502 LYS353-TYR501 2.688 Å 2.788 Å 2.905 Å 3.132 Å 2.695 Å 2.778 Å 3.244 Å 2.707 Å 3.084 Å 2.784 Å 2.766 Å 3 B.1.617 (Indian) variant 10 GLN24-ASN487 TYR83-ASN487 ASP30-LYS417 GLU35-GLN493 ASP38-TYR449 GLN42-TYR449 GLN42-GLy446 TYR41-THR500 LYS353-GLY496 LYS353-GLY502 2.688 Å 2.788 Å 2.905 Å 3.132 Å 2.695 Å 2.788 Å 3.244 Å 2.707 Å 3.084 Å 2.784 Å Molecular docking performed by the HDOCK server revealed that B.1.617.2 has a higher binding affinity (Fig. 1B), which was further analyzed using the MM/GBSA technique. The study found increased molecular interactivity between hACE2-RBD binding of B.1.1.7 and B.1.617.2 compared to the Wuhan reference strain. In this study, the HDOCK server was used to dock the RBD and hACE2 residues of B.1.1.7. and B.1.617.2. HDOCK server is used to perform protein-protein docking very frequently by researchers. Recently, Calcagnile et al. have used the HDOCK server in their molecular docking study to understand the hACE2 gene polymorphism, which can interfere with the interaction landscape between the SARS-CoV-2 spike and hACE2 (Calcagnile et al., 2021). Bhattacharya et al. have used the HDOCK server in several studies to perform the protein-protein interaction (Bhattacharya et al., 2020a; Bhattacharya et al., 2020b; Bhattacharya et al., 2021b; Bhattacharya et al., 2022a). 3.9 BFE calculation The HawkDock is a server for online docking online. A freely available docking server uses the ATTRACT docking algorithm to calculate the BFE using the MM/GBSA technique. The server describes the free energy of binding of hACE2 and RBD of the three strains. The predicted negative score of binding free energy indicates that the binding affinity of the hACE2 and RBD residues is vital for these two variants. The comparative free energies of the binding complexes of the three variants' hACE2 residues and RBD residues are shown in Table 2 . It was noted that the lowest negative binding free energy for B.1.1.7 during the molecular association of the hACE2 residues and RBD residues is ΔGbind = −64.27 and ΔEvdw = −104.01, respectively. Data also revealed an increased molecular association between hACE2-RBD binding of these two variants compared to the Wuhan reference strain.Table 2 Comparative evaluation of binding free energy using MM/GBSA method. Here we used one control and two variants of the RBD-hACE2 complex. Table 2Energy components (kcal/mol) Wuhan spike_ variant_hACE2 B.1.617 spike_ variant_hACE2 B.1.1.7 spike_ variant_hACE2 Binding free energy (ΔGbind) −59.71 −64 −64.27 Van der Waal energy (ΔEvdw) −99.84 −99.6 −104.01 Electrostatic energy (ΔEelectrostatic) −662.07 −1084.18 −672.78 Electrostatic solvation energy (ΔGGB) 715.29 1132.7 725.93 Non-electrostatic solvation energy(ΔGSA) −13.09 −12.93 −13.41 In this study, we used the HawkDock server for Binding free energy (BFE) calculation to understand the binding affinity of the hACE2 residues and RBD residues of B.1.1.7 and B.1.617.2 variants. Several scientists have used the HawkDock server to calculate BFE. Akachar et al. have used the HawkDock server to understand the binding efficiency of newly designed peptides inhibitors against spike RBD SARS-CoV-2 (Akachar et al., 2020). Parate et al. (2021) have applied the HawkDock server to illustrate the BFE of the binding interaction of the inhibitory protein of Raf Kinase (Parate et al., 2021). Tallei et al. used the HawkDock server to evaluate the BFE to understand the attachment of a fruit-derived small peptide. It has been noted that the small peptide can restrain the attachment of variants of SARS-CoV-2 RBD to hACE2 (Tallei et al., 2022). 3.10 Molecular dynamics simulation The RMSF, RMSD, and the quantity of hydrogen bonding were estimated using the MDS trajectories of all three variants. The Wuhan strain RBD-hACE2 complex and the developed and equilibrated RBDs of these two variants were simulated for 20 ns. The average RMSF of the Wuhan RBD-hACE2 complex was 1.5 ± 0.85 nm, and the average RMSF of B.1.1.7 variant RBD-hACE2 complex protein was 1.68 ± 0.99 nm, the average RMSF of B.1.617.2 variant RBD-hACE2 complex was 1.26 ± 0.4 nm (Fig. 10A). During the evaluation process of RMSD, it was observed that all these complexes showed a parallel deviation of about 1.1 nm from the initial point to the endpoint (20 ns of simulations). The results of RMSD of RBD-hACE2 of Wuhan strain (1.2 ± 0.34 nm), RBD-hACE2 of B.1.17 variant (1.0 ± 0.25 nm), and RBD-hACE2 of B.1.617.2 variant (1.1 ± 0.12 nm), are shown in Fig. 10B. After analyzing the deviations and fluctuations within the complex, the hydrogen bond count was evaluated for the RBD-hACE2 complex for all three variants. We observed that the Wuhan strain formed a hydrogen bond between RBD-hACE2 (10 ± 3), followed by the B.1.617.2 variant (13 ± 2), the highest number of hydrogen bonds present in this analysis, and the B.1.1.7 variant formed 11 ± 4 hydrogen bonds from all the three MDS analyses, respectively, (Fig. 10C). A similar hydrogen bond pattern was noted between the RBD-hACE2 formations in these two variants.Fig. 10 The MDS evaluation of the three complexes.(A) The RMSF plot of one control and two variants. The RMSF plot reveals the flexibility of three complexes. The mutated residues show lesser fluctuations and are comparable in all three structures. (B) The RMSD plot of one control and two variants. The RMSD plot represents similar deviations from the simulated structure. (C) MDS analysis shows hydrogen bonds of one control and two variants. The number of hydrogen bonds (H-bonds) count indicates that mutated variants have more H-bonds than the Wuhan strain. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Fig. 10 Here, we used the NAMD server to perform the MD simulation and illustrate the molecular dynamics of B.1.617.2 RBD-hACE2 and B.1.1.7 RBD-hACE2. Several researchers are using the NAMD server to perform the MD simulation during the simulation performance. Baral et al. have also performed an MD simulation of B.1.617.2 RBD-neutralizing Abs and RBD-ACE2 to understand the RBD-neutralizing Abs interactions and RBD-ACE2 interactions of the Delta variant using the NAMD server (Baral et al., 2021). Chakraborty et al. have used the MD simulation to access the molecular interaction prototype of Spike RBD and hACE2 of SARS-CoV-2 using the NAMD server (Chakraborty et al., 2021h). 4 Discussion Several variants have been generated from time to time and circulated worldwide. The significant variants are B.1.1.7 (Alpha; originated from the UK), B.1.617.2 (Delta; originated from India), P1 (Gamma; originated from Brazil), and B.1.351 (Beta; originated from South Africa). However, several other variants have been reported in other regions (Chakraborty et al., 2021a; Chakraborty et al., 2021f; Chakraborty et al., 2021g; Chakraborty et al., 2021h; Bhattacharya et al., 2021b).The recent one is Omicron which was first observed in South Africa (Mohapatra et al., 2022). However, among these variants, two variants (Delta and Alpha) have a significant role at a certain time point of the pandemic period and increased the infection, which generated a new wave in both the countries (UK and India) (Fig. 1c). B.1.1.7 was noted to increase SARS-CoV-2 infection in the UK during 2020 September/October, which is circulating in different places in the country (Wilton et al., 2021). It also got transferred to other countries like the USA (Galloway et al., 2020). The mutated variants may show significant differences in epidemiological patterns, such as reinfection possibility, severity, and transmissibility. The variant was observed with some significant mutations in spike protein (Cherian et al., 2021).Comparative genomics assessment of these two variants has revealed specific unique mutations, such as N501Y, D614G L452R, E484Q, and P681R in the S-glycoprotein. Several scientists have illustrated several mutations from time to time and described their significance in the pandemic situation (Harvey et al., 2021; Mohammadi et al., 2021). N501Y, D614G L452R, and E484Q mutations are reported as significant mutations in different variants in SARS-CoV-2. Some mutations are responsible for immune escape, antibody escape, and partial vaccine escape (Chakraborty et al., 2022a: Chakraborty et al., 2022b). One of the significant mutations is D614G which is responsible for infectivity and re-infectivity. The mutation also increases the ACE2 receptor binding capacity of S-glycoprotein and also augments the fitness of SARS-CoV-2 variants (Bhattacharya et al., 2021a: Plante et al., 2021). D614G is found in all VOIs and VOCs, and the mutation might be the outcome of a positive selection (Chakraborty et al., 2021j). Our analyses focused on country-wise and region-wise genome sequences and their submission patterns, molecular phylogenetics, scatter plots of the cluster evaluation, country-wise geographical distribution and transmission pattern, frequencies, and entropy diversity mutational landscape of B.1.1.7 and B.1.617.2. Finally, we evaluated the comparative receptor binding (hACE2) pattern with the Wuhan strain, VOC B.1.1.7 variant (RBD N501Y mutation), and VOI B.1.617.2 variant of the spike protein (RBD double mutations L452R, E484Q). Our study illustrated a similar binding pattern of these two variants, which might play a crucial role in the entry of these viruses into the cell and thereby infection. A stable binding pattern was noted for the binding of the B.1.617.2 variant, which may help explain the more infectious property of the variant. Therefore, this study will help understand the variants' infectivity and their public health importance. Molecular dynamics (MD) simulation patterns and structural mutational landscape data help understand the variants' infectivity(Khan et al., 2021). Our previous study analyzed the MD simulation of the interaction of human ACE2 and the RBD region of S-glycoprotein with the Wuhan strain (Chakraborty et al., 2021k). In this study, the MD simulation of the human ACE2 and the RBD of the Wuhan strain showed a very similar pattern as observed previously (Chakraborty et al., 2021k). The MD simulation pattern of the B.1.617.2 variant showed a more stable binding than the Alpha variant. The RMSF results showed that the variant was more stable in most residues; however, some fluctuations were noted from 412 to 472. Furthermore, the RMSD showed more or less stability throughout the simulation. However, one small peak was noted between 13 and 14 ns. The Delta variant showed a few more H-bonds than the Alpha variant and Wuhan strain during MD simulation, and this may provide a more stable binding in the B.1.617.2 variant. Simultaneously, these variants formed 11H-bonds during residue interactions with RBD and receptor binding (hACE2). This variant formed one more hydrogen bond during the interaction than the Wuhan strain (10H-bonds). Similarly, the B.1.1.7 variant formed one more hydrogen bond during the interaction (12H-bonds) than the B.1.617.2 variant. The free energy calculations confirmed the increased binding pattern of ACE2 and spike RBD to these two variants. Our comparative binding study of RBD and receptor binding (hACE2) will help comprehend all binding properties and further understand epidemiological patterns such as transmissibility and reinfection possibility. The L452R mutation, found in the B.1.617.2 variant, is linked with increased transmissibility. This mutation was detected in other variants, B.1.429/ B.1.427 (VOI), which is also related to augmented transmissibility. This mutation shows a moderate reduction in neutralization in post-vaccination sera and the generation of a few monoclonal antibodies (Control and Prevention, 2021; WHO 2021). We also analyzed the genome submission pattern of the B.1.617.2 variant from the GISAID server. The increased sequence submission pattern of the Delta variant suggests that this variant is more dominant and has a higher growth rate than other circulating variants in India (Rambaut et al., 2020: O'Toole et al., 2020;). The variant with increased transmissibility might be one of the significant factors contributing to the second wave generation in India, along with other factors (Chakraborty et al., 2021i). Molecular recognition of proteins is essential in molecular biology, which helps administer protein-ligand interactions. Therefore, the entropy landscape of a protein is an important factor in understanding the receptor-ligand binding properties of a protein (Caro et al., 2017). Researchers are trying to understand the entropy of the RNA genome. It has been observed that conformational entropy is useful to develop the RNA secondary structures from the RNA sequences (Garcia-Martin and Clote, 2015; Manzourolajdad et al., 2015). Our previous studies have described the entropy at different positions of the nucleotides in the wild strain of SARS CoV-2 and recent variants like Omicron (Chakraborty et al., 2021f; Bhattacharya et al., 2022b). In this study, we found more entropy at different positions in the B.1.617.2 variant nucleotides compared to B.1.1.7. Like, RNA entropy, the entropy of a protein also helps to understand several protein characteristics like protein-ligand interaction. Consecutively, higher entropy was noted in various places in the nucleotides of S-glycoprotein in the B.1.617.2 variant.. More entropy or higher entropy-enthalpy in the protein helps to interact with ligand molecules, and the whole entropy is linked with protein binding (Du et al., 2016; Fenley et al., 2012). The augmented entropy condition of the S-glycoprotein of the B.1.617.2 variant may support the binding of ACE2 and spike RBD, and this property supports the infective nature of this variant. This study also performed a comparative genomics analysis of these two variants. The distribution pattern and country-wise genome sequence submission pattern in the depository server showed the further spread of these two significant variants worldwide. This evolutionary epidemiology study directs phylogenetic analysis associated with the additional adaptation of these two variants. It may help in the future to continue evidence-based research on the evolutionary changes of these two variants, which may be substantial for public health measures. Recently, Day et al. described the evolutionary epidemiology of SARS-CoV-2, where they described the importance of genome-wide analysis of mutational events and genome-wide analysis of entropy calculations (Day et al., 2020). We can also understand the individual variants shared mutations with other variants using bioinformatics and molecular phylogenetics. This will help us understand the epidemiological characteristics and activity of circulating variants and guide us to formulate strategies for pandemic response (Bauer et al., 2020). 5 Limitation of the study The study has performed a comprehensive analysis of comparative genomics, evolutionary epidemiology, and RBD-hACE2 receptor binding pattern of two significant variants, Alpha and Delta. Moreover, the study revealed the pandemic response in UK and India due to the variants. The study was performed through computational biology approaches. In this study, we have performed molecular dynamics (MD) simulations. We could not achieve very high resolution (200 ns or above) due to resource invalidity, which is a limitation of our study. However, to curtail the COVID-19 pandemic, any kind of data produced, even through bioinformatics methods for Alpha and Delta variants, is necessary for future researchers and society. These two variants have been marked as VOI by WHO and made the pandemic more serious in UK and India. From this point of view, our data is very significant and highly beneficial for society. However, all the data needs to be validated through in vitro and in vivo methods, another limitation of the study. We urgently urge future researchers to validate our data through in vitro and in vivo analysis to end the pandemic crisis and prepare for a future pandemic. 6 Future perspectives This study of the evolutionary epidemiology of these two SARS-CoV-2 variants will help understand the emergence and their geographic spread. It can also be related to the epidemiological pattern and infectivity. The evolution pattern of these two variants might explain the dynamic behavior patterns of these two new emerging variants. With the help of these primary data, one can comprehend the epidemiological dynamics, transmission rates, asymptomatic infections, disease progression, and virulence patterns. The data will help better understand future researchers to illustrate the evolutionary mechanism and develop an epidemiological model in the future. Conversely, the genome sequences of B.1.1.7 and B.1.617.2 were deposited in more than 100 countries. Therefore, these variants have already spread in countries other than the country of origin, which is a severe concern. More data are urgently required to understand the epidemiological characteristics, including reinfection possibility, severity, transmissibility, activity against the nAb, nAb escape phenomena, and partial vaccine escape event of B.1.1.7 and B.1.617.2 variant. Understanding the molecular mechanism might help to fight against the pandemic in a stipulated time. 7 Conclusion In conclusion, we found the increased molecular interactivity between hACE2-RBD binding of these two variants compared to the Wuhan reference strain. The augmented receptor binding pattern of B.1.1.7 and B.1.617.2 might help to increase the infectivity compared to the Wuhan reference strain. Therefore, these two variants are accountable for the second wave generation in the UK and India. Some researchers have attempted to identify different properties, such as transmissibility and reinfection possibility of the newly emerged.1.1.7 variant. Our study has provided a better understanding of molecular phylogenetics, the cluster of evaluation, and these two variants' country-wise geographical distribution patterns. Furthermore, we have attempted to understand comparative genomics and comprehend the different mutations related to viral replication rate in terms of infectivity. We also attempted to understand the significant mutations in their structural and functional landscape. In the future, researchers might develop intervention strategies for these essential mutations to stop the spread of these variants, which has public health implications to control the wave of the pandemic period. Finally, we explored the binding association of the RBD and hACE2 of these new variants compared to the Wuhan strain. Consequently, the structural study of significant mutations will provide a strong foundation for structure-based drug design using these mutations against emerging variants to fight against the COVID-19 pandemic. Informed consent statement Not applicable. Data availability statement Not applicable. Author contributions CC: Concept development, Data collection, analysis, review, writing the draft, and editing the final manuscript. ARS: discussion, editing, and reviewing the final manuscript. MB, BM, SSN: Some part of the analysis and scientific discussion of the final manuscript. SSL: Validation and supervised the study. Uncited references ECDC, 2021b Galloway et al., 2021 Manzourolajdad and Arnold, 2015 Phillips et al., 2020 Portelli et al., 2020 SARS-CoV-2 Variant Classifications and Definitions, 2021 Yan et al., 2020 Zhang et al., 2020 Declaration of Competing Interest The authors have no conflict of interest to declare. ==== Refs References Afrin S.Z. Paul S.K. Begum J.A. Nasreen S.A. Ahmed S. Ahmad F.U. Aziz M.A. Parvin R. Aung M.S. Kobayashi N. Extensive genetic diversity with novel mutations in spike glycoprotein of SARS-CoV-2, Bangladesh in late 2020 New Microbes New Infect. 2021 100889 33936746 Akachar J. Bouricha E.M. Hakmi M. Belyamani L. El Jaoudi R. Ibrahimi A. Identifying epitopes for cluster of differentiation and design of new peptides inhibitors against human SARS-CoV-2 spike RBD by an in-silico approach Heliyon 6 2020 e05739 Alai S. Gujar N. Joshi M. Gautam M. Gairola S. Pan-India novel coronavirus SARS-CoV-2 genomics and global diversity analysis in spike protein Heliyon 7 2021 e06564 Alam S. Kamal T.B. Sarker M.M.R. Zhou J.R. Rahman S.A. Mohamed I.N. Therapeutic effectiveness and safety of repurposing drugs for the treatment of COVID-19: position standing in 2021 Front. Pharmacol. 12 2021 10.3389/fphar.2021.659577 Andrew Banchich Á.O.T. Lineage B.1.1.7. PANGO Lineages https://cov-lineages.org/global_report_B.1.1.7.html 2021 Accessed on September 15, 2021 Baral P. Bhattarai N. Hossen M.L. Stebliankin V. Gerstman B.S. Narasimhan G. Chapagain P.P. Mutation-induced changes in the receptor-binding interface of the SARS-CoV-2 Delta variant B. 1.617. 2 and implications for immune evasion Biochem. Biophys. Res. Commun. 574 2021 14 19 34425281 Bauer D.C. Tay A.P. Wilson L.O. Reti D. Hosking C. McAuley A.J. Pharo E. Todd S. Stevens V. Neave M.J. Supporting pandemic response using genomics and bioinformatics: a case study on the emergent SARS-CoV-2 outbreak Transbound. Emerg. Dis. 67 2020 1453 1462 32306500 Bhattacharya M. Sharma A.R. Mallick B. Sharma G. Lee S.S. Chakraborty C. Immunoinformatics approach to understand molecular interaction between multi-epitopic regions of SARS-CoV-2 spike-protein with TLR4/MD-2 complex Infect. Genet. Evol. 85 2020 104587 33039603 Bhattacharya M. Sharma A.R. Patra P. Ghosh P. Sharma G. Patra B.C. Saha R.P. Lee S.S. Chakraborty C. A SARS-CoV-2 vaccine candidate: In-silico cloning and validation Inform. Med. Unlocked 20 2020 100394 32835079 Bhattacharya M. Chatterjee S. Sharma A.R. Agoramoorthy G. Chakraborty C. D614G mutation and SARS-CoV-2: impact on S-protein structure, function, infectivity, and immunity Appl. Microbiol. Biotechnol. 105 2021 9035 9045 34755213 Bhattacharya M. Sharma A.R. Ghosh P. Lee S.S. Chakraborty C. A next-generation vaccine candidate using alternative epitopes to protect against Wuhan and all significant mutant variants of SARS-CoV-2: an immunoinformatics approach Aging Dis. 12 2021 2173 34881093 Bhattacharya M. Sharma A.R. Ghosh P. Patra P. Mallick B. Patra B.C. Lee S.S. Chakraborty C. TN strain proteome mediated therapeutic target mapping and multi-epitopic peptide-based vaccine development for Mycobacterium leprae Infect. Genet. Evol. 2022 105245 10.1016/j.meegid.2022.105245 35150891 Bhattacharya M. Sharma A.R. Dhama K. Agoramoorthy G. Chakraborty C. Omicron variant (B. 1.1. 529) of SARS-CoV-2: understanding mutations in the genome, S-glycoprotein, and antibody-binding regions GeroScience 2022 1 19 10.1007/s11357-022-00532-4 34292477 Calcagnile M. Forgez P. Iannelli A. Bucci C. Alifano M. Alifano P. Molecular docking simulation reveals ACE2 polymorphisms that may increase the affinity of ACE2 with the SARS-CoV-2 Spike protein Biochimie 180 2021 143 148 33181224 Cao L.-R. Zhang C.-Y. Zhang D.-L. Chu H.-Y. Zhang Y.-B. Li G.-H. Recent developments in using molecular dynamics simulation techniques to study biomolecules Acta Phys. -Chim. Sin. 33 2017 1354 1365 Caro J.A. Harpole K.W. Kasinath V. Lim J. Granja J. Valentine K.G. Sharp K.A. Wand A.J. Entropy in molecular recognition by proteins Proc. Natl. Acad. Sci. 114 2017 6563 6568 28584100 Chakraborty C. Bhattacharya M. Sharma A.R. Present Variants of Concern and Variants of Interest of Severe Acute Respiratory Syndrome Coronavirus 2: Their Significant Mutations in S-Glycoprotein, Infectivity, Re-infectivity, Immune Escape and Vaccines Activity Wiley Online Library. e2270 2021 10.1002/rmv.2270 Chakraborty C. Sharma A.R. Bhattacharya M. Agoramoorthy G. Lee S.S. The drug repurposing for COVID-19 clinical trials provide very effective therapeutic combinations: lessons learned from major clinical studies Front. Pharmacol. 12 2021 10.3389/fphar.2021.704205 Chakraborty C. Sharma A.R. Bhattacharya M. Lee S.S. Lessons learned from cutting-edge immunoinformatics on next-generation COVID-19 vaccine research Int. J. Pept. Res. Ther. 27 2021 2303 2311 34276266 Chakraborty C. Sharma A.R. Bhattacharya M. Agoramoorthy G. Lee S.S. Asian-origin approved COVID-19 vaccines and current status of COVID-19 vaccination program in Asia: a critical analysis Vaccines 9 2021 600 34199995 Chakraborty C. Sharma A.R. Bhattacharya M. Sharma G. Saha R.P. Lee S.S. Ongoing clinical trials of vaccines to fight against COVID-19 pandemic Immune Netw. 21 1 2021 10.4110/in.2021.21.e5 Chakraborty C. Sharma A.R. Bhattacharya M. Agoramoorthy G. Lee S.S. Evolution, mode of transmission, and mutational landscape of newly emerging SARS-CoV-2 variants Mbio 12 2021 e01140–21 Chakraborty C. Bhattacharya M. Sharma A.R. Lee S.S. Agoramoorthy G. SARS-CoV-2 Brazil variants in Latin America: more serious research urgently needed on public health and vaccine protection Ann. Med. Surg. 66 2021 102428 Chakraborty C. Ranjan A. Bhattacharya M. Agoramoorthy G. Lee S.-S. A paradigm shift in the combination changes of SARS-CoV-2 variants and increased spread of delta variant (B.1.617.2) across the world Aging Dis. 2021 10.14336/AD.2021.1117 Chakraborty C. Ranjan A. Bhattacharya M. Agoramoorthy G. Lee S.-S. The current second wave and COVID-19 vaccination status in India Brain Behav. Immun. 96 2021 1 4 34022371 Chakraborty C. Saha A. Sharma A.R. Bhattacharya M. Lee S.S. Agoramoorthy G. D614G mutation eventuates in all VOI and VOC in SARS-CoV-2: is it part of the positive selection pioneered by Darwin? Mol. Ther. Nucleic Acids 26 2021 237 241 34484868 Chakraborty C. Sharma A. Mallick B. Bhattacharya M. Sharma G. Lee S. Evaluation of molecular interaction, physicochemical parameters and conserved pattern of SARS-CoV-2 Spike RBD and hACE2: in silico and molecular dynamics approach Eur. Rev. Med. Pharmacol. Sci. 25 2021 1708 1723 33629340 Chakraborty C. Bhattacharya M. Sharma A.R. Emerging mutations in the SARS-CoV-2 variants and their role in antibody escape to small molecule-based therapeutic resistance Curr. Opin. Pharmacol. 62 2022 64 73 34920267 Chakraborty C. Sharma A.R. Bhattacharya M. Lee S.S. A detailed overview of immune escape, antibody escape, partial vaccine escape of SARS-CoV-2 and their emerging variants with escape mutations Front. Immunol. 2022 53 10.3389/fimmu.2022.801522 Chen F. Liu H. Sun H. Pan P. Li Y. Li D. Hou T. Assessing the performance of the MM/PBSA and MM/GBSA methods. 6. Capability to predict protein–protein binding free energies and re-rank binding poses generated by protein–protein docking Phys. Chem. Chem. Phys. 18 2016 22129 22139 27444142 Cheng L. Song S. Zhou B. Ge X. Yu J. Zhang M. Ju B. Zhang Z. Impact of the N501Y substitution of SARS-CoV-2 Spike on neutralizing monoclonal antibodies targeting diverse epitopes Virol. J. 18 2021 1 6 33397387 Cherian S. Potdar V. Jadhav S. Yadav P. Gupta N. Das M. Rakshit P. Singh S. Abraham P. Panda S. Team, NIC SARS-CoV-2 spike mutations, L452R, T478K, E484Q and P681R, in the second wave of COVID-19 in Maharashtra, India Microorganisms 9 2021 1542 34361977 Day T. Gandon S. Lion S. Otto S.P. On the evolutionary epidemiology of SARS-CoV-2 Curr. Biol. 30 2020 R849 R857 32750338 De Vries S.J. Van D.M. Bonvin A.M. The HADDOCK web server for data-driven biomolecular docking Nat. Protoc. 5 5 2010 883 897 20431534 Deng X. Garcia-Knight M.A. Khalid M.M. Servellita V. Wang C. Morris M.K. Sotomayor-González A. Glasner D.R. Reyes K.R. Gliwa A.S. Transmission, infectivity, and antibody neutralization of an emerging SARS-CoV-2 variant in California carrying a L452R spike protein mutation medRxiv 2021 10.1101/2021.03.07.21252647 Du X. Li Y. Xia Y.-L. Ai S.-M. Liang J. Sang P. Ji X.-L. Liu S.-Q. Insights into protein–ligand interactions: mechanisms, models, and methods Int. J. Mol. Sci. 17 2016 144 ECDC Emergence of SARS-CoV-2 B.1.617 Variantsin India and Situation in the EU/EEA May 11 2021 2021 ECDC Stockholm 1 12 ECDC SARS-CoV-2 Variants of concern as of May 11 2021. 2021 European Centre for Disease Prevention and Control https://www.ecdc.europa.eu/en/covid-19/variants-concern (Accessed on September 15, 2021) ECDC SARS-CoV-2 Variants of Concern Pose a Higher Risk for Hospitalisation and Intensive Care Admission 2021 European Centre for Disease Prevention and Control https://www.ecdc.europa.eu/en/news-events/sars-cov-2-variants-concern-pose-higher-risk-hospitalisation-and-intensive-care (Accessed on September 15, 2021) Farooq R.K. Rehman S.U. Ashiq M. Siddique N. Ahmad S. Bibliometric analysis of coronavirus disease (COVID-19) literature published in web of science 2019–2020 J. Fam. Community Med. 28 2021 1 7 Fenley A.T. Muddana H.S. Gilson M.K. Entropy–enthalpy transduction caused by conformational shifts can obscure the forces driving protein–ligand binding Proc. Natl. Acad. Sci. 109 2012 20006 20011 23150595 Galloway S.E. Paul P. MacCannell D.R. Johansson M.A. Brooks J.T. MacNeil A. Slayton R.B. Tong S. Silk B.J. Armstrong G.L. Biggerstaff M. Emergence of SARS-CoV-2 b. 1.1. 7 lineage—united states, December 29, 2020–January 12, 2021 Morb. Mortal. Wkly Rep. 70 2021 95 Garcia-Martin J.A. Clote P. RNA thermodynamic structural entropy PLoS One 10 11 2015 e0137859 Graham M.S. Sudre C.H. May A. Antonelli M. Murray B. Varsavsky T. Kläser K. Canas L.S. Molteni E. Modat M. Changes in symptomatology, reinfection, and transmissibility associated with the SARS-CoV-2 variant B. 1.1. 7: an ecological study Lancet Public Health 6 2021 e335 e345 33857453 Grint D.J. Wing K. Williamson E. McDonald H.I. Bhaskaran K. Evans D. Evans S.J. Walker A.J. Hickman G. Nightingale E. Case fatality risk of the SARS-CoV-2 variant of concern B. 1.1. 7 in England, November 16 to February 5 Eurosurveillance 26 2021 2100256 Hadfield J. Megill C. Bell S.M. Huddleston J. Potter B. Callender C. Sagulenko P. Bedford T. Neher R.A. Nextstrain: real-time tracking of pathogen evolution Bioinformatics 34 2018 4121 4123 29790939 Hammer Ø. Harper D.A. Ryan P.D. PAST: paleontological statistics software package for education and data analysis Palaeontol. Electron. 4 2001 9 Harvey W.T. Carabelli A.M. Jackson B. Gupta R.K. Thomson E.C. Harrison E.M. Ludden C. Reeve R. Rambaut A. Peacock S.J. Robertson D.L. SARS-CoV-2 variants, spike mutations and immune escape Nat. Rev. Microbiol. 19 2021 409 424 34075212 Humphrey W. Dalke A. Schulten K. VMD: visual molecular dynamics J. Mol. Graph. 14 1996 33 38 8744570 Islam O.K. Al-Emran H.M. Hasan M.S. Anwar A. Jahid M.I.K. Hossain M.A. Emergence of European and North American mutant variants of SARS-CoV-2 in South-East Asia Transbound. Emerg. Dis. 68 2021 824 832 32701194 Jacob J.J. Vasudevan K. Pragasam A.K. Gunasekaran K. Veeraraghavan B. Mutreja A. Evolutionary tracking of SARS-CoV-2 genetic variants highlights an intricate balance of stabilizing and destabilizing mutations Mbio 12 2020 e01188–21 Jangra S. Ye C. Rathnasinghe R. Stadlbauer D. Alshammary H. Amoako A.A. Awawda M.H. Beach K.F. Bermúdez-González M.C. Chernet R.L. SARS-CoV-2 spike E484K mutation reduces antibody neutralisation Lancet Microbe 2 2021 e283 e284 33846703 Kalé L. Skeel R. Bhandarkar M. Brunner R. Gursoy A. Krawetz N. Phillips J. Shinozaki A. Varadarajan K. Schulten K. NAMD2: greater scalability for parallel molecular dynamics J. Comput. Phys. 151 1999 283 312 Kalia K. Saberwal G. Sharma G. The lag in SARS-CoV-2 genome submissions to GISAID Nat. Biotechnol. 39 2021 1058 1060 34376850 Khan A. Zia T. Suleman M. Khan T. Ali S.S. Abbasi A.A. Mohammad A. Wei D.Q. Higher infectivity of the SARS-CoV-2 new variants is associated with K417N/T, E484K, and N501Y mutants: an insight from structural data J. Cell. Physiol. 2021 10.1002/jcp.30367 Khare S. Gurry C. Freitas L. Schultz M.B. Bach G. Diallo A. Akite N. Ho J. Lee R.T. Yeo W. Team, G.C.C GISAID's role in pandemic response China CDC Wkly. 3 2021 1049 34934514 Kouranov A. Xie L. de la Cruz J. Chen L. Westbrook J. Bourne P.E. Berman H.M. The RCSB PDB information portal for structural genomics Nucleic Acids Res. 34 2006 D302 D305 16381872 Liu A. Two weeks of “COVID-19” search on PubMed. Gov Acta Bio Med. Atenei Parmensis 91 2020 e2020199 Lubinski B. Tang T. Daniel S. Jaimes J.A. Whittaker G. Functional evaluation of proteolytic activation for the SARS-CoV-2 variant B. 1.1. 7: role of the P681H mutation bioRxiv 2021 10.1101/2021.04.06.438731 MacKerell A.D. Jr. Banavali N. Foloppe N. Development and current status of the CHARMM force field for nucleic acids Biopolymers Orig. Res. Biomol. 56 2000 257 265 Mallapaty S. India's massive COVID surge puzzles scientists Nature 592 2021 667 668 33883710 Manzourolajdad A. Arnold J. Secondary structural entropy in RNA switch (riboswitch) identification BMC Bioinforma. 16 1 2015 1 77 MathWorks, Inc MATLAB, High-Performance Numeric Computation and Visualization Software: User's Guide: for UNIX Workstations 1992 MathWorks Melo M.C. Bernardi R.C. Rudack T. Scheurer M. Riplinger C. Phillips J.C. Maia J.D. Rocha G.B. Ribeiro J.V. Stone J.E. Neese F. NAMD Goes Quantum: An Integrative Suite for Hybrid Simulations 15 2018 351 354 Mohammadi M. Shayestehpour M. Mirzaei H. The impact of spike mutated variants of SARS-CoV2 [alpha, beta, gamma, delta, and lambda] on the efficacy of subunit recombinant vaccines Braz. J. Infect. Dis. 25 2021 10.1016/j.bjid.2021.101606 Mohapatra R.K. Tiwari R. Sarangi A.K. Islam D.R. Chakraborty C. Dhama K. COMMENTARY omicron (B. 1.1. 529) variant of SARS-CoV-2–concerns, challenges and recent updates J. Med. Virol. 2022 10.1002/jmv.27633 Nextstrain Genomic epidemiology of novel coronavirus-global subsampling Nextstrain. Org 1 2020 https://nextstrain.org/ncov/global?dmax=2020-04-08 (Accessed on September 15, 2021) Ostrov D.A. Structural consequences of variation in SARS-CoV-2 B. 1.1. 7 J. Cell. Immunol. 3 2021 103 33969357 O'Toole Á. McCrone J. Software Package for Assigning SARS-CoV-2 Genome Sequences to Global Lineages 2020 GitHub https://github.com/hCoV-2019/pangolin (Accessed on September 15, 2021) O'Toole A. Scher E. Underwood A. Jackson B. Hill V. McCrone J. Ruis C. Abu-Dahab K. Taylor B. Yeats C. Pangolin: lineage assignment in an emerging pandemic as an epidemiological tool Virus Evol. 7 2020 veab064 Parate S. Rampogu S. Lee G. Hong J.C. Lee K.W. Exploring the binding interaction of Raf kinase inhibitory protein with the N-terminal of C-Raf through molecular docking and molecular dynamics simulation Front. Mol. Biosci. 8 2021 496 Phillips J.C. Braun R. Wang W. Gumbart J. Tajkhorshid E. Villa E. Chipot C. Skeel R.D. Kale L. Schulten K. Scalable molecular dynamics with NAMD J. Comput. Chem. 26 2005 1781 1802 16222654 Phillips J.C. Hardy D.J. Maia J.D. Stone J.E. Ribeiro J.V. Bernardi R.C. Buch R. Fiorin G. Hénin J. Jiang W. McGreevy R. Scalable molecular dynamics on CPU and GPU architectures with NAMD J. Chem. Phys. 153 4 2020 044130 Plante J.A. Liu Y. Liu J. Xia H. Johnson B.A. Lokugamage K.G. Zhang X. Muruato A.E. Zou J. Fontes-Garfias C.R. Mirchandani D. Spike mutation D614G alters SARS-CoV-2 fitness Nature 592 2021 116 121 33106671 Portelli S. Olshansky M. Rodrigues C.H. D'Souza E.N. Myung Y. Silk M. Alavi A. Pires D.E. Ascher D.B. Exploring the structural distribution of genetic variation in SARS-CoV-2 with the COVID-3D online resource Nat. Genet. 52 2020 999 1001 32908256 Rambaut A. Holmes E.C. O'Toole Á. Hill V. McCrone J.T. Ruis C. du Plessis L. Pybus O.G. A dynamic nomenclature proposal for SARS-CoV-2 lineages to assist genomic epidemiology Nat. Microbiol. 5 2020 1403 1407 32669681 Rubin E.J. Baden L.R. Udwadia Z.F. Morrissey S. Audio interview: India's Covid-19 crisis N. Engl. J. Med. 384 2021 e84 Control, C.F.D., Prevention SARS-CoV-2 Variant Classifications and Definitions 2021 Retrieved March 16, 2020 Tallei T.E. Adam A.A. Elseehy M.M. El-Shehawi A.M. Mahmoud E.A. Tania A.D. Niode N.J. Kusumawaty D. Rahimah S. Effendi Y. Idroes R. Fruit bromelain-derived peptide potentially restrains the attachment of SARS-CoV-2 variants to hACE2: a pharmacoinformatics approach Molecules 27 2022 260 35011492 Thiagarajan K. Why is India having a covid-19 surge? Brit. Med. J. Publ. Group 374 2021 n2005 Vaidyanathan G. Coronavirus variants are spreading in India-what scientists know so far Nature 593 2021 321 322 33976409 Velazquez A. Bustria M. Ouyang Y. Moshiri N. An analysis of clinical and geographical metadata of over 75,000 records in the GISAID COVID-19 database medRxiv 2020 10.1101/2020.09.22.20199497 Volz E. Mishra S. Chand M. Barrett J.C. Johnson R. Geidelberg L. Hinsley W.R. Laydon D.J. Dabrera G. O'Toole Á. Assessing transmissibility of SARS-CoV-2 lineage B. 1.1. 7 in England Nature 593 2021 266 269 33767447 Weng G. Wang E. Wang Z. Liu H. Zhu F. Li D. Hou T. HawkDock: a web server to predict and analyze the protein–protein complex based on computational docking and MM/GBSA Nucleic Acids Res. 47 2019 W322 W330 31106357 WHO Weekly Epidemiological Update–October 27 579 2020 WHO Geneva 580 Widera M. Mühlemann B. Corman V.M. Toptan T. Beheim-Schwarzbach J. Kohmer N. Schneider J. Berger A. Veith T. Pallas C. Surveillance of SARS-CoV-2 in Frankfurt am Main from October to December 2020 reveals high viral diversity including spike mutation N501Y in B. 1.1. 70 and B. 1.1. 7 Microorganisms 9 2021 748 33918332 Wilton T. Bujaki E. Klapsa D. Majumdar M. Zambon M. Fritzsche M. Mate R. Martin J. Rapid increase of SARS-CoV-2 variant B. 1.1. 7 detected in sewage samples from England between October 2020 and January 2021 Msystems 6 2021 e00353–21 Yan Y. Zhang D. Zhou P. Li B. Huang S.-Y. HDOCK: a web server for protein–protein and protein–DNA/RNA docking based on a hybrid strategy Nucleic Acids Res. 45 W1 2017 W365 W373 28521030 Yan Y. Tao H. He J. Huang S.-Y. The HDOCK server for integrated protein–protein docking Nat. Protoc. 15 2020 1829 1852 32269383 Zelenova M. Ivanova A. Semyonov S. Gankin Y. Analysis of 329,942 SARS-CoV-2 records retrieved from GISAID database Comput. Biol. Med. 139 2021 104981 34735950 Zhang L. Jackson C.B. Mou H. Ojha A. Peng H. Quinlan B.D. Rangarajan E.S. Pan A. Vanderheiden A. Suthar M.S. SARS-CoV-2 spike-protein D614G mutation increases virion spike density and infectivity Nat. Commun. 11 2020 1 9 31911652 Zuo X. Chen Y. Ohno-Machado L. Xu H. How do we share data in COVID-19 research? A systematic review of COVID-19 datasets in PubMed Central Articles Brief. Bioinform. 22 2021 800 811 33757278
PMC009xxxxxx/PMC9005230.txt	==== Front Digital Chinese Medicine 2096-479X 2589-3777 S2589-3777(22)00011-8 10.1016/j.dcmed.2022.03.003 Article Network pharmacology and molecular docking analysis on molecular targets and mechanism prediction of Huanglian Jiedu Decoction in the treatment of COVID-19 Xinyi Xu 1† Leping Liu 23† Xueshuai Cao 2 Xi Long 4 Sujuan Peng 5 Guomin Zhang 4∗ 1 School of Chinese Medicine, Hunan University of Chinese Medicine, Changsha, Hunan, 410208, China 2 School of Medicine, Hunan University of Chinese Medicine, Changsha, Hunan, 410208, China 3 Department of Blood Transfusion, The Third Xiangya Hospital of Central South University, Changsha, Hunan, 410013, China 4 Graduate School, Hunan University of Chinese Medicine, Changsha, Hunan, 410208, China 5 Department of Respiratory Medicine, The Second Affiliated Hospital of Hunan University of Chinese Medicine, Changsha, Hunan, 410005, China ∗ Corresponding author. † These authors contributed equally. 13 4 2022 3 2022 13 4 2022 5 1 1832 3 11 2021 23 2 2022 . 2022 Digital Chinese Medicine Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Objective To investigate and predict the molecular targets and mechanism of Huanglian Jiedu Decoction (黄连解毒汤, HLJDD) in the treatment of Corona Virus Disease 2019 (COVID-19) through network pharmacology and molecular docking analysis. Methods The chemical constituents and action targets of HLJDD were retrieved on Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP), SymMap v2, Encyclopedia of Traditional Chinese Medicine (ETCM), a High-throughput Experiment- and Reference-guided Database of Traditional Chinese Medicine (HERB), and Traditional Chinese Medicine Integrated Database (TCMID). UniProt and GeneCards were used to query the target genes that corresponding to the active compounds, and then a compound-target network was constructed using Cytoscape 3.7.2. Gene Ontology (GO) database was used to annotate GO functions. Kyoto Encyclopedia of Genes and Genomes (KEGG) was used to predict the possible mechanisms of active compounds. The Database for Annotation, Visualization and Integrated Discovery (DAVID) was used to analysis the tissue enrichment. The main active compounds in HLJDD are molecularly docked with their corresponding related targets. Results Seventy-six compounds were screened and 458 corresponding targets in the network were obtained. Gene annotation showed that the targets were involved mainly in 1953 biological processes. 884 signaling pathways was enriched, involving signaling by interleukins, cytokine signaling in immune system, generic transcription pathway, and RNA polymerase II transcription. The targets mainly distributed in the lung, liver, and placenta, involving a variety of immune cells, such as T cells and B cells. The molecular docking results showed that core compounds such as wogonin, berberine, and baicalein had high affinity with tumor necrosis factor (TNF), insulin (INS), and tumor protein 53 (TP53). Conclusion The active compounds in HLJDD may have a therapeutic effect on COVID-19 through regulating multiple signal pathways by targeting genes such as vascular endothelial growth factor A (VEGFA), INS, interleukin-6 (IL-6), TNF, caspase-3 , TP53, and mitogen-activated protein kinase 3 (MAPK3). Key words Huanglian Jiedu decoction (黄连解毒汤 HLJDD) Active compounds Corona virus disease 2019 (COVID-19) Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Network pharmacology Molecular docking ==== Body pmc1 Introduction Corona Virus Disease 2019 (COVID-19) pneumonia is an acute respiratory infectious disease with a long incubation period, strong contagiousness and pathogenicity, and general susceptibility to the population. The seriousness of novel coronavirus pneumonia is a heavy economic burden on the countries suffering from widespread infection through its long treatment cycle and high consumption of materials 1 , 2. The main clinical symptoms are fever, cough, and asthma. Some patients also experience gastrointestinal symptoms, while some patients enter a severe stage where they develop respiratory failure or even die [ 3 ]. In China, in the treatment for COVID-19, Chinese medicine is a major player, and equal importance is attached to both traditional Chinese medicine (TCM) and western medicine. In order to contend the epidemic, so far, the National Health Commission and the National Administration of Traditional Chinese Medicine have published eight versions of a new coronavirus diagnosis and treatment plan. In the protocol, Huanglian Jiedu Decoction (黄连解毒汤, HLJDD) is used to treat patients with the following clinical symptoms: high fever, cough, little sputum, or yellow sputum, chest tightness, shortness of breath, bloating, and constipation. This prescription is well-known for treating heat-syndrome in China and was first mentioned in the Medical Secrets of an Official (Wai Tai Mi Yao, 《外台秘要》) by WANG Tao, a medical scientist in the Tang Dynasty. It is an aqueous extract of four herbal materials with the ratio of 3∶2∶2∶3 in Huanglian (Coptidis Rhizoma), Huangqin (Scutellariae Radix), Huangbo (Phellodendri Chinensis Cortex), and Zhizi (Gardeniae Fructus). This formula has been used historically and widely in clinical practice [ 4 ]. In gastrointestinal diseases, inflammation, cardiovascular diseases, and Alzheimer’s disease, HLJDD has shown positive clinical effects [ [5], [6], [7] ]. A modern pharmacological study has also elucidated the pharmacokinetics and pharmacodynamics of HLJDD, and findings suggest that the main compounds of iridoids, flavonoids and alkaloids in HLJDD can have an anti-inflammatory effect [ 8 ]. In the current study, HLJDD can exert its anti-inflammatory effect by interfering with the MAPKs/NF-κB pathway [ 9 ]. LI et al. [ 10 ] studied the effect of HLJDD on the urine metabolomics of healthy people and found seven potential biomarkers, including 2-(formylamino)benzoic acid, which has proven the mechanism of treating heat syndrome from pharmacology. However, the mechanism of HLJDD in treating COVID-19 is unclear and needs further investigation. Network pharmacology is a new discipline that combines the functions of drug compounds, disease targets, and biological signaling pathways based on computer network analysis 11 , 12, which is suitable to analyze TCM, owing to the multiple targets affected by the multi-components. Network pharmacology is able to visualize, systematize, and informatize the principles of the process of treating diseases using TCM. The molecular docking technology predicts the binding mode and affinity between two molecules by analyzing the physical and chemical properties of the molecules, as well as by computer simulation [ 13 ]. Molecular docking plays an important role in detecting the mechanism of active compounds and target proteins of TCM. The crystal structure of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has been determined by Shanghai Tech University (PDB 6LU7) [ 14 ]. SARS-CoV-2 invades cells by binding the angiotensin-converting enzyme 2 (ACE2) receptor on the surface of human cells with the S protein of its spinous [ 15 ]. It was recently discovered, by German scientist Markus Hoffmann, that SARS-CoV-2 requires the help of the transmembrane protease serine 2 (TMPRSS2) protein to enter cells [ 16 ]. This article intends to analyze the active compounds and target genes in HLJDD through network pharmacology, and dock the main active compounds with their related targets to provide a theoretical basis for its clinical application. 2 Materials and methods 2.1 Components collection and screening in HLJDD and their corresponding targets This research was based on the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP, https://old.tcmsp-e.com/tcmsp.php) [ 17 ], SymMap v2 (http://www.symmap.org/), The Encyclopedia of Traditional Chinese Medicine (ETCM, http://www.tcmip.cn/ETCM/index.php/Home/), a High-throughput Experiment- and Reference-guided Database of Traditional Chinese Medicine (HERB) (http://herb.ac.cn/), and Traditional Chinese Medicine Integrated Database (TCMID, http://www.megabionet.org/tcmid/). The keywords “Huanglian (Coptidis Rhizoma)” “Huangqin (Scutellariae Radix)” “Huangbo (Phellodendri Chinensis Cortex)”, and “Zhizi (Gardeniae Fructus)” were used to obtain all compounds and their targets. The obtained protein and gene information was normalized through the Uniport database. In this study, oral bioavailability (OB) ≥ 30% and drug-likeness (DL) ≥ 0.18 were used to screen the components of Huanglian (Coptidis Rhizoma), Huangqin (Scutellariae Radix), Huangbo (Phellodendri Chinensis Cortex), and Zhizi (Gardeniae Fructus) to obtain the more active components [ 18 ]. Bioavailability refers to the relative amount of drugs that is absorbed into the systemic blood circulation and metabolized after being administered via an extravascular route. Drug-like properties are usually used to evaluate the possible failure characteristics of a compound. The significance of this standard lies in the bioavailability; the higher the degree of drug-like properties, the more potential research significance the human body presents [ 19 ]. 2.2 Establishing the compound-target network The collected compounds and targets are sorted and imported into the Cytoscape 3.7.2 software (http://www.cytoscape.org/) [ 20 ] to construct a network of active compounds-target interactions in HLJDD. Visualize the pharmacological action mechanism of HLJDD. 2.3 Collection of disease targets Based on GeneCards database (https://www.genecards.org/), pharmGKB database (https://www.pharmgkb.org/), and DisGeNet database (http://www.disgenet.org/home/), “coronavirus” was searched as the keyword, and supplemented targets through literature search to collect the targets of COVID-19. 2.4 Establishment of protein-protein interaction (PPI) network The collected COVID-19 targets were imported into the search tool for the retrieval of interacting genes/proteins (STRING) database (https://string-db.org/) to obtain the PPI network. It was imported into the Cytoscape 3.7.2 software, then merged with the component-target network for intersection. Following this, the target proteins of HLJDD acting on COVID-19 were obtained. The target proteins were imported into the STRING database to obtain the PPI network of target proteins for COVID-19 treatment with HLJDD. Finally, the network was imported into Cytoscape to observe and analyze the topological properties. 2.5 Target pathway analysis The target proteins obtained after weight reduction of the predicted target point was imported into the Gene Ontology (GO) database (http://geneontology.org/) and the threshold was set at FDR <0.05. After annotating the GO function, an analysis of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway was made through the Reactome Pathway Database (https://reactome.org/Pathway Browser/). The pathways related to HLJDD for the treatment of novel coronavirus pneumonia were obtained by consulting the literature and the KEGG database. OmicShare Tools (http://www.omicshare.com/tools/index.php/) was used to visualize the enrichment analysis results. Further tissue enrichment analysis on target protein through the Database for Annotation, Visualization and Integrated Discovery (DAVID) (https://david.ncifcrf.gov/) was carried out. 2.6 Component-target molecular docking The ZINC Is Not Commercial (ZINC) database was used to collect the “.mol2” format of the structures of the first seven core compounds obtained from the analysis, and then the Protein Data Bank (PDB) database was used to download the “.pdb” format of the corresponding targets [ 21 ]. The target proteins were dewatered and hydrogenated using the PyMOL software, and the compounds and the target proteins were converted to “.pdbqt” format by AutoDock software [ 22 ]. Binding energy less than 0 indicates that the ligand molecule and the receptor can bind spontaneously. There is no standard for target screening of active molecules, and according to the literature [ 23 ], the binding energy ≤﹣5.00 kJ/mol was selected here as the basis for screening the active compound. The docking results were visualized in PyMOL. The workflow is demonstrated in Figure 1 .Figure 1 The analysis process of this study. Figure 1 3 Results 3.1 Active compounds screening and collection in HLJDD A total of 429 compounds were obtained from TCMSP, SymMap v2, ETCM, HERB, and TCMID databases and related literature. Among them, there were 143 compounds for Huangqin (Scutellariae Radix), 48 compounds for Huanglian (Coptidis Rhizoma), 140 compounds for Huangbo (Phellodendri Chinensis Cortex), and 98 compounds for Zhizi (Gardeniae Fructus). With OB ≥ 30% and DL ≥ 0.18 as screening criteria, 102 compounds were obtained (Table 1 ). After removing the duplicates, there were 76 main compounds in HLJDD.Table 1 Active compounds in HLJDD Table 1Source MOL ID Compound Molecular weight OB (%) DL Huangqin(Scutellariae Radix) MOL000073 Ent-epicatechin 290.29 48.96 0.24 MOL000173 Wogonin 284.28 30.68 0.23 MOL000228 (2R)-7-Hydroxy-5-methoxy-2-phenylchroman-4-one 270.30 55.23 0.20 MOL000358 Beta-sitosterol 414.79 36.91 0.75 MOL000359 Sitosterol 414.79 36.91 0.75 MOL000449 Stigmasterol 412.77 43.83 0.76 MOL000525 Norwogonin 270.25 39.40 0.21 MOL000552 5,2'-Dihydroxy-6,7,8-trimethoxyflavone 344.34 31.71 0.35 MOL001458 Coptisine 320.34 30.67 0.86 MOL001490 Bis[(2s)-2-ethylhexyl] benzene-1,2-dicarboxylate 390.62 43.59 0.35 MOL001506 Supraene 410.80 33.55 0.42 MOL001689 Acacetin 284.28 34.97 0.24 MOL002714 Baicalein 270.25 33.52 0.21 MOL002879 Diop 390.62 43.59 0.39 MOL002897 Epiberberine 336.39 43.09 0.78 MOL002908 5,8,2'-Trihydroxy-7-methoxyflavone 300.28 37.01 0.27 MOL002909 5,7,2,5-Tetrahydroxy-8,6-dimethoxyflavone 376.34 33.82 0.45 MOL002910 Carthamidin 288.27 41.15 0.24 MOL002911 2,6,2',4'-Tetrahydroxy-6'-methoxychaleone 302.30 69.04 0.22 MOL002913 Dihydrobaicalin_qt 272.27 40.04 0.21 MOL002914 Eriodyctiol (flavanone) 288.27 41.35 0.24 MOL002915 Salvigenin 328.34 49.07 0.33 Huangqin(Scutellariae Radix) MOL002917 5,2′,6′-Trihydroxy-7,8-dimethoxyflavone 330.31 45.05 0.33 MOL002925 5,7,2',6'-Tetrahydroxyflavone 286.25 37.01 0.24 MOL002926 Dihydrooroxylin A 286.30 38.72 0.23 MOL002927 Skullcapflavone II 374.37 69.51 0.44 MOL002928 Oroxylin A 284.28 41.37 0.23 MOL002932 Panicolin 314.31 76.26 0.29 MOL002933 5,7,4'-Trihydroxy-8-methoxyflavone 300.28 36.56 0.27 MOL002934 Neobaicalein 374.37 104.34 0.44 MOL002937 Dihydrooroxylin 286.30 66.06 0.23 MOL008206 Moslosooflavone 298.31 44.09 0.25 MOL010415 11,13-Eicosadienoic acid, methyl ester 322.59 39.28 0.23 MOL012245 5,7,4'-Trihydroxy-6-methoxyflavanone 302.30 36.63 0.27 MOL012246 5,7,4'-Trihydroxy-8-methoxyflavanone 302.30 74.24 0.26 MOL012266 Rivularin 344.34 37.94 0.37 Huanglian(Coptidis Rhizoma) MOL000098 Quercetin 302.25 46.43 0.28 MOL000622 Magnograndiolide 266.37 63.71 0.19 MOL000762 Palmidin A 510.52 35.36 0.65 MOL000785 Palmatine 352.44 64.60 0.65 MOL001454 Berberine 336.39 36.86 0.78 MOL001458 Coptisine 320.34 30.67 0.86 MOL002668 Worenine 334.37 45.83 0.87 MOL002894 Berberrubine 322.36 35.74 0.73 MOL002897 Epiberberine 336.39 43.09 0.78 MOL002903 (R)-Canadine 339.42 55.37 0.77 MOL002904 Berlambine 351.38 36.68 0.82 MOL002907 Corchoroside A_qt 404.55 104.95 0.78 MOL008647 Moupinamide 313.38 86.71 0.26 MOL013352 Obacunone 454.56 43.29 0.77 Huangbo(Phellodendri Chinensis Cortex) MOL000098 Quercetin 302.25 46.43 0.28 MOL000358 Beta-sitosterol 414.79 36.91 0.75 MOL000449 Stigmasterol 412.77 43.83 0.76 MOL000622 Magnograndiolide 266.37 63.71 0.19 MOL000762 Palmidin A 510.52 35.36 0.65 MOL000785 Palmatine 352.44 64.60 0.65 MOL000787 Fumarine 353.40 59.26 0.83 MOL000790 Isocorypalmine 341.44 35.77 0.59 MOL001131 Phellamurin_qt 356.40 56.60 0.39 MOL001454 Berberine 336.39 36.86 0.78 MOL001455 (S)-Canadine 339.42 53.83 0.77 Huangbo(Phellodendri Chinensis Cortex) MOL001458 Coptisine 320.34 30.67 0.86 MOL001771 Poriferast-5-en-3beta-ol 414.79 36.91 0.75 MOL002636 Kihadalactone A 512.70 34.21 0.82 MOL002641 Phellavin_qt 374.42 35.86 0.44 MOL002643 Delta 7-Stigmastenol 414.79 37.42 0.75 MOL002644 Phellopterin 300.33 40.19 0.28 MOL002651 Dehydrotanshinone II A 292.35 43.76 0.40 MOL002652 Delta7-Dehydrosophoramine 242.35 54.45 0.25 MOL002656 Dihydroniloticin 458.80 36.43 0.81 MOL002659 Kihadanin A 486.56 31.60 0.70 MOL002660 Niloticin 456.78 41.41 0.82 MOL002662 Rutaecarpine 287.34 40.30 0.60 MOL002663 Skimmianin 259.28 40.14 0.20 MOL002666 Chelerythrine 332.37 34.18 0.78 MOL002668 Worenine 334.37 45.83 0.87 MOL002670 Cavidine 353.45 35.64 0.81 MOL002671 Candletoxin A 608.79 31.81 0.69 MOL002672 Hericenone H 580.88 39.00 0.63 MOL002673 Hispidone 472.78 36.18 0.83 MOL002894 Berberrubine 322.36 35.74 0.73 MOL005438 Campesterol 400.76 37.58 0.71 MOL006392 Dihydroniloticin 458.80 36.43 0.82 MOL006401 Melianone 470.76 40.53 0.78 MOL006413 Phellochin 488.83 35.41 0.82 MOL006422 Thalifendine 322.36 44.41 0.73 MOL013352 Obacunone 454.56 43.29 0.77 Zhizi (Gardeniae Fructus) MOL000098 Quercetin 302.25 46.43 0.28 MOL000358 Beta-sitosterol 414.79 36.91 0.75 MOL000422 Kaempferol 286.25 41.88 0.24 MOL000449 Stigmasterol 412.77 43.83 0.76 MOL001406 Crocetin 328.44 35.30 0.26 MOL001494 Mandenol 308.56 42.00 0.19 MOL001506 Supraene 410.80 33.55 0.42 MOL001663 (4aS,6aR,6aS,6bR,8aR,10R,12aR,14bS)-10-Hydroxy-2,2,6a,6b,9,9,12a-heptamethyl-1,3,4,5,6,6a,7,8,8a,10,11,12,13,14b-Tetradecahydropicene-4a-carboxylic acid 456.78 32.03 0.76 MOL001941 Ammidin 270.30 34.55 0.22 MOL001942 Isoimperatorin 270.30 45.46 0.23 Zhizi (Gardeniae Fructus) MOL002883 Ethyl oleate (NF) 310.58 32.40 0.19 MOL003095 5-Hydroxy-7-methoxy-2-(3,4,5-trimethoxyphe-nyl)chromone 358.37 51.96 0.41 MOL004561 Sudan III 352.42 84.07 0.59 MOL007245 3-Methylkempferol 300.28 60.16 0.26 MOL009038 GBGB 550.57 45.58 0.83 3.2 Component-target network of HLJDD The corresponding targets of the main compounds in HLJDD were collected in the databases, and the results were imported into Cytoscape 3.7.2. The active compound-prediction target network was constructed, and 534 nodes (76 active compound nodes and 458 predicted target nodes) and 2749 interaction relationships were obtained, as shown in Figure 2 .Figure 2 The target network diagram of the compounds in HLJDD. Figure 2 Purple represents the compounds of Huangqin (Scutellariae Radix); red represents the compounds of Huanglian (Coptidis Rhizoma); dark blue represents the compounds of Huangbo (Phellodendri Chinresis Cortex); green represents the compounds of Zhizi (Gardeniae Fructus); light blue represents predicted targets. The size of the nodes represents the degree; and the edges between the nodes represent the interrelations of the active compounds and targets. The core target network diagram of the compound of HLJDD (Figure 3 ) shows the higher degree of cross-linking between compounds in HLJDD and targets, including 69 core compound nodes and 191 main target nodes. The top-ranked compound nodes are beta-sitosterol, stigmasterol, rivularin, and wogonin. The correlation degrees are 169, 113, 68, and 64, respectively; and the top five target nodes are prostaglandin-endoperoxide synthase 2 (PTGS2), androgen receptor (AR), estrogen receptor 1 (ESR1), prostaglandin-endoperoxide synthase 1 (PTGS1), and nitric oxide synthase 2 (NOS2). Their correlation degrees are 62, 61, 57, 52, and 52, respectively.Figure 3 The core target network diagram of the active compounds in HLJDD. Figure 3 Purple represents the compounds of Huangqin (Scutellariae Radix); red represents the compounds of Huanglian (Coptidis Rhizoma); dark blue represents the compounds of Huangbo (Phellodendri Chinresis Cortex); and green represents the compounds of Zhizi (Gardeniae Fructus); light blue represents predicted target. The size of the nodes represents the degree; and the edges between the nodes represent the interrelations of the active compounds and targets. 3.3 Intersecting compounds in HLJDD From the previous result, it is illustrated that the same compound exists in different drugs in HLJDD, which can be obtained from Funrich’s Venn diagram (Figure 4 ). Huanglian (Coptidis Rhizoma) and Huangbo (Phellodendri Chinensis Cortex) both include MOL001454-berberine, MOL002894-berberrubine, MOL000622-magnograndiolide, MOL000785-palmatine, MOL000762-palmidinA, MOL002668-worenine, and MOL001458-coptisine. MOL001458-coptisine, MOL008583-beta-sitosterol, and MOL000449-stigmasterol are common to Huangqin (Scutellariae Radix) and Huangbo (Phellodendri Chinresis Cortex). Meanwhile, MOL000098-quercetin, MOL008583-beta-sitosterol, and MOL000449-stigmasterol are common to Huangbo (Phellodendri Chinresis Cortex) and Zhizi (Gardeniae Fructus). MOL008583-beta-sitosterol, MOL000449-stigmasterol, and MOL001506-supraene are common to Huangqin (Scutellariae Radix) and Zhizi (Gardeniae Fructus), while MOL008583-beta-sitosterol, MOL000449-stigmasterol, and MOL001506-supraene are common to Zhizi (Gardeniae Fructus) and Huangqin (Scutellariae Radix). Finally MOL006393-epiberberine and MOL001458-coptisine are common to Huangqin (Scutellariae Radix) and Huanglian (Coptidis Rhizoma).Figure 4 Distribution of active compounds of HLJDD. Figure 4 3.4 GO, KEGG, and tissue enrichment analysis The GO database was used to annotate the GO functional annotation of HLJDD and the pathway analysis of reactome. GO function annotation is used to annotate and classify genes through biological processes (BP), cell components (CC), and molecular function (MF), as shown in Figure 5 .Figure 5 GO enrichment analysis of HLJDD targets. Figure 5 Biological process, cellular component, and molecular function categories are represented by red, blue, and green bars, respectively. The height of the bar graph represents the number of genes in which the annotated genes overlap. Biological regulation, stress response, and metabolic processes are highly relevant to biological processes. The proportion of membrane, cytosol and endomembrane system in the cell components are relatively high, and protein binding, ion binding, and transferase activity have a great impact on molecular functions. In KEGG enrichment pathway analysis, 458 signaling pathways are screened using a P < 0.01, and with an FDR <0.05. The top 20 items are ranked in descending order by the number of related genes involved in the pathway, and are visualized using OmicShare Tools (Figure 6 ). Targets with a high degree of cross-linking such as interleukin(IL)-10 , IL-6, IL-1β, and tumor necrosis factor (TNF) are involved in IL-10 signaling.Figure 6 Visualization of KEGG enrichment pathway. Figure 6 Further tissue enrichment analysis was carried out on targets. As shown in Figure 7 , tissue enrichment reveals that target expression sites are mainly distributed in lung tissue, liver, and placenta, and involve a variety of immune cells, such as T cells and B cells. It shows that the key targets of the active compounds in HLJDD are mainly expressed in lung tissue and immune cells.Figure 7 Tissue enrichment analysis of HLJDD targets. Figure 7 3.5 Component-target molecular docking In theory, the lower the energy, the more stable conformation of the ligand-receptor binding, the more likely the interaction. It is generally believed that the lower the energy, the more stable the conformation of ligand-receptor binding and the higher the possibility of action. The molecular docking results (Table 2 ) show that all of the molecular docking affinity of the core active compounds in HLJDD and their corresponding related targets are less than 5.00 kJ/mol, which indicates that the core active compounds in HLJDD have good binding activity to their related targets. The results of the molecular docking study show that the binding energy of TNF to wogonin is the lowest at ﹣6.24 kKJ/mol, indicating that this ligand has the most stable conformation with the receptor. The docking results are shown in Figure 8 .Table 2 The binding energy values of the core compounds in HLJDD and their corresponding targets Table 2Gene Ligand Binding energy (kJ/mol) IL-6 Berberine ﹣4.82 IL-6 Oroxylin A ﹣4.28 IL-6 Wogonin ﹣4.08 IL-6 Quercetin ﹣3.64 INS Berberine ﹣5.75 MAPK3 Baicalein ﹣3.59 MAPK3 Quercetin ﹣2.82 TNF Wogonin ﹣6.24 TNF Berberine ﹣6.18 TNF Baicalein ﹣5.10 TNF Quercetin ﹣3.73 TP53 Baicalein ﹣5.39 TP53 Berberine ﹣5.26 TP53 Wogonin ﹣4.84 TP53 Acacetin ﹣4.36 TP53 Quercetin ﹣4.04 VEGFA Berberine ﹣5.07 VEGFA Baicalein ﹣3.26 VEGFA Quercetin ﹣2.80 Figure 8 A diagram of molecular docking. Figure 8 A, INS-berberine docking. B, TNF-baicalein docking. C, TNF-berberine docking. D, TNF-wogonin docking. E, TP53-baicalein docking. F, TP53-berberine docking. G, VEGFA-berberine docking. Black box shows the docking site. 4 Discussion 4.1 Foundations of HLJDD as a treatment for COVID-19 in TCM Since the outbreak of the COVID-19 pneumonia, the National Health Commission and other relevant units in various regions have successively issued a number of diagnosis and treatment plans. Among them, the recommended prescription for severe stages of COVID-19 pneumonia, HLJDD is in line with the blazing of both Qi and Ying Phases in “Novel Coronavirus Pneumonia Diagnosis and Treatment Plan (Trial Operation Seventh Edition)” [ 18 ]. ZOU et al. [ 24 ] analyzed the contents of Chinese medicine in the “Diagnosis and Treatment of Novel Coronavirus Pneumonia” issued by 24 provinces, cities, and autonomous regions. Among the 17 types of Chinese medicine formulas, 13 types of Chinese patent medicines, and 25 types of unnamed prescriptions for severe and critical illnesses, HLJDD appeared 9 times. WANG et al. [ 25 ] analyzed 33 COVID-19 TCM diagnosis and treatment plans (including one national plan and 32 regional plans) released before February 19, 2020. According to the statistical analysis of 65 types of Chinese patent medicines, HLJDD appeared six times, and all of these were used for treatment in the severe stage of the disease. HLJDD is an effective remedy for acute heat syndrome, as it is excess-cold and bitter in nature and can clear the pattern of excess heat-toxicity in triple-jiao thoroughly. The main points of clinical application are a fever with vexation and thirst, dry mouth and dry throat, a red tongue with a yellow coating, as well as a rapid and strong pulse [ 17 ]. The pathogenicity of COVID-19 is similar to the one of the “epidemic Qi” of TCM [ 18 ]. It penetrates the triple-jiao from the exterior to the interior, and can also reverse transmission into the pericardium. It is also characteristic of strong contagiousness and high fatality rate, and it manifests itself in different periods of disease development. The symptoms are slightly different: the main manifestations of the intermediate stage are fever, cough, excessive sputum, general fatigue, headache, wheezing, diarrhea, red urine, constipation; or dry mouth, bitterness, red and dry tongue, with a yellow or greasy coating, and slippery; or soft, rapid pulse [ 19 ]. The main pathogenesis is heat-toxin blockage in the lungs, and dysfunction of Qi in the fu-organs. Therefore, HLJDD can be prescribed, as it has the effects of clearing heat in the lungs and fu-organs, and dispersing lung Qi. 4.2 Therapeutic effects of HLJDD in modern pharmacology In modern clinical research, HLJDD is widely used in various departments, and has significant therapeutic effects on many viral and bacterial infectious diseases [ [26], [27], [28] ]. GAO et al. [ 29 ] found that the effective rate of HLJDD in treating high fever in children was as high as 94.55%, which was not significantly different from the western medicine treatment group (92.93%). The Huangqin (Scutellariae Radix) in HLJDD has a positive, protective effect on cells, and a significant effect of inhibiting the virus [ 30 ]. Huanglian (Coptidis Rhizoma) can also inhibit a variety of influenza viruses [ 31 ]. Huanglian (Coptidis Rhizoma) can effectively inhibit the expression level of the influenza virus mRNA in lung cancer A549 cells, reduce inflammation, and significantly increase Th1/Th2 and Th17/Treg values [ 32 ]. Another clinical study has shown that HLJDD can significantly exert antiviral, anti-inflammatory and antipyretic, antioxidant, immune regulation, antibacterial, and tissue protection pharmacological effects, and reduce the risk of COVID-19 turning into a severe condition [ 18 ]. HLJDD can also reduce blood pressure, hemostasis, and prevents thrombosis [ 33 ]. 4.3 The positive curative effects of HLJDD on COVID-19 HLJDD has a positive curative effect on novel coronavirus pneumonia. It can reduce the incidence of complications of COVID-19 pneumonia in many ways, improve the treatment efficiency of patients, improve the prognosis of patients, improve medical resource cost-effectiveness, as well as reduce the burden on the country, society, and individuals, which are all of great significance to hasten the ending of the epidemic [ 34 ]. Therefore, studying the active compounds, therapeutic targets, and molecular docking mechanisms of HLJDD can provide a theoretical basis for the treatment of a large number of patients with moderate or severe COVID-19. In this study, after analyzing the active compounds through the network pharmacology method, 458 potential targets, 1953 biological processes, 130 molecular functions, and 458 KEGG pathways were obtained. After preliminary clinical observation, the common clinical symptoms of the new COVID-19 strain are dyspnea, and severe cases will have a significant increase in proinflammatory cytokines such as IL-6, TNF-α, Interferon-γ (IFN-γ), which has the characteristics of cytokine storm [ 35 ]. The cytokine storm, also known as the “inflammatory storm”, is actually an important node in the transition from mild patients to severe patients, and it is also a cause of death of severe patients [ [36], [37], [38], [39] ]. Once an inflammatory storm is formed, the immune system kills the virus, but it will also kill a large number of normal cells in the lung, severely destroying the lung's ventilation function, leading to respiratory failure until hypoxia and death. IL-6 is a pro-inflammatory factor, and its main function is to accelerate the alveolar inflammation in the early stage of pulmonary fibrosis through chemotactic inflammatory cell aggregation and promote inflammatory cell infiltration, and then mediate the occurrence of idiopathic pulmonary fibrosis 38 , 40. The latest researches found that IL-6 is an important inflammatory marker that induces the inflammatory storm of COVID-19 pneumonia 40 , 41. Based on the preliminary understanding of the mechanism and the KEGG analysis results, it is speculated that the core active compounds in HLJDD may regulate cytokine signaling and IL signaling in the immune system by acting on targets such as IL-6. Vascular endothelial growth factor (VEGF) is an important factor that promotes angiogenesis. It mainly exerts its physiological function by binding with receptors VEGF Receptor 1 (VEGFR1), VEGF Receptor 2 (VEGFR2), etc. [ 42 ]. Under pathological conditions, their combination can inhibit the apoptosis of vascular endothelial cells, promote their proliferation, migration and differentiation, increase vascular permeability, and stimulate neovascularization in the body [ 43 ]. Therefore, it is speculated that HLJDD may inhibit VEGF signal transduction by acting on VEGFA reduce pulmonary fibrosis, and play a role in treating COVID-19. Tumor protein 53 (TP53) is a known target of several viral on coproteins, including SARS-CoV-2. Studies have observed that human coronaviruses antagonize the viral inhibitor p53 by stabilizing CHY-zinc finger domain-containing 1 (RCHY1) as an interaction partner of the viral SARS-CoV-2 unique structural domain and promoting RCHY1-mediated degradation of p53 [ [44], [45], [46] ]. Potentially, viruses can use its downregulation to aid their own replication, and pharmacological rescue of p53 function can be explored to monitor viruses [ 47 ]. The results of molecular docking showed that the conformation of TNF and wogonin was the most stable and the possibility of their action was the greatest. This indicates that wogonin in Huangqin (Scutellariae Radix) plays a more important role in the treatment of COVID-19. Baicalein and berberine also showed good binding activity to TNF. Study has shown that wogonin and baicalein have inhibiting inflammatory mediators, regulating immunity, and eliminating free radical effects [ 48 ]. Some studies have shown that berberine has certain anti-inflammatory effects 48 , 49, and can be combined with pneumolysin cholesterol binding site to prevent the toxin from binding to membrane, playing a competitive antagonistic role, thus to have an anti-infective effect [ 50 ]. Therefore, it is speculated that HLJDD may play a better role in the treatment of COVID-19 through its inhibiting inflammatory mediators, regulating immunity, and eliminating free radical effects. 5 Conclusion Overall, this study used network pharmacology and molecular docking analysis to explore the chemical constituents, action targets, and the core active compounds in HLJDD. The active compounds such as berberine, baicalein, and wogonin in HLJDD may have a therapeutic effect on COVID-19 through regulating multiple signaling pathways by targeting genes such as VEGF, IL-6, TNF, TP53, etc. However, since this study is mainly discussed at the theoretical level, further experimental research on pharmacodynamic evaluation, metabolomics, and clinical efficacy is needed to provide a solid basis for the treatment and drug development of COVID-19. Funding National Natural Science Foundation of China (81973670), Natural Science Foundation of Hunan Province (2018JJ2297), Key Program of Scientific Research Fund of Hunan Provincial Education Department (19A370), Domestic First-class Cultivation Discipline Integrated Traditional Chinese and Western Medicine Discipline Project of Hunan Province (2021ZXYJH10), and College Student Innovation and Entrepreneurship Training Program of Hunan Province (S201910541046). Competing interests The authors declare no conflict of interest. ==== Refs References 1 Chen Q.Q. Wang F.X. Cai Y.Y. Untargeted metabolomics and lipidomics uncovering the cardioprotective effects of Huanglian Jiedu Decoction on pathological cardiac hypertrophy and remodeling Journal of Ethnopharmacology 270 2020 113646 33264659 2 DU W. Zhan M.X. Wu S.L. Analysis of the feasibility of applying Artemisia annua L. to COVID-19 based on network pharmacology strategy Journal of Southwest University (Natural Science Edition) 44 2 2022 69 75 3 Chan J.F. Kok K.H. Zhu Z. Genomic characterization of the 2019 novel human-pathogenic coronavirus isolated from a patient with atypical pneumonia after visiting Wuhan Emerging Microbes & Infections 9 1 2020 221 236 31987001 4 Wang R.Y. Lin J.T. Analysis of the mechanism of Zhichuanling oral liquid in treating bronchial asthma based on network pharmacology Evidence-based Complementary and Alternative Medicine 2020 2020 1875980 32015750 5 Li M. Shang H. Wang T. Huanglian decoction suppresses the growth of hepatocellular carcinoma cells by reducing CCNB1 expression World Journal of Gastroenterology 27 10 2021 939 958 33776365 6 Tanaka Y. Ito T. Tsuji G. Baicalein inhibits benzo[a]pyrene-induced toxic response by downregulating src phosphorylation and by upregulating NRF2-HMOX1 system Antioxidants (Basel) 9 6 2020 507 7 Qi Y. Zhang Q. Zhu H. Huang-lian jie-du decoction: a review on phytochemical, pharmacological and pharmacokinetic investigations Chinese Medicine 14 1 2019 1 22 30636970 8 Li C.L. Pan J.J. Xu C. A preliminary inquiry into the potential mechanism of Huang-Lian-Jie-Du Decoction in treating rheumatoid arthritis via network pharmacology and molecular docking Frontiers in Cell and Developmental Biology 9 2021 740266 35127697 9 Fan H.J. Zhao X.S. Tan Z.B. Effects and mechanism of action of Huang-Lian-Jie-Du-Tang in atopic dermatitis-like skin dysfunction in vivo and in vitro Journal of Ethnopharmacology 240 2019 111937 31075381 10 Li Y. Xie J. Li Y. Literature data based systems pharmacology uncovers the essence of "body fire" in traditional Chinese medicine: a case by Huang-Lian-Jie-Du-Tang Journal of Ethnopharmacology 237 2019 266 285 30922854 11 Luo T.T. Lu Y. Yan S.K. Network pharmacology in research of Chinese medicine formula: methodology, application and prospective Chinese Journal of Integrative Medicine 26 1 2020 72 80 30941682 12 Wang X. Wang Z.Y. Zheng J.H. TCM network pharmacology: a new trend towards combining computational, experimental and clinical approaches Chinese Journal of Nature Medicines 19 1 2021 1 11 13 Saikia S. Bordoloi M. Molecular docking: challenges, advances and its use in drug discovery perspective Current Drug Target 20 5 2019 501 521 14 Jin Z. DU X. Xu Y. Structure of Mpro from SARS-CoV-2 and discovery of its inhibitors Nature 582 7811 2020 289 293 32272481 15 Jackson C.B. Farzan M. Chen B. Mechanisms of SARS-CoV-2 entry into cells Nature Reviews Molecular Cell Biology 23 1 2022 3 20 34611326 16 Hoffmann M. Kleine W.H. Schroeder S. SARS-CoV-2 Cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor Cell 181 2 2020 271 280 32142651 17 Huang L.L. Wang J.A. Xu R. Study on mechanism of Huanglian Jiedu Decoction in treating novel coronavirus pneumonia based on network pharmacology Journal of Chinese Medicinal Materials 43 3 2020 779 785 18 Guo X. Sun R. Prescription-syndrome relationship in traditional Chinese medicine and pharmacology research progress of Huanglian Jiedu Decoction in treatment of COVID-19 with severe syndrome of dual blaze of Qi and nutrient Chinese Traditional and Herbal Drugs 51 11 2020 3070 3077 19 Li X.Y. Song B. Lei Y. Interpretation of traditional Chinese medicine diagnosis and treatment program of diagnosis and treatment program of COVID-19 (trial version 6) Jilin Journal of Chinese Medicine 40 6 2020 701 708 20 Jiang Y.B. Zhong M. Long F. Deciphering the active ingredients and molecular mechanisms of Tripterygium hypoglaucum (Levl.) Hutch against rheumatoid arthritis based on network pharmacology Evidence-based Complementary and Alternative Medicine 2020 2020 2361865 32015751 21 Velankar S. Burley S.K. Kurisu G. The protein data bank archive Methods in Molecular Biology 2305 2021 3 21 33950382 22 Goodsell D.S. Sanner M.F. Olson A.J. The AutoDock suite at 30 Protein Science 30 1 2021 31 43 32808340 23 Alfonso F. Floriana M. Morena F. The value of matrix metalloproteinase-9 and vascular endothelial growth factor receptor 1 pathway in diagnosing indeterminate pleural effusion Interactive Cardiovascular and Thoracic Surgery 16 3 2013 263 269 23190621 24 Zhang L.Q. Miao S.Y. Xia M.T. Analysis and thinking on traditional Chinese medicine in preventing and treating severe cases of novel coronavirus pneumonia Academic Journal of Shanghai University of Traditional Chinese Medicine 34 2 2020 11 16 25 Wang C.C. Wu S. Jiang L.J. Comprehensive analysis of TCM diagnosis and treatment schemes for COVID-19 in all regions of China World Science and Technology/Modernization of Traditional Chinese Medicine and Materia Medica 22 2 2020 257 263 26 Li X. Tang H. Tang Q. Decoding the mechanism of Huanglian Jiedu Decoction in treating pneumonia based on network pharmacology and molecular docking Frontiers in Cell and Developmental Biology 9 2021 638366 33681222 27 Li X. Wei S. Ma X. Huanglian Jiedu Decoction exerts antipyretic effect by inhibiting MAPK signaling pathway Evidence-based Complementary and Alternative Medicine 2021 2021 2209574 35003291 28 Zhou J. Gu X. Fan X. Anti-inflammatory and regulatory effects of Huanglian Jiedu Decoction on lipid homeostasis and the TLR4/MyD88 signaling pathway in LPS-induced zebrafish Frontiers in Physiology 10 2019 1241 31616320 29 Gao L.Y. Clinical observation on Huanglian Jiedu Decoction in treating infantile high fever World Latest Medicine Information 17 90 2017 101 30 Guo Y. Analysis of the chemical constituents and pharmacological effects of Scutellaria baicalensis Georgi Electronic Journal of Clinical Medical Literature 6 63 2019 137 31 Fan T.T. Cheng B.L. Fang X.M. Application of Chinese medicine in the management of critical conditions: a review on sepsis American Journal of Chinese Medicine 48 6 2020 1315 1330 32907362 32 Yan Y.Q. Fu Y.J. Wu S. Anti-influenza activity of berberine improves prognosis by reducing viral replication in mice Phytotherapy Research 32 12 2018 2560 2567 30306659 33 Wang K.X. Gao Y. Gong W.X. A novel strategy for decoding and validating the combination principles of Huanglian Jiedu Decoction from multi-scale perspective Frontiers in Pharmacology 11 2020 567088 33424585 34 Liu W. Zeng Y. Exploring the potential targets and mechanisms of Huang lian jie du decoction in the treatment of coronavirus disease 2019 based on network pharmacology International Journal of General Medicine 14 2021 9873 9885 34938107 35 Costela R.V.J. Lllescas M.R. Puerta J.M. SARS-CoV-2 infection: the role of cytokines in COVID-19 disease Cytokine and Growth Factor Reviews 54 2020 62 75 32513566 36 Ge Q. Chen L. Tang M. Analysis of mulberry leaf components in the treatment of diabetes using network pharmacology European Journal of Pharmacology 833 2018 50 62 29782863 37 Huang K.J. Su J. Theron M. An interferon-gamma-related cytokine storm in SARS patients Journal of Medical Virology 75 2 2005 185 194 15602737 38 Li L.J. Fan A.R. Ge D.Y. Astragalus Angelica ratio of drug doses and on IPF mice survival condition and TGF-β, IL-6, Foxp3, ROR Gamma the influence of the level of gene expression Journal of Liaoning University of Traditional Chinese Medicine 17 7 2015 42 46 39 Fara A. Mitrev Z. Rosalia R.A. Cytokine storm and COVID-19: a chronicle of pro-inflammatory cytokines Open Biology 10 9 2020 200160 32961074 40 Ma Q.H. Huang W.B. Zhao J. Liu Shen Wan inhibits influenza a virus and excessive virus-induced inflammatory response via suppression of TLR4/NF-κB signaling pathway in vitro and in vivo Journal of Ethnopharmacology 252 2020 112584 31972325 41 Huang C. Wang Y. Li X. Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China The Lancet 395 10223 2020 497 506 42 Sahebnasagh A. Nabavi S.M. Kashani H.R.K. Anti-VEGF agents: as appealing targets in the setting of COVID-19 treatment in critically ill patients International Immunopharmacology 101 B 2021 108257 34673299 43 Korobelnik J.F. Loewenstein A. Eldem B. Anti-VEGF intravitreal injections in the era of COVID-19: responding to different levels of epidemic pressure Graefes Archive for Clinical and Experimental Ophthalmology 259 3 2021 567 574 44 Ma L.Y. Carbajo L.J. Hein M.Y. p53 down-regulates SARS coronavirus replication and is targeted by the SARS-unique domain and PLpro via E3 ubiquitin ligase RCHY1 Proceedings of the National Academy of Sciences of the United States of America 113 35 2016 e5192 e5201 27519799 45 Ramaiah M.J. mTOR inhibition and p53 activation, microRNAs: the possible therapy against pandemic COVID-19 Gene Reports 20 2020 100765 32835132 46 Yan S. Wu G. Spatial and temporal roles of SARS-CoV PLpro -A snapshot FASEB Journal 35 1 2021 e21197 47 Mishra A. Chanchal S. Ashraf M.Z. Host-viral interactions revealed among shared transcriptomics signatures of ARDS and thrombosis: a clue into COVID-19 pathogenesis TH Open 4 4 2020 e403 e412 33354650 48 Huang Y.F. Bai C. He F. Review on the potential action mechanisms of Chinese medicines in treating Coronavirus Disease 2019 (COVID-19) Pharmacological Research 158 2020 104939 32445956 49 Tong T. Wu Y.Q. Ni W.J. The potential insights of traditional Chinese medicine on treatment of COVID-19 Chinese Medicine 15 2020 51 32483409 50 Goździcka J.A. Warowicka A. Nawrot R. GO Antiviral activity of berberine Archives of Virology 165 9 2020 1935 1945 32594322
PMC009xxxxxx/PMC9005239.txt	==== Front Disabil Health J Disabil Health J Disability and Health Journal 1936-6574 1876-7583 The Author(s). Published by Elsevier Inc. S1936-6574(22)00065-6 10.1016/j.dhjo.2022.101325 101325 Original Article COVID-19 vaccine website accessibility dashboard Jo Grace B.A. Undergraduate Senior a1 Habib Daniel B.A. Undergraduate Senior a1 Varadaraj Varshini M.D. Postdoctoral Research Fellow ab Smith Jared M.S. Associate Director d Epstein Sabrina B.A. Undergraduate Senior a Zhu Jiafeng M.S. Research Associate e Yenokyan Gayane M.D., Ph.D. Associate Scientist e Ayers Kara Ph.D., M.P.H. Associate Professor af Swenor Bonnielin K. Ph.D Associate Professor abc∗ a Johns Hopkins Disability Health Research Center, Johns Hopkins University, 3400 N. Charles Street, Baltimore, MD, 21218, USA b The Wilmer Eye Institute, Johns Hopkins School of Medicine, 600 N. Wolfe Street, Baltimore, MD, 21287, USA c Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, 615 N. Wolfe Street, Baltimore, MD, 21205, USA d Institute for Disability Research, Policy, and Practice, Utah State University, 6807 Old Main Hill, Logan, UT, 84322, USA e Johns Hopkins Biostatistics Center, Johns Hopkins Bloomberg School of Public Health, 615 N. Wolfe Street, Baltimore, MD, 21205, USA f University of Cincinnati, College of Medicine, Department of Pediatrics, Cincinnati Children's Hospital Medical Center 3333 Burnet Avenue, Cincinnati, OH, 45229, USA ∗ Corresponding author. Johns Hopkins Disability Health Research Center, Johns Hopkins University, 600 N. Wolfe Street, Wilmer 116, Baltimore, MD, 21287, USA. 1 Equally contributing first authors. 12 4 2022 12 4 2022 10132529 6 2021 26 2 2022 4 4 2022 © 2022 The Author(s) 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background The COVID-19 pandemic has exacerbated historical inequities for people with disabilities including barriers in accessing online information and healthcare appointment websites. These barriers were brought to the foreground during the vaccine rollout and registration process. Objective This cross-sectional study aimed to examine accessibility of U.S. state and territory COVID-19 information and registration centralized websites. Methods The Johns Hopkins Disability Health Research Center created a COVID-19 Vaccine Dashboard compiling COVID-19 information and vaccine registration web pages from 56 states and territories in the United States (U.S.) reviewed between March 30 through April 5, 2021 and analyzed accessibility using WAVE Web Accessibility Evaluation Tool (WAVE). WAVE identifies website accessibility barriers, including insufficient contrast, alternative text, unlabeled buttons, total number of errors, and error density. Web pages were ranked and grouped into three groups by number of errors, creating comparisons between states on accessibility barriers for people with disabilities. Results All 56 U.S states and territories had COVID-19 information web pages and 29 states had centralized state vaccine registration web pages. Total errors, error density, and alert data were utilized to generate accessibility scores for each web page, the median score was 259 (range = 14 to 536 and IQR = 237) for information pages, and 146 (range = 10 to 281 and IQR = 105) for registration pages. Conclusions These results highlight barriers people with disabilities may encounter when accessing information and registering for the COVID-19 vaccine, which underscore inequities in the pandemic response for the disability community and elevate the need to prioritize accessibility of public health information. Keywords COVID-19 vaccine Health services for persons with disabilities Barriers to accessing information and registering for COVID-19 vaccine Website accessibility of public health information Disability inclusion ==== Body pmcThe COVID-19 pandemic has disproportionately impacted the disability community, compounding already existing barriers that individuals with disabilities face in accessing healthcare, as well as public health information.1, 2, 3, 4, 5 These gaps in access create inequities in accessing COVID-19 testing, treatment, and vaccination.1 , 2 People with disabilities are more likely to have underlying conditions that impact the immune system and make them exponentially more susceptible to severe and negative COVID-19 outcomes than the general population.2 , 4 This may result in some individuals with disabilities requiring personal care assistance, and the disruption of caregiving and support services due to the need for social distancing may also increase the risk for health complications unrelated to COVID-19.2 In particular, individuals with disabilities have a multitude of unique obstacles within the health care system including, but not limited to, difficulties with transportation and barriers in accessing information that have been exacerbated by COVID-19 regulations such as the social distancing mandates.6 People with disabilities are also more likely than people without disabilities to live in nursing homes or congregate care settings, which have been associated with much higher risk of COVID-19 exposure.2, 3, 4 In addition, for certain disabilities (such as sensory disabilities), there can be differential societal implications of social measures taken to combat the spread of COVID-19. For example, someone who is Blind cannot tell if the people around them are wearing a mask or maintaining social distancing. For these combinations of reasons, people with disabilities are more at risk and differentially impacted by the pandemic, and therefore prioritized for COVID-19 vaccine allocation. The disability community has historically faced barriers to accessing information.1, 2, 3, 4, 5 The impact of these gaps in access have been elevated during the pandemic.1 , 2 Inaccessibility of COVID-19 information poses significant barriers, as individuals with disabilities may face longer delays in obtaining information about COVID-19 and receiving COVID-19 vaccine, creating inequities in health and daily life. Gaps in accessibility are compounded by the COVID-19 vaccine registration websites, potentially resulting in greater challenges to register for vaccination for people with disabilities.7, 8, 9, 10 For example, one study analyzing data collected from U.S. state health department websites reported multiple website accessibility barriers including lack of mobile access to COVID-19 vaccine websites, absence of non-English language options, poor readability, and failure to disclose critical information regarding web-based scheduling or availability.10 While this article is focused on U.S. COVID-19 website, gaps in website accessibility are widespread. One Australian study assessed a variety of WCAG violations, such as low background contrast or lack of alternative or descriptive text, on local ACT COVID-19 information websites while a Canadian article evaluated the potential impact between WCAG score errors in Ontario public health websites and the ease in which those with disabilities had access to COVID-19 vaccine appointments.11 , 12 This project focused on website accessibility, as under the Americans with Disabilities Act (ADA) and the Rehabilitation Act of 1973, states have a responsibility to provide accessible information to people with disabilities. The ADA mandates against discrimination on the basis of disability. Although Title III of the ADA focuses on “public accommodations and commercial facilities”,13 internet websites are not explicitly mentioned. The Department of Justice has taken the position to consider websites public domain, and as a result, website accessibility is protected under Title III of the ADA. Therefore, under the ADA, state COVID-19 information and registration websites need to be accessible to individuals with disabilities.13 In addition to the ADA, Section 504 of the Rehabilitation Act of 1973 requires all recipients of federal government aid to ensure their activities, programs, and services are accessible to people with disabilities, including websites.14 In 2018, Section 508 of the Rehabilitation Act adopted the internationally accepted Website Content Accessibility Guidelines (WCAG) which legally requires federal agencies (or parties doing business with a federal agency) to ensure all pages of their websites are accessible to individuals with disabilities.14 This project aimed to examine and quantify the accessibility of U.S. COVID-19 state vaccine information and registration websites and disseminate this information via the Johns Hopkins Disability Health Research Center COVID-19 Vaccine Dashboard. Although website errors and accessibility have been reliably measured by the Web Accessibility Evaluation Tool (WAVE), which has been used to assess website accessibility in many settings15, 16, 17 including a study on two local Australian government COVID-19 websites, this project aimed to track the accessibility of COVID-19 vaccine information across the U.S.11 Methods Data collection This cross-sectional study utilized data on website accessibility metrics determined from state/territory public health department COVID-19 vaccine web pages across all 50 U.S. states, the District of Columbia, and 5 territories collected by searching the internet for the most up-to-date web page Uniform Resource Locator (URL). As the data was publicly available and did not involve human subjects, IRB approval was not required. Web pages were monitored weekly to ensure that each URL was still active and to update the states and territories that created new web pages. The entire website was not scored. Two types of web pages were analyzed: COVID-19 information and COVID-19 vaccine registration. COVID-19 information web pages were determined based on containing state/territory-wide guidelines on business regulations, social distancing policies, vaccine eligibility, etc. Our analysis focused on the main COVID-19 state information webpage or landing page. As a result, no additional webpages were studied (only one COVID-19 information webpage URL was analyzed for each U.S. state or territory). A registration web page was determined based on containing a centralized portal for residents to register for COVID-19 vaccine appointments. Our analysis was limited to single state/territory-level vaccine pre-registration or registration government web pages and did not include data from states or territories without a centralized state-level registration vaccine portal. For example, states or territories that simply listed the contact information of each locale's health department or displayed an interactive map of which providers and commercial pharmacies had vaccines available were not captured. In addition, states or territories that listed multiple providers, each offering independent pathways for registration, were not included. These web pages were excluded, despite being centralized (i.e., hosted by state- and territory-wide departments of health), because these web pages only served as intermediary information hubs between the resident and vaccine provider. During data collection, state COVID-19 vaccine websites were split evenly between two primary coders, G.J and D.H. Each person had 4 days to collect the data, after which checks were completed, graphs created, and the data was published in the Johns Hopkins COVID-19 Vaccine Dashboard each Wednesday. Data from reviews conducted between March 30 until April 5, 2021 are included in these analyses. Assessment of web page accessibility The accessibility scores and rankings for COVID-19 web pages were generated from automatic testing data provided by WAVE, a web accessibility evaluation tool administered by WebAIM at Utah State University. WAVE is a commonly utilized accessibility testing tool that has been in development since 2001. WAVE analyzes patterns in web site codes to identify accessibility barriers such as insufficient contrast, alternative text, unlabeled buttons, total number of errors, and error density. A free, online version of WAVE is publicly available for download at https://wave.webaim.org/. Browser extensions for Chrome and Firefox are also available. The WAVE API was utilized for automated testing of the COVID-19 web pages.18 The number of detected accessibility errors, the density of those errors on the page (errors divided by number of page elements), and the number of likely/potential accessibility issues “(“Alerts” within WAVE) were considered when generating the rankings. Automatically detectable accessibility issues do not represent all accessibility issues, but typically indicate barriers for users with disabilities and non-compliance with accessibility guidelines and best practices. Accessibility scores were generated for each web page by rank ordering the 56 COVID-19 information pages (and 29 vaccine registration pages) on number of errors, error density, and likely/potential issues. Weightings were then applied to each of the rank orders (ties were treated equally) to calculate an overall access score—with error ranks being assigned a score of 6, error density ranks being assigned a score of 3, and likely/potential issues assigned a score of 1. This access score is not necessarily a good measure of overall accessibility but provides a normalized metric for comparisons across the web pages. The dashboard presented pages ordered or ranked by the overall access score. For information web pages, WAVE was run automatically, while for registration web pages, accessibility scores were calculated by navigating to the first web page of the registration site where user-specific information (e.g., name, age, etc.) was entered. Rather than capturing the entire website, the first relevant information or registration page was considered reflective of all web pages of each site. WAVE was run automatically if the URL was linked to the correct web page of the vaccine portal. However, WAVE was run manually (using a browser extension) on the next page of the registration portal if account creation or CAPTCHA (completely automated public turing test to tell computers and humans apart) was required on the first page before entering user-specific data for 9 of the 20 COVID-19 vaccine registration web pages (leaving 20 of the 29 being scores using automated WAVE methods). In manual assessments, testers captured total errors, error density, and alert data—the same data collected via the automated process for pages that could be directly accessed by WAVE. However, automated WAVE data captured more in-depth analyses of the types of errors, such as contrast errors, empty links/buttons, images without alternative text, and unlabeled form inputs. If a state or territory required the creation of a new account to register for vaccination, the account creation page was used to assess accessibility: no new accounts were created by the researchers during this process. Dashboard These data were used to create the Johns Hopkins Disability Health Research Center COVID-19 Vaccine Dashboard.19 The dashboard was made publicly available, focused on ensuring accessibility of information, and was updated weekly to both enhance accessibility to vaccine registration URLs and highlight important considerations about mitigating disparities in COVID-19 information access for people with disabilities.19 The initial dashboard release focused on COVID-19 vaccine prioritization for people with disabilities, but was expanded to also examine state-level COVID-19 vaccine information and registration website accessibility. These analyses present a cross-sectional snapshot of the accessibility data for the week of March 30 to April 5, 2021. This week was selected as it is the “baseline” when all state website accessibility data began to be tracked. Statistical analysis Accessibility rankings were assigned based on an inverse relationship with accessibility score, as higher accessibility rankings corresponded to lower frequency of errors (i.e., lower accessibility score). Contrast errors, error density, total errors, accessibility scores, rankings, and URLs of COVID-19 vaccine information and registration web pages were tabulated. Additional properties of registration web pages—account requirement, pre-registration status, and CAPTCHA implementation—were also tabulated. For both information and registration web pages, the mean, standard deviation, median, range, and interquartile range (IQR) were calculated and tabulated for accessibility scores, total errors, error density, insufficient contrast errors, empty links/buttons, images without alternative text, and unlabeled form inputs. Only those registration web pages that did not preclude automatic WAVE testing due to CAPTCHA or account creation before accessing the main page were included in the table for combined statistics. Accessibility scores for information and registration web pages in the states and territories that hosted both types were plotted as bar graphs. Information and registration web pages were divided into three groups based on their accessibility scores to compare and contrast centralized web pages in various U.S. states and territories. The top third contained the fewest accessibility errors while the bottom third contained the most accessibility errors. Maps color-coded by group were created for easy visualization and comparison of information and registration web page accessibility. The correlation between state information and registration pages was calculated using Pearson's correlation coefficient (and a paired t-test to obtain the p-value) for the subset of 29 states/territories that had centralized registration portals. Statistical analyses were complete using R statistical software (version 3.6.3; The R Foundation). Results Accessibility scores were calculated for 56 U.S. states and territories with COVID-19 information web pages (Table 1 ), and the 29 states that had a centralized (i.e., state/territory-wide) COVID-19 registration web page as of April 5, 2021 (Table 2 ). Total accessibility scores for each state and territory information and registration web pages were examined (Fig. 1 a, b, c). The within-state correlation between information and registration page accessibility scores was 0.304 (p < 0.001). States with information and registration web pages that were both in the top third of states with the fewest accessibility errors were California, New Jersey, Maine, and Ohio. States (and one territory) with both types of web pages in the bottom third with the most accessibility errors were New York, Nebraska, Northern Mariana Islands, Florida, Idaho, New Mexico, and Illinois. The most common issues were: (1) text with insufficient contrast to the background (which is more difficult to read, particularly for users with certain visual disabilities), (2) empty links and buttons that would be announced by screen readers but lack descriptions, (3) images without descriptive alternative text for screen reader users, and (4) form inputs not properly labeled with descriptive text (Table 3 ). Other accessibility issue types, such as if a page code did not define the natural language of the document or if headings or ARIA code were used incorrectly, were also considered in the accessibility scores.Table 1 COVID-19 information web page accessibility ranking by United States state/territory (data as of April 5, 2021). Table 1Rank State/territory Score Total errors Error density Contrast errors URL 1 Minnesota (MN) 14 0 0.00% 0 https://www.health.state.mn.us/diseases/coronavirus/vaccine/plan.html 2 Kansas (KA) 20 0 0.00% 0 https://www.kansasvaccine.gov/ 3 Louisiana (LA) 31 0 0.00% 0 https://ldh.la.gov/covidvaccine/ 4 California (CA) 45 0 0.00% 0 https://covid19.ca.gov/vaccines/ 5 Arkansas (AR) 53 0 0.00% 0 https://www.healthy.arkansas.gov/programs-services/topics/covid-19-vaccination-plan 6 Washington (WA) 54 0 0.00% 0 https://www.doh.wa.gov/Emergencies/COVID19/vaccine 7 Maryland (MD) 55 0 0.00% 0 https://covidlink.maryland.gov/content/vaccine/ 8 New Jersey (NJ) 75 1 0.12% 0 https://covid19.nj.gov/faqs/nj-information/slowing-the-spread/who-is-eligible-for-vaccination-in-new-jersey-who-is-included-in-the-vaccination-phases 9 Maine (ME) 96 1 0.36% 0 https://www.maine.gov/covid19/vaccines 10 New Hampshire (NH) 97 1 0.16% 0 https://www.vaccines.nh.gov/ 10 Vermont (VT) 97 1 0.15% 1 https://www.healthvermont.gov/covid-19/vaccine 12 North Dakota (ND) 141 3 0.18% 1 https://www.health.nd.gov/covid-19-vaccine-information 13 Rhode Island (RI) 145 2 0.20% 0 https://covid.ri.gov/vaccination 14 Ohio (OH) 151 3 0.40% 2 https://coronavirus.ohio.gov/wps/portal/gov/covid-19/covid-19-vaccination-program 15 Georgia (GA) 179 5 0.38% 5 https://dph.georgia.gov/covid-vaccine 15 Michigan (MI) 179 3 0.51% 1 https://www.michigan.gov/coronavirus/0,9753,7-406-98178_103214---,00.html 17 Connecticut (CT) 188 3 0.97% 3 https://portal.ct.gov/Coronavirus/COVID-19-Vaccination---Phases 18 U.S. Virgin Islands (VI) 202 4 0.93% 4 https://www.covid19usvi.com/vaccines 19 Pennsylvania (PA) 205 5 0.46% 0 https://www.health.pa.gov/topics/disease/coronavirus/Vaccine/Pages/Vaccine.aspx 20 Indiana (IN) 207 5 0.15% 2 https://www.coronavirus.in.gov/vaccine/index.htm 20 Montana (MT) 207 3 1.10% 3 https://dphhs.mt.gov/covid19vaccine 22 Texas (TX) 212 4 0.70% 1 https://www.dshs.state.tx.us/coronavirus/immunize/vaccine.aspx 23 Wisconsin (WI) 232 5 0.43% 2 https://www.dhs.wisconsin.gov/covid-19/vaccine-about.htm 24 Alabama (AL) 234 4 1.10% 1 https://www.alabamapublichealth.gov/covid19vaccine/index.html 24 Nevada (NV) 234 5 0.85% 3 https://www.immunizenevada.org/nv-covid-fighter 26 Puerto Rico (PR) 235 5 1.14% 5 https://www.vacunatepr.com/covid-19 27 Guam (GU) 246 5 0.96% 3 https://dphss.guam.gov/covid-19-vaccinations/ 28 Oregon (OR) 247 5 0.54% 1 https://covidvaccine.oregon.gov/ 29 Kentucky (KY) 271 7 0.38% 5 https://govstatus.egov.com/ky-covid-vaccine 30 North Carolina (NC) 286 6 0.65% 6 https://covid19.ncdhhs.gov/vaccines 31 Colorado (CO) 287 7 0.39% 2 https://covid19.colorado.gov/for-coloradans/vaccine/vaccine-for-coloradans 32 Utah (UT) 304 7 0.51% 5 https://coronavirus.utah.gov/vaccine-distribution/ 33 Massachusetts (MA) 305 9 0.87% 6 https://www.mass.gov/info-details/massachusetts-covid-19-vaccination-phases 34 Tennessee (TN) 324 7 1.24% 3 https://www.tn.gov/health/cedep/ncov/covid-19-vaccine-information.html 35 Oklahoma (OK) 334 6 3.80% 0 https://vaccinate.oklahoma.gov/ 36 South Carolina (SC) 344 11 0.60% 6 https://scdhec.gov/covid19/covid-19-vaccine 37 West Virginia (WV) 359 11 2.38% 10 https://dhhr.wv.gov/COVID-19/Pages/Vaccine.aspx 38 Alaska (AL) 364 11 1.72% 10 http://dhss.alaska.gov/dph/epi/id/pages/COVID-19/vaccine.aspx 39 New York (NY) 379 12 2.17% 8 https://covid19vaccine.health.ny.gov/ 40 Nebraska (NE) 380 14 0.96% 6 http://dhhs.ne.gov/Pages/COVID-19-Vaccine-Information.aspx 40 Northern Mariana Islands (MP) 380 15 2.09% 12 https://www.vaccinatecnmi.com/ 42 Wyoming (WI) 407 17 2.24% 17 https://health.wyo.gov/publichealth/immunization/wyoming-covid-19-vaccine-information/ 43 Delaware (DE) 416 16 2.57% 13 https://coronavirus.delaware.gov/vaccine/ 44 Florida (FL) 417 15 2.63% 15 https://floridahealthcovid19.gov/covid-19-vaccines-in-florida/ 45 Iowa (IA) 419 22 1.30% 14 https://coronavirus.iowa.gov/pages/vaccineinformation 46 Missouri (MO) 427 17 2.56% 17 https://covidvaccine.mo.gov/ 47 Hawaii (HI) 433 19 1.87% 17 https://hawaiicovid19.com/vaccine/ 48 South Dakota (SD) 450 14 4.47% 8 https://doh.sd.gov/COVID/Vaccine/ 49 District of Columbia (DC) 456 17 2.49% 13 https://coronavirus.dc.gov/vaccine 50 Idaho (ID) 459 18 2.38% 16 https://coronavirus.idaho.gov/covid-19-vaccine/ 51 New Mexico (NM) 468 20 2.49% 20 https://cv.nmhealth.org/covid-vaccine/ 52 American Samoa (AS) 504 33 2.71% 30 https://www.facebook.com/asdoh.hotline/ 53 Virginia (VA) 512 51 3.30% 42 https://www.vdh.virginia.gov/covid-19-vaccine/ 54 Illinois (IL) 523 110 9.67% 104 https://www.dph.illinois.gov/covid19/vaccine-faq 55 Arizona (AZ) 530 56 5.62% 29 https://azdhs.gov/ 56 Mississippi (MS) 536 81 3.03% 77 https://msdh.ms.gov/msdhsite/_static/14,0,420,976.html Data as of April 5, 2021. Table 2 COVID-19 vaccine registration web page accessibility ranking by United States state/territory. Table 2Rank State/territory Score Total errors Error density Requires account creation/log-in Pre-registration Has CAPTCHAa CAPTCHAa audio option URL 1 Nevada (NV) 10 0 0.00% No No No No https://vax4nv.nv.gov/patient/s/ 2 California (CA) 14 0 0.00% No No No No https://myturn.ca.gov/screening 2 Maine (ME) 14 0 0.00% No No No No https://vaccinateme.maine.gov/screening 2 Massachusetts (MA) 14 0 0.00% No Yes No No https://vaccinesignup.mass.gov/#/ 5 Indiana (IN) 17 0 0.00% No No No No https://vaccine.coronavirus.in.gov/ 6 Virginia (VA) 20 0 0.00% No Yes No No https://vax.preregister.virginia.gov/#/ 7 Alabama (AL) 22 0 0.00% No No No No https://govstatus.egov.com/vaccine-eligibility-form 8 District of Columbia (DC) 99 1 0.23% No Yes No No https://v51r5a21s.dc.gov/cvq_dc/ 8 New Jersey (NJ) 99 1 0.15% No Yes Yes Yes https://covidvaccine.nj.gov/ 10 Ohio (OH) 110 1 1.12% No No No No https://gettheshot.coronavirus.ohio.gov/screening 11 Maryland (MD) 111 1 0.83% Yes Yes No No https://onestop.md.gov/users/sign_up?registration_context=%2Fpreregistration 12 Tennessee (TN) 130 2 0.36% No No Yes Yes https://vaccinate.tn.gov 13 Arizona (AZ) 131 2 1.38% Yes No No No https://podvaccine.azdhs.gov/signup 14 Oklahoma (OK) 139 2 1.00% No No Yes Yes https://vaccinate.oklahoma.gov/en-US/ 15 Delaware (DE) 146 2 1.56% No No Yes No https://vaccinerequest.delaware.gov/s/de-vms-screening?language=en_US 15 Vermont (VT) 146 2 2.44% Yes No No No https://vermont.force.com/events/s/selfregistration 17 Missouri (MO) 154 2 1.79% No No No No https://modhss.iad1.qualtrics.com/jfe/form/SV_231d5TxZxkGedCt 18 Minnesota (MN) 166 3 0.13% No No No No https://vaccineconnector.mn.gov/covid-19%20vaccine/ 19 West Virginia (WV) 175 4 0.35% Yes No No No https://member.everbridge.net/747122446041089/new 20 Wisconsin (WI) 182 3 0.73% No No Yes Yes https://vaccinate.wi.gov/en-US/ 21 New York (NY) 200 4 1.44% No No Yes No https://am-i-eligible.covid19vaccine.health.ny.gov/Public/prescreener 22 Idaho (ID) 204 5 0.62% No Yes No No https://covidvaccine.idaho.gov/register 23 Illinois (IL) 221 7 2.95% No No No No https://covidvaccination.dph.illinois.gov/questionnaire 24 Nebraska (NE) 223 3 3.49% No No Yes Yes https://vaccinate.ne.gov/en-US/ 25 Georgia (GA) 229 8 1.43% No Yes No No https://myvaccinegeorgia.com/en/site/ALL 26 Northern Mariana Islands (MP) 233 17 0.70% No Yes No No https://www.vaccinatecnmi.com/covid-19/registration/ 27 New Hampshire (NH) 250 12 3.20% No No No No https://sonh-community.force.com/providers/s/ 28 Florida (FL) 274 32 8.99% No Yes No No https://myvaccine.fl.gov/#/RegistrationForm 29 New Mexico (NM) 281 57 3.44% No No Yes No https://cvvaccine.nmhealth.org Abbreviations: CAPTCHA=Completely automated public turing test to tell computers and humans apart a Data as of April 5, 2021. Fig. 1 United States COVID-19 information and vaccine registration web page accessibility (data as of April 5, 2021). (a) COVID-19 information and (b) COVID-19 vaccine registration pages ranked in three groups by number of accessibility errors. (c) COVID-19 information and COVID-19 vaccine registration web page accessibility scores for the 29 states and territories with both types. Fig. 1 Table 3 Statistics for common COVID-19 information web page accessibility errors across the United States. Table 3Issue type Number of issues detected (% of all errors) Mean issue per web page (SD) Insufficient text contrast 549 (78.2%) 9.80 (18.06) Empty links and buttons 55 (7.8%) 0.98 (1.69) Missing alternative text 46 (6.6%) 0.82 (2.91) Missing input labels 18 (2.6%) 0.34 (0.72) COVID-19 information web page accessibility scores for all 56 states and territories ranged from 14 (Minnesota) to 536 (Mississippi) with a median score of 259 (Table 4 ). Errors ranged from zero (Minnesota, Kansas, Louisiana, California, Arkansas, Washington, and Maryland) to 110 (Illinois) with a median value of 5.50. Mississippi (the state with the lowest accessibility score) had 81 total errors. Error density ranged from 0% to 9.67% with a median value of 0.90%. Insufficient contrast errors ranged from 0 to 104 with a median of 3.50. Errors with empty links and buttons had a range of 0–8 with a median value of 0. Images without alternative text ranged from 0 to 21 with a median of 0, and errors with form inputs not labeled had a median value of 0 with a range of 0–3.Table 4 Statistics for COVID-19 information web page accessibility data across the United States. Table 4Accessibility Mean (SD) Median (range) Interquartile range Accessibility scores 274.20 (151.37) 259 (14, 536) 172, 409.25 Total errors 12.54 (19.72) 5.50 (0, 110) 3, 15 Error density, % 1.41 (1.68) 0.90 (0, 9.67) 0.38, 2.28 Insufficient contrast errors 9.80 (18.06) 3.50 (0, 104) 1, 12.25 Empty links/buttons 0.98 (1.69) 0 (0, 8) 0, 1 Images without alternative text 0.82 (2.91) 0 (0, 21) 0, 1 Form inputs not labeled 0.34 (0.72) 0 (0, 3) 0, 0 Abbreviations: SD = standard deviation Data from 50 US States, the District of Columbia, and 5 U.S. territories, as of April 5, 2021. Of the 29 states and territories that hosted both types of websites, accessibility scores from the 29 centralized COVID-19 information web pages ranged from 14 to 530 with a median score of 324. Errors ranged from zero to 110 with a median of 7 and a mean of 14.66 (SD 22.66). Error density had a median of 0.01 (range of 0–0.10). The accessibility scores for state COVID-19 vaccine registration web pages ranged from 10 (Nevada) to 281 (New Mexico), with a median score of 146 (Table 5 ). Error density ranged from 0 (Nevada, California, Maine, Massachusetts, Indiana, Indiana, Virginia, and Alabama) to 0.09 (Florida) with a median percentage of 0.01. Seven of the 29 states had zero total errors (Nevada, California, Maine, Massachusetts, Indiana, Virginia, and Alabama) and New Mexico (the state with the lowest accessibility ranking) had 57 total errors. Errors for the 29 COVID-19 vaccine registration web pages had a median of 2 (Table 5). Four states (Maryland, Arizona, Vermont, and West Virginia) required account creation for access and eleven states (Massachusetts, Virginia, District of Columbia, New Jersey, Maryland, Idaho, Georgia, Northern Mariana Islands, and Florida) were open for pre-registration of the vaccine (Table 2). There were eight state COVID-19 registration web pages with CAPTCHA (New Jersey, Tennessee, Oklahoma, Delaware, Wisconsin, New York, Nebraska, and New Mexico), of which five (New Jersey, Tennessee, Oklahoma, Wisconsin, and Nebraska) had an audio option for completion, which enhances accessibility for individuals with visual impairments (Table 2).Table 5 Statistics for registration web page accessibility scores, total errors, error density, contrast errors, empty links/buttons, images without alternative text, and unlabeled form inputs. Table 5All 29 registration web pages Mean (SD) Median (range) Interquartile range Accessibility scores 138.41 (85.52) 146 (10, 281) 99, 204 Total errors 5.90 (11.84) 2 (0, 57) 1, 4) Error density, % 1.77 (2.06) 1.10 (0, 9.67) 0.36, 2.49 20 registration web pages Mean (SD) Median (range) IQR Insufficient contrast errors 0.95 (1.82) 0 (0, 6) 0, 1 Empty links/buttons 0.35 (0.81) 0 (0, 3) 0, 0 Images without alternative text 0.45 (0.83) 0 (0, 3) 0, 1 Form inputs not labeled 2.40 (6.44) 0.50 (0, 29) 0, 1.25 Abbreviations: SD = standard deviation Data from 20 of the 29 registration web pages that have details on types of errors (as the remainder were extracted manually using the WAVE browser plug-in), as of April 5, 2021. Of the 29 registration web pages, 20 were assessed using automatic WAVE testing while the other 9 (California, Maine, New Jersey, Tennessee, Oklahoma, Wisconsin, Idaho, Nebraska, and New Mexico) exhibited barriers that made manual extraction necessary (Table 5). For these 20 states, insufficient contrast errors ranged from 0 to 6 with a median of 0. The median value of errors regarding empty links and buttons was 0 with a range of 0–3. Web pages containing images without alternative text ranged from 0 to 3 with a median of 0, and errors with form inputs not labeled ranged from 0 to 29 with a median value of 0.50. Discussion By creating a dashboard comparing COVID-19 information and registration website accessibility, we found inequities in access to this critical information in the U.S. These results highlight important gaps in the opportunity to access COVID-19 information and vaccines for people with disabilities. The mean number of errors detected across these web pages (12.50 per page) was lower than the mean accessibility errors (51.40 per page) detected on a sample of 1 million web pages from across the internet.20 This lower mean number of accessibility errors may due to the fact this dashboard only examined state government websites, which generally have fewer error when compared to private webpages.20 Even though the accessibility of COVID-19 web pages scored better than most web pages, the importance of the information on these pages demands that any accessibility barriers be addressed, as disparities in COVID-19 outcomes and vaccination rates among the disability community may be compounded by these gaps in information access. Website errors and accessibility have been reliably measured by the Web Accessibility Evaluation Tool (WAVE) in many settings,15, 16, 17 including a study on two local Australian government COVID-19 websites, but not to the scale of encompassing all territory-wide websites.11 The majority of state rankings did not have consistent accessibility rankings across COVID-19 vaccine information and registration websites. Only a few states had within-state consistency in these rankings, such as California where the COVID-19 vaccine information website was ranked 4th, and the COVID-19 vaccine registration ranked 2nd. The reasons for the within-state inconsistency in accessibility ranks are unknown. It is possible there are within-state variations in the allocation of resources and personnel focusing on website accessibility that may contribute to these differences. It is also possible that COVID-19 vaccine information websites may have been created earlier than registration websites, allowing more time for states to improve information website accessibility based on the public's feedback. States without centralized vaccine registration websites also had discrepancies between information and registration rankings, as no ranking was possible for registration websites in these cases. It is important to consider how the disability community is impacted by barriers in accessing COVID-19 vaccine information. For instance, CAPTCHA usually requires visual input from the user, creating difficulties for individuals who are blind or are visually impaired. CAPTCHA may therefore create additional delays in or prevent vaccine registration. Delays in vaccine registration that may have resulted from inaccessible websites may result in greater risk for people with disabilities to contract COVID-19, a disease which has been proven to have disproportionally negative outcomes for this population group.2 Acknowledging barriers for website accessibility, the federal government has recently devoted substantial funding towards reducing barriers in accessing COVID-19 vaccines for the people with disabilities by providing access to direct phone numbers for vaccine registration and providing transportation to vaccine appointments.21 Information on these efforts, however, is primarily available via the state COVID-19 websites. While this critical funding will likely go far towards closing gaps in accessing vaccines, there remains limited resources and efforts towards ensuring COVID-19 information and registration is actually accessible on state public health web pages. It is important for web designers to prioritize accessibility and include disability advocates and people with disabilities at the outset of website design. Only by partnering with the disability community can web designers create public health websites which are accessible and informative for everyone. While website accessibility is an important contributing factor in barriers to COVID-19 vaccination uptake, this is only one component of the COVID-19 inequities impacting the disability community. Further data is needed to identify gaps in accessibility of all aspects of COVID-19 vaccine distribution, including transportation to the COVID-19 vaccine appointment and accessibility at the vaccine clinic site. Despite its potential impact, the limitations of this project must be considered. While this project and resulting Dashboard focus on website accessibility, there remains important work in collecting data documenting and identifying gaps in accessibility of the physical vaccine locations and quantifying vaccine rates among people with disabilities to determine inequities. Additionally, the accessibility score/ranking system is primarily designed around people with visual disabilities, and there are barriers to online accessibility for people with other types of disabilities that were not captured. These may include, for example, video captioning for people with hearing loss, or plain language summaries for people with intellectual, executive and cognitive disabilities. Further, some websites collect more information per page (i.e., long format) while other websites collect less information per page but include multiple pages (i.e., wide format). A website with a wide format may have fewer total errors per page but may have a higher percentage of errors with respect to how much information is on a single page. Information web pages were more often in long format, likely accounting for their notably higher accessibility scores compared to those for registration web pages (Table 3, Table 4). To mitigate this, the accessibility scores included error density—the number of errors by web page elements (i.e., how dense errors are within the page content and functionality). Our method measured accessibility scores for the first page of the registration website, which is indicative of patterns in subsequent pages; however, users may abandon the process if significant accessibility barriers are encountered on subsequent pages, and future research should focus on developing a metric for capturing accessibility data across many pages of the same website rather than just the first page. Some COVID-19 vaccine registration web pages required pre-registration, including entering driver's license credentialing, or creating an account. This limitation may have created bias in our sample as we did not gain access to websites requiring pre-registration to avoid entering false information and burdening the system. While WAVE was used in this project, it is only one type of method to evaluate website accessibility. This tool analyzes patterns in web site code and design that align with accessibility compliance failures with a very high level of reliability. However, WAVE, like all automated tools, cannot fully assess all aspects of end user accessibility, and the issues it does analyze are primarily focused on users with visual disabilities. Analyzing plain language or assessing video captioning presence/quality or testing for keyboard accessibility issues would necessitate manual testing that was not conducted as part of this study. Additional aspects of accessibility, such as options for end users to register for a vaccine or get additional supports via phone, were not considered, primarily because web sites are a primary mechanism for accessing such information and functionality. Despite these limitations, the issues detected by WAVE nearly always align with negative impact on users with disabilities and thus provide a useful measure of accessibility. In addition, the COVID-19 vaccine roll-out is rapidly evolving, and some states or territories may have adjusted their information and vaccine registration portals since initial data collection. Conclusion This project highlights the barriers that people with disabilities may have encountered when accessing state information and registration COVID-19 vaccine websites in the U.S during the spring of 2021. Our results underscore the addressable, yet persistent inequities in the pandemic response for the disability community.22 The accessibility of public health information must be prioritized and supported in order to ensure that the disability community is no longer left behind. Funding American Association of People with Disabilities (AAPD). Disclaimer The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the AAPD. Prior presentation of abstract No prior presentation or submissions of abstract. Contributors DH and GJ contributed to data collection, manuscript writing, and interpretation of results. VV contributed to conception of the work, data collection, manuscript writing, and interpretation of results. JS contributed to data analysis, and interpretation of results. SE contributed to the conception of the work and data collection. JZ and GY contributed to data analysis and visualization. KA contributed to the conception of the work. BKS contributed to conception of the work, manuscript writing, and interpretation of results. All authors had full access to all data in the study, contributed edits of the manuscript, and had final responsibility for the decision to submit for publication. Data sharing The COVID-19 website accessibility scores and rankings used in this study are publicly available on the Johns Hopkins Disability Research Center Website. Declaration of competing interest All authors declare no competing interests. Acknowledgments We thank the following team members for their work on the COVID-19 Dashboard Project: Caroline Cerilli, Jessica Campanile, Leah Smith, Fiona Sweeney, Maureen van Stone, Tracy Waller. ==== Refs References 1 Epstein S. Campanile J. Cerilli C. Gajwani P. Varadaraj V. Swenor B.K. New obstacles and widening gaps: a qualitative study of the effects of the COVID-19 pandemic on U.S. adults with disabilities Disability and Health Journal March 2021 101103 10.1016/j.dhjo.2021.101103 33840617 2 Lund E.M. Forber-Pratt A.J. Wilson C. Mona L.R. The COVID-19 pandemic, stress, and trauma in the disability community: a call to action Rehabil Psychol 65 4 2020 313 322 10.1037/rep0000368 33119381 3 Sabatello M. Burke T.B. McDonald K.E. Appelbaum P.S. Disability, ethics, and health care in the COVID-19 pandemic Am J Publ Health 110 10 2020 1523 1527 10.2105/ajph.2020.305837 4 Pulrang A. The Disability Community Fights Deadly Discrimination amid the COVID-19 Pandemic. Forbes https://www.forbes.com/sites/andrewpulrang/2020/04/14/the-disability-community-fights-deadly-discrimination-amid-the-covid-19-pandemic/?sh=48083850309c 5 Okoro C.A. Hollis N.T.D. Cyrus A.C. Griffin-Blake S. Prevalence of disabilities and health care access by disability status and type Among adults — United States, 2016 MMWR Morbidity and Mortality Weekly Report 67 32 2018 882 887 10.15585/mmwr.mm6732a3 30114005 6 Shalaby W.S. Odayappan A. Venkatesh R. The impact of COVID-19 on individuals across the spectrum of visual impairment Am J Ophthalmol 227 2021 53 65 10.1016/j.ajo.2021.03.016 33781768 7 Epstein S. Ayers K. Swenor B.K. COVID-19 vaccine prioritisation for people with disabilities Lancet Public Health 6 6 2021 10.1016/s2468-2667(21)00093-1 8 Kayla Hui M.P.H. People with Disabilities Are Experiencing Barriers to Covid-19 Vaccination. Verywell Health https://www.verywellhealth.com/people-with-disabilities-barriers-covid-19-vaccine-5115883 9 Keegan J. Lecher C. We ran tests on every state's COVID-19 vaccine website – the markup https://themarkup.org/blacklight/2021/03/24/we-ran-tests-on-every-states-covid-19-vaccine-website 10 Howe J.L. Young C.R. Parau C.A. Trafton J.G. Ratwani R.M. Accessibility and usability of state health department COVID-19 vaccine websites JAMA Netw Open 4 5 2021 10.1001/jamanetworkopen.2021.14861 11 Yu S.Y. A review of the accessibility of ACT COVID-19 information portals Technol Soc 64 2021 101467 10.1016/j.techsoc.2020.101467 33324025 12 Rotenberg S. Downer M.B. Brown H.K. COVID-19 vaccination for people with disabilities Science Briefs of the Ontario COVID-19 Science Advisory Table 2 35 2021 10.47326/ocsat.2021.02.35.1.0 13 U.S. Department of JusticeCivil Rights DivisionPublic Access Section The Americans with Disabilities Act : Title III Regulations : Part 36 Nondiscrimination on the Basis of Disability in Public Accommodations and Commercial Facilities 1991 U.S. Dept. of Justice, Civil Rights Division, Public Access Section Washington, D.C. 14 (WAI), W. C. W. A. I Web Content Accessibility Guidelines (WCAG) Overview November 13, 2021 Web Accessibility Initiative (WAI) (n.d.) https://www.w3.org/WAI/standards-guidelines/wcag/ 15 Ahmi A. Mohamad R. Evaluating accessibility of Malaysian public universities websites using AChecker and WAVE J Inf Commun Technol 15 2 2016 10.2139/ssrn.3550314 16 Adepoju S.A. Shehu I.S. Usability Evaluation of Academic Websites Using Automated Tools September 2014 2014 3rd International Conference on User Science and Engineering (i-USEr) 10.1109/iuser.2014.7002700 17 Lujan-Mora S. Navarrete R. Penafiel M. Egovernment and Web Accessibility in South America April 2014 2014 First International Conference on eDemocracy & eGovernment (ICEDEG) 77 82 10.1109/icedeg.2014.6819953 18 WAVE API. WAVE Web Accessibility Evaluation Tool https://wave.webaim.org/api/ 19 Vaccine Prioritization Dashboard The Johns Hopkins Disability Health Research Center https://disabilityhealth.jhu.edu/vaccine-2/ 20 The WebAIM MillionAn Annual Accessibility Analysis of the Top 1,000,000 Home Pages 2021 WebAIM https://webaim.org/projects/million/#errors 21 U.S. Department of Health and Human Services HHS to Expand Access to COVID-19 Vaccines for Older Adults and People with Disabilities. HHS.gov https://www.hhs.gov/about/news/2021/03/29/hhs-to-expand-access-to-covid-19-vaccines-for-older-adults-and-people-with-disabilities.html 22 Swenor B.K. Including disability in all health equity efforts: an urgent call to action Lancet Public Health 6 6 2021 10.1016/s2468-2667(21)00115-8
PMC009xxxxxx/PMC9005241.txt	==== Front Mult Scler Relat Disord Mult Scler Relat Disord Multiple Sclerosis and Related Disorders 2211-0348 2211-0356 Elsevier B.V. S2211-0348(22)00312-1 10.1016/j.msard.2022.103800 103800 Article Long term persistence of SARS-CoV-2 humoral response in multiple sclerosis subjects Maniscalco Giorgia Teresa ab⁎# Ferrara Anne Lise cd# Liotti Antonietta c Manzo Valentino a Di Battista Maria Elena ab Salvatore Simona ab Graziano Daniela e Viola Assunta f Amato Gerardino g Moreggia Ornella b Di Giulio Cesare Daniele b Alfieri Gennaro a Di Iorio Walter a Della Rocca Gennaro a Andreone Vincenzo aǂ De Rosa Veronica cǂ⁎ǂ a Neurological Clinic and Stroke Unit, "A. Cardarelli" Hospital, Via A. Cardarelli 9, Naples 80131, Italy b Multiple Sclerosis Center, "A. Cardarelli" Hospital, Via A. Cardarelli 9, Naples 80131, Italy c Institute of Experimental Endocrinology and Oncology (IEOS-CNR), Via S. Pansini 5, Naples 80131, Italy d Department of Translational Medical Science and Center for Basic and Clinical Immunology Research (CISI), University of Naples "Federico II", Via S. Pansini 5, Naples 80131, Italy e Unit of Trasfusional Medicine, SIMT, "A. Cardarelli" Hospital, Via A. Cardarelli 9, Naples 80131, Italy f Molecular Biology Laboratory, Hematology and Transplantation CSE, "A. Cardarelli" Hospital, Via A. Cardarelli 9, Naples 80131, Italy g Clinical Pathology and Microbiology Laboratory “A. Cardarelli" Hospital, Via A. Cardarelli 9, Naples 80131, Italy ⁎ Corresponding author. # GTM and ALF contributed equally to this work. ǂ VA and VDR contributed equally as senior authors to this work. 12 4 2022 6 2022 12 4 2022 62 103800103800 9 3 2022 18 3 2022 8 4 2022 © 2022 Elsevier B.V. All rights reserved. 2022 Elsevier B.V. Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background & objectives The persistence of the severe acute respiratory syndrome coronavirus (SARS-CoV)-2 pandemic, partly due to the appearance of highly infectious variants, has made booster vaccinations necessary for vulnerable groups. Here, we present data regarding the decline of the SARS-CoV-2 BNT162b2 mRNA vaccine-induced humoral immune response in a monocentric cohort of MS patients. Methods 96 MS patients undergoing eight different DMTs, all without previous SARS-CoV-2 infection, were evaluated for anti-Spike IgG levels, 21 days (T1) and 5–6 months (T2) after the second SARS-CoV-2 BNT162b2 mRNA vaccine dose. The anti-Spike IgG titre from MS subjects was compared with 21 age- and sex-matched healthy controls (HC). Results When compared with SARS-CoV-2 IgG levels at T2 in HC, we observed comparable levels in interferon-β 1a-, dimethyl fumarate-, teriflunomide- and natalizumab-treated MS subjects, but an impaired humoral response in MS subjects undergoing glatiramer acetate-, cladribine-, fingolimod- and ocrelizumab-treatments. Moreover, comparison between SARS-CoV-2 IgG Spike titre at T1 and T2 revealed a faster decline of the humoral response in patients undergoing dimethyl fumarate-, interferon-β 1a- and glatiramer acetate-therapies, while those receiving teriflunomide and natalizumab showed higher persistence compared to healthy controls. Conclusion The prominent decline in humoral response in MS subjects undergoing dimethyl fumarate-, interferon-β 1a- and glatiramer acetate-therapies should be considered when formulating booster regimens as these subjects would benefit of early booster vaccinations. Keywords BNT162b2-mRNA vaccine Coronavirus-19 Disease modifying therapies Multiple sclerosis Humoral persistence SARS-CoV-2 spike protein Abbreviations CLAD, Cladribine COVID-19, Coronavirus disease 2019 DMF, Dimethyl fumarate DMTs, Disease modifying therapies EDSS, Expanded disability status scale FTY, Fingolimod GA, Glatiramer acetate HC, Healthy controls IgG, Immunoglobulin G INF, Interferon-β 1a IQR, Interquartile range MS, Multiple sclerosis NAT, Natalizumab OCRE, Ocrelizumab PPMS, Primary progressive multiple sclerosis RRMS, Relapsing remitting multiple sclerosis SARS-CoV-2, severe acute respiratory syndrome coronavirus-2 SPMS, Secondary progressive multiple sclerosis TERI, Teriflunomide ==== Body pmc1 Introduction Vaccination against the severe acute respiratory syndrome coronavirus (SARS-CoV)−2 has become a global priority as coronavirus disease 2019 (COVID-19) presents a severe or life-threatening disease course up to 5–10% of cases (Gavriatopoulou et al., 2020). Among all the available vaccines, BNT162b2 mRNA has been shown to induce high levels of the anti-SARS-CoV-2 Spike-receptor binding domain (RBD) IgG neutralizing antibodies (NAbs) that strongly correlate with clinical immune protection from COVID-19 in healthy subjects (Borobia et al., 2021; Walsh et al., 2020; Khoury et al., 2021). Furthermore, NAbs also induce a good level of seroprotection in Multiple Sclerosis (MS) patients, although with different effectiveness depending on the disease-modifying therapy (DMT). Indeed, while MS patients treated with immunosuppressive therapies such as cladribine (CLAD), fingolimod (FTY) and ocrelizumab (OCRE) exhibited an impaired secretion of SARS-CoV-2 anti-Spike IgG (Tortorella et al., 2021; Sormani et al., 2021; Maniscalco et al., 2021), an increased humoral vaccine response has been reported in interferon-β 1a (IFN)-treated MS subjects (Maniscalco et al., 2021). In light of the current discussion regarding SARS-CoV-2 booster vaccinations, several studies have evaluated the duration of anti-SARS-CoV-2 immune protection in healthy individuals (A Achiron et al., 2021; Andrews et al., 2022); however clinical data suggest an attenuated short-term humoral response to SARS-CoV-2 vaccines in patients with MS receiving DMTs (Tortorella et al., 2021; Capuano et al., 2022). Here we present a prospective study aimed at assessing the six-months persistence of the humoral response in a monocentric cohort of MS patients treated with eight different DMTs compared with healthy subjects, all receiving BNT162b2 mRNA vaccine and without previous SARS-CoV-2 infection. 2 Materials and methods 2.1 Subjects and study design This is a prospective monocentric study to evaluate the kinetics of SARS-CoV-2 IgG Spike titre 5–6 months after the second dose of BNT162b2 vaccine in MS subjects undergoing vaccination at the Multiple Sclerosis centre of the Cardarelli Hospital (Naples, Italy) from March to November 2021. All human subjects were enrolled after obtaining informed consent. The study was approved by the Institutional Review Board of the Cardarelli Hospital. We enrolled 96 MS and 21 healthy controls receiving the two doses of BNT162b2 vaccine according to the recommendations of Italian Authority of Health, all without previous SARS-CoV-2 infection. MS subjects were vaccinated according to specific timing; more in detail, MS subjects treated with IFN, GA, TERI, DMF, FTY and NAT were vaccinated without any interruption of immunomodulatory treatment, while CLAD- and OCRE-treated MS subjects were vaccinated at least 1 or 3 months respectively after the last therapeutic administration, according to the recommendations of Italian Authority of Health. Blood samples were collected at 9:00AM into heparinized Vacutainers (BD Biosciences) and processed within the following 4 h. Demographic and clinical characteristics of the study cohort are shown in Table 1 . Inclusion criteria were patients aged between 18 and 65 years, diagnosed with multiple sclerosis treated with DMTs for at least 6 months. Exclusion criteria were previous SARS-CoV-2 infection (antibody screening), any relapse and/or steroid use in the last 30 days before enrolment. Healthy subjects were matched for age and sex and had no history of inflammation, endocrine or autoimmune disease. The ethnic distribution among the groups was comparable, with all participants being Caucasian.Table 1 Clinical and demographic characteristics of the study cohort. Table 1 MS patients (N = 96) Healthy Controls (N = 21) Gender, n (%) ▒Female 57 (59.4) 11 (52.4) ▒Male 39 (40.6) 10 (47.6) Age, years ▒Mean age (±SD) 40.7 (±10.5) 48.3 (±10.8) ▒IQR (25–75) 14.5 20.5 EDSS, mean (range) 1.9 (0–6.5) DMTs duration, mean (months ± SD) 50.2 (±41.5) MS type, n (%) ▒RRMS 91 (94.8) ▒PPMS 1 (1.04) ▒SPMS 4 (4.2) DMTs, n (%) ▒Interferon -β 1a 19 (19.8) ▒Natalizumab 14 (14.6) ▒Dimethyl fumarate 14 (14.6) ▒Ocrelizumab 13 (13.5) ▒Fingolimod 13 (13.5) ▒Teriflunomide 8 (8.3) ▒Cladribine 8 (8.3) ▒Glatiramer acetate 7 (7.4) DMTs: Disease modifying therapy; EDSS: Expanded disability status scale; IQR: Interquartile range; MS: Multiple sclerosis; PPMS: Primary progressive multiple sclerosis; RRMS: Relapsing remitting multiple sclerosis; SPMS: Secondary progressive multiple sclerosis. 2.2 SARS-CoV-2 IgG spike detection Quantitative determination of antibodies to the SARS-CoV-2 Spike protein was carried out by Roche Elecsys® Anti‑SARS‑CoV‑2 S assay (Roche Diagnostics International Ltd, Rotkreuz, Switzerland). The assay was performed using a recombinant protein representing the RBD of the S antigen leading to a double-antigen sandwich assay complex which favors detection of high affinity antibodies against SARS-CoV-2 (range between 0.4 to 250 U/mL), resulting in a sensitivity of 98.8% (95% CI: 98.1 – 99.3%). 2.3 Ethics (standard protocol approvals, registrations and patient consents) The study was conducted according the Good Clinical Practice guidelines and the ethical principles of the Declaration of Helsinki. Investigators obtained ethic committee approval for the study protocol and amendments by the local Ethic Committee of A.O.R.N. A. Cardarelli/Santobono-Pausilipon (protocol number 2821). All subjects give written informed consent to participate to the study. 2.4 Statistical analysis Descriptive analyses were presented as mean (± standard error of the mean), median and interquartile range (IQR). Categorial variables were described as frequency and percentage. A Shapiro-wilk test was performed to assess the normal distribution of data. In case of not-normal distribution appropriate non-parametric tests were performed (Wilcoxon test). The fold persistence of SARS-CoV-2 Spike IgG titre was calculated as the percentage ratio between IgG titre at T2 and T1, relative to HC (100%). P-value less of 0.05 indicated significance. A multilinear regression model was used to compare the antibody levels across subjects treated with differerent DMTs after adjusting for age, sex, EDSS levels, disease and DMT duration and antibody levels in the pre-booster samples. Data analyses were performed using Graphpad Prism (version 8). 3 Results 3.1 Study cohort Data were collected from March to November 2021. In this prospective monocentric study, we excluded subjects previously infected with SARS-CoV-2, through the measurement of nucleocapsid-specific antibodies. After assessment, 96 MS patients and 21 HC were evaluated for anti-Spike IgG levels, 21 days (T1) and 5–6 months (T2) after the second SARS-CoV-2 BNT162b2 mRNA vaccine dose. The demographic and clinical characteristics are reported in Table 1. In the MS group, 57 were female (59.4%) and 39 male (40.6%) and the mean age was 40.7 ± 10.5 (mean ± SD) years. In the control group (HC), 11 subjects were females (52.4%) and 10 males (47.6%), with a mean age of 48.3 ± 10.8 (mean ± SD) years. In the MS cohort, 91 had relapsing-remitting (RR) (94.8%), 1 primary-progressive (PP) (1.04%) and 4 secondary-progressive (SP) (4.2%) MS. Different types of disease-modifying therapies (DMTs) were: interferon-β 1a (IFN) (n = 19; 19.8%), natalizumab (NAT) (n = 14; 14.6%), dimethyl fumarate (DMF) (n = 14; 14.6%), ocrelizumab (OCRE) (n = 13; 13.5%), fingolimod (FTY) (n = 13; 13.5%), teriflunomide (TERI) (n = 8; 8.3%), cladribine (CLAD) (n = 8; 8.3%), and glatiramer acetate (GA) (n = 7; 7.4%). 3.2 SARS-CoV-2 - specific humoral response declines differently in MS subjects undergoing distinct disease modifying therapies (DMTs) We previously reported that IFN-treated MS subjects showed a significant increase of anti-Spike IgG levels compared to HC, while humoral response was profoundly affected in MS subjects undergoing CLAD, FTY and OCRE, 21 days after the second BNT162b2 mRNA vaccine dose (Sormani et al., 2021). We aimed at evaluating the persistence of the SARS-CoV-2 IgG titre 5–6 months after the second vaccine dose (T2), in our monocentric cohort of MS subjects undergoing different DMTs, respect to age- and sex-matched HC. We found that SARS-CoV-2 IgG levels in MS subjects under treatment with IFN (median 566.6 (IQR: 427.5–1182) U/mL), DMF (median 694.6 (IQR: 438.3.5–1103) U/mL) TERI (median 507.8 (IQR: 278.7–1514) U/mL) and NAT (median 456.4 (IQR: 198.6–1156) U/mL) were comparable to HC (median 533.8 (IQR: 366.4–1008) U/mL) (Fig. 1 A and Table 2 ). On the contrary, humoral response at T2 was reduced in GA- (median 232.9 (IQR: 167.5–440.4) U/mL), CLAD- (median 246.5 (IQR: 145.3–363.3) U/mL) and FTY- (median 70.71 (IQR: 13.22–157.3) U/mL) and almost undetectable in OCRE- (median 0.4 (IQR: 0.4–4.3) U/mL) treated MS subjects (Fig. 1 A and Table 2). Follow-up of our monocentric cohort at 6 months revealed a significant drop in anti-Spike IgG titre compared to the relative T1 IgG values in HC (median 533.8 (IQR: 366.4 – 1008) vs 1087 (IQR:640.3 – 1542) U/mL), IFN- (median 566.6 (IQR: 427.5 – 1182) vs 1909 (IQR: 1073 – 2758) U/mL), DMF- (median 694.6 (IQR: 438.3 – 1103) vs 1587 (IQR:767.7 – 3472) U/mL), NAT- (median 456.4 (IQR: 198.6 – 1156) vs 735.1 (IQR: 351.3 – 2151) U/mL) and GA- (median 232.9 (IQR: 167.5 – 440.4) vs 446.5 (IQR: 417.7 – 3034) U/mL) treated MS patients (Table 2). The reduction was not significant in TERI-, CLAD- and OCRE-treated MS groups, while subjects under FTY showed a slight - albeit irrelevant - increase of anti-Spike IgG at T2 compared to their T1 levels (Fig. 1 B e Table 2). Finally, in those groups of DMT-treated MS subjects having shown a significant antibody production after the second SARS-CoV-2 BNT162b2 mRNA vaccine dose, we evaluated the persistence of the humoral response at 6 months, expressed as the percentage ratio between IgG titre at T2 and T1, relative to HC (Fig. 1 C). When compared to healthy controls, we observed a faster decline of anti-Spike IgG levels in patients undergoing DMF- (75.7 ± 11.5%), IFN- (62.4 ± 6.7%) and GA- (54.3 ± 11.3%) therapies, while those receiving TERI and NAT showed an increased persistence (130.9 ± 34.4% and 127.2 ± 32.6%, respectively), still not reaching the statistical significance (Fig. 1 C). We performed linear regression analyses to rule out that individual factors or clinical variables (age, sex, EDSS, DMT duration, time to last therapeutic administration) could affect the SARS-CoV-2 vaccine humoral response at T2, but we did not find any significant correlation between these parameters and IgG-Spike titre in the different DMT groups (data not shown). Our data highlight how DMTs could differentially promote the persistence of protective SARS-CoV-2 humoral response in MS subjects.Fig. 1 (A) SARS-CoV-2 IgG Spike titre (median with IQR) in MS subjects treated with different DMTs, 5–6 months after the second vaccine dose (T2) compared to healthy controls (HC). Wilcoxon two-tailed test vs HC subjects was performed and a p-value less than 0.05 was considered statistically significant. * p<0.05; p<0.01; *p<0.0001. (B) SARS-CoV-2 IgG Spike titre kinetics (mean ± SEM) of MS subjects treated with different DMTs at 21 days (T1) and 5–6 months (T2) after the second vaccine dose, compared to HC. Wilcoxon two-tailed test was performed to compare T1 and T2 levels and a p-value less than 0.05 was considered statistically significant. p<0.05; p<0.01; * p<0.005; ***p<0.0001. (C) Fold percentage of SARS-CoV-2 IgG Spike titre persistance (mean ± SEM) of MS subjects treated with different DMTs at T2, relative to HC (100%). Wilcoxon two-tailed test vs HC subjects was performed and a p-value less than 0.05 was considered statistically significant. p<0.05; ****p<0.0001. Fig 1 Table 2 Kinetics of SARS-CoV-2 IgG Spike titre at T1 ant T2, respectively 21 days and 5–6 months after the second vaccine dose in HC and MS subjects treated with different DMTs. Table 2DMTs SARS-CoV-2 IgG Spike titre (U/mL) p-value Median (IQR) T1 Median (IQR) T2 Healthy Controls 1087 (640.3 – 1542) 533.8 (366.4 – 1008) 0.003 Interferon-β 1a 1909 (1073 – 2758) 566.6 (427.5 – 1182) <0.0001 Dimethyl fumarate 1587 (767.7 – 3472) 694.6 (438.3 – 1103) 0.009 Teriflunomide 1312 (288.4 – 1920) 507.8 (278.7 – 1514) 0.38 Natalizumab 735.1 (351.3 – 2151) 456.4 (198.6 – 1156) 0.002 Glatiramer Acetate 446.5 (417.7 – 3034) 232.9 (167.5 – 440.4) 0.015 Cladribine 468.2 (65.5 – 1021) 246.5 (145.3 – 363.3) 0.31 Fingolimod 20.6 (5.9 – 456.3) 70.7 (13.2 – 157.3) 0.94 Ocrelizumab 1.2 (0.4 – 119.4) 0.4 (0.4 – 4.3) 0.10 Wilcoxon two-tailed test was performed to compare SARS-CoV-2 IgG Spike titre at T2 and T1 in HC and MS subjects treated with different DMTs. Data are reported as median and IQR at T1 and T2. A p-value less than 0.05 was considered statistically significant. DMTs: Disease modifying therapies; IQR: Interquartile range. 4 Discussion Our study shows that antibody responses to SARS-CoV-2 BNT162b2 mRNA vaccine are broadly distributed and declined substantially in most individuals over time. Antibodies are key immune effectors that confer protection against pathogens (Chen et al., 2020). The longevity of the antibody response to SARS-CoV-2 vaccine in multiple sclerosis (MS) subjects are still not well defined. In this context, decisions regarding initiation or continuation of specific disease modifying therapy (DMT) have to consider the potential relevance to the pandemic. Understanding the possible distinctive effects of each therapeutic agent on the immune response to the vaccine is essential during this special time. In this monocentric study, we charted longitudinal anti-Spike specific IgG antibody response to SARS-CoV-2 BNT162b2 mRNA vaccine in MS subjects undergoing eight different DMTs, compared to healthy controls (HC). 96 MS subjects followed longitudinally from day 21 (T1) to 5–6 months (T2) after the second SARS-CoV-2 BNT162b2 vaccine dose showed marked heterogeneity in antibody duration dynamics. Anti-Spike IgG decayed substantially in most individuals, whereas distinct groups had stable or increasing antibody levels in the same time frame (T2), despite the initial antibody magnitudes at T1. More in detail, anti-Spike IgG levels at T2 were comparable among IFN-, DMF-, TERI- and NAT-treated MS subjects and reflect those of HC. Conversely, antibody levels were significantly reduced in MS subjects under GA-, CLAD-, FTY-treatments and almost absent in OCRE-treated MS groups. Vaccination of cladribine-treated subjects occurred at least one month after the last therapy, in line with the Italian Authority of Health recommendations (Centonze et al., 2021); as these subjects have reduced anti-Spike IgG levels, this aspect should be considered in the selection of the booster time administration. However, based on their anti-Spike IgG levels at T1, we evaluated the antibody persistence in those DMT-groups exhibiting an adequate humoral response after vaccination. We observed that, when compared to HC, TERI- and NAT-treated MS subjects showed an increased persistence of anti-Spike specific IgG antibody response to SARS-CoV-2 BNT162b2 mRNA vaccine after 5–6 months, while DMF-, IFN- and GA-therapies affected the retention of the humoral response overtime. An important implication of our data is the possible existence of an efficient SARS-CoV-2 vaccine ‘‘holder’’ phenotype, defined as individuals who experience relatively sustained anti-SARS-CoV-2 IgG production. This seems to be the case of MS under teriflunomide, which selectively inhibits dihydro-orotate dehydrogenase (DHODH), an important mitochondrial enzyme in the de novo pyrimidine synthesis pathway. The downstream effect is the reduced proliferation of rapidly dividing cells, including activated T and B lymphocytes. It has been reported that TERI-treated MS patients developed effective immunity to seasonal influenza after vaccination. In addition, there is emerging evidence suggesting a direct antiviral effect for teriflunomide and other DHODH inhibitors against a range of viruses such as Theiler's, respiratory syncytial, Ebola, cytomegalovirus, Epstein–Barr, and picornavirus (Maghzi et al., 2020). Moreover, our data are in line with the successful development of anti-SARS-CoV-2 antibodies described in patients under teriflunomide after COVID-19 infection (Bollo et al., 2020). Regarding natalizumab, it acts as an antagonist to alpha-4 integrin on leukocyte surface, blocking their interaction with vascular cell adhesion molecules and preventing leukocyte migration to the CNS. Data from influenza vaccine studies strongly support its treatment under the COVID-19 pandemic; however, some concerns have been raised regarding the potential increase of viral shedding due to the reduction of lymphocyte trafficking in the lungs (Aguirre et al., 2020). Since our data unveiled a slight increase in both T1 and T2 anti-Spike IgG levels compared to HC and a good humoral persistence, this could suggest a delayed booster timing for NAT-treated MS subjects. The same holds true also for TERI-treated MS subjects which, despite a lower antibody production at T1, retained an excellent humoral reservoir at T2. According to the literature, our study shows that different humoral persistence could be observed among immuno-modulating DMTs (Tortorella et al., 2021; Capuano et al., 2022; A Achiron et al., 2021); this offers new evidence that, if confirmed in a larger cohort of patients, should be considered when formulating booster regimens in MS subjects. 5 Data access, responsibility, and analysis The corresponding author had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Author contributions G.T. Maniscalco, A.L. Ferrara, V. De Rosa: Conceptualization; A.L. Ferrara, A. Liotti, D. Di Giulio Cesare, O. Moreggia: Data curation; A.L. Ferrara, A. Liotti, D. Di Giulio Cesare: Formal analysis; G.T. Maniscalco, V. De Rosa: Investigation; G.T. Maniscalco, V. De Rosa: Methodology; V. Manzo, M. E. Di Battista, S. Salvatore, D. Graziano, A. Viola, G. Amato, G. Alfieri, W. Di Iorio, G. Della Rocca, V. Andreone: Project administration; D. Graziano, A. Viola, G. Amato: Resources; A.L. Ferrara, A. Liotti, D. Di Giulio Cesare, O. Moreggia: Software; G.T. Maniscalco, V. De Rosa, V. Andreone: Supervision; V. Manzo, M. E. Di Battista, S. Salvatore, D. Graziano, A. Viola, G. Amato, O. Moreggia, G. Alfieri, W. Di Iorio, G. Della Rocca, V. Andreone: Validation; V. Manzo, M. E. Di Battista, S. Salvatore, D. Graziano, A. Viola, G. Amato, G. Alfieri, W. Di Iorio, G. Della Rocca, V. Andreone: Visualization; G.T. Maniscalco, A.L. Ferrara, V. De Rosa: Roles/Writing - original draft; A. Liotti, V. Manzo, M. E. Di Battista, S. Salvatore, D. Graziano, A. Viola, G. Amato, G. Alfieri, W. Di Iorio, G. Della Rocca, V. Andreone: Writing - review & editing. Funding This work was supported by FISM 2018/R/4 from Fondazione Italiana Sclerosi Multipla, STAR Program Linea 1-2018 by UniNA and by Compagnia di San Paolo from the Università degli Studi di Napoli “Federico II”, Bando PRIN 2017 Prot. 2017K7FSYB from Ministry of Education, University and Research (MIUR). Ethical standards All procedures were performed in accordance with the institutional ethics committee and the Declaration of Helsinki. CRediT authorship contribution statement Giorgia Teresa Maniscalco: Conceptualization, Investigation, Methodology, Supervision, Writing – original draft. Anne Lise Ferrara: Conceptualization, Data curation, Formal analysis, Software, Writing – original draft. Antonietta Liotti: Data curation, Formal analysis, Software, Writing – review & editing. Valentino Manzo: Project administration, Validation, Visualization, Writing – review & editing. Maria Elena Di Battista: Project administration, Validation, Visualization, Writing – review & editing. Simona Salvatore: Project administration, Validation, Visualization, Writing – review & editing. Daniela Graziano: Project administration, Resources, Validation, Visualization, Writing – review & editing. Assunta Viola: Project administration, Resources, Validation, Visualization, Writing – review & editing. Gerardino Amato: Project administration, Resources, Validation, Visualization, Writing – review & editing. Ornella Moreggia: Data curation, Software, Validation. Daniele Di Giulio Cesare: Data curation, Formal analysis, Software. Gennaro Alfieri: Project administration, Validation, Visualization, Writing – review & editing. Walter Di Iorio: Project administration, Validation, Visualization, Writing – review & editing. Gennaro Della Rocca: Project administration, Validation, Visualization, Writing – review & editing. Vincenzo Andreone: Project administration, Supervision, Validation, Visualization, Writing – review & editing. Veronica De Rosa: Conceptualization, Investigation, Methodology, Supervision, Writing – original draft. Declaration of Competing Interest G.T. Maniscalco received personal compensations from Serono, Biogen, Novartis, Roche and TEVA for public speaking and advisory boards. The other authors have nothing to disclose. ==== Refs References Gavriatopoulou M. Korompoki E. Fotiou D. Ntanasis-Stathopoulos I. Psaltopoulou T. Kastritis E. Terpos E. Dimopoulos M.A. Organ-specific manifestations of COVID-19 infection Clin. Exp. Med. 20 4 2020 493 506 10.1007/s10238-020-00648-x 32720223 Borobia A.M. Carcas A.J. Perez-Olmeda M. Castano L. Bertran M.J. Garcia-Perez J. Campins M. Portoles A. Gonzalez-Perez M. Garcia Morales M.T. Arana-Arri E. Aldea M. Diez-Fuertes F. Fuentes I. Ascaso A. Lora D. Imaz-Ayo N. Baron-Mira L.E. Agusti A. Perez-Ingidua C. Gomez de la Camara A. Arribas J.R. Ochando J. Alcami J. Belda-Iniesta C. Frias J. CombiVac S.S.G. Immunogenicity and reactogenicity of BNT162b2 booster in ChAdOx1-S-primed participants (CombiVacS): a multicentre, open-label, randomised, controlled, phase 2 trial Lancet 398 10295 2021 121 130 10.1016/S01406736(21)01420-3 34181880 Walsh E.E. Frenck R.W. Jr. Falsey A.R. Kitchin N. Absalon J. Gurtman A. Lockhart S. Neuzil K. Mulligan M.J. Bailey R. Swanson K.A. Li P. Koury K. Kalina W. Cooper D. Fontes-Garfias C. Shi P.Y. Tureci O. Tompkins K.R. Lyke K.E. Raabe V. Dormitzer P.R. Jansen K.U. Sahin U. Gruber W.C. Safety and immunogenicity of two RNA-based Covid-19 vaccine candidates N. Engl. J. Med. 383 25 2020 2439 2450 10.1056/NEJMoa2027906 33053279 Khoury D.S. Cromer D. Reynaldi A. Schlub T.E. Wheatley A.K. Juno J.A. Subbarao K. Kent S.J. Triccas J.A. Davenport M.P. Neutralizing antibody levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection Nat. Med. 27 7 2021 1205 1211 10.1038/s41591-021-01377-8 34002089 Tortorella C. Aiello A. Gasperini C. Agrati C. Castilletti C. Ruggieri S. Meschi S. Matusali G. Colavita F. Farroni C. Cuzzi G. Cimini E. Tartaglia E. Vanini V. Prosperini L. Haggiag S. Galgani S. Quartuccio M.E. Salmi A. Repele F. Altera A.M.G. Cristofanelli F. D'Abramo A. Bevilacqua N. Corpolongo A. Puro V. Vaia F. Capobianchi M.R. Ippolito G. Nicastri E. Goletti D. INMI COVID-19 Vaccine Study Group Humoral- and T-cell-specific immune responses to SARS-CoV-2 mRNA vaccination in patients with MS using different disease-modifying therapies Neurol 98 5 2021 e541 e554 10.1212/WNL.0000000000013108 1 Sormani M.P. Inglese M. Schiavetti I. Carmisciano L. Laroni A. Lapucci C. Da Rin G. Serrati C. Gandoglia I. Tassinari T. Perego G. Brichetto G. Gazzola P. Mannironi A. Stromillo M.L. Cordioli C. Landi D. Clerico M. Signoriello E. Frau J. Ferro M.T. Di Sapio A. Pasquali L. Ulivelli M. Marinelli F. Callari G. Iodice R. Liberatore G. Caleri F. Repice A.M. Cordera S. Battaglia M.A. Salvetti M. Franciotta D. Uccelli A. M. S. study group on behalf of the Italian Covid-Alliance in M. S. CovaXi Effect of SARS-CoV-2 Mrna vaccination in MS patients treated with disease modifying therapies EBioMedicine 72 2021 103581 10.1016/j.ebiom.2021.103581 Maniscalco G.T. Manzo V. Ferrara A.L. Perrella A. Di Battista M. Salvatore S. Graziano D. Viola A. Amato G. Moreggia O. Di Giulio Cesare D. Barbato S. Servillo G. Longo K. Di Giovanni M. Scarpati B. Muggianu S.M. Longo G. Russo G. Andreone V. De Rosa V. Interferon Beta-1a treatment promotes SARS-CoV-2 mRNA vaccine response in multiple sclerosis subjects Mult. Scler. Relat. Disord. 58 2021 103455 10.1016/j.msard.2021.103455 Achiron A. Mandel M. Dreyer-Alster S. Harari G. Gurevich M. Humoral SARS-COV-2 IgG decay within 6 months in COVID-19 healthy vaccinees: the need for a booster vaccine dose? Eur. J. Intern. Med. 94 2021 105 107 10.1016/j.ejim.2021.10.027 34742628 Andrews N. Stowe J. Kirsebom F. Toffa S. Rickeard T. Gallagher E. Gower C. Kall M. Groves N. O'Connell A.M. Simons D. Blomquist P.B. Zaidi A. Nash S. Iwani Binti Abdul Aziz N. Thelwall S. Dabrera G. Myers R. Amirthalingam G. Gharbia S. Barrett J.C. Elson R. Ladhani S.N. Ferguson N. Zambon M. Campbell C.N.J. Brown K. Hopkins S. Chand M. Ramsay M. Lopez Bernal J. Covid-19 vaccine effectiveness against the Omicron (B.1.1.529) variant N. Engl. J. Med. 2022 10.1056/NEJMoa2119451 Capuano R. Bisecco A. Conte M. Donnarumma G. Altieri M. Grimaldi E. Franci G. Chianese A. Galdiero M. Coppola N. Tedeschi G. Gallo A. Six-month humoral response to mRNA SARS-CoV-2 vaccination in patients with multiple sclerosis treated with ocrelizumab and fingolimod Mult. Scler. Relat. Disord. 4 60 2022 103724 10.1016/j.msard.2022.103724 Chen Y. Zuiani A. Fischinger S. Mullur J. Atyeo C. Travers M. Lelis F.J.N. Pullen K.M. Martin H. Tong P. Gautam A. Habibi S. Bensko J. Gakpo D. Feldman J. Hauser B.M. Caradonna T.M. Cai Y. Burke J.S. Lin J. Lederer J.A. Lam E.C. Lavine C.L. Seaman M.S. Chen B. Schmidt A.G. Balazs A.B. Lauffenburger D.A. Alter G. Wesemann D.R. Quick COVID-19 healers sustain anti-SARS-CoV-2 antibody production Cell. Dec 10 6 2020 1496 1507 10.1016/j.cell.2020.10.051 183e16 Centonze D. Rocca M.A. Gasperini C. Kappos L. Hartung H.P. Magyari M. Oreja-Guevara C. Trojano M. Wiendl H. Filippi M. Disease-modifying therapies and SARS-CoV-2 vaccination in multiple sclerosis: an expert consensus J. Neurol. 268 11 2021 3961 3968 10.1007/s00415-021-10545-2 33844056 Maghzi A.H. Houtchens M.K. Preziosa P. Ionete C. Beretich B.D. Stankiewicz J.M. Tauhid S. Cabot A. Berriosmorales I. Schwartz T.H.W. Sloane J.A. Freedman M.S. Filippi M. Weiner H.L. Bakshi R. COVID-19 in teriflunomide-treated patients with multiple sclerosis J. Neurol. 267 10 2020 2790 2796 10.1007/s00415-020-09944-8 32494856 Bollo L. Guerra T. Bavaro D.F. Monno L. Saracino A. Angarano G. Paolicelli D. Trojano M. Iaffaldano P. Seroconversion and indolent course of COVID-19 in patients with multiple sclerosis treated with fingolimod and teriflunomide J. Neurol. Sci. 416 2020 117011 10.1016/j.jns.2020.117011 Aguirre C. Meca-Lallana V. Barrios-Blandino A. Del Rio B. Vivancos J. Covid-19 in a patient with multiple sclerosis treated with natalizumab: may the blockade of integrins have a protective role? Mult. Scler. Relat. Disord. 44 2020 102250 10.1016/j.msard.2020.102250 Achiron A. Mandel M. Dreyer-Alster S. Harari G. Dolev M. Menascu S. Magalashvili D. Flechter S. Givon U. Guber D. Sonis P. Zilkha-Falb R. Gurevich M. Humoral immune response in multiple sclerosis patients following PfizerBNT162b2 COVID19 vaccination: up to 6 months cross-sectional study J. Neuroimmunol. 361 2021 577746 10.1016/j.jneuroim.2021.577746
PMC009xxxxxx/PMC9005242.txt	==== Front Behav Res Ther Behav Res Ther Behaviour Research and Therapy 0005-7967 1873-622X The Authors. Published by Elsevier Ltd. S0005-7967(22)00066-3 10.1016/j.brat.2022.104095 104095 Article Social cognition theories and behavior change in COVID-19: A conceptual review Hagger Martin S. abcd∗ Hamilton Kyra bde a Department of Psychological Sciences, University of California, Merced, 5200 N. Lake Rd., Merced, CA, 95343, USA b Health Sciences Research Institute, University of California, Merced, 5200 N. Lake Rd., Merced, CA, 95343, USA c Faculty of Sport and Health Sciences, University of Jyväskylä, PO Box 35, FI-40014, Jyväskylä, Finland d School of Applied Psychology, Griffith University, Mt. Gravatt Campus, 176 Messines Ridge Rd, Mt. Gravatt, QLD, 4122, Australia e Menzies Health Institute Queensland, Griffith University, G40 Griffith Health Centre, Level 8.86, Gold Coast Campus, QLD, 4222, Australia ∗ Corresponding author. Department of Psychological Sciences and Health Sciences Research Institute, University of California, Merced, 5200 N. Lake Rd., Merced, CA, 95343, USA. 12 4 2022 7 2022 12 4 2022 154 104095104095 31 12 2021 31 3 2022 8 4 2022 © 2022 The Authors 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. The COVID-19 pandemic has had unprecedented health, economic, and social consequences worldwide. Although contact reductions and wearing face coverings have reduced infection rates, and vaccines have reduced illness severity, emergence of new variants of the coronavirus that causes COVID-19, and the shift from pandemic to endemic patterns of infection, highlights the importance of ongoing preventive behavior adherence to manage future outbreaks. Research applying social cognition theories may assist in explaining variance in these behaviors and inform the development of efficacious behavior change interventions to promote adherence. In the present article, we summarize research applying these theories to identify modifiable determinants of COVID-19 preventive behaviors and the mechanisms involved, and their utility in informing interventions. We identify limitations of these applications (e.g., overreliance on correlational data, lack of long-term behavioral follow-up), and suggest how they can be addressed. We demonstrate the virtue of augmenting theories with additional constructs (e.g., moral norms, anticipated regret) and processes (e.g., multiple action phases, automatic processes) to provide comprehensive, parsimonious behavioral explanations. We also outline how the theories contribute to testing mechanisms of action of behavioral interventions. Finally, we recommend future studies applying these theories to inform and test interventions to promote COVID-19 preventive behavior adherence. Keywords COVID-19 preventive behaviors Integrated models Motivation Reflective processes Automatic processes Mechanism of action ==== Body pmc1 Introduction Since the COVID-19 outbreak was declared a global pandemic in March 2020 (WHO, 2020), the disease has contributed substantially to excess deaths globally (CDC, 2021b), resulted in damaging economic outcomes (e.g., unemployment and redundancy, increased numbers of people falling below the poverty line, business bankruptcy and closure; Jackson et al., 2021), placed immense strain on healthcare services (e.g., reduced capacity to provide elective services and long-term care, increased stress on staff, elevated health risks for frontline workers; Bartsch et al., 2020), and led to deleterious social consequences (e.g., increased incidence of mental health difficulties including depression, suicidality, loneliness, and isolation; Holmes et al., 2020). Lockdown and other legislated mitigation procedures together with the rapid development and administration of vaccines have made important contributions to curbing infection rates, particularly severe cases, as well reducing economic, healthcare, and social strains, such that people worldwide have begun to envision a post-pandemic world (Kashte, Gulbake, El-Amin, & Gupta, 2021). However, the emergent threat of new highly-virulent variants of the coronavirus that causes COVID-19 infections, such as the delta and omicron variants, highlights the reality that the pandemic is far from over, and that there will likely be need to maintain mitigation and management procedures for some time in the future to bring infection rates under control (Wu et al., 2021). Furthermore, even if mitigation procedures lead to COVID-19 infection rates to fall below pandemic levels, complete eradication of the virus is unlikely (Phillips, 2021). Rather, it is more likely to become endemic, similar to other viral infections such as influenza and the common cold. This means that people worldwide may need to be prepared to manage localized and seasonal outbreaks in future (Phillips, 2021). Throughout the pandemic, behavior change has been central to the effective management of COVID-19 infection rates (Michie, 2020; West, Michie, Rubin, & Amlôt, 2020). Populations worldwide have become eminently familiar with a raft of COVID-19 preventive behaviors including physical distancing, wearing face coverings, avoiding large gatherings particularly indoors, hand sanitization, and adherence to transmission prevention guidelines (e.g., self-isolation, quarantine). Further key preventive behaviors have also become cornerstone in the mitigation of coronavirus transmission and minimizing infection severity, including uptake of COVID-19 vaccines and booster shots (Bar-On et al., 2021; CDC, 2021a; JVCI, 2021) and rapid antigen testing (Crozier, Rajan, Buchan, & McKee, 2021). However, despite extensive messaging and advocacy of these COVID-19 preventive behaviors by government agencies and public health services, engagement in these behaviors has been highly variable, and often falls short of the levels necessary to bring infection rates under control, particularly when they are at their peak (Mathieu et al., 2021). This has particularly been the case for vaccine uptake, where adherence rates vary considerably but often fall below those necessary to attain the widespread immunity to halt subsequent waves of infection (Mathieu et al., 2021). Further, increased documentation of ‘breakthrough’ infections among the vaccinated population (Bergwerk et al., 2021), and the need to offer protection to those vulnerable to serious cases of COVID-19 (e.g., vaccine allergic individuals, individuals with underlying conditions, and the immunosuppressed), means that other COVID-19 preventive behaviors such as physical distancing and wearing face coverings are still necessary even among those who have been vaccinated. Given the ongoing pandemic and its future management necessitates continuity of interventions promoting uptake and maintenance of COVID-19 preventive behaviors, the need for effective messaging and health communication has come to the fore. Governmental and public health agencies have turned to behavioral science to identify strategies that will promote attention to, assimilation of, and responsivity to messaging interventions around COVID-preventive behaviors (Albarracín & Jung, 2021; Bonell et al., 2020; Michie, Rubin, & Amlôt, 2020; Michie et al., 2020; West et al., 2020). Importantly, drawing from a substantive body of prior research on behavior change, behavioral scientists have highlighted the imperative for messaging interventions to be based on a fundamental understanding of human behavior (Hagger, Cameron, et al., 2020; Michie et al., 2018). Central to this understanding is the need to base interventions on behavioral theory (Hagger, Moyers, McAnally, & McKinley, 2020; Prestwich, Webb, & Conner, 2015), a contention predicated on evidence that theory-based behavioral interventions demonstrate greater efficacy and efficiency (Hagger & Weed, 2019; McEwan et al., 2019; Prestwich et al., 2014). Theory provides a basis for providing a priori predictions on how interventions work, that is, the extent to which the intervention is expected to change behavior and the mechanisms involved, and provides a means to falsify predictions relating to efficacy and mechanism against observation (Rothman, Klein, & Sheeran, 2020; Sheeran, Klein, & Rothman, 2017). In addition, basing interventions on theory enables specification of the inter- and intra-personal, socio-structural, and contextual conditions that are expected to enhance or mitigate their efficacy (Hagger, Gucciardi, & Chatzisarantis, 2017; Rothman & Sheeran, 2020; Trafimow, 2012). Social cognition theories feature prominently among theories that have been applied to predict behavior and behavior change, and have made important contributions to identifying health behavior determinants and the mechanisms involved (Conner and Norman, 2015a, Conner and Norman, 2015b; Fishbein et al., 2001). Social cognition theories adopt a reasoned action approach, assuming that engagement in a given target behavior is a function of individuals' beliefs or judgements with respect to their future performance of the behavior (Fishbein & Ajzen, 2010). Importantly, the beliefs are considered potentially modifiable through information-provision strategies presented in behavioral interventions (e.g., messages highlighting the utility of, available support for, and personal capacity to perform the behavior; Ajzen & Schmidt, 2020; Hamilton & Johnson, 2020). The theories also specify the processes by which these strategies ‘work’ to change behavior, known as mechanisms of action (Hagger, Cameron, et al., 2020; Hagger, Moyers, et al., 2020; Sheeran et al., 2017). A large corpus of research applying social cognition approaches in health contexts has supported their predictions and demonstrated that their constructs account for non-trivial variance in behavior (e.g., Carpenter, 2010; McEachan, Conner, Taylor, & Lawton, 2011; Milne, Sheeran, & Orbell, 2000). In the context of the COVID-19 pandemic, social cognition theories have featured prominently in studies predicting COVID-19 preventive behaviors (for a review see Albarracín & Jung, 2021). These studies have been successful in accounting for unique variance in behavior, and have assisted in identifying relevant processes (e.g., Bogg & Milad, 2020; Chu & Liu, 2021; Hagger, Smith, Keech, Moyers, & Hamilton, 2020; Hamilton, Smith, Keech, Moyers, & Hagger, 2020; Jang, Kim, & Kwon, 2021; Norman, Wilding, & Conner, 2020; Tong, He, Wu, Dang, & Chen, 2021). In addition, their application has also demonstrated utility in informing the development of behavioral interventions in this context – interventions based on these theories have demonstrated efficacy in changing behavior (e.g., Ahn, Hu, & Vega, 2021; Keller et al., 2021; Okuhara, Okada, & Kiuchi, 2020; Smith, Hagger, Keech, Moyers, & Hamilton, 2021). Despite these successful applications, numerous limitations and knowledge gaps remain in the application of these theories in this context such as a preponderance of evidence based on cross-sectional and correlational data; a lack of behavioral data and, particularly, data on long-term behavioral prediction and the prediction of change in behavior; a lack of experimental and intervention research; and few tests of the theory-based mechanisms by which the interventions change behavior. In this article, we aim to summarize the contribution of research applying social cognition theories to predict and change behavior in the context of COVID-19. Specifically, we outline the value of social cognition theories in identifying the determinants and possible targets for intervention in health behavior, focusing on COVID-19 preventive behaviors (e.g., maintaining physical distancing, wearing face coverings, avoiding group gatherings; getting vaccinated, participating in rapid testing); provide a critique of the application of these theories including identifying some of their prominent limitations and boundary conditions (e.g., prediction of intentions rather than behavior, confinement to short term prediction, lack of specificity in beliefs, overuse of correlational designs), and their implications for COVID-19 prevention; summarize some emergent innovations in the research in this context (e.g., integration of other constructs, testing of mechanisms of action) that seek to address these limitations and advance knowledge on behavioral determinants and behavior change in this context; and identify evidence gaps (e.g., the transition from pandemic to endemic illness management, accounting for disparities and inequality in health behavior, dealing with traits and individual differences) and provide some suggested future advancements that may contribute to addressing these gaps. 2 Social cognition theories and COVID-19 preventive behaviors 2.1 Social cognition theories: basic assumptions and supporting evidence Social cognition theories have their origins in reasoned action approaches that assume individuals’ decisions on whether or not to engage in a given target behavior are based on their processing of the available information relating to the behavior and their evaluation of it (Ajzen, 1991; Conner, 2015; Fishbein & Ajzen, 2010; Fishbein et al., 2001). Specifically, the theories predict that individuals form belief-based judgements or expectations that represent or summarize information about the behavior and make their decisions on action accordingly. Social cognitive theory (Bandura, 1986), the health belief model (Rosenstock, 1974), protection motivation theory (Rogers, 1975), and the theories of reasoned action (Ajzen & Fishbein, 1980) and planned behavior (Ajzen, 1991) are leading examples of social cognition theories, and they have been widely applied. The belief-based behavioral determinants specified in social cognition theories are summarized in sets of constructs. Common to many of these theories are constructs that represent beliefs regarding the value or utility of the behavior in producing outcomes (e.g., outcome evaluations in social cognitive theory, perceived benefits in the health belief model, response efficacy in protection motivation theory, attitudes in the theories of reasoned action and planned behavior); risk perceptions (e.g., threat appraisals in the health belief model and protection motivation theory); capacity beliefs (e.g., self-efficacy beliefs in social cognitive theory and protection motivation theory, perceived behavioral control in the theory of planned behavior), and social norms (e.g., subjective norms in the theories of reasoned action and planned behavior). A number of theories also specify dispositions to act (e.g., intention in the theories of reasoned action and planned behavior, protection motivation in protection motivation theory), which reflect a readiness to act and willingness to invest effort to pursue the behavior in future, often operationalized as a mediator of effects of belief-based constructs on behavior, a mechanistic prediction (Ajzen, 1991; Fishbein & Ajzen, 2010; Rogers, 1975; Fishbein et al., 2001, McMillan and Conner, 2007). For example, a prototypical social cognition theory, the theory of planned behavior, predicts that effects of attitudes, subjective norms, and perceived behavioral control with respect to future performance of the target behavior on behavior are mediated by intentions (Ajzen, 1991). The basic premises of social cognition theories in the context of health behaviors has been supported in multiple predictive studies (e.g., Carpenter, 2010; McEachan et al., 2011; Milne et al., 2000). Studies typically tap the belief-based constructs from the theories using validated psychometric instruments in which individuals from a target population are prompted to estimate their beliefs with respect to performing a target behavior at some specified time in the future. Valid tests of the theories necessitate subsequent collection of behavior measures within the specified timeframe, and evaluate the extent to which the belief-based constructs predict the behavior using a prospective design. Studies adopting such designs have provided consistent support for the theories in explaining variance in behavior in multiple health behaviors, populations, and contexts. For example, meta-analytic syntheses of these studies demonstrate small-to-medium sized effects of theory constructs on action dispositions and behavior, and have also supported key mechanistic predictions, such as the mediation of the effects of belief-based constructs on behavior by dispositions to act (e.g., Albarracín, Johnson, Fishbein, & Muellerleile, 2001; Hamilton, van Dongen, & Hagger, 2020; McEachan et al., 2011). This meta-analytic research has also been extended to account for other important predictions and auxiliary assumptions of these theories such as the capacity of the theory to account for unique variance in behavior when controlling for past behavior (e.g., Chatzisarantis, Hagger, Smith, & Phoenix, 2004; Hagger, Polet, & Lintunen, 2018), a test of theory sufficiency, and also for moderator effects among the constructs themselves, such as the moderation of the intention-behavior relationship by perceived behavioral control (Hagger, Cheung, Ajzen, & Hamilton, 2022) or the properties of attitudes (Cooke & Sheeran, 2004) in the theory of planned behavior. Taken together, research has provided basic evidence in support of theory hypotheses with regard to the prediction of behavior and some key proposed mechanisms within the theories. 2.2 Applying social cognition theories in COVID-19 The demonstrated efficacy of social cognition theories in accounting for behavior provides impetus for their application to predict COVID-19 preventive behaviors. Specifically, leading social cognition theories have been adopted to predict intention toward, and actual participation in, general COVID-19 preventive behaviors (e.g., Clark, Davila, Regis, & Kraus, 2020; Norman et al., 2020; Peterson, Helweg-Larsen, & DiMuccio, 2021; Rabin & Dutra, 2021), or specific preventive behaviors such as social or physical distancing (Adiyoso & Wilopo, 2021; Das, Abdul Kader Jilani, Uddin, Uddin, & Ghosh, 2021; Gibson, Magnan, Kramer, & Bryan, 2021; Yu, Lau, & Lau, 2021), wearing face coverings (e.g., Barile et al., 2020; Irfan et al., 2021), hand hygiene (e.g., Derksen, Keller, & Lippke, 2020; Luszczynska et al., 2021), and COVID-19 testing adherence (e.g., McElfish, Purvis, James, Willis, & Andersen, 2021; Vandrevala, Montague, Terry, & Fielder, 2022). General trends from this research suggest that beliefs about utility such as attitudes and response efficacy (e.g., Clark et al., 2020; Rabin & Dutra, 2021; Yu et al., 2021), normative beliefs such as subjective and descriptive norms (e.g., Das et al., 2021; Gibson et al., 2021; Peterson et al., 2021), and beliefs about capacity such as self-efficacy and perceived behavioral control (e.g., Adiyoso & Wilopo, 2021; Das et al., 2021; Norman et al., 2020) account for unique variance in intentions or behavior in these behavioral contexts. Beliefs about threat or risk from COVID-19, such as risk perceptions or perceived severity and susceptibility, have also been shown to have unique effects on intentions and behavior for these behaviors (e.g., Betsch et al., 2021; Vandrevala et al., 2022), but effect sizes tend to be modest by comparison, or even non-significant, when included as predictors in parallel other theory constructs (Adiyoso & Wilopo, 2021; Derksen et al., 2020; Hamilton, Smith, et al., 2020; Rabin & Dutra, 2021). This is consistent with research indicating that beliefs reflecting behavioral engagement tend to account for substantially more variance in intention and behavior than beliefs about risk from the conditions they are purported to prevent (Hagger & Orbell, 2021) – likely a consequence of the close correspondence between the measures of the beliefs and the behavior, but also because decisions to engage in these preventive behaviors are seldom focused solely on risk reduction. The theories have also been applied to predict COVID-19 vaccine intentions (e.g., Chu & Liu, 2021; Matute, Palau-Saumell, Meyer, Derqui, & Jiménez-Asenjo, 2021; Sherman et al., 2021; Shiloh, Peleg, & Nudelman, 2021). Similar patterns of effects for key social cognition constructs on vaccine intentions have been observed as those found for other COVID-19 preventive behaviors. In a marked deviation, however, trends across these studies suggest that beliefs about the risks of the vaccines themselves and concerns about general vaccine administration such as injections, rather than risks related to COVID-19, are key correlates of vaccine intentions (Chu & Liu, 2021; Hamilton & Hagger, 2022; Matute et al., 2021; Sherman et al., 2021). However, to date, no research has provided a formal comparison of the relative effects of risk perceptions related to COVID-19 and risk perceptions relating to the vaccine itself on COVID-19 vaccination intentions. Futhermore, there a dearth of research on the social cognition correlates of actual vaccine uptake (Shiloh et al., 2021). Taken together, application of social cognition theories has provided initial evidence of the belief-based correlates of COVID-19 preventive behaviors and contribute to an initial evidence base of potentially modifiable constructs that may be targeted in interventions. 2.3 Social cognition theories in COVID-19: limitations and solutions Studies applying social cognition theories have provided evidence of the constructs associated with COVID-19 preventive behaviors and the mechanisms involved. However, this growing but relatively new research literature has a number of prominent limitations that restrict the extent to which inferences can be drawn from their findings and the degree to which they can be generalized broadly. The limitations include a focus on COVID-19 preventive behavioral intentions rather than behavior itself, with few studies adopting long-term behavioral follow up; a focus on direct or global measures of social cognition constructs, which neglects measures of specific sets of beliefs relevant to COVID-19 preventive behaviors; an over-reliance on cross-sectional, correlational design studies which limit inference of directionality, causality, and change in constructs, particularly behavior, over time; and an exclusive focus on the individual which eschews effects of broader socio-structural factors and environmental context on behavior. In this section we outline the ramifications of these limitations and outline current and future research that may address these limitations. 2.3.1 A focus on intention and short-term prediction Many studies applying social cognition theories in the context of COVID-19 preventive behaviors have focused on the prediction of intention, with relatively few providing follow-up measures of behavior (e.g., Clark et al., 2020; Das et al., 2021; Derksen et al., 2020). This is particularly the case for vaccine behavior, likely due to the challenges of collecting behavioral data and the relatively short time vaccines have been made available (for an exception, see Shiloh et al., 2021). A sole focus on intention is problematic because although intentions are an important theoretical antecedent of behavior, and are often closely associated with behavior in research on social cognition theories (e.g., Carpenter, 2010; McEachan et al., 2011; Milne et al., 2000), the relationship is far from perfect with modest effect sizes observed across studies (Sheeran & Webb, 2016). This intention-behavior ‘gap’ indicates that, for many individuals, intentions may be a necessary but not sufficient condition for behavioral enactment. Including a behavioral follow up in research applying social cognition constructs to predict COVID-19 preventive behaviors is, therefore, important as it not only permits measurement of the variance in behavior accounted for by the theory constructs, but also allows tests of the extent to which social cognition constructs of the theory are mediated by intention and the extent of intention-behavior gap. In addition, few studies applying social cognition theories have provided long-term follow-up of behavior for any COVID-19 preventive behavior. To date, no study in this context has predicted behavior beyond a few months. This is a substantive evidence gap considering the importance of identifying the determinants of sustained participation in preventive behaviors to minimizing infection rates in the long-term. As a consequence, there is an urgent need for studies that predict behavior over time, and test the extent to which social cognition theories are able to account for sustained engagement in COVID-19 behaviors. Such research should consider adoption of multiple measures of behavior at follow-up and over time periods that extend to a year or more rather than a few weeks. Such predictive studies would have important implications for the potential sustainability of messaging interventions that target change in these beliefs. 2.3.2 Specific beliefs A further limitation of research applying social cognition theories to COVID-19 behaviors is that they have tended not to account for specific beliefs relating for COVID-19. For example, few studies have investigated specific sets of behavioral, normative, and control beliefs purported to underpin social cognition constructs such as attitudes, subjective norms, and perceived behavioral control in the theory of planned behavior (Ajzen, 1991), or examined specific outcome expectancies or self-efficacy in the face of specific barriers in social cognitive theory (Bandura, 1986). Similarly, there is relatively little of research focusing on self-efficacy beliefs in the face of specific barriers or facilitating factors to the targeted COVID-19 preventive behavior, or specific expected outcomes with respect to performing the behavior, versions of the self-efficacy and outcome expectancy constructs identified as most potent in accounting for variance in future behavior in the original conceptualization of social cognitive theory (Bandura, 1986). This is problematic given that strategies to change behavior applying the theory necessitates forming persuasive communications that target specific beliefs with respect to the target behavior (Ajzen & Schmidt, 2020; Fishbein & Ajzen, 2010; Hamilton & Johnson, 2020). Resolution lies in performing the necessary belief elicitation research to develop expectancy-value indirect measures of constructs in the case of the theory of planned behavior (Ajzen, 1991), or identification of salient outcomes and barriers or facilitating factors in the case of social cognitive theory (see DuCharme & Brawley, 1995), and include them as unique predictors of COVID-19 preventive behaviors in predictive studies. Research aimed at identifying specific beliefs in research on COVID-19 preventive behaviors is being conducted, including formal elicitation of beliefs based on social cognition theory guidelines (e.g., Varol et al., 2021). In addition, researchers have included additional measures in predictive studies that encompass some specific beliefs. For example, recognition of the moral imperative for engaging in COVID-19 behaviors to protect others from infection, particularly those vulnerable to serious infection such as the elderly and immunocompromised, researchers have included moral norms as an additional predictor alongside other social cognition constructs in predictive studies. While subjective norms reflect perceived social pressure from significant others to perform a COVID-19 preventive behavior, moral norms reflect beliefs that performing the behavior is the socially responsible course of action. Results indicate an important role for this construct as an additional predictor of physical distancing intentions and behavior (e.g., Hagger, Smith, et al., 2020; Hagger, Smith, Keech, Moyers, & Hamilton, 2021) and general COVID-19 preventive behaviors (Kojan, Burbach, Ziefle, & Calero Valdez, 2021), intention to avoid COVID-19 (Raza, Ali, & Hussain, 2021), and COVID-19 vaccination intentions (Matute et al., 2021) and behavior (Shiloh et al., 2021). These studies have progressed knowledge on the social cognition correlates of COVID-19 preventive behaviors, and we look to future research that further elicits specific beliefs concerning COVID-19 prevention and examines their effects on preventive behaviors. 2.3.3 Study design and inferences A further limitation is an overreliance on studies adopting cross-sectional and correlational designs to test the predictive validity of these theories in the context of COVID-19 preventive behaviors. Such designs are limited because they do not model change in the theory constructs and in behavior, and preclude inference of causality (Weinstein, 2007). While proposed models specifying the proposed effects of a given social cognition theory may fit well with cross-sectional data measuring theory constructs, equally plausible models that specify a different pattern of effects may fit the data equally well, even if such models may be contraindicated theoretically. In addition, such tests do not rule out that the estimated relations among the constructs on a model could be accounted for by other, unmeasured variables (e.g., other social cognition constructs such as moral norms and anticipated regret, implicit attitudes and motives, dispositional and individual difference constructs). Including a follow-up measure of behavior so as to model past behavior effects on future behavior alongside theory constructs may provide a test of theory sufficiency (Ajzen, 1991; Chatzisarantis et al., 2004), but is not informative of these extraneous factors because past behavior is not a psychological construct (Hagger et al., 2018; Ouellette & Wood, 1998). As a consequence, predictive studies of this kind provide limited evidence as means to test theoretical predictions, and additional evidence is required to permit more elaborate inferences. The adoption of cross-lagged panel designs – longitudinal studies in which all constructs and outcomes are measured simultaneously across multiple time points – yield data that permit more elaborate inferences (e.g., Gollob & Reichardt, 1987; Liska, Felson, Chamlin, & Baccaglini, 1984). Such data allow researchers to control for a certain type of change in constructs over time, called covariance stability, and the extent to which relations among constructs at any given time points vary over time, known as stationarity. These models also enable examination of reciprocal effects among theory constructs, which permits inference of whether constructs predict outcomes, or whether the relationship occurs in the opposite direction, or in both directions over time, for example the effects of behavior on attitudes in addition to the effects of attitudes on behavior (Albarracín, 2021). Research in the context of COVID-19 preventive has demonstrated consistency of effects on social cognition constructs over time (e.g., Hagger et al., 2021). However more research is needed, which also needs to take the important step of controlling for effects of localized restrictions, which may affect individuals' beliefs. For example, individuals’ norms or outcome expectancies may vary depending on whether or not mitigation behaviors like wearing face coverings is required – they might feel confident and secure in wearing masks when it is mandated to do so, but might be embarrassed or afraid to do so when restrictions have been dropped. In addition, researchers should also heed calls to adopt optimal approaches to analyzing panel designs (Usami, 2021). In sum, longitudinal panel designs have advantages over cross-sectional or prospective designs by permitting inference on the direction of predicted effects in a theory, and the extent to which it accounts for temporal change in behavior or other outcomes over time. However, panel designs still do not permit inference of causal effects. An effective means to evaluate causality is to adopt experimental or randomized controlled intervention designs in which the effect of change in a social cognition theory construct as a result of a manipulation of intervention strategy (e.g., provision of information on the benefits or advantages of physical distancing using a persuasive communication to target attitude change) on change in a target behavioral outcome is evaluated (Imai, Keele, Tingley, & Yamamoto, 2011). In such designs, groups of individuals from the target population are randomized to receive the intervention, while groups of individuals randomized in a comparison or control group do not receive the intervention. Intervention effects are evaluated through observed differences in the behavior measured post-intervention across the two groups while simultaneously accounting for pre-intervention variation in behavior across the two groups. Such designs better allow for causal inferences, assuming that randomization was effective and the intervention strategy or method activated change in the targeted theory construct. There are relatively few studies that have adopted experimental and intervention research to test predictions of social cognition theories in the context of COVID-19 predictive behaviors, which is consistent with research applying social cognition theories more broadly, a trend likely attributable to the greater financial and time cost of research adopting these designs (Hagger, Cameron, et al., 2020). However, there are examples of studies demonstrating that intervention strategies targeting social cognition constructs (e.g., persuasive communication, information provision) have changed intentions toward, and actual participation in, hand hygiene behaviors and staying at home during lockdown (e.g., Capasso, Caso, & Conner, 2021; Okuhara et al., 2020; Smith et al., 2021). These studies provide preliminary evidence that persuasive communications targeting key constructs in social cognition theories lead to changes in preventive behaviors, and consistent with results of intervention studies observed for other health behaviors (Ajzen & Schmidt, 2020; Norman et al., 2018; Sheeran et al., 2016). However, the literature applying these designs is under developed, with few studies systematically demonstrating concomitant change in social cognition theory constructs and behavior, and or utilizing such designs to verify theoretical predictions, highlighting the need for further research in the area. 2.3.4 An exclusive focus on the individual A further critique leveled at social cognition theories concerns the almost exclusive focus on individuals' beliefs and how they impact decision making and subsequent behavior. By contrast, relatively little consideration is given to group influences, and the socio-structural and socio-environmental constructs that may also serve to line-up individuals' behavior. This is an important omission considering that individuals’ behavior is not merely a function of behavioral beliefs, but also responses to salient others and group influences (e.g., the pervasiveness of normative influences), and the social environment and context in which the behavior is performed. This is highly pertinent in the context of COVID-19 given that many preventive behaviors are dependent on extraneous factors, such as the presence of lockdown and mask-wearing mandates, availability of face coverings, or layout and available space for physical distancing in indoor areas. Although there have been attempts to account for the effects of these extraneous factors in tests of social cognition theories by specifying and measuring constructs that reflect individuals' belief-based responses to their social and physical environment, such as norms, these still reflect individuals' beliefs rather than specific group processes that may alter behavioral responses or the extent to which the context places actual constraints on behavior, both of which may not be sufficiently reflected in individuals’ beliefs (Albarracín, 2021). This is in contrast to other theoretical perspectives, such as ecological models of health behavior, which explicitly outline how behavior is a function of factors that operate at multiple levels including the individual, social, and environmental levels (Sallis, Owen, & Fisher, 2015; Salmon, Hesketh, Arundell, Downing, & Biddle, 2020). While such models have been widely applied in multiple health behavior contexts, there have been few applications in the context of COVID-19 preventive behaviors with considerable scope for their adoption in future research in this context (e.g., Latkin et al., 2021). In addition, there have been attempts to integrate constructs and predictions from social cognition theories with the broader factors derived from ecological models to provide more comprehensive explanations of health behavior and the process involved (e.g., Rhodes, Saelens, & Sauvage-Mar, 2018). We return to integrated approaches that incorporate socio-structural variables within social cognition theories and highlight their value in the context of COVID-19 prevention later in the article. 3 Integrated theories: developing comprehensive but parsimonious models There is general acknowledgment within the behavioral science community that social cognition theories have boundary conditions that delimit the extent to which their inferences apply. As a consequence, theorists and researchers have sought to augment or further develop the theories to expand their range of prediction and increase the comprehensiveness of the behavior explanations they offer (Hagger, 2009; Montaño & Kasprzyk, 2015). Foremost among these efforts has been the development of integrated theories that draw constructs and processes from multiple theories and models of human behavior and aim to address boundary conditions of unitary theories. Integrated theories offer two important advances on existing theory: They provide a means to highlight and eliminate redundancy in the constructs and processes used across theories toward the goal of identifying a core set of constructs that have optimal distinctiveness conceptually, and demonstrable discriminant and predictive validity empirically; and they enable the introduction of additional constructs and processes that assist in addressing the boundary conditions that delimit the explanations offered by existing theories (Hagger, 2009; Hagger & Hamilton, 2020). Integrated theories, therefore, provide the opportunity to expand the range of falsifiable predictions of existing social cognition theories while retaining optimal parsimony, two key features of a ‘good’ or ‘strong’ theory (Davis, Campbell, Hildon, Hobbs, & Michie, 2015). 3.1 Theory integration: a rationale A primary goal of theory integration is to promote theory parsimony by identifying a core set of constructs that are sufficiently unique, conceptually and empirically, that form the basis of theory predictions. This has value to those interested in changing behavior by assisting in identifying a core set of psychological constructs that optimally capture the mental processes central to decision making (McMillan & Conner, 2007), and can be activated or changed through intervention (Avishai, Brewer, Mendel, & Sheeran, 2021). However, considerable redundancy has been observed in the definition, conceptualization, operationalization, and measurement of the vast array of social cognition constructs across theories. This is not a new phenomenon, and it has been recognized as a perennial problem in psychology, often referred to as a ‘jangle’ fallacy: multiple constructs with similar content but differing labels (Block, 1995; Hagger, 2014; Kelley, 1927). Theorists wary of this fallacy have sought to develop means to address this redundancy by developing schemes to analyze the content of constructs across theories in order to identify commonalities and redundancy. In the context of social cognition theories applied in health contexts, researchers have developed similar schemes and have identified core constructs (e.g., McMillan & Conner, 2007; Protogerou, Johnson, & Hagger, 2018). For example, McMillan and Conner (2007) identified five core constructs across an analysis of multiple social cognition theories: dispositions to act (e.g., protection motivation, intentions), attitudes (e.g., outcome expectancies, affective attitudes), norms (e.g., subjective norms, social support), self-perceptions (e.g., self-esteem), and control (e.g., self-efficacy, perceived behavioral control). These methods are exemplary of an important process that is necessary when integrating theories to minimize redundancy in the constructs adopted and ensure that the constructs represented are likely unique and, therefore, are adequate in capturing the constructs likely to account for variance in behavior. The major contribution of theory integration to advancing knowledge on individuals' behavior, however, is the introduction of additional processes that address the boundary conditions of extant theories (Hagger, 2009; Hagger & Hamilton, 2020; Jacobs, Hagger, Streukens, De Bourdeaudhuij, & Claes, 2011). The integrated theories aim to provide more comprehensive behavioral explanations in terms of the mechanisms involved or account for more variance in behavior, and may also pave the way for theory-based interventions that have greater efficacy and efficiency in changing behavior or greater scope in changing behavior across contexts or populations. Of course, development of these integrated approaches needs to be tested against observation in studies adopting appropriate designs to test the predictions of the ‘new’ theory (Hagger, Gucciardi, & Chatzisarantis, 2017). Next, we outline two prominent examples of theory integration using social cognition theories applied to health behavior contexts: integration of dual-phase approaches and planning, and integration of dual-process models to account for non-conscious, automatic processes. We illustrate how these examples have yielded more comprehensive theories and better behavioral explanation. We also demonstrate how these integrations have demonstrated utility advancing the prediction of engagement COVID-19 preventive behaviors. 3.2 Augmenting theories to account for action phases Integrated theories have been instrumental in addressing the intention-behavior ‘gap’ observed in social cognition theories. The modest association between intention and behavior across social cognition theories applied in health contexts suggests individuals do not necessarily readily act on their intentions (Orbell & Sheeran, 1998; Sheeran & Webb, 2016). Researchers have addressed this problem by integrating processes from other theories on self-regulation (Leventhal, Meyer, & Nerenz, 1980), and, particularly, dual-phase models of action, such as the model of action phases (Heckhausen & Gollwitzer, 1987), into traditional social cognition theories like the theories of reasoned action and planned behavior (Orbell, Hodgkins, & Sheeran, 1997). Dual-phase models propose different phases of action, and, in particular, distinguish between an intentional or motivational phase in which intentions to perform a behavior to attain an outcome are formed, and an implemental or volitional phase in which intentions are augmented with specific action plans in order to enact the behavior. It is predicted that the extent to which individuals form plans, particularly plans that link the behavior with salient environmental cues that line up the behavior, enhances intention enactment and strengthens the intention-behavior relationship. Some integrated theories, such as the health action process approach (HAPA; Schwarzer, 2008), include multiple phases by design and encompass different types of planning as a conduit between intentions and behavior (for comprehensive descriptions of the HAPA see Schwarzer & Hamilton, 2020; Zhang, Zhang, Schwarzer, & Hagger, 2019). Tests of social cognition theories that integrate predictions from dual-phase models have demonstrated that planning constructs account for significant variance in intentions (e.g., Zhang et al., 2019), and moderate the intention-behavior association upward (e.g., de Bruijn, Rhodes, & van Osch, 2012). In addition, intervention and experimental studies have shown that individuals prompted to form action plans are more likely to follow-through on their intentions (e.g., Armitage, 2004; Hagger et al., 2012; Orbell et al., 1997). Research illustrates that such planning interventions work by promoting greater recall of the intended action (Orbell et al., 1997), and more efficient or ‘automatic’ behavioral enactment (Martiny-Huenger, Martiny, Parks-Stamm, Pfeiffer, & Gollwitzer, 2017) – key mechanisms by which plan formation leads to behavior change. Broad support for these planning interventions has been demonstrated in research syntheses (e.g., Gollwitzer & Sheeran, 2006), illustrating the applied value of integrating a dual-phase approach within social cognition theories in the context of behavior change. Dual-phase approaches have demonstrable applicability in the context of COVID-19 preventive behaviors (Harvey, Armstrong, Callaway, Gumport, & Gasperetti, 2021). Specifically, research adopting the HAPA has demonstrated associations between intention and planning constructs in studies targeting general COVID-19 preventive behaviors (Lin et al., 2020), and physical distancing (Beeckman et al., 2020; Hamilton, Smith, et al., 2020), hand washing (Lao, Li, Zhao, Gou, & Zhou, 2021; Luszczynska et al., 2021), and wearing face coverings (Lao et al., 2021) behaviors. Furthermore, direct effects of planning constructs on behavior were found in some of the studies (Beeckman et al., 2020; Lao et al., 2021; Lin et al., 2020; Luszczynska et al., 2021), and that they mediated intention effects on behavior (Beeckman et al., 2020; Lin et al., 2020; Luszczynska et al., 2021), implicating planning in the process by which intentions are enacted. A specific form of planning, known as implementation intentions or “if-then” plans (Gollwitzer, 1999), has also been proposed as moderator the intention-behavior relationship, suggesting that individuals who plan may be more likely to follow through on their intentions. Studies have indicated that compliance with interventions using if-then plans leads to greater adherence to physical distancing guidelines (Ahn et al., 2021; Bieleke, Martarelli, & Wolff, 2021). These findings illustrate that integration of planning from dual-phase models in social cognition theories broadens their capacity to explain intention-behavior relations and inform interventions to promote COVID-19 preventive behaviors. However, currently available research applying these integrated approaches is relatively sparse and confined to only a few COVID-19 preventive behaviors. In addition, there are very few studies adopting experimental and randomized-controlled designs examining effects of planning strategies on COVID-19 preventive behaviors. There are also a number of outstanding questions that need to be addressed, such as the specific mechanism by which planning leads to behavioral enactment – a mediating process in which planning forms part of the decision-making process, as specified in the HAPA (Schwarzer, 2008), or a moderating process in which planning promotes intention enactment, an explanation consistent with the model of action phases (Heckhausen & Gollwitzer, 1987). There is also little research examinng some of the potential mediators of planning on these behaviors such as recall of the intended behavior and greater accessibility of the behavior when the cue or codnition stated in the plan is presented. We therefore call for studies that test planning effects in a broader range of COVID-19 preventive behaviors, adopt intervention or experimental designs, and mediation and moderation effects. 3.3 Augmenting theories to account for automatic processes A fundamental assumption of social cognition theories is that actions are a function of reasoned, rational decision making based on available information (Conner and Norman, 2015a, Conner and Norman, 2015b; Fishbein et al., 2001). An alternative, but not incongruous perspective, is offered by theories of implicit cognition, which outline how stored knowledge structures that link behaviors, contexts, and evaluations lead to behavioral initiation and enactment beyond an individual's awareness (Hagger, 2016; Sheeran, Gollwitzer, & Bargh, 2013). These implicit or automatic processes tend to control many everyday behaviors for which elaborate, reasoned decision making is both unnecessary and inefficient, and particularly mundane behaviors with which individuals have copious prior experience (Gardner, Lally, & Wardle, 2012; Wood, Quinn, & Kashy, 2002). According to these approaches, individuals learn over time to associate behavioral responses with concomitantly-experienced information such as cues or initiating events or evaluations in memory in organized knowledge structures or schema. Subsequent presentation of a triggering event or evaluation linked to the behavioral response leads to rapid, efficient activation of the behavior. Accordingly, dual-process theories of social cognition have been proposed aimed at providing more comprehensive explanations of behavior and to evaluate the extent to which target behaviors are determined by reasoned and automatic processes, and the conditions that might determine when each process predominates (Strack & Deutsch, 2004; Wood, Labrecque, Lin, & Rünger, 2014). In predictive tests of dual process theories, reasoned processes are represented through the effects of constructs traditionally specified in social cognition theories such as attitudes, norms, risk perceptions, and self-efficacy (e.g., Bandura, 1986; Fishbein & Ajzen, 2010). In contrast, automatic processes tend to be inferred from constructs representing different types of implicit cognition such as implicit attitudes and motives and measures of habit, (e.g., Hagger, Trost, Keech, Chan, & Hamilton, 2017; Hamilton, Gibbs, Keech, & Hagger, 2020), or affective desires or behavioral prepotency that reflect impulse-related tendencies, internal drive states, and cue salience (e.g., Gibbons, Gerrard, Blanton, & Russell, 1998; Hall & Fong, 2007). Incorporating these types of construct in predictive studies of social cognition theories enables researchers to estimate the relative contribution of each to explaining variance in a target behavior. For example, given that implicit cognition is assumed to bypass the reasoned processes that lead to behavior, effects of constructs representing implicit processes are expected to relate to behavior directly, unmediated by intentions. Studies adopting this integrated approach have demonstrated unique effects of implicit attitudes and motives on behavior independent of intentions and other social cognition constructs (e.g., Hamilton, Gibbs, et al., 2020; Keatley, Clarke, & Hagger, 2012). A similar pattern of effects has been found for measures of habit, a specific form of automaticity that is cue- or context- dependent but unrelated to goal pursuit (Brown, Hagger, & Hamilton, 2020; Hamilton, Kirkpatrick, Rebar, & Hagger, 2017; Verplanken & Orbell, 2003). In addition, there is research that has illustrated that certain types of behavior, such as those that are less complex, have high propensity to be formed as habits, or demand less deliberative or reasoned consideration, are more likely to be impacted by implicit cognition and habit (Verplanken, 2006). In addition, effects of traditional social cognition constructs have smaller effects on behavior when individuals report higher levels of implicit attitudes or habit. For example, habits and implicit attitudes have been shown to moderate the intention-behavior relationship (Divine, Berry, Rodgers, & Hall, 2021; Gardner, Lally, & Rebar, 2020). Taken together, predictive studies applying integrated approaches that incorporate constructs that represent automatic processes have provided useful information on the contribution of non-conscious processes that lead to behavior and the moderating conditions. Predictive studies examining dual process effects are expected to provide important insight into potentially efficacious intervention strategies. In behavioral contexts where implicit cognition or habit account for non-trivial variance in an undesired behavior, interventionists can select strategies that block or counter cues and initiating events, or promote skills that assist in overriding or managing the behavioral response (Duckworth, Gendler, & Gross, 2016; Gardner, Rebar, & Lally, 2020). These include strategies such as environmental or context restructuring, which aim to limit or dampen the salience of the cues (e.g., altering the layout of a grocery store so as to reduce capacity or density of checkout queues or lines), or cue identification, monitoring, and management that aim to provide individuals with an awareness of the conditions that lead to the unwanted behavior (e.g., signage or notices illustrating how viruses spread when an unmasked individual coughs, floor markers showing recommended social distance, provision of hand sanitizer at points of entry to venues), capacity to identify when the cues occur (e.g., facilitating mental imagery of a typical trip to a grocery store and the frequency of contact with people on the way), and knowledge of an appropriate alternative response (e.g., prompting practice with booking a rapid antigen test in the event of exposure). In contexts where implicit cognition or habit explain substantive variance in a desired behavior, strategies that enhance habits or cue-dependent responding are appropriate. These might include habit-forming strategies such as repetition of the behavior in stable contexts and providing reinforcement for successful behavioral performance. In the context of COVID-19 preventive behaviors, studies integrating constructs from dual-process models have provided preliminary evidence to indicate the relevance of automatic processes in predicting behavior and elucidating the mechanisms involved. For example, research adopting prospective and longitudinal designs has indicated that habits account for unique variance in physical distancing behavior in two national samples, effects which hold when accounting for past behavior (Hagger, Smith, et al., 2020; Hagger et al., 2021). Importantly, these results also indicate, unsurpisingly, that there is substantive stability in habits across three time points, and consistency in their effects on behavior unmediated by intentions, congruent with proposals of dual-process theories. However, it must be stressed that effect sizes for the habit effects in these samples and for this behavior were relatively modest, particularly relative to effects of intentions and the other social cognition constructs. To speculate, it may be that physical distancing requirements vary from context to context, which means the cues to this behavior may not be sufficiently consistent, necessitating reasoned consideration on the part of the actor when making decisions to act in future. It is also a relatively ‘new’ behavior for many, so there may have been less opportunity for many to form habits, especially given time to form a habit varies considerably across individuals (Lally, van Jaarsveld, Potts, & Wardle, 2010). Research is needed to explore potential moderators of habit effects in research on physical distancing in the context of COVID-19, such as the extent to which individuals have had the opportunity to practice physical distancing in the same context. More broadly, research is needed to corroborate habit effects for other COVID-19 preventive behaviors. 4 Mechanisms of action While researchers have long recognized the value of theory in informing behavior change interventions, many interventions do not have a clear basis in theory despite claims to the contrary by intervention developers (e.g., Michie et al., 2018). A major issue for these interventions is the lack of systematic mapping of theoretical principles on to the content and components of the intervention resulting in ‘theory inspired’ rather than ‘theory based’ interventions (Michie et al., 2018). Advances in the scientific study of behavior change has sought to formalize the processes of development and subsequent description of the theoretical constructs targeted in behavioral interventions, the strategies or techniques that form the content of the intervention and how it is delivered (Kok et al., 2016; Michie et al., 2013), and the links between them that specify how the intervention ‘works’, that is, the process or mechanism by which the intervention acts to change behavior (Rothman et al., 2020; Sheeran et al., 2017). To this end, researchers have used expert consensus and reviews of the extant literature on behavioral interventions to develop organized characterizations or taxonomies of behavior change techniques, and specified links between techniques from taxonomies and the psychological constructs, many originating from social cognition theories, they are purported to change or activate in order to change behavior, known as mechanisms of action (Avishai et al., 2021; Hagger, Moyers, et al., 2020; Rothman et al., 2020; Sheeran et al., 2017). Research is beginning to provide indication of associations between the identified behavior change techniques and theory-based constructs implicated in their mechanisms of action through expert consensus and reviews of stated associations in the extant literature (Carey et al., 2019; Connell et al., 2019). There is also research that has identified how theory-based techniques targeting change in social cognition constructs change intention and behavior (Knittle et al., 2018; Sheeran et al., 2016). Together this work has advanced knowledge by providing a common, formal nomenclature to describe the techniques that form the content of behavioral interventions, and enabled researchers to better specify how interventions work in changing behavior through change or activation of the theory-based constructs involved. Importantly, it has highlighted the value of developing interventions based on theory, the importance of formal specification of the mechanism of action involved, and facilitated greater clarity and precision in intervention description that enables more effective research syntheses and conceptual replications of interventions. Providing strong evidence to support the mechanisms of action of behavioral interventions necessitates randomized controlled trials or experimental research designs that demonstrate the effect of behavior change techniques that form the content of the interventions on change in the target theoretical constructs and concomitant change in behavior (Hagger, Moyers, et al., 2020; Rothman et al., 2020; Sheeran et al., 2017). Verification is derived from testing the mediation of the effect of the intervention on behavior change by change in the theoretical construct. This test of a mechanism of action is illustrated in Fig. 1 (Hagger, 2019). The effect of the intervention on change in the theoretical construct (path a, Fig. 1) and the effect of change in the construct on behavior change (path b, Fig. 1) constitutes the indirect or mediated effect of the intervention on behavior that should fully or, at least, partially account for the direct effect (path c, Fig. 1) on behavior. From an analytic perspective, the mediated effect is tested, the residual effect of the intervention on behavior (path c’, Fig. 1) should be attenuated to zero for full or complete mediation, or to a significantly smaller value for partial mediation. Such tests provide clear evidence in support of the proposed mechanism of action of a behavior change intervention. It is important to note that the change in the mediator should occur in advance of the changes in the behavior, otherwise the mediator may just signal or be broadly indicative of the mechanism (Kazdin, 2007). Testing for the effects of extraneous moderators, such as interpersonal or intrapersonal characteristics, or contextual or environmental variables, of the mediated effect is also indicated by these tests, and would elucidate the extent to which the mechanism accounts for behavior change across such contexts. It is also important that such tests utilize factorial designs that enable isolation of the effects of specific techniques and their mechanisms, which would resolve a perennial problem in cases where interventions use multiple techniques in a single intervention.Fig. 1 Diagram of a behavior change mechanism of action (Hagger, M. S. (2019). Basic model of a behavior change mechanism of action. PsyArXiv. https://doi.org/10.31234/osf.io/9a5k6). Fig. 1 A proliferating body of research has conducted formal tests of the mechanism of action of behavioral interventions adopting theory-based behavior change techniques to change behavior. For example, studies have demonstrated that interventions adopting randomized controlled designs and applying techniques targeting change in attitudes (Chatzisarantis & Hagger, 2005), social support (Quaresma, Palmeira, Martins, Minderico, & Sardinha, 2014), and self-efficacy (Larsen et al., 2021) changed physical activity behavior in the target populations through the mediation of change in the measures of the targeted constructs. However, such tests are not routinely conducted – a recent set of meta-reviews indicated that evidence supporting mechanisms of action using mediation tests is sparse and has not advanced significantly over the years (Suls et al., 2020). Meta-analytic research syntheses of behavioral intervention studies have offered some insight demonstrating that interventions that change constructs from social cognition theories, such as attitudes, subjective norms, and self-efficacy, also result in concomitant change in behavior (Sheeran et al., 2021; Webb & Sheeran, 2006). However, the change observed using these meta-analytic designs still do not provide definitive evidence for mediation. Alternative approaches to research synthesis have formally tested mediation demonstrating indirect intervention effects on behavior through theory-based constructs across studies (Rhodes, Boudreau, Weman Josefsson, & Ivarsson, 2020; Sheeran et al., 2020). However, these syntheses are still sub-optimal means to test the mechanisms as the synthesized data did not provide evidence for the effect of change in the theoretical constructs on change in behavior, both as a result of the intervention, so that association may still be affected by extraneous variables. Specifying the theoretical basis for behavioral interventions and testing their mechanisms of action in the context of COVID-19 preventive behaviors holds considerable promise in advancing knowledge on how interventions work and their breadth of application. Such an approach together with judicious testing of contextual attributes such as location and target population may provide important evidence of the contexts in which interventions targeting social cognition constructs, such as attitudes and risk perceptions, are likely to have optimal efficacy. However, although a small number of randomized controlled designed interventions adopting techniques aimed at changing behavior through changes in social cognition theory constructs have been reported (e.g., Ahn et al., 2021; Keller et al., 2021; Okuhara et al., 2020; Smith et al., 2021), none have tested mechanisms of action using the proposed mediated effects. Given the paucity of this research we recommend the need for interventions based on social cognition theories that test mechanisms of action. Such interventions should using optimal methods including randomized controlled designs and sufficient measures to model behavior change through putative mediators of intervention techniques. These interventions should also be pre-registered consistent with open science guidelines (Hagger, 2022). 5 Future directions In previous sections we outlined the value of application of social cognition theories in the prediction of COVID-19 preventive behaviors, and their utility in informing behavioral interventions that are optimally efficacious in changing these behaviors. We provided an overview of the current state of the research literature applying these theories in this contexts, and, along the way, highlighted some important avenues for future research, including the need for predictive studies adopting cross-lagged panel and experimental designs to better infer directionality and causality in theory effects; the need for more research testing the prediction of social cognition theories that integrate additional salient processes from other theories to provide comprehensive, optimally parsimonious predictions of these behaviors; and the need for more theory-based interventions testing effects of specific behavior change techniques with formal mediation analyses to test their mechanisms of action. Next, we outline additional research directions that will make important contributions to knowledge on the determinants of preventive behaviors as the global COVID-19 pandemic evolves, and will provide important formative knowledge to inform ongoing intervention and messaging efforts toward the pandemic ‘endgame’ and a shift toward COVID-19 as an endemic illness with threat levels commensurate with other endemic viruses like influenza and the common cold. 5.1 Moving from pandemic to endemic management of COVID-19 There is a need for data on application of social cognition theories to predict preventive behaviors under conditions of endemic management of COVID-19. The theories may provide insight into the determinants of preventive behaviors in isolated outbreaks, short-term ‘circuit breaker’ lockdown measures, and ongoing ‘booster’ vaccination, measures that are likely needed in the foreseeable future to manage the infections. Unsurprisingly, research to date has focused exclusively on preventive behaviors under global pandemic conditions, and ongoing waves of infection including breakthrough infections among the vaccinated population suggest that the pandemic is currently far from over. Continuing research focusing on social cognition determinants of preventive behaviors under current conditions, therefore, has immediate value and is essential to ongoing management. However, preparation for long-term management of COVID-19 infections in the transition from pandemic to endemic conditions needs an evidence base. This can be provided through studies in which hypothetical future COVID-19 conditions are proposed (e.g., the advent of a localized outbreak) followed by measures of social cognition beliefs and intentions to engage in preventive behaviors as a response. Such research may inform messaging interventions relevant to ongoing infection management as pandemic-level infection rates subside and are replaced by isolated outbreaks and seasonal waves. With the recognition that the immunity afforded by COVID-19 vaccines wanes over time (Naaber et al., 2021), governmental health agencies have approved and recommended the administration of an additional vaccine dose to boost immunity (CDC, 2021a; JVCI, 2021). These ‘booster’ vaccines are likely to become an ongoing requirement to maintain immunity and minimize infection transmission, particularly to those who are most vulnerable to serious bouts of the illness (Krause et al., 2021). Social cognition theories may offer insight into the determinants likely associated with the uptake of these ‘booster’ vaccine doses, with a view to informing public health messages aimed at promoting adherence to booster vaccine recommendations. The theories may contribute to identification of the beliefs salient to booster vaccine intentions and behavior, such as apathy, fatigue, and decreased perceptions of vulnerability and severity, which may lead the previously vaccinated to fail to get a booster vaccine (Hagger & Hamilton, 2022). Identification of these beliefs may signal potential strategies to counter such perceptions. 5.2 Social cognition, health disparities, and COVID-19 From the onsetof the pandemic, studies have demonstrated considerable disparities in COVID-19 infection rates and outcomes related to the illness including serious cases, ‘long COVID’, and mortality rates in minority groups, particularly racial and ethnic groups that have historically been underserved and those from low incomes and educational backgrounds (CDC, 2020). There is also evidence to suggest disparities in COVID-19 preventive behaviors in these groups, particularly for vaccination (Ndugga, Hill, Artiga, & Haldar, 2021). Recent research has suggested that such behavioral disparities in health contexts may be manifested in the social cognition constructs that line up these behaviors. For example, studies have indicated that constructs from social cognition theories mediate the associations between socio-structural variables that indicate health disparities (e.g., race and ethnicity, income, education) on health behavior participation (Hagger & Hamilton, 2021; Orbell, Szczepura, Weller, Gumber, & Hagger, 2017). These data suggest that individuals from underserved minority groups, on low incomes, or with lower levels of education may be less likely to view illnesses and other risky conditions as threatening, and report lower self-efficacy with respect to health behaviors, which manifests in lower participation in health behaviors. Furthermore, these socio-structural variables also moderate relations among constructs and behavior in social cognition theories (Schüz, Brick, Wilding, & Conner, 2020). For example, individuals from low income and education backgrounds are less likely to act on their intentions, which may be the result of lower expectation of control over their behavior due to experiences of healthcare inaccessibility or disenfranchisement from healthcare services. Recent research has corroborated these moderating effects in eight COVID-19 preventive behaviors including physical distancing, restricting time outside the home, and wearing facemasks (Schüz et al., 2021). Larger intention-COVID-19 preventive behavior relations were observed in less socio-economically deprived groups. The research provides preliminary information on the mechanisms to which observed disparities in preventive behaviors could be attributed. It may also signal potential targets for interventions that may be effective in promoting behavior change in these groups when the source of disparities are difficult to modify or change in the short term. 5.3 Traits in social cognition theories While there is emerging research examining effects of traits or trait-like constructs such as personality and other intrapersonal individual difference constructs on COVID-19 preventive behaviors (e.g., Nofal, Cacciotti, & Lee, 2020), there is little research that has explored the effects of these dispositional constructs in the context of social cognition theories. Incorporating trait-like constructs into social cognition theories may indicate mechanistic explanations for how these dispositions relate to behavior. A central premise of many social cognition theories is that trait-like constructs serve as sources of information in the decision-making process and inform individuals’ beliefs with respect to performing a target behavior in future (Ajzen, 1991; Bandura, 1986). Empirically, therefore, social cognition constructs should act as mediators of effects of dispositions on behavior. Accordingly, verifying that such constructs serve as distal correlates of behavior, will provide potential information on the constructs that should be targeted to change behavior, particularly among groups with specific traits. Tests of these proposed mediation effects in health contexts abound. For example, researchers have demonstrated that dispositional constructs that reflect better capacity to pursue goal-directed behaviors, such as the conscientiousness personality trait and trait self-control, are associated with greater participation in health behavior mediated by social cognition constructs such as attitudes and self-efficacy (e.g., Bogg, 2008; Conner & Abraham, 2001; Hagger et al., 2019). The mediation of traits on behavior in social cognition theories has been tested in the context of COVID-19 preventive behaviors. As an illustrative example, personality traits openness to experience and agreeableness personality traits on adherence to COVID-19 prevention guidelines were mediated by attitudes, and both attitudes and perceived norms, respectively (Bogg & Milad, 2020). A finding that suggests that individuals with tendencies to be more socially agreeable and open-minded are more likely to hold beliefs in the utility of COVID-19 preventive behaviors, and view their social environments as supportive, which may be implicated in their decisions to engage in those behaviors. Analogously, highly politicized context of COVID-19 prevention, political orientation, a summary of individuals' political beliefs as liberal or conservative, is a trait-like construct that would be expected to affect individuals' decision making with respect to COVID-19 preventive behaviors. For example, studies have indicated that political orientation has been shown to be associated with COVID-19 vaccine intentions, with conservatives less likely to intend to perform behaviors such as mask wearing and getting vaccinated (e.g., Hamilton & Hagger, 2022; Huynh, Zsila, & Martinez-Berman, 2022). Further, effects of political orientation on booster vaccine intentions has been shown to be fully mediated by social cognition constructs (attitudes, subjective norms, perceived behavioral control, risk perceptions), with more conservative beliefs related to lower intentions (Hagger & Hamilton, 2022). These findings highlight the importance of considering political beliefs as a trait that informs individuals’ estimates of their intentions in this context. Incorporating measures of political orientation in social cognition theories may, therefore, assist in explaining the mechanisms underpinning these effects. A further, seldom-investigated role for traits or trait-like constructs in social cognition theories is their role in moderating the effects of social cognition constructs on intention and behaviors. Traits may magnify or diminish effects of key social cognition constructs on intention and behavior, and may, therefore, be directly implicated in the decision-making process. For example, research indicates that conscientiousness and extraversion personality traits moderate the intention-behavior relationship in the domain of physical activity, with larger intention-behavior relations observed at higher levels of these traits (Rhodes, Courneya, & Hayduk, 2002). This finding indicates that individuals with greater work ethic and those who have greater tendency to explore new opportunities tend to be more effective in acting on their exercise intentions. Similarly, research demonstrates that conscientiousness moderates the effect of affective attitudes on intention, such that the affective attitude-intention relationship was smaller among those reporting high conscientiousness (Rhodes et al., 2002). This suggests that individuals scoring higher on conscientiousness are less likely to base their intentions on a consideration of affective, impulse-related outcomes (e.g., “exercise makes me feel good) and more likely to base them on cognitive, utility-based outcomes (e.g., “exercise will make me fitter”). Numerous traits have been explored as determinants of COVID-19 preventive behaviors, including personality constructs (Zettler et al., 2022) and need for cognition (Xu & Cheng, 2021), but relatively few as moderators of intention-behavior relations or effects of social cognition constructs on intentions. As in the mediation analyses reported previously, political orientation would be expected to impact individuals’ decision making in the context of COVID-19 preventive behaviors. As an illustration, political orientation has been shown to moderate effects of perceived determining factors like catastrophic potential, dread, and moral nature on risk perceptions, a social cognition construct (Ju & You, 2022). However, there has been few investigations examining political orientation as a moderator within social cognition models, particularly its role as a moderator of the intention-behavior relationship. To speculate, individuals with more conservative political beliefs may not only be less likely to form intentions to perform behaviors viewed as overly-restrictive, or associated with government overreach, such as face mask wearing, but may also be less likely to act on their intentions because their beliefs may place limits on their desire to act non-normatively. As such, political orientation would be expected to moderate the intention-behavior relationship. We look to future research to explore such effects. Taken together, knowledge that traits and enduring beliefs such as personality and political orientation are both associated with individuals’ specific beliefs regarding performing COVID-19 preventive behaviors, and may moderate their effects on behavior, highlights the need for tailored interventions, or dedicated messages, aimed at countering or challenging those behavioral beliefs. Future research should seek to verify these predictions and test the efficacy of messaging aimed at promoting COVID-19 preventive behaviors in groups defined by specific dispositional characteristics. 5.4 Affective processes Incorporation of constructs that represent affective processes in social cognition theories has made an important contribution to explaining variance in health behavior. For example, numerous studies have incorporated constructs representing anticipated affect, including the affective component of attitude and constructs such as anticipated regret and negative emotion, into predictive tests of these theories in health behavior contexts (e.g., Rivis, Sheeran, & Armitage, 2009). Generally, anticipated affective responses have been shown to predict behavior via the mediation of intentions, but also directly, suggesting that affective considerations not only inform decision making, but may also lead to impulsive or automatic behavioral engagement (e.g., Conner, McEachan, Taylor, O'Hara, & Lawton, 2015). Exploring effects of these constructs may also provide salient information on the determinants of COVID-19 preventive behaviors. For example, anticipated affect constructs such as regret have been shown to be related to physical distancing behavior, but effects are small and tend to be usurped by other constructs that have greater salience, such as moral norms (Hagger, Smith, et al., 2020). Similarly, fear and worry of COVID-19 infection has also been shown to predict intentions to perform COVID-19 preventive behaviors (Coifman et al., 2021), and have been shown to be mediated by constructs from the theory of planned behavior (Yahaghi et al., 2021). In such cases, fear perceptions serve as an information source in decision making toward these preventive behaviors, a process consistent with the premise from social cognition theories that threat perceptions are likely to compel individuals to select a course of behavior to manage the threat. These findings notwithstanding, there is a relative dearth of research examining affective processes in the context of social cognition theories in COVID-19. Such approaches have much potential to inform knowledge given that affective beliefs are likely to be highly salient to making decisions for some COVID-19 preventive behaviors, such as vaccination (Chou & Budenz, 2020). Similarly, researchers should also consider examining the moderating conditions that determine whether affective beliefs, such as affective attitudes, are likely to be most salient when predicting intentions to perform these behaviors alongside utilitarian beliefs, such as cognitive attitudes. Such research may provide further insight into the relative contribution of affective processes in decision making, and a basis for the development of targeted messages in behavior change interventions that promote engagement in preventive COVID-19 behaviors. 6 Conclusion Minimizing the transmission of COVID-19 infections requires widespread adherence to COVID-19 preventive behaviors that include mitigation measures (e.g., physical distancing, wearing face coverings, avoiding gathering in large groups, engaging in sanitization behaviors, participating in rapid testing) and immunization through vaccination. Adherence, however, is sub-optimal for many of these behaviors, necessitating health authorities to intervene to promote uptake and maintenance of these behaviors going forward. Such behavior change interventions should be informed by theory and evidence derived from behavioral science. Social cognition theories have made substantive contributions to the identification of the belief-based determinants of health behavior (e.g., attitudes, norms, self-efficacy, risk perceptions). In addition, research based on these theories has led to the specification of links between theoretical constructs and behavior change techniques purported to change them, known as mechanisms of action. Application of social cognition theories in the context of COVID-19 has identified the salient predictors of multiple preventive behaviors. Furthermore, integrated approaches have also demonstrated the salience of additional constructs, like moral norms and anticipated affect, and processes, such as planning and habits, in accounting for unique variance in these behaviors. Importantly, these data have begun to inform behavior change interventions adopting techniques targeting the social cognition constructs shown to be related to COVID-19 preventive behaviors. However, few interventions in this context have a strong basis in theory, and none have tested theory-based mechanisms of action. Considering these evidence gaps in research on COVID-19 preventive behaviors, future studies should test social cognition theory predictions using cross-lagged panel, randomized controlled, and experimental designs that better enable inference of directionality and causality. There is also the need to apply these theories to predict emergent preventive behaviors such as booster vaccination, and to the management of COVID-19 as an endemic illness. We also call for research that incorporates socio-structural variables, individual difference constructs, and constructs representing affective processes into existing theories to address key questions on the determinants of COVID-19 preventive behaviors. These recommendations will broaden the contribution that applying social cognition theories makes to an evidence base to inform efficacious and efficient behavior change interventions in the context of COVID-19 prevention. CRediT authorship contribution statement Martin S. Hagger: Conceptualization, Writing – original draft, Writing – review & editing. Kyra Hamilton: Conceptualization, Writing – original draft, Writing – review & editing. Declarations of competing interest The authors declare no conflicts of interest with the work conducted in this study. All views expressed are solely those of the authors. The other authors have no financial disclosures. ==== Refs References Adiyoso W. Wilopo Social distancing intentions to reduce the spread of COVID-19: The extended theory of planned behavior BMC Public Health 21 1 2021 1836 10.1186/s12889-021-11884-5 34635071 Ahn J.N. Hu D. Vega M. Changing pace: Using implementation intentions to enhance social distancing behavior during the COVID-19 pandemic Journal of Experimental Psychology: Applied 2021 10.1037/xap0000385 No Pagination Specified-No Pagination Specified Ajzen I. The theory of planned behavior Organizational Behavior and Human Decision Processes 50 2 1991 179 211 10.1016/0749-5978(91)90020-T Ajzen I. Fishbein M. Understanding attitudes and predicting social behavior 1980 Prentice Hall Ajzen I. Schmidt P. Changing behavior using the theory of planned behavior Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 17 31 10.1017/97811086773180.002 Albarracín D. Action and inaction in a social world: Predicting and changing attitudes and behaviors 2021 Cambridge University Press Albarracín D. Johnson B.T. Fishbein M. Muellerleile P.A. Theories of reasoned action and planned behavior as models of condom use: A meta-analysis Psychological Bulletin 127 1 2001 142 161 10.1037/0033-2909.127.1.142 11271752 Albarracín D. Jung H. A research agenda for the post-COVID-19 world: Theory and research in social psychology Asian Journal of Social Psychology 24 1 2021 10 17 10.1111/ajsp.12469 33821136 Armitage C.J. Evidence that implementation intentions reduce dietary fat intake: A randomized trial Health Psychology 23 2004 319 323 15099174 Avishai A. Brewer N.T. Mendel J.R. Sheeran P. Expanding the analysis of mechanisms of action in behavioral interventions: Cognitive change versus cognitive activation Psychology and Health 2021 1 20 10.1080/08870446.2021.1969021 Bandura A. Social foundations of thought and action: A social-cognitive theory 1986 Prentice-Hall Bar-On Y.M. Goldberg Y. Mandel M. Bodenheimer O. Freedman L. Kalkstein N. Protection of BNT162b2 vaccine booster against Covid-19 in Israel New England Journal of Medicine 385 15 2021 1393 1400 10.1056/NEJMoa2114255 34525275 Barile J.P. Guerin R.J. Fisher K.A. Tian L.H. Okun A.H. Vanden Esschert K.L. Theory-based behavioral predictors of self-reported use of face coverings in public settings during the COVID-19 pandemic in the United States Annals of Behavioral Medicine 2020 10.1093/abm/kaaa109 Bartsch S.M. Ferguson M.C. McKinnell J.A. O'Shea K.J. Wedlock P.T. Siegmund S.S. The potential health care costs and resource use associated with COVID-19 in the United States Health Affairs 39 6 2020 927 935 10.1377/hlthaff.2020.00426 32324428 Beeckman M. De Paepe A. Van Alboom M. Maes S. Wauters A. Baert F. Adherence to the physical distancing measures during the COVID-19 pandemic: A HAPA-based perspective Applied Psychology: Health and Well-Being 12 4 2020 1224 1243 10.1111/aphw.12242 33052008 Bergwerk M. Gonen T. Lustig Y. Amit S. Lipsitch M. Cohen C. Covid-19 breakthrough infections in vaccinated health care workers New England Journal of Medicine 385 16 2021 1474 1484 10.1056/NEJMoa2109072 34320281 Betsch C. Korn L. Burgard T. Gaissmaier W. Felgendreff L. Eitze S. The four weeks before lockdown during the COVID-19 pandemic in Germany: A weekly serial cross-sectional survey on risk perceptions, knowledge, public trust and behaviour, 3 to 25 March 2020 Euro Surveillance 26 42 2021 2001900 10.2807/1560-7917.ES.2021.26.42.2001900 Bieleke M. Martarelli C.S. Wolff W. If-then planning, self-control, and boredom as predictors of adherence to social distancing guidelines: Evidence from a two-wave longitudinal study with a behavioral intervention Current Psychology 2021 10.1007/s12144-021-02106-7 Block J. A contrarian view of the five-factor approach to personality description Psychological Bulletin 117 1995 187 215 10.1037/0033-2909.117.2.187 7724687 Bogg T. Conscientiousness, the transtheoretical model of change, and exercise: A neo-socioanalytic integration of trait and social-cognitive frameworks in the prediction of behavior Journal of Personality 76 4 2008 775 802 10.1111/j.1467-6494.2008.00504.x 18482356 Bogg T. Milad E. Demographic, personality, and social cognition correlates of coronavirus guideline adherence in a U.S. sample Health Psychology 39 12 2020 1026 1036 10.1037/hea0000891 33252928 Bonell C. Michie S. Reicher S. West R. Bear L. Yardley L. Harnessing behavioural science in public health campaigns to maintain ‘social distancing’ in response to the COVID-19 pandemic: Key principles Journal of Epidemiology & Community Health 74 2020 617 619 10.1136/jech-2020-214290 32385125 Brown D.J. Hagger M.S. Hamilton K. The mediating role of constructs representing reasoned-action and automatic processes on the past behavior-future behavior relationship Social Science & Medicine 258 2020 113085 10.1016/j.socscimed.2020.113085 de Bruijn G.-J. Rhodes R.E. van Osch L. Does action planning moderate the intention-habit interaction in the exercise domain? A three-way interaction analysis investigation Journal of Behavioral Medicine 35 5 2012 509 519 10.1007/s10865-011-9380-2 21979328 Capasso M. Caso D. Conner M. Anticipating pride or regret? Effects of anticipated affect focused persuasive messages on intention to get vaccinated against COVID-19 Social Science & Medicine 289 2021 114416 10.1016/j.socscimed.2021.114416 Carey R.N. Connell L.E. Johnston M. Rothman A.J. de Bruin M. Kelly M.P. Behavior change techniques and their mechanisms of action: A synthesis of links described in published intervention literature Annals of Behavioral Medicine 53 8 2019 693 707 10.1093/abm/kay078 30304386 Carpenter C.J. A meta-analysis of the effectiveness of health belief model variables in predicting behavior Health Communication 25 8 2010 661 669 10.1080/10410236.2010.521906 21153982 CDC COVID-19 in racial and ethnic minority groups 2020 from https://www.cdc.gov/coronavirus/2019-ncov/need-extra-precautions/racial-ethnic-minorities.html CDC CDC expands eligibility for COVID-19 booster shots to all adults 2021 from https://www.cdc.gov/media/releases/2021/s1119-booster-shots.html CDC Excess deaths associated with COVID-19 2021 from https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm Chatzisarantis N.L.D. Hagger M.S. Effects of a brief intervention based on the theory of planned behavior on leisure time physical activity participation Journal of Sport & Exercise Psychology 27 2005 470 487 10.1123/jsep.27.4.470 Chatzisarantis N.L.D. Hagger M.S. Smith B. Phoenix C. The influences of continuation intentions on the execution of social behaviour within the theory of planned behaviour British Journal of Social Psychology 43 4 2004 551 583 10.1348/0144666042565399 15601509 Chou W.-Y.S. Budenz A. Considering emotion in COVID-19 vaccine communication: Addressing vaccine hesitancy and fostering vaccine confidence Health Communication 35 14 2020 1718 1722 10.1080/10410236.2020.1838096 33124475 Chu H. Liu S. Integrating health behavior theories to predict American's intention to receive a COVID-19 vaccine Patient Education and Counseling 104 8 2021 1878 1886 10.1016/j.pec.2021.02.031 33632632 Clark C. Davila A. Regis M. Kraus S. Predictors of COVID-19 voluntary compliance behaviors: An international investigation Global Transitions 2 2020 76 82 10.1016/j.glt.2020.06.003 32835202 Coifman K.G. Disabato D.J. Aurora P. Seah T.H.S. Mitchell B. Simonovic N. What drives preventive health behavior during a global pandemic? Emotion and worry Annals of Behavioral Medicine 55 8 2021 791 804 10.1093/abm/kaab048 34165145 Connell L.E. Carey R.N. de Bruin M. Rothman A.J. Johnston M. Kelly M.P. Links between behavior change techniques and mechanisms of action: An expert consensus study Annals of Behavioral Medicine 53 8 2019 708 720 10.1093/abm/kay082 30452535 Conner M.T. Abraham C. Conscientiousness and the theory of planned behavior: Toward a more complete model of the antecedents of intentions and behavior Personality and Social Psychology Bulletin 27 11 2001 1547 1561 10.1177/01461672012711014 Conner M.T. McEachan R. Taylor N. O'Hara J. Lawton R. Role of affective attitudes and anticipated affective reactions in predicting health behaviors Health Psychology 34 6 2015 642 652 10.1037/hea0000143 25222083 Conner M.T. Norman P. Predicting and changing health behaviour: A social cognition approach Conner M.T. Norman P. Predicting and changing health behaviour: Research and practice with social cognition models 3rd ed. 2015 Open University Press 1 29 Conner M.T. Norman P. Predicting and changing health behaviour: Research and practice with social cognition models 3rd ed. 2015 Open University Press Cooke R. Sheeran P. Moderation of cognition-intention and cognition-behaviour relations: A meta-analysis of properties of variables from the theory of planned behaviour British Journal of Social Psychology 43 2 2004 159 186 10.1348/0144666041501688 15285829 Crozier A. Rajan S. Buchan I. McKee M. Put to the test: Ase of rapid testing technologies for covid-19 BMJ 372 2021 n208 10.1136/bmj.n208 33536228 Das A.K. Abdul Kader Jilani M.M. Uddin M.S. Uddin M.A. Ghosh A.K. Fighting ahead: Adoption of social distancing in COVID-19 outbreak through the lens of theory of planned behavior Journal of Human Behavior in the Social Environment 31 1–4 2021 373 393 10.1080/10911359.2020.1833804 Davis R. Campbell R. Hildon Z. Hobbs L. Michie S. Theories of behaviour and behaviour change across the social and behavioural sciences: A scoping review Health Psychology Review 9 3 2015 323 344 10.1080/17437199.2014.941722 25104107 Derksen C. Keller F.M. Lippke S. Obstetric healthcare workers' adherence to hand hygiene recommendations during the COVID-19 pandemic: Observations and social-cognitive determinants Applied Psychology: Health and Well-Being 12 4 2020 1286 1305 10.1111/aphw.12240 33016518 Divine A. Berry T. Rodgers W. Hall C. The relationship of self-efficacy and explicit and implicit associations on the intention–behavior gap Journal of Physical Activity and Health 18 1 2021 29 36 10.1123/jpah.2019-0033 33338987 DuCharme K.A. Brawley L.R. Predicting the intentions and behavior of exercise initiates using two forms of self-efficacy Journal of Behavioral Medicine 18 1995 479 497 10.1007/BF01904775 8847716 Duckworth A.L. Gendler T.S. Gross J.J. Situational strategies for self-control Perspectives on Psychological Science 11 1 2016 35 55 10.1177/1745691615623247 26817725 Fishbein M. Ajzen I. Predicting and changing behavior: The reasoned action approach 2010 Psychology Press 10.4324/9780203838020 Fishbein M. Triandis H.C. Kanfer F.H. Becker M. Middlestadt S.E. Eichler A. Factors influencing behavior and behavior change Baum A. Revenson T.A. Singer J.E. Handbook of health psychology 2001 Lawrence Erlbaum 3 17 Gardner B. Lally P. Rebar A.L. Does habit weaken the relationship between intention and behaviour? Revisiting the habit-intention interaction hypothesis Social and Personality Psychology Compass 14 8 2020 e12553 10.1111/spc3.12553 Gardner B. Lally P. Wardle J. Making health habitual: The psychology of ‘habit-formation’ and general practice British Journal of General Practice 62 605 2012 664 Gardner B. Rebar A. Lally P. Habit interventions Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 599 616 10.1017/97811086773180.041 Gibbons F.X. Gerrard M. Blanton H. Russell D.W. Reasoned action and social reaction: Willingness and intention as independent predictors of health risk Journal of Personality and Social Psychology 74 1998 1164 1180 9599437 Gibson L.P. Magnan R.E. Kramer E.B. Bryan A.D. Theory of planned behavior analysis of social distancing during the COVID-19 pandemic: Focusing on the intention–behavior gap Annals of Behavioral Medicine 55 8 2021 805 812 10.1093/abm/kaab041 34228112 Gollob H.F. Reichardt C.S. Taking account of time lags in causal models Child Development 58 1987 80 92 10.2307/1130293 3816351 Gollwitzer P.M. Implementation intentions: Strong effects of simple plans American Psychologist 54 7 1999 493 503 10.1037/0003-066X.54.7.493 Gollwitzer P.M. Sheeran P. Implementation intentions and goal achievement: A meta-analysis of effects and processes Advances in Experimental Social Psychology 38 2006 69 119 10.1013/S0065-2601(06)38002-1 Hagger M.S. Theoretical integration in health psychology: Unifying ideas and complimentary explanations British Journal of Health Psychology 14 2 2009 189 194 10.1348/135910708X397034 19236795 Hagger M.S. Avoiding the ‘déjà-variable’ phenomenon: Social psychology needs more guides to constructs Frontiers in Psychology 5 2014 52 10.3389/fpsyg.2014.00052 24550871 Hagger M.S. Non-conscious processes and dual-process theories in health psychology Health Psychology Review 10 4 2016 375 380 10.1080/17437199.2016.1244647 27718880 Hagger M.S. Basic model of a behavior change mechanism of action PsyArXiv 10.31234/osf.io/9a5k6 2019 Hagger M.S. Developing an open science ‘mindset Health Psychology and Behavioral Medicine 10 1 2022 1 21 10.1080/21642850.2021.2012474 34993003 Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 10.1017/9781108677318 Hagger M.S. Cheung M.W.L. Ajzen I. Hamilton K. Perceived behavioral control moderating effects in the theory of planned behavior: A meta-analysis Health Psychology 41 2 2022 155 167 10.1037/hea0001153 35143225 Hagger M.S. Gucciardi D.F. Chatzisarantis N.L.D. On nomological validity and auxiliary assumptions: The importance of simultaneously testing effects in social cognitive theories applied to health behavior and some guidelines Frontiers in Psychology 8 2017 1933 10.3389/fpsyg.2017.01933 29163307 Hagger M.S. Hamilton K. Changing behavior using integrated theories Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 208 224 10.1017/97811086773180.015 Hagger M.S. Hamilton K. Effects of socio-structural variables in the theory of planned behavior: A mediation model in multiple samples and behaviors Psychology and Health 36 3 2021 307 333 10.1080/08870446.2020.1784420 32608265 Hagger M.S. Hamilton K. Predicting COVID-19 booster vaccine intentions Applied Psychology: Health and Well-Being 2022 10.1111/aphw.12349 Hagger M.S. Hankonen N. Kangro E.-M. Lintunen T. Pagaduan J. Polet J. Trait self-control, social cognition constructs, and intentions: Correlational evidence for mediation and moderation effects in diverse health behaviors Applied Psychology: Health and Well-Being 11 3 2019 407 437 10.1111/aphw.12153 30724028 Hagger M.S. Lonsdale A. Koka A. Hein V. Pasi H. Lintunen T. An intervention to reduce alcohol consumption in undergraduate students using implementation intentions and mental simulations: A cross-national study International Journal of Behavioral Medicine 19 1 2012 82 96 10.1007/s12529-011-9163-8 21562782 Hagger M.S. Moyers S. McAnally K. McKinley L.E. Known knowns and known unknowns on behavior change interventions and mechanisms of action Health Psychology Review 14 1 2020 199 212 10.1080/17437199.2020.1719184 31964227 Hagger M.S. Orbell S. The common sense model of illness self-regulation: A conceptual review and proposed extended model Health Psychology Review 2021 10.1080/17437199.2021.1878050 Hagger M.S. Polet J. Lintunen T. The reasoned action approach applied to health behavior: Role of past behavior and test of some key moderators using meta-analytic structural equation modeling Social Science & Medicine 213 2018 85 94 10.1016/j.socscimed.2018.07.038 30064092 Hagger M.S. Smith S.R. Keech J.J. Moyers S.A. Hamilton K. Predicting social distancing intention and behavior during the COVID-19 pandemic: An integrated social cognition model Annals of Behavioral Medicine 54 10 2020 713 727 10.1093/abm/kaaa073 32914831 Hagger M.S. Smith S.R. Keech J.J. Moyers S.A. Hamilton K. Predicting physical distancing over time during COVID-19: Testing an integrated model Psychology and Health 2021 10.1080/08870446.2021.1968397 Hagger M.S. Trost N. Keech J. Chan D.K.C. Hamilton K. Predicting sugar consumption: Application of an integrated dual-process, dual-phase model Appetite 116 2017 147 156 10.1016/j.appet.2017.04.032 28461198 Hagger M.S. Weed M.E. Debate: Do behavioral interventions work in the real world? International Journal of Behavioral Nutrition and Physical Activity 16 2019 36 10.1186/s12966-019-0795-4 31023328 Hall P.A. Fong G.T. Temporal self-regulation theory: A model for individual health behavior Health Psychology Review 1 1 2007 6 52 10.1080/17437190701492437 Hamilton K. Gibbs I. Keech J.J. Hagger M.S. Reasoned and implicit processes in heavy episodic drinking: An integrated dual process model British Journal of Health Psychology 25 1 2020 189 209 10.1111/BJHP.12401 31876984 Hamilton K. Hagger M.S. The vaccination concerns in COVID-19 scale (VaCCS): Development and validation PLoS One 17 3 2022 0264784 10.1371/journal.pone.0264784 Hamilton K. Johnson B.T. Attitude and persuasive communication interventions Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 445 460 10.1017/97811086773180.031 Hamilton K. Kirkpatrick A. Rebar A. Hagger M.S. Child sun safety: Application of an integrated behavior change model Health Psychology 36 9 2017 916 926 10.1037/hea0000533 28726470 Hamilton K. Smith S.R. Keech J.J. Moyers S.A. Hagger M.S. Application of the health action process approach to social distancing behavior during COVID‐19 Applied Psychology: Health and Well-Being 12 4 2020 1244 1269 10.1111/aphw.12231 33006814 Hamilton K. van Dongen A. Hagger M.S. An extended theory of planned behavior for parent-for-child health behaviors: A meta-analysis Health Psychology 39 10 2020 863 878 10.1037/hea0000940 32597678 Harvey A.G. Armstrong C.C. Callaway C.A. Gumport N.B. Gasperetti C.E. COVID-19 prevention via the science of habit formation Current Directions in Psychological Science 30 2 2021 174 180 10.1177/0963721421992028 Heckhausen H. Gollwitzer P.M. Thought contents and cognitive functioning in motivational and volitional states of mind Motivation and Emotion 11 1987 101 120 10.1007/BF00992338 Holmes E.A. O'Connor R.C. Perry V.H. Tracey I. Wessely S. Arseneault L. … Multidisciplinary research priorities for the COVID-19 pandemic: A call for action for mental health science The Lancet Psychiatry 7 6 2020 547 560 10.1016/S2215-0366(20)30168-1 32304649 Huynh H.P. Zsila Á Martinez-Berman L. Psychosocial predictors of intention to vaccinate against the coronavirus (COVID-19) Behavioral Medicine 2022 10.1080/08964289.2021.1990006 Imai K. Keele L. Tingley D. Yamamoto T. Unpacking the black box of causality: Learning about causal mechanisms from experimental and observational studies American Political Science Review 105 4 2011 765 789 10.2307/23275352 Irfan M. Akhtar N. Ahmad M. Shahzad F. Elavarasan R.M. Wu H. Assessing public willingness to wear face masks during the COVID-19 pandemic: Fresh insights from the theory of planned behavior International Journal of Environmental Research and Public Health 18 9 2021 4577 33925929 Jackson J.K. Weiss M.A. Schwarzenberg A.B. Nelson R.M. Sutter K.M. Sutherland M.D. Economic effects of COVID-19 2021 Congressional Research Services Washington DC from https://sgp.fas.org/crs/row/R46270.pdf Jacobs N. Hagger M.S. Streukens S. De Bourdeaudhuij I. Claes N. Testing an integrated model of the theory of planned behaviour and self-determination theory for different energy-balance related behaviours and intervention intensities British Journal of Health Psychology 16 1 2011 113 134 10.1348/135910710X519305 21226787 Jang D. Kim I. Kwon S. Motivation and intention toward physical activity during the COVID-19 pandemic: Perspectives from integrated model of self-determination and planned behavior theories Frontiers in Psychology 12 2021 3124 10.3389/fpsyg.2021.714865 Ju Y. You M. It's politics, isn't it? Investigating direct and indirect influences of political orientation on risk perception of COVID-19 Risk Analysis 42 1 2022 56 68 10.1111/risa.13801 34459000 JVCI JCVI statement, september 2021: COVID-19 booster vaccine programme for winter 2021 to 2022 2021 Department of Health and Social Care UK Retrieved November 19 from https://www.gov.uk/government/publications/jcvi-statement-september-2021-covid-19-booster-vaccine-programme-for-winter-2021-to-2022 Kashte S. Gulbake A. El-Amin S.F. Iii Gupta A. COVID-19 vaccines: Rapid development, implications, challenges and future prospects Human Cell 34 3 2021 711 733 10.1007/s13577-021-00512-4 33677814 Kazdin A.E. Mediators and mechanisms of change in psychotherapy research Annual Review of Clinical Psychology 3 1 2007 1 27 10.1146/annurev.clinpsy.3.022806.091432 Keatley D.A. Clarke D.D. Hagger M.S. Investigating the predictive validity of implicit and explicit measures of motivation on condom use, physical activity, and healthy eating Psychology and Health 27 5 2012 550 569 10.1080/08870446.2011.605451 21895458 Keller J. Kwasnicka D. Wilhelm L.O. Lorbeer N. Pauly T. Domke A. Hand washing and related cognitions following a brief behavior change intervention during the COVID-19 pandemic: A pre-post analysis International Journal of Behavioral Medicine 2021 10.1007/s12529-021-10042-w Kelley T.L. Interpretation of educational measurements 1927 World Book Co Knittle K. Nurmi J. Crutzen R. Hankonen N. Beattie M. Dombrowski S.U. How can interventions increase motivation for physical activity? A systematic review and meta-analysis Health Psychology Review 12 3 2018 211 230 10.1080/17437199.2018.1435299 29385950 Kojan L. Burbach L. Ziefle M. Calero Valdez A. Perceptions of behaviour efficacy, not perceptions of threat, are drivers of COVID-19 protective behaviour in Germany from 10.31219/osf.io/fm69j 2021 Kok G. Gottlieb N.H. Peters G.-J.Y. Mullen P.D. Parcel G.S. Ruiter R.A.C. A taxonomy of behavior change methods: An intervention mapping approach Health Psychology Review 10 3 2016 297 312 10.1080/17437199.2015.1077155 26262912 Krause P.R. Fleming T.R. Peto R. Longini I.M. Figueroa J.P. Sterne J.A.C. Considerations in boosting COVID-19 vaccine immune responses The Lancet 398 10308 2021 1377 1380 10.1016/S0140-6736(21)02046-8 Lally P. van Jaarsveld C.H.M. Potts H.W.W. Wardle J. How are habits formed: Modelling habit formation in the real world European Journal of Social Psychology 40 2010 998 1009 10.1002/ejsp.674 Lao C.K. Li X. Zhao N. Gou M. Zhou G. Using the health action process approach to predict facemask use and hand washing in the early stages of the COVID-19 pandemic in China Current Psychology 2021 10.1007/s12144-021-01985-0 Larsen B. Dunsiger S.I. Pekmezi D. Linke S. Hartman S.J. Marcus B.H. Psychosocial mediators of physical activity change in a web-based intervention for Latinas Health Psychology 40 1 2021 21 29 10.1037/hea0001041 33370154 Latkin C. Dayton L.A. Yi G. Konstantopoulos A. Park J. Maulsby C. COVID-19 vaccine intentions in the United States, a social-ecological framework Vaccine 39 16 2021 2288 2294 10.1016/j.vaccine.2021.02.058 33771392 Leventhal H. Meyer D. Nerenz D. The common sense model of illness danger Rachman S. Medical psychology Vol. II 1980 Pergamon Press 7 30 Lin C.-Y. Imani V. Ghasemi Z. Majd N.R. Griffiths M.D. Hamilton K. Using an integrated social cognition model to predict COVID-19 preventive behaviors British Journal of Health Psychology 25 4 2020 981 1005 10.1111/bjhp.12465 32780891 Liska A.E. Felson R.B. Chamlin M. Baccaglini W. Estimating attitude-behavior reciprocal effects within a theoretical specification Social Psychology Quarterly 47 1984 15 23 10.2307/3033884 Luszczynska A. Szczuka Z. Abraham C. Baban A. Brooks S. Cipolletta S. … The interplay between strictness of policies and individuals’ self-regulatory efforts: Associations with handwashing during the COVID-19 pandemic Annals of Behavioral Medicine 2021 10.1093/abm/kaab102 Martiny-Huenger T. Martiny S.E. Parks-Stamm E.J. Pfeiffer E. Gollwitzer P.M. From conscious thought to automatic action: A simulation account of action planning Journal of Experimental Psychology: General 146 10 2017 1513 1525 10.1037/xge0000344 28703618 Mathieu E. Ritchie H. Ortiz-Ospina E. Roser M. Hasell J. Appel C. A global database of COVID-19 vaccinations Nature Human Behaviour 5 7 2021 947 953 10.1038/s41562-021-01122-8 Matute J. Palau-Saumell R. Meyer J. Derqui B. Jiménez-Asenjo N. Are you getting it? Integrating theories to explain intentions to get vaccinated against COVID-19 in Spain Journal of Risk Research 2021 1 20 10.1080/13669877.2021.1958044 McEachan R.R.C. Conner M.T. Taylor N. Lawton R.J. Prospective prediction of health-related behaviors with the theory of planned behavior: A meta-analysis Health Psychology Review 5 2 2011 97 144 10.1080/17437199.2010.521684 McElfish P.A. Purvis R. James L.P. Willis D.E. Andersen J.A. Perceived barriers to COVID-19 testing International Journal of Environmental Research and Public Health 18 5 2021 2278 33668958 McEwan D. Beauchamp M.R. Kouvousis C. Ray C.M. Wyrough A. Rhodes R.E. Examining the active ingredients of physical activity interventions underpinned by theory versus no stated theory: A meta-analysis Health Psychology Review 13 1 2019 1 17 10.1080/17437199.2018.1547120 30412685 McMillan B. Conner M. Health cognition assessment Ayers A.B.S. McManus C. Newman S. Wallston K. Weinman J. West R. Cambridge handbook of psychology, health and medicine 2nd ed. 2007 Cambridge University Press 260 266 Michie S. Behavioural strategies for reducing covid-19 transmission in the general population 2020 2020, March 3, from https://blogs.bmj.com/bmj/2020/03/03/behavioural-strategies-for-reducing-covid-19-transmission-in-the-general-population/ Michie S. Carey R.N. Johnston M. Rothman A.J. de Bruin M. Kelly M.P. From theory-inspired to theory-based interventions: A protocol for developing and testing a methodology for linking behaviour change techniques to theoretical mechanisms of action Annals of Behavioral Medicine 52 6 2018 501 512 10.1007/s12160-016-9816-6 27401001 Michie S. Richardson M. Johnston M. Abraham C. Francis J. Hardeman W. The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: Building an international consensus for the reporting of behavior change interventions Annals of Behavioral Medicine 46 1 2013 81 95 10.1007/s12160-013-9486-6 23512568 Michie S. Rubin G.J. Amlôt R. Behavioural science must be at the heart of the public health response to COVID-19 2020 from https://blogs.bmj.com/bmj/2020/02/28/behavioural-science-must-be-at-the-heart-of-the-public-health-response-to-covid-19/ Michie S. West R. Rogers M.B. Bonell C. Rubin G.J. Amlôt R. Reducing SARS-CoV-2 transmission in the UK: A behavioural science approach to identifying options for increasing adherence to social distancing and shielding vulnerable people British Journal of Health Psychology 2020 10.1111/bjhp.12428 Milne S. Sheeran P. Orbell S. Prediction and intervention in health-related behavior: A meta-analytic review of protection motivation theory Journal of Applied Social Psychology 30 2000 106 143 10.1111/j.1559-1816.2000.tb02308.x Montaño D.E. Kasprzyk D. Theory of reasoned action, theory of planned behavior, and the integrated behavioral model Glanz K. Rimer B.K. Viswanath K. Health behavior and health education: Theory, research, and practice 5th ed. 2015 Jossey-Bass 95 124 Naaber P. Tserel L. Kangro K. Sepp E. Jürjenson V. Adamson A. Dynamics of antibody response to BNT162b2 vaccine after six months: A longitudinal prospective study The Lancet Regional Health - Europe 10 2021 100208 10.1016/j.lanepe.2021.100208 Ndugga N. Hill L. Artiga S. Haldar S. Latest data on COVID-19 vaccinations by race/ethnicity from https://www.kff.org/coronavirus-covid-19/issue-brief/latest-data-on-covid-19-vaccinations-by-race-ethnicity/ 2021 Nofal A.M. Cacciotti G. Lee N. Who complies with COVID-19 transmission mitigation behavioral guidelines? PLoS One 15 10 2020 e0240396 10.1371/journal.pone.0240396 Norman P. Cameron D. Epton T. Webb T.L. Harris P.R. Millings A. A randomized controlled trial of a brief online intervention to reduce alcohol consumption in new university students: Combining self-affirmation, theory of planned behaviour messages, and implementation intentions British Journal of Health Psychology 23 1 2018 108 127 10.1111/bjhp.12277 28941040 Norman P. Wilding S. Conner M.T. Reasoned action approach and compliance with recommended behaviours to prevent the transmission of the SARS-CoV-2 virus in the UK British Journal of Health Psychology 25 4 2020 1006 1019 10.1111/bjhp.12474 33007143 Okuhara T. Okada H. Kiuchi T. Examining persuasive message type to encourage staying at home during the COVID-19 pandemic and social lockdown: A randomized controlled study in Japan Patient Education and Counseling 103 12 2020 2588 2593 10.1016/j.pec.2020.08.016 Orbell S. Hodgkins S. Sheeran P. Implementation intentions and the theory of planned behavior Personality and Social Psychology Bulletin 23 9 1997 945 954 10.1177/0146167297239004 29506445 Orbell S. Sheeran P. Inclined abstainers': A problem for predicting health related behaviour British Journal of Social Psychology 37 2 1998 151 165 10.1111/j.2044-8309.1998.tb01162.x 9639861 Orbell S. Szczepura A. Weller D. Gumber A. Hagger M.S. South Asian ethnicity, socio-economic status and psychological mediators of faecal occult blood colorectal screening participation: A prospective test of a process model Health Psychology 36 12 2017 1161 1172 10.1037/hea0000525 28726477 Ouellette J.A. Wood W. Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior Psychological Bulletin 124 1 1998 54 74 10.1037/0033-2909.124.1.54 Peterson L.M. Helweg-Larsen M. DiMuccio S. Descriptive norms and prototypes predict COVID-19 prevention cognitions and behaviors in the United States: Applying the prototype willingness model to pandemic mitigation Annals of Behavioral Medicine 55 11 2021 1089 1103 10.1093/abm/kaab075 34487142 Phillips N. The coronavirus is here to stay — here's what that means Nature 590 2021 382 384 10.1038/d41586-021-00396-2 33594289 Prestwich A. Sniehotta F.F. Whittington C. Dombrowski S.U. Rogers L. Michie S. Does theory influence the effectiveness of health behavior interventions? Meta-analysis Health Psychology 33 5 2014 465 474 10.1037/a0032853 23730717 Prestwich A. Webb T.L. Conner M. Using theory to develop and test interventions to promote changes in health behaviour: Evidence, issues, and recommendations Current Opinion in Psychology 5 2015 1 5 10.1016/j.copsyc.2015.02.011 Protogerou C. Johnson B.T. Hagger M.S. An integrated model of condom use in sub-saharan African youth: A meta-analysis Health Psychology 37 6 2018 586 602 10.1037/hea0000604 29708390 Quaresma A.M. Palmeira A.L. Martins S.S. Minderico C.S. Sardinha L.B. Effect of a school-based intervention on physical activity and quality of life through serial mediation of social support and exercise motivation: The PESSOA program Health Education Research 29 6 2014 906 917 10.1093/her/cyu056 25274722 Rabin C. Dutra S. Predicting engagement in behaviors to reduce the spread of COVID-19: The roles of the health belief model and political party affiliation Psychology Health & Medicine 2021 10.1080/13548506.2021.1921229 Raza A. Ali Q. Hussain T. Role of knowledge, behavior, norms, and e-guidelines in controlling the spread of COVID-19: Evidence from Pakistan Environmental Science and Pollution Research 28 30 2021 40329 40345 10.1007/s11356-020-10931-9 33011950 Rhodes R.E. Boudreau F. Weman Josefsson K. Ivarsson A. Mediators of physical activity behavior change interventions among adults: A systematic review and meta-analysis Health Psychology Review 15 2 2020 272 286 10.1080/17437199.2019.1706614 31875768 Rhodes R.E. Courneya K.S. Hayduk L.A. Does personality moderate the theory of planned behavior in the exercise domain? Journal of Sport & Exercise Psychology 24 2002 120 132 10.1123/jsep.24.2.120 Rhodes R.E. Saelens B.E. Sauvage-Mar C. Understanding physical activity through interactions between the built environment and social cognition: A systematic review Sports Medicine 48 8 2018 1893 1912 10.1007/s40279-018-0934-0 29767384 Rivis A. Sheeran P. Armitage C.J. Expanding the affective and normative components of the theory of planned behavior: A meta-analysis of anticipated affect and moral norms Journal of Applied Social Psychology 39 12 2009 2985 3019 10.1111/j.1559-1816.2009.00558.x Rogers R.W. A protection motivation theory of fear appeals and attitude change Journal of Psychology 91 1 1975 93 114 10.1080/00223980.1975.9915803 28136248 Rosenstock I.M. Historical origins of the health belief model Health Education Monographs 2 1974 328 335 10.1177/109019817400200403 Rothman A.J. Klein W.M.P. Sheeran P. Moving from theoretical principles to intervention strategies: Applying the experimental medicine approach Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 285 299 10.1017/97811086773180.020 Rothman A.J. Sheeran P. The operating conditions framework: Integrating mechanisms and moderators in health behavior interventions Health Psychology 2020 10.1037/hea0001026 Sallis J.F. Owen N. Fisher E.B. Ecological models of health behavior Glanz K. Rimer B.K. Viswanath K. Health behavior and health education: Theory, research, and practice 5th ed. 2015 Jossey-Bass 43 64 Salmon J. Hesketh K.D. Arundell L. Downing K.L. Biddle S.J.H. Changing behavior using ecological models Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. The handbook of behavior change 2020 Cambridge University Press 237 250 10.1017/97811086773180.017 Schüz B. Brick C. Wilding S. Conner M.T. Socioeconomic status moderates the effects of health cognitions on health behaviors within participants: Two multibehavior studies Annals of Behavioral Medicine 54 1 2020 36 48 10.1093/abm/kaz023 31260512 Schüz B. Conner M. Wilding S. Alhawtan R. Prestwich A. Norman P. Do socio-structural factors moderate the effects of health cognitions on COVID-19 protection behaviours? Social Science & Medicine 285 2021 114261 10.1016/j.socscimed.2021.114261 Schwarzer R. Modeling health behaviour change: How to predict and modify the adoption and maintenance of health behaviors Applied Psychology: International Review 57 1 2008 1 29 10.1111/j.1464-0597.2007.00325.x Schwarzer R. Hamilton K. Changing behavior using the health action process approach Hagger M.S. Cameron L.D. Hamilton K. Hankonen N. Lintunen T. Handbook of behavior change 2020 Cambridge University Press 89 103 10.1017/97811086773180.007 Sheeran P. Gollwitzer P.M. Bargh J.A. Nonconscious processes and health Health Psychology 32 5 2013 460 473 10.1037/a0029203 22888816 Sheeran P. Klein W.M.P. Rothman A.J. Health behavior change: Moving from observation to intervention Annual Review of Psychology 68 1 2017 573 600 10.1146/annurev-psych-010416-044007 Sheeran P. Maki A. Montanaro E. Avishai-Yitshak A. Bryan A. Klein W.M.P. The impact of changing attitudes, norms, and self-efficacy on health-related intentions and behavior: A meta-analysis Health Psychology 35 11 2016 1178 1188 10.1037/hea0000387 27280365 Sheeran P. Webb T.L. The intention–behavior gap Social and Personality Psychology Compass 10 9 2016 503 518 10.1111/spc3.12265 Sheeran P. Wright C.E. Avishai A. Villegas M.E. Lindemans J.W. Klein W.M.P. Self-determination theory interventions for health behavior change: Meta-analysis and meta-analytic structural equation modeling of randomized controlled trials Journal of Consulting and Clinical Psychology 88 8 2020 726 737 10.1037/ccp0000501 32437175 Sheeran P. Wright C.E. Avishai A. Villegas M.E. Rothman A.J. Klein W.M.P. Does increasing autonomous motivation or perceived competence lead to health behavior change? A meta-analysis Health Psychology 40 10 2021 706 716 10.1037/hea0001111 34881939 Sherman S.M. Smith L.E. Sim J. Amlôt R. Cutts M. Dasch H. COVID-19 vaccination intention in the UK: Results from the COVID-19 vaccination acceptability study (CoVAccS), a nationally representative cross-sectional survey Human Vaccines & Immunotherapeutics 17 6 2021 1612 1621 10.1080/21645515.2020.1846397 33242386 Shiloh S. Peleg S. Nudelman G. Vaccination against COVID-19: A longitudinal trans-theoretical study to determine factors that predict intentions and behavior Annals of Behavioral Medicine 2021 10.1093/abm/kaab101 Smith S.R. Hagger M.S. Keech J.J. Moyers S.A. Hamilton K. Improving hand hygiene behavior using a novel theory-based intervention during the COVID-19 pandemic from 10.31219/osf.io/uzhvx 2021 Strack F. Deutsch R. Reflective and impulsive determinants of social behavior Personality and Social Psychology Review 8 2004 220 247 10.1207/s15327957pspr0803_1 15454347 Suls J. Mogavero J.N. Falzon L. Pescatello L.S. Hennessy E.A. Davidson K.W. Health behaviour change in cardiovascular disease prevention and management: Meta-review of behavior change techniques to affect self-regulation Health Psychology Review 14 1 2020 43 65 10.1080/17437199.2019.1691622 31707938 Tong K.K. He M. Wu A.M.S. Dang L. Chen J.H. Cognitive factors influencing COVID-19 vaccination intentions: An application of the protection motivation theory using a probability community sample Vaccines 9 10 2021 1170 34696278 Trafimow D. The role of auxiliary assumptions for the validity of manipulations and measures Theory & Psychology 22 4 2012 486 498 10.1177/0959354311429996 Usami Satoshi On the differences between general cross-lagged panel model and random-intercept cross-lagged panel model: Interpretation of cross-lagged parameters and model choice Structural Equation Modeling: A Multidisciplinary Journal 28 3 2021 331 344 10.1080/10705511.2020.1821690 Vandrevala T. Montague A. Terry P. Fielder M.D. Willingness of the UK public to volunteer for testing in relation to the COVID-19 pandemic BMC Public Health 22 1 2022 565 10.1186/s12889-022-12848-z 35317756 Varol T. Crutzen R. Schneider F. Mesters I. Ruiter R.A.C. Kok G. Selection of determinants of students' adherence to COVID-19 guidelines and translation into a brief intervention Acta Psychologica 219 2021 103400 10.1016/j.actpsy.2021.103400 Verplanken B. Beyond frequency: Habit as mental construct British Journal of Social Psychology 45 3 2006 639 656 10.1348/014466605X49122 16984725 Verplanken B. Orbell S. Reflections on past behavior: A self-report index of habit strength Journal of Applied Social Psychology 33 6 2003 1313 1330 10.1111/j.1559-1816.2003.tb01951.x Webb T.L. Sheeran P. Does changing behavioral intentions engender behavior change? A meta-analysis of the experimental evidence Psychological Bulletin 132 2 2006 249 268 10.1037/0033-2909.132.2.249 16536643 Weinstein N.D. Misleading tests of health behavior theories Annals of Behavioral Medicine 33 2007 1 10 10.1207/s15324796abm3301_1 17291165 West R. Michie S. Rubin G.J. Amlôt R. Applying principles of behaviour change to reduce SARS-CoV-2 transmission Nature Human Behaviour 4 5 2020 451 459 10.1038/s41562-020-0887-9 WHO WHO Director-General's opening remarks at the media briefing on COVID-19 - 11 March 2020 2020 Retrieved March 26, 2022, from https://www.who.int/director-general/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19---11-march-2020 Wood W. Labrecque J.S. Lin P.-Y. Rünger D. Habits in dual process models Sherman J.W. Gawronski B. Trope Y. Dual-process theories of the social mind 2014 Guildford Press 371 385 Wood W. Quinn J.M. Kashy D.A. Habits in everyday life: Thought, emotion, and action Journal of Personality and Social Psychology 83 6 2002 1281 1297 10.1037/0022-3514.83.6.1281 12500811 Wu S. Neill R. De Foo C. Chua A.Q. Jung A.-S. Haldane V. Aggressive containment, suppression, and mitigation of covid-19: Lessons learnt from eight countries BMJ 375 2021 e067508 10.1136/bmj-2021-067508 Xu P. Cheng J. Individual differences in social distancing and mask-wearing in the pandemic of COVID-19: The role of need for cognition, self-control and risk attitude Personality and Individual Differences 175 2021 110706 10.1016/j.paid.2021.110706 Yahaghi R. Ahmadizade S. Fotuhi R. Taherkhani E. Ranjbaran M. Buchali Z. Fear of COVID-19 and perceived COVID-19 infectability supplement theory of planned behavior to explain Iranians' intention to get COVID-19 vaccinated Vaccines 9 7 2021 684 10.3390/vaccines9070684 34206226 Yu Y. Lau J.T.F. Lau M.M.C. Levels and factors of social and physical distancing based on the Theory of Planned Behavior during the COVID-19 pandemic among Chinese adults Translational Behavioral Medicine 11 5 2021 1179 1186 10.1093/tbm/ibaa146 33598679 Zettler I. Schild C. Lilleholt L. Kroencke L. Utesch T. Moshagen M. The role of personality in COVID-19-related perceptions, evaluations, and behaviors: Findings across five samples, nine traits, and 17 criteria Social Psychological and Personality Science 13 1 2022 299 310 10.1177/19485506211001680 Zhang C.Q. Zhang R. Schwarzer R. Hagger M.S. A meta-analysis of the health action process approach Health Psychology 38 7 2019 623 637 10.1037/hea0000728 30973747
PMC009xxxxxx/PMC9005316.txt	==== Front SN Compr Clin Med SN Compr Clin Med Sn Comprehensive Clinical Medicine 2523-8973 Springer International Publishing Cham 35434525 1177 10.1007/s42399-022-01177-2 Case Report Non-traumatic Myositis Ossificans as Unusual Cause of Neck Pain During COVID-19 Pandemic: a Case Report Vitale Valerio valerio.vitale@aulss8.veneto.it 1 Bleve Cosimo 2 Mansour Mariam 1 De Corti Federica 3 Giarraputo Leonardo 4 Brugiolo Alessandra 5 Affinita Maria Carmen 6 Santoro Luisa 7 Chiarenza Salvatore Fabio 2 Iannucci Giuseppe 1 1 grid.416303.3 0000 0004 1758 2035 Department of Neurosciences, Neuroradiology Unit, San Bortolo Hospital, AULSS 8 Berica, Vicenza, Italy 2 grid.416303.3 0000 0004 1758 2035 Department of Pediatric Surgery and Pediatric Minimally Invasive Surgery and New Technologies, San Bortolo Hospital, AULSS 8 Berica, Vicenza, Italy 3 grid.411474.3 0000 0004 1760 2630 Pediatric Surgery Unit, Department of Women’s and Children’s Health, Padua University Hospital, Padua, Italy 4 grid.416303.3 0000 0004 1758 2035 Radiology Unit, San Bortolo Hospital, AULSS 8 Berica, Vicenza, Italy 5 grid.416303.3 0000 0004 1758 2035 Pediatrics Unit, San Bortolo Hospital, AULSS 8 Berica, Vicenza, Italy 6 grid.5608.b 0000 0004 1757 3470 Hematology and Oncology Unit, Department of Women’s and Children’s Health, University of Padua, Padua, Italy 7 grid.411474.3 0000 0004 1760 2630 Pathology Unit, University Hospital of Padua, Padua, Italy 13 4 2022 2022 4 1 964 4 2022 © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Myositis ossificans circumscripta (MOC) is a benign disease characterized by localized heterotopic bone formation within muscles or soft tissue, usually interesting great muscles of extremities. We report a rare case of unusual location in the neck not associated with previous trauma, mimicking a solid tumor, with well-documented diagnostic imaging features. During COVID-19 pandemic outbreak in Italy, in May 2020, a 14-year-old boy developed a progressive and persistent neck pain on the right side, without known history of trauma. Initial therapy with non-steroid anti-inflammatory drugs and physiokinetic therapy gave only a slight improvement. A neck ultrasound showed an inhomogeneous right neck mass, with posterior shadowing due to calcifications. Computed tomography and magnetic resonance imaging confirmed a huge right neck mass, located in the paravertebral space with peripheral calcifications and mild central contrast enhancement. After surgical excision of the lesion, pathology revealed the presence of muscular tissue mixed with fibroblastic/myofibroblastic proliferation and ossification areas consistent with myositis ossificans. A careful analysis of clinical and radiological features is very important to manage young patients showing progressive pain and swelling of the neck, since MOC can mimic soft tissue or bone tumors, and it should be suspected even in the absence of a known history of trauma. Keywords Myositis ossificans Head-neck MRI Tumor Case report issue-copyright-statement© Springer Nature Switzerland AG 2022 ==== Body pmcIntroduction Myositis ossificans circumscripta (MOC) is a benign disease characterized by localized heterotopic bone formation within muscles or soft tissue [1, 2]. Usually, it occurs in the extremities and after history of trauma, but pathogenesis is still unclear in non-traumatic cases [3, 4]. Men and young athletes are most commonly affected [3, 5]. The evolution of MOC is a dynamic process that can last from weeks to several months to complete maturation of the lesion [3, 5]. We describe a rare case of non-traumatic neck MOC in an otherwise healthy adolescent, mimicking a solid tumor, with well-documented diagnostic imaging features. Case Description A 14-year-old boy, during COVID-19 pandemic outbreak in Italy in May 2020, developed a progressive and persistent neck pain on the right side, without known history of trauma. During that period, all sport activities were suspended and patient stayed home for distance learning. No relevant past health history was reported and no abnormalities of the cervical spine were found on radiography, executed 1 month after clinical onset. The persistence of pain and slight swelling led to a progressive movement limitation. Therefore, he was treated with non-steroid anti-inflammatory drugs and physiokinetic therapy with only slight improvement. Four months later, he underwent a neck ultrasound (US) showing an inhomogeneous right neck mass, with posterior shadowing (Fig. 1A) indicating the presence of calcifications. Further in-depth analysis was carried out using computed tomography (CT) and magnetic resonance imaging (MRI).Fig. 1 Ultrasound scan showing a right-sided neck mass with posterior shadowing (caliper) (A); axial and coronal CT scans before (B, D) and after (C, E) contrast injection, showing the right inhomogeneous mass located between scalene and paraspinals muscles, with prevalent peripheral calcifications and central contrast enhancement. Axial and coronal bone window (F, G) showing close relationship with transverse process of C7 and the first right rib (white arrows in G); note also irregular sclerotic reaction (white arrow in F) CT scan images (64 channels General Electric, Boston, Massachusetts) confirmed the presence of a huge right neck mass, located in the paravertebral space between the scalene muscles anteriorly and paraspinal muscles posteriorly (i.e., longissimus capitis muscle, longissimus cervicis muscle, semispinalis capitis muscle, and splenius cervicis muscle), and closely related to the right transverse process of C7 and the first rib, which showed cortical bone irregularity (Fig. 1B–G). The lesion was also characterized by peripheral calcifications and mild central contrast enhancement (CE). At MRI (3 Tesla Skyra Siemens, Erlangen, Germany), the lesion was hypointense on T1-weighted (w) and hyperintense on T2w images, with mild signal inhomogeneity corresponding to calcifications, showed a more vivid CE compared to CT, and a peripheral tissue alteration was revealed (Fig. 2); no relationship with right C6–C7 neuroforamina or nerve roots was found.Fig. 2 MRI scan: T2-weighted (w) sequences without (A, axial) and with (B–C, axial and coronal) fat saturation showing the lobulated right neck mass, inhomogeneously hyperintense with swelling of surrounding tissues (white arrow); there is no relationship with right neuroforamina. 3D T1-w (VIBE sequence) before (D, axial) and after (E–F, axial and coronal) contrast injection: note contrast enhancement of the lesion and peripheral dark rim related to calcifications (white arrow in D) Retrospective revision of spine X-ray evidenced subtle swelling of right cervical soft tissues (Fig. 3).Fig. 3 X-ray scans: note right soft tissue neck opacity compared with left side, indicating swelling (A, white arrow) Percutanous US-guided biopsy, performed to rule out the benign or malignant nature of the lesion, was unsuccessful. Therefore, patient underwent a partial surgical excision of the lesion, which was strongly attached to the right transverse process of C7. Pathology revealed the presence of muscular tissue mixed with fibroblastic/myofibroblastic proliferation and ossification areas consistent with myositis ossificans (Fig. 4); COL1A1-USP6 fusion transcript upon molecular investigation by RT-PCR method was found.Fig. 4 A–B Histologically, the lesion shows a zonation pattern characterized by hypercellular spindle areas surrounding progressively maturing woven and well-formed/trabecular bone. Molecular analysis revealed the presence of COL1A1-USP6 transcript After surgery, the patient’s course was regular, with no sequelae. Discussion MOC is a non-neoplastic ossification, appearing as a mass, involving soft tissue [7]. Commonly, it can be consequence of a muscular lesion. In fact, 80% of cases are located in the extremities [1, 3]. A neck location is quite infrequent particularly within paraspinal muscles, as it is in our case [1, 2, 4, 5]. In our case, patient experienced neck pain and swelling with movement impairment, consistent with other reported cases associated with MOC [1]. Differential diagnosis in a child or adolescent can be very difficult in the absence of known history of trauma, including musculoskeletal neoplasm or infection [1, 3]. In our case, patient and parents were asked several times if previous traumas occurred, but they always denied. Moreover, the absence of trauma was indirectly confirmed by the on-going COVID-19 pandemic lockdown between March and May 2020, in which all sport activities were suppressed and there was mandatory home confinement of the population. Also, poor response to conservative therapy was an additional negative predictor in the patient’s medical history. In this context, multimodality imaging can have a crucial role in suspecting MOC, particularly showing progressive evolution of the lesion [3]. In our case, US, CT, and MRI were performed after months from clinical onset, and by then, the lesion was in its last stage, with central “soft” part and prominent peripheral floccular calcifications, signs which were absent on the first cervical spine X-ray, which showed only a slight soft tissue swelling. Although these features and the evolution of the lesion may be suspect of MOC, the presence of contrast enhancement, peripheral tissue swelling, and irregular sclerotic bone reaction could also mimic a tumoral soft tissue or bone lesion. Synovial sarcoma, rhabdomyosarcoma or osteosarcoma, could be counted among the differential diagnosis; these lesions often show calcifications in the central portion [3]. In the initial phase of the disease, infections or hematomas should be excluded [4]. To rule out the nature of the lesion, in our case, a first approach with percutanous US-guided biopsy was chosen; but it was unsuccessful confirming some difficulties in diagnose MOC with this less invasive method [5]; so patient underwent surgery. In addition to histological features consistent with MOC, PCR analysis identified the transcript COL1A1-USP6, recently reported in some cases of MOC [6]. MOC is a benign lesion, and it can be managed conservatively; nevertheless if other lesions are suspected, a surgical approach should be considered. In fact, cases of recurrence after surgery or malignant transformation have been described [7]. Conclusion MOC could be a rare cause of neck mass associated with prolonged pain and progressive movement impairment over the weeks in child or adolescent. Since it can mimic soft tissue or bone tumors, a careful analysis of clinical history and radiological features is very important to manage these young patients; MOC should be suspected even in the absence of history of trauma or if the anatomical location of the lesion is unusual. Author Contribution All authors have contributed significantly to this paper and agree with the content of the manuscript. Data Availability Data available upon reasonable request. Code Availability Not applicable. Declarations Ethics Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Consent to Participate Not applicable. Consent for Publication The patient itself and his parents signed the consent for publication and the consent is held by our institution. Conflict of Interest The authors declare no competing interests. This article is part of the Topical Collection on Imaging Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Dubuisson A Lombard A Otto B Pseudomalignant myositis ossificans of the neck in a child: case report and review of the literature World Neurosurg 2019 130 95 97 10.1016/j.wneu.2019.06.165 31260851 2. Sarac S Sennaroglu L Hosal AS Sozeri B Myositis ossificans in the neck Eur Arch Otorhinolaryngol 1999 256 199 201 10.1007/s004050050139 10337511 3. Wang XL Malghem J Parizel PM Gielen JL Vanhoenacker F De Schepper AMA Pictorial essay. Myositis ossificans circumscripta JBR-BTR 2003 86 5 278 285 14651084 4. Kokkosis AA Balsam D Lee TK Schreiber ZJ Pediatric nontraumatic myositis ossificans of the neck Pediatr Radiol 2009 39 409 412 10.1007/s00247-009-1165-1 19229531 5. Mann SS Som PM Gumprecht JP The difficulties of diagnosing myositis ossificans circumscripta in the paraspinal muscles of a human immunodeficiency virus-positive man: magnetic resonance imaging and temporal computed tomographic findings Arch Otolaryngol Head Neck Surg 2000 126 6 785 788 10.1001/archotol.126.6.785 10864118 6. Flucke U Bekers EM Creytens D van Gorp JM COL1A1 is a fusionpartner of USP6 in myositis ossificans—FISH analysis of six cases Ann Diagn Pathol 2018 36 61 62 10.1016/j.anndiagpath.2018.06.009 29980413 7. Saussez S Blaivie C Lemortn M Chantrain G Non-traumatic myositis ossificans in the paraspinal muscles Eur Arch Otorhinolaryngol 2006 263 331 335 10.1007/s00405-005-0997-z 16133463
PMC009xxxxxx/PMC9005331.txt	==== Front Early Child Res Q Early Child Res Q Early Childhood Research Quarterly 0885-2006 0885-2006 Elsevier Inc. S0885-2006(22)00022-9 10.1016/j.ecresq.2022.03.001 Article Caregivers’ perceived changes in engaged time with preschool-aged children during COVID-19: Familial correlates and relations to children's learning behavior and emotional distress Zhang Xiao ⁎ Faculty of Education, The University of Hong Kong, Hong Kong ⁎ Corresponding author. 13 4 2022 3rd Quarter 2022 13 4 2022 60 319331 26 2 2021 5 1 2022 9 3 2022 © 2022 Elsevier Inc. All rights reserved. 2022 Elsevier Inc. Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. The COVID-19 pandemic and its resulting containment measures have forced many children and their caregivers around the world to spend unprecedented amounts of time at home. Based on a sample of 764 households with preschool-aged children in Wuhan, China, where the pandemic began, this study examined how primary caregivers perceived changes in the amount of time spent engaging with their children (i.e., engaged time) from the start of the pandemic and whether these changes were associated with children's learning behavior and emotional distress. The results showed that primary caregivers generally perceived increases in the amount of engaged time spent on indoor activities with their children but decreases in the amount of engaged time spent playing with their children outdoors. A bigger family size and greater loss of family income during the pandemic were associated with bigger perceived increases in caregivers’ engaged time spent on indoor activities, whilst a higher level of parental education was associated with bigger perceived decreases in engaged time spent playing with children outdoors. The family's poorer physical health and higher levels of chaos during the pandemic were related to smaller perceived increases in caregivers’ engaged time spent on educational activities. Finally, although bigger perceived increases in caregivers’ indoor engaged time (e.g., time spent on educational activities) were associated with higher levels of positive learning behavior and fewer symptoms of anxiety and withdrawal in the children, bigger perceived decreases in outdoor play time were associated with fewer symptoms of anxiety and withdrawal. These findings offer valuable insights into caregivers’ allocation of engaged time with their preschool-aged children during the COVID-19 pandemic. Keywords time use engaged time caregiver involvement learning behavior emotional distress COVID-19 Abbreviations COVID-19, coronavirus disease 2019 SARS, severe acute respiratory syndrome coronavirus MERS, Middle East respiratory syndrome ==== Body pmcIntroduction The COVID-19 pandemic and its resulting containment measures have forced many schools to close and many adults to work from home around the world. Unsurprisingly, the amount of time during which parents and caregivers are physically available to their children, often referred to as ‘accessible time’ in the literature (Milkie et al., 2015), has been found to be much higher during the pandemic than prior to it (Craig & Churchill, 2021). Yet little is known about how the pandemic has affected the amount of ‘engaged time’ caregivers have spent with their children on specific types of activities (e.g., playing, telling stories). It is important to study caregivers’ engaged time spent with their children because it has important implications for children's learning (Zhang et al., 2019) and emotional health (Fiorini & Keane, 2014; Milkie et al., 2015). Based on a sample of 764 households with preschool-aged children in Wuhan, China, where the COVID-19 pandemic started, this study investigated whether the pandemic led primary caregivers to change the amount of time spent engaging with their children. It also examined which family characteristics were associated with changes in caregivers’ engaged time and whether these changes had implications for preschool-aged children's learning behavior and emotional distress. I chose to study households in Wuhan because they had unique suffering during the COVID-19 pandemic. First, people in Wuhan knew little about how contagious and dangerous this coronavirus is and they experienced a lack of protective and medical equipment at the start of the pandemic (Kupferschmidt & Cohen, 2020). Second, Wuhan was hit the hardest by the pandemic and had the largest number of confirmed COVID-19 cases and deaths among all cities in China (50,340 confirmed cases and 3,869 deaths as of 31 May 2020; Hubei Health Commission, 2020). Third, Wuhan imposed one of the toughest lockdowns in the world (Chong et al., 2020). On 23 January 2020, the central government of China instituted a complete city lockdown in Wuhan, suspended all public transport, and prohibited residents of Wuhan from leaving the city (Chong et al., 2020). Although at first people were allowed out, from mid-February, almost all residents were prohibited from leaving their residential communities and all non-essential businesses were shut down (Chong et al., 2020). Consequently, caregivers in Wuhan might have had a unique pattern in their allocation of engaged time with their children during the pandemic. Understanding the Impact of COVID-19 on Children and Families via a Life Course Lens To understand how the COVID-19 pandemic and its resulting containment measures have affected caregivers and children in Wuhan, I took a life course perspective (Elder, 1998; Elder et al., 2003). The life course approach takes a historical perspective on individual and family development and looks at how macro-level sociohistorical events affect individual lives and family life. Emerging from studies on the impact of the Great Depression on children, life course theory posits that ‘historical forces shape the social trajectories of family, education, and work, and they in turn influence behavior and particular lines of development’ (Elder, 1998, p. 2). Research on sociohistorical events, such as natural, health, and socioeconomic disasters, suggests that they had both immediate and prolonged influences on human development (Elder, 1998; see a review in Benner & Mistry, 2020). For example, it has been found that school-age children who experienced previous pandemics (e.g., SARS in 2003, H1N1 in 2009) had high levels of emotional health problems, such as depression, anxiety, and posttraumatic stress disorders (PTSD; Ko, Yen, Yen, & Yang, 2006, Sprang & Silman, 2013). Life course theory emphasizes the interdependence of individual lives and takes the perspective that sociohistorical events affect interactions between family members (Bengtson & Allen, 1993; Elder et al., 2003). For example, the Great Recession, which started in December 2007 in the United States and was associated with high levels of unemployment, was found to cause an increase in the incidence rates of maternal spanking (Brooks-Gunn et al., 2013). Likewise, the 2014 outbreak of Ebola virus disease (EVD) was demonstrated to lead parents in Liberia to increase their harsh parenting and household conflict (Green et al., 2018). Life course theory also emphasizes that family characteristics (e.g., social and economic resources, physical health) differentially affect individual lives and family life during times of sociohistorical events, which may contribute to social stratification and rising inequality (Elder, 1998). In the Great Recession, for example, direct cash payments to low- and middle-income families in Australia helped mitigate the negative impact of the recession on the vulnerable families (UNICEF-IRC, 2014). In the 2014 outbreak of EVD in Liberia, parents from households that experienced EVD-related sickness or death reported more household conflict and were more likely to use harsh parenting than their counterparts from households that had no such experience (Green et al., 2018). Finally, according to life course theory, changes in proximal environments, such as family life, serve as the intervening mechanism through which a sociohistorical event affects individual lives (Elder, 1998; Elder et al., 2003). In a longitudinal study of children and parents who experienced the 2011 Missouri and Alabama tornadoes, Bountress et al. (2020) found that parental distress and parent-child conflict prospectively predicted children's PTSD. In a retrospective study of children and parents with SARS and H1N1 pandemic-related containment experiences (e.g., quarantine, isolation), Sprang and Silman (2013) showed that most parents diagnosed with PTSD had children who were similarly diagnosed. The COVID-19 pandemic is one of the deadliest pandemics in history (Feehan, 2021). As of 4 August 2021, it had caused over 200 million confirmed infections and over 4 million deaths (Johns Hopkins University, 2021). It has led to stringent containment measures, such as school closures, home quarantine, and work-from-home arrangements, in many countries (Zhang, 2021). From a life course perspective, the pandemic has affected many aspects of family life, including caregivers’ allocation of engaged time to their children. There is also good reason to believe that family circumstances, some of which may have already existed before the emergence of COVID-19 while others altered by COVID-19, would be associated with changes in caregivers’ allocation of engaged time to their children during the pandemic. Finally, theoretically, any changes in caregivers’ time allocation to their children that are caused by the pandemic should subsequently exert an influence on the children's learning behavior and emotional health. COVID-19 and Caregivers’ Engaged Time with Preschool-aged Children Prior research on large-scale sociohistorical events, such as earthquakes and financial crises, has documented that caregivers’ engaged time, defined as the amount of time caregivers spend engaging in specific types of activities (e.g., reading) with their children (Milkie et al., 2015), dramatically shifted during the events (e.g., Cano, 2019). For example, Kaneko and Noguchi (2020) found that the Great East Japan Earthquake led to perceived increases in parents’ engaged time spent with their young children in the most seriously affected regions. However, caregivers in families with children of various ages may respond to the same sociohistorical event in different ways. For instance, Bauer and Sonchak's (2017) study of the Great Recession showed that an increase in state-level unemployment rates led mothers to decrease and fathers to increase the amount of engaged time spent with children aged 0–4 years. In contrast, mothers’ and fathers’ engaged time spent with older children (i.e., 5-17 years of age) was invariant to macroeconomic fluctuations. The age-related differences in caregivers’ engaged time spent with their children might be related to the general declining pattern in the amount of time caregivers spend in childcare as their children age (Kendig & Bianchi, 2008). Although all children demand time, preschool-aged children are in greater need of caregivers’ care, help, teaching and supervision than older children (Milkie et al., 2004). It is thus possible that sociohistorical crises, such as the COVID-19 pandemic, have created additional challenges to the engaged time of caregivers whose children are too young to be left unsupervised. I reviewed the literature and did not locate any research on the impact of previous pandemics (e.g., SARS, H1N1, MERS) on caregivers’ engaged time. Nevertheless, one may speculate that the COVID-19 pandemic has led caregivers to increase the amount of engaged time spent with their preschool-aged children because preschools have closed and unemployment, home quarantine, and work-from-home arrangements have given caregivers more time at home. During the pandemic, caregivers may have experienced a ‘time windfall’ as a result of home quarantine, flexible working schedules or fewer working hours under the work-from-home arrangement, or changes in their employment status. The literature has shown that parents who worked fewer hours spent more time engaging with their preschool-aged children (Roeters et al., 2010), and unemployed parents were found to devote more engaged time to their preschool-aged children than employed parents (Kendig & Bianchi, 2008). However, the evidence is mixed regarding whether and how the COVID-19 pandemic has led caregivers to change the amount of engaged time devoted to their young children. In a study of households with children aged 0 to 11 years in Germany, both mothers and fathers reported that they spent more time on childrearing tasks during the pandemic than before it, with the largest increases observed in parents of children aged 3 to 5 (Kreyenfeld & Zinn, 2021). Similarly, mothers and fathers in the United Kingdom reported that they spent approximately twice as much time on childcare during the lockdown as they did before the pandemic, with the largest increases observed in parents of children aged 0 to 5 (Zhou et al., 2020). By contrast, in a study of households with children aged 0 to 12 in the United States, most parents reported that they engaged more often in reading, playing, and going on walks with their children during the pandemic than prior to it, but less often or the same amount in telling stories and singing songs (Lee et al., 2021). In a study of families with young children (age: M ± SD = 5.7 ± 2.0 years) in Canada, Carroll et al. (2020) found that most of the families reported increases in screen time (for 74% of the mothers, 61% of the fathers, and 87% of the children) and decreases in physical activity (for 59% of the mothers, 52% of the fathers, and 52% of the children). These mixed results may have various antecedents, such as differences in geographic locations and sociocultural backgrounds and in the type of activities caregivers engage in with their young children. For example, in the city of Wuhan, the 11 weeks of strict lockdown during the pandemic, when nearly all residents were prohibited from leaving their communities (Chong et al., 2020), likely led caregivers to decrease the amount of engaged time spent playing with their young children outdoors (i.e., outdoor engaged time) but increase the amount of time spent engaging in indoor activities such as reading and telling stories (i.e., indoor engaged time). In contrast, in countries and cities where no lockdown was imposed, caregivers might have spent more time engaging their preschool-aged children in both indoor and outdoor activities. By recruiting households in Wuhan, this study provides an opportunity to understand the impact of the COVID-19 pandemic on caregivers’ engaged time with their preschool-aged children in the original epicenter with a unique sociocultural background. The Role of Family Characteristics in Caregivers’ Engaged Time with Preschool-aged Children Although the COVID-19 pandemic might have led to a general increasing or decreasing pattern in the amount of time spent undertaking a certain activity, there might have been substantial variations in the changes in caregivers’ engaged time with their preschool-aged children. In the literature, a number of family characteristics, such as single parenthood (Kendig & Bianchi, 2008), lower family socioeconomic status (e.g., lower levels of family income and parental education; England & Strivastava, 2013), and larger family size (Esping-Andersen, 2009), have been found to limit the amount of engaged time caregivers spend with their children. These family characteristics may reflect insufficient or diluted family resources (e.g., economic and educational resources), preventing caregivers from investing sufficient engaged time with their children. I speculate that the family characteristics described above, which may have already existed before the emergence of the COVID-19 pandemic, would continue to be associated with a reduction in the engaged time that caregivers reported spending with their preschool-aged children during the pandemic. Moreover, I hypothesize that family members’ physical health problems, family income instability, and household chaos resulting from the pandemic may have been potential family barriers that constrained caregivers’ engaged time with their preschool-aged children. The pandemic has brought multiple health, financial, and psychosocial challenges to many aspects of family life. First, the large number of COVID-19 infections and the resulting overwhelmed health systems have made it challenging for people to maintain physical health (Chong et al., 2020). Second, as many countries have shut down their non-essential businesses to contain the pandemic, the elevated unemployment rate has forced many families into poverty or at least income instability (United Nations, 2020). Third, containment measures such as school closures and work-from-home arrangements have disrupted family routines, created challenges for caregivers to balance multiple duties (e.g., managing their jobs, chores, and childcare), and led to disorder and chaos among many families (Zhang, 2022). These family circumstances may have impeded caregivers’ engaged time spent with their preschool-aged children. For example, the demands on caregivers to care for sick family members (e.g., those infected with SARS-CoV-2) and organize their households may have resulted in little available time to engage with their preschool-aged children. Income instability may have impeded positive parenting practices and limited the amount of attention and engaged time that caregivers could give to their preschool-aged children (Monahan, 2020). Changes in Caregivers’ Engaged Time in Relation to Preschool-aged Children's Learning Behavior and Emotional Health Changes in caregivers’ allocation of engaged time with their preschool-aged children that are a result of the COVID-19 pandemic may subsequently influence the children's learning behavior and emotional health. Learning behavior, also known as approaches to learning, has been identified as a core component of preschool children's school readiness that predicts their reading and math achievement in the later years of schooling (Li-Grining et al., 2010). It refers to a broad set of skills and behaviors that reflect children's enthusiasm, curiosity, persistence, and engagement in learning settings (Hyson, 2008). Emotional health is an umbrella term covering a range of emotional states, from diagnosable emotional distress and disorders at one end of the spectrum to emotional well-being at the other. In this study, children's emotional health is indexed by four symptoms of emotional distress (Saylor et al., 1999), namely anxiety and withdrawal (i.e., displaying anxious and solitary behavior), fear (i.e., being afraid of something or someone), acting out (i.e., exhibiting out-of-control aggressive behavior in order to relieve tension), and COVID-19-related trauma (i.e., displaying intense emotional responses to COVID-19). Caregivers’ engaged time with their preschool-aged children is considered to be critically important because it facilitates children's learning through practice, instruction, and scaffolding (Barger et al., 2019). Moreover, the time caregivers spend engaging with their children often reflects their love and care for their children, which may protect children from developing emotional problems (Del Bono et al., 2016). The benefits of indoor and outdoor engaged time with caregivers for preschool-aged children's learning behavior and emotional health during normal situations have been documented (e.g., Fiorini & Keane, 2014; Fomby & Musick, 2018; Milkie et al., 2015). For example, in a study of 1,127 children aged 0 to 12 years in the United States, Hsin and Felfe (2014) showed that the amount of time mothers spent engaging with their children in educational and structured activities positively predicted children's later levels of cognitive ability (verbal and math scores) and positive behavior (persistence and social skills). Moreover, these effects were stronger for children younger than 6 years old than for children 6 years and older. In a study of 15,077 children interviewed at age 3 in the United Kingdom, Del Bono et al. (2016) demonstrated that the amount of engaged time mothers and children spent together on educational and recreational activities was positively associated with the children's cognitive skills and emotional health measured at age 7. The extent to which findings obtained from studies during normal situations can be generalized to a health crisis situation such as the COVID-19 pandemic is unknown. It has been reported that the pandemic and its associated lockdown measures have resulted in extensive caregiving and mental health burdens for parents and caregivers (Wu et al., 2020), which have been linked to negative relationships with their children (Russell et al., 2020). While more time indoors in confinement may mean intense caretaking, teaching, and potentially stressful activities, outdoor engagement in confinement may mean more relaxed time, at least in Euro-American countries such as the United States (Lee et al., 2021). However, other countries or regions, such as Wuhan, implemented home quarantine and/or isolation rules and discouraged or even prohibited outdoor activities during the pandemic (Chong et al., 2020). To achieve this policy objective, the mass media in China further conveyed to people a message about the risk of catching and spreading coronavirus outdoors. So outdoor engagement in confinement may also mean an elevated risk of contracting COVID-19. It is unclear whether or how changes in the amount of outdoor and indoor time caregivers spend engaging with their preschool-aged children in such a situation affect children's learning behavior and emotional health, especially for children who are experiencing school closures and cannot benefit from usual relationships with peers and teachers (e.g., Liu et al., 2020). The Present Study Drawing on life course theory (Elder, 1998; Elder et al., 2003), this study aimed to address three research questions: (1) Did primary caregivers in Wuhan perceive increases or decreases in the amount of time they spent engaging with their preschool-aged children indoors and outdoors from the start of the pandemic? (2) Which family characteristics were related to caregivers’ perceived changes in their engaged time with their preschool-aged children? (3) Were caregivers’ perceived changes in the amount of engaged time associated with preschool-aged children's learning behavior and emotional distress? To achieve this goal, I recruited a sample of preschool children aged 3 to 6 years and their primary caregivers (i.e., the person primarily responsible for the care and upbringing of the child) in Wuhan. Based on the literature reviewed above, we hypothesize that (1) primary caregivers from Wuhan would perceive increases in the amount of indoor time spent engaging with their preschool-aged children but decreases in the amount of outdoor engaged time from the start of the pandemic; (2) single parenthood, lower family socioeconomic status (parental education, parental occupation, and family income in 2019), larger family size, being in poorer physical health, having greater income loss and more household chaos during the pandemic would be associated with smaller increases in both indoor and outdoor engaged time perceived by caregivers; and (3) greater increases in indoor and outdoor engaged time would be associated with preschool-aged children's higher levels of positive learning behavior and fewer symptoms of emotional distress reported by caregivers. Because child age and sex have been found to relate to caregivers’ engaged time (Kendig & Bianchi, 2008; Sun et al., 2019) and preschool-aged children's learning behavior and emotional distress (Cai, 2015; Saylor et al., 1999), we included them as control variables in this study. The present study contributes to the literature in important ways. First, it focuses on a crisis situation. Crisis situations such as a pandemic can include life-altering events such as being quarantined, being seriously ill or hospitalised, losing a family member, losing a job, or experiencing financial hardship (Brown et al., 2020). As such, they might lead to a big change in family life (Prime et al., 2020). The COVID-19 pandemic provides a valuable opportunity to assess whether a health crisis situation can change caregivers’ allocation of engaged time to their preschool-aged children and whether these changes are associated with children's learning behavior and emotional distress. Findings of this study will bring interesting complements to the life course theory and potential challenges to the existing theories of caregiver engagement and time allocation. Second, I examined a large pool of familial correlates of caregivers’ time investments during the COVID-19 pandemic. It is crucial to identify families who are most likely to decrease the amount of engaged time spent with their preschool-aged children during a pandemic and provide them with specific intervention services aimed at preventing such a decrease. Third, this study was conducted in Wuhan, the epicenter of China's coronavirus outbreak. Life course theory's principle of time and place emphasizes that individual and family development are embedded in and shaped by the geographic places where a sociohistorical event is experienced (Elder et al., 2003). In other words, the geographic location where people are living can determine how much a sociohistorical event such as the COVID-19 pandemic affects individual lives and family life, including caregivers’ time allocation to engage their preschool-aged children in different activities. As discussed above, the experience in Wuhan was very unique, and investigating the extreme situations of preschool-aged children and their families in Wuhan is another unique contribution of this study. Method Participants and Procedure The participants in this study were the primary caregivers of 764 preschool-aged children (403 boys and 361 girls; age: M ± SD = 59.07 ± 12.28 months) recruited from nine preschools in the city of Wuhan, China. Using a purposive sampling technique, namely the typical case sampling method, I initially contacted 22 preschools located in the three major districts of Wuhan, and nine agreed to participate. The typical case sampling method, in which the typical cases of preschool children and their caregivers in Wuhan were identified, permits an easier generalization about the sample than random sampling, which was almost impossible during the pandemic. To meet the requirement of the typical case sampling method that researchers must have prior knowledge of the sample, three professors of early childhood in three different universities in Wuhan, who knew local preschools and residents very well, were responsible for recruiting schools. When recruiting schools, we considered the major attributes of the preschool programs (i.e., public vs private, school size) and the city population (i.e., income, education). According to Wuhan Statistical Bureau (2020), the sample was generally representative of the population of Wuhan in terms of household income (see online Appendix 1). In China, most preschools accommodate children aged 3 to 6, who are typically grouped into three levels by age (Zhang, 2011): junior class (3 to 4 years old), middle class (4 to 5 years old), and senior class (5 to 6 years old). In 2019, the preschool gross enrolment rate reached 83.4% in China (Ministry of Education of China, 2020a). In this study, 310 (40.6%), 197 (25.8%), and 257 (33.6%) children were from junior, middle, and senior preschool classes, respectively. All of them were in preschool before the pandemic. Most of the participating children were the only child in their families (63.4%), followed by the last born child (19.6%), the first born child (13.7%), and the middle child (2.4%). Most parents (97.1%) were married or cohabiting and 2.9% were separated, divorced, or single; 29.5% of the mothers and 32.3% of the fathers had a Bachelor's degree or above; 53.6% of the families had an annual income of RMB100,000 (US$14,290) or above in 2019. The average family size was 4.87 (SD = 1.75). Data were collected between mid-June and mid-July 2020, when Wuhan was gradually reopening after the 11-week strict lockdown, but all preschools remained closed and children remained out of school. Because the Ministry of Education of China (2020b) strictly prohibited preschools from conducting online teaching during COVID-19 school closures, I assumed that preschools did not conduct online teaching at the time of the study. Nevertheless, some children might have engaged in online learning assigned by their caregivers. A total of 1,200 consent forms and questionnaires were mailed to the nine principals, who then temporarily opened their schools to allow caregivers to participate. In two preschools, children's primary caregivers were asked to complete consent forms and questionnaires at the school within one hour. In the other seven preschools, primary caregivers were asked to pick up consent forms and questionnaires from the school, fill them out at home, and return the completed questionnaires to the school. Of the 790 questionnaires (65.8%) returned, 26 were invalid. The Human Research Ethics Committee of the authors’ university approved the data collection procedures. Most of the participating caregivers were the children's mothers (n = 517), followed by fathers (n = 198), grandparents (n = 20), and other relatives (n = 5); the remaining caregivers (n = 24) did not indicate their relationship to the child. Most of the caregivers had a full-time job at the time of the study (62.0%), followed by being housewives or househusbands (19.1%), having a part-time job (7.7%), being out of work and looking for work (3.1%) and being out of work but not looking for work (3.1%); the remaining caregivers did not provide information about their job status. I did not collection information about whether the caregivers worked at home or online at the time of the study. Measures Changes in Caregivers’ Engaged Time with Children Primary caregivers rated the change in the amount of time they spent engaging with their children in six activities (Table 1 ) from the start of the pandemic (January 20, 2020) using the Time Spent by Mothers and Others Scale (Singletary & Schmeer, 2020). A 5-point Likert scale was used (1 = increased a lot, 2 = increased a little, 3 = stayed the same, 4 = decreased a little, 5 = decreased a lot). I reverse-coded the items, so that higher scores indicated greater increases in caregivers’ engaged time with children. Principal factor analysis revealed the presence of two factors with eigenvalues exceeding 1: Items 2–6 belonged to Factor 1 (factor loadings ≥ .63), and Item 1 belonged to Factor 2 (factor loading = .97). Factor 2 was named ‘changes in outdoor engaged time’. Although Items 3–6 could occur both indoors and outdoors, Wuhan's complete city lockdown, in which nearly all people were prohibited from leaving their communities, is likely to have caused the activities described in these items to have been conducted mainly indoors. I thus named Factor 1 ‘changes in indoor engaged time’. For Factor 1, Cronbach's α was .81, and average scores were calculated.Table 1 Changes in primary caregivers’ engaged time with children. Table 1Item One-sample t (specified value = 3) Cohen's d M SD 1. Since 20 January, how has the amount of time you spend playing with your child outdoors changed? 1.87 1.28 −23.88* −.88 2. Since 20 January, how has the amount of time you spend playing with your child indoors changed? 4.04 1.17 23.93* .89 3. Since 20 January, how has the amount of time you spend demonstrating or showing your child how to do something changed? 3.38 1.01 9.96* .38 4. Since 20 January, how has the amount of time you spend reading or telling stories to the child changed? 3.58 1.07 14.66* .54 5. Since 20 January, how has the amount of time you spend helping the child with school work or learning tasks changed? 3.58 1.19 13.19* .49 6. Since 20 January, how has the amount of time you spend disciplining the child changed? 3.84 1.11 20.55* .76 Changes in indoor engaged time: Composite score of Items 2–6 3.68 .86 21.76* .79 Note. * p < 0.001. Children's Learning Behavior To measure children's learning behavior, primary caregivers completed the 6-item Chinese version of the Parent-report Preschool Children's Approaches to Learning Scale: Short Form (Cai, 2015) on a 4-point Likert scale (1 = almost never, 2 = sometimes, 3 = often, and 4 = very often). The scale measures children's initiative and curiosity, persistence, attentiveness, reflection and explanation, expressiveness, and creativity. Specifically, the six items ask whether the child takes the initiative to approach new things, finish a specific task persistently, focus on his or her work in a noisy environment, recognize mistakes and explain the reasons, express his or her own ideas and opinions in various ways, and make his or her own toys. Cronbach's α was .83 in Cai (2015) and .81 in this study. Composite scores were calculated. Higher scores indicated more positive learning behavior. Children's Emotional Distress To measure children's emotional distress, primary caregivers completed the 21-item Pediatric Emotional Distress Scale (PEDS) (Saylor et al., 1999). The PEDS measures four symptoms of emotional distress, namely anxiety and withdrawal (e.g., “seems sad and withdrawn”; α = .74), fear (e.g., “seems fearful without good reason”; α = .65), acting out (e.g., “acts aggressively”; α = .71), and trauma (e.g., “create games, stories, or pictures about COVID-19”; α = .61). A 4-point Likert scale was used (1 = almost never, 2 = sometimes, 3 = often, and 4 = very often). In Saylor et al.’s (1999) study, the Cronbach's αs of the scales ranged from .72 to .78. Composite scores were calculated separately for each symptom. Higher scores indicated more symptoms of emotional distress. Household Chaos Primary caregivers rated the chaos and confusion in their families from the start of the pandemic using the Chinese version of the Confusion, Hubbub, and Order Scale (CHAOS): Short Form (Matheny et al., 1995). The Chinese version contained five items rated on a 5-point Likert scale (1 = definitely untrue, 2 = somewhat untrue, 3 = neutral, 4 = somewhat true, and 5 = definitely true). It differed from the English version in that one item (‘Usually a television is turned on somewhere in our home’) was removed due to its low factor loading. A sample item in the Chinese version was “You can't hear yourself think in our home.” Cronbach's α was .79 in Matheny et al.’s (1995) study and .70 in this study. Composite scores were calculated. Higher scores indicated higher levels of household chaos. Physical Health of the Family Primary caregivers rated their own physical health status and that of their child using two items (‘Since 20 January 2020 [the start of the pandemic], how would you rate your overall physical health?’ and ‘Since 20 January 2020, how would you rate your child's overall physical health?’) on a 5-point Likert scale (1 = excellent, 2 = very good, 3 = good, 4 = fair, and 5 = poor). Composite scores were calculated as an indicator of the physical health of the family. Cronbach's α was .91. We reverse-coded the items, so that higher scores indicated better physical health. Economic Instability Primary caregivers rated their household economic stability on one item (‘How has your monthly household income changed since the start of the pandemic?’) using a 5-point Likert scale (1 = increased a lot, 2 = increased a little, 3 = stayed the same, 4 = decreased a little, 5 = decreased a lot). I reverse-coded the item, so that higher scores indicated greater increases in household income and lower scores indicated greater income reductions. Demographic Variables Primary caregivers also reported a large pool of demographic variables, including the child's sex and birth date, family size (i.e., the number of people living in the home), their own marital status, their annual family income in 2019, and the education level and occupation of the child's mother and father (see online Appendix 1 for the numeric codes of family characteristics). Because strong correlations were observed between paternal and maternal education (r = 0.73, p < 0.001) and between paternal and maternal occupation (r = 0.47, p < 0.001), the highest education level and the most prestigious occupation in the household were used as indicators of education and occupation, respectively. This decision made it possible to use education and occupation data for almost all family structures, thereby reducing the amount of missing data. Data Analysis To examine whether primary caregivers perceived increases or decreases in the amount of time spent engaging with their preschool-aged children from the start of the pandemic (RQ1), I computed the average score for each item of the Time Spent by Mothers and Others Scale and the average score for its Factor 1 (i.e., changes in indoor engaged time) and then compared these scores with a specified value of 3 (‘stayed the same’) using a series of one-sample t tests. A score significantly greater than 3 would indicate perceived increases in engaged time spent on a certain activity, whereas a score significantly smaller than 3 would indicate perceived decreases in engaged time. Cohen's ds were calculated to examine whether the increase/decrease was large or not. To examine whether perceived changes in caregivers’ engaged time with preschool-aged children predicted children's learning behavior and emotional distress (RQ3), and whether they were predicted by family variables (RQ2), I calculated the zero-order correlations among these variables and estimated a path model. In the model, perceived changes in caregivers’ indoor engaged time and those in outdoor engaged time were regressed on all of the demographic variables (i.e., child's age and sex, caregivers’ marital status, family size, parental education, parental occupation, and family income in 2019) and pandemic-specific family variables (i.e., financial instability, physical health, and household chaos during the pandemic). Children's learning behavior and four aspects of emotional distress were regressed on perceived changes in caregivers’ indoor engaged time and outdoor engaged time as well as on all of the demographic and pandemic-specific family variables. Finally, I conducted a more nuanced, item-by-item analysis of the family correlates and developmental implications of perceived changes in caregivers’ engaged time in individual activities. To do this, I reran the path model described above by replacing the composite of perceived changes in caregivers’ indoor engaged time with the score on each of the five indoor activities (Items 2−6). The data analyses were performed using SPSS 26.0 and Mplus 7.0. Results Caregivers’ Perceived Changes in Engaged Time Spent with Their Children (RQ1) As shown in Table 1, results from the one-sample t tests indicate that primary caregivers generally perceived increases in the amount of indoor engaged time with their preschool-aged children but decreases in the amount of outdoor engaged time from the start of the pandemic. As indicated by the effect size (Cohen's d), the largest perceived increase in caregivers’ engaged time was in playing with their child indoors, followed by disciplining the child. The effect size for the perceived decrease in caregivers’ outdoor engaged time (i.e., playing with the child outdoors) was also large. The Relation of Family Characteristics with Caregivers’ Perceived Changes in Engaged Time (RQ2) Table 2 presents the descriptive statistics and intercorrelations among the study variables. As shown in Table 2, perceived changes in indoor engaged time were positively associated with parental education and annual income in 2019. Perceived changes in outdoor engaged time were negatively associated with parental education.Table 2 Means, standard deviations, and intercorrelations between study variables. Table 2Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. Child's age in months ‒ 2. Child's sex .05 ‒ 3. Primary caregivers’ marital status .05 ‒.03 ‒ 4. Family size .11 .06 ‒.14 ‒ 5. Parental education ‒.27 ‒.13 .05 ‒.16 ‒ 6. Parental occupation ‒.16 ‒.06 ‒.04 ‒.06 .49 ‒ 7. Annual household income in 2019 ‒.28 ‒.06 ‒.04 ‒.12 .38 .39 ‒ 8. Changes in monthly household income during the pandemic ‒.01 ‒.00 ‒.01 ‒.05 .22 .18 .09 ‒ 9. Physical health of family during the pandemic ‒.00 ‒.02 ‒.00 ‒.04 ‒.06 ‒.01 .04 .11 ‒ 10. Household chaos during the pandemic .00 .01 .09 .11 ‒.04 ‒.07 ‒.10 ‒.16 ‒.13 ‒ 11. Changes in indoor engaged time ‒.09 ‒.11 ‒.01 .05 .07 .04 .09 ‒.06 .05 ‒.05 ‒ 12. Changes in outdoor engaged time ‒.05 .02 .01 .04 ‒.12 ‒.05 ‒.03 ‒.03 .05 ‒.05 .11 ‒ 13. Learning behavior ‒.01 ‒.06 .02 ‒.02 .13 .09 .14 .09 .13 ‒.23 .19 .02 ‒ 14. Anxiety/withdrawal .00 .04 .06 .06 ‒.11 ‒.12 ‒.13 ‒.07 ‒.11 .14 ‒.12 .09 ‒.02 ‒ 15. Fear ‒.13 ‒.01 .03 ‒.05 .06 .01 ‒.01 ‒.08 ‒.12 .17 ‒.02 ‒.05 .01 .55 ‒ 16. Acting out ‒.03 .02 .08 ‒.01 .04 ‒.05 ‒.03 ‒.05 ‒.02 .16 ‒.03 ‒.06 .10 .50 .51 ‒ 17. COVID-19-related trauma .05 .04 ‒.02 ‒.03 ‒.02 ‒.03 ‒.05 ‒.03 .05 ‒.01 ‒.03 ‒.02 .20 .38 .23 .30 ‒ M 59.06 .53 .03 4.87 4.03 3.06 11.36 2.03 8.99 12.43 3.68 1.87 14.27 9.69 8.82 11.06 6.80 SD 12.27 .50 .17 1.75 1.08 1.09 6.36 .94 1.43 3.83 .86 1.28 2.97 2.52 2.33 2.45 1.72 Note. N = 498–764. Sex: 0 = girl, 1 = boy. Marital status: 0 = couple, 1 = single. * p < 0.05, if r ≥ 0.07; p < 0.01, if r ≥ 0.10; * p < 0.001, if r ≥ 0.13. Table 3 and Fig. 1 show the path model described above. As shown in Table 3 and Fig. 1, perceived changes in caregivers’ indoor engaged time with their preschool-aged children were negatively predicted by changes in monthly household income during the pandemic, and positively predicted by family size. Perceived changes in caregivers’ outdoor engaged time were negatively predicted by parental education. That is, primary caregivers were more likely to perceive increases in the amount of time spent engaging with their preschool-aged children indoors when (1) there were more people living in the household and (3) they experienced a greater loss of family income during the pandemic. Primary caregivers were more likely to perceive decreases in the amount of time spent playing with their children outdoors when parental education level was higher. The model explained 4.1% (p = 0.006) and 3.4% (p = 0.013) of the variance in caregivers’ perceived changes in indoor and outdoor engaged time with children, respectively.Table 3 Standardized path coefficients in the path model predicting child outcomes and changes in primary caregivers’ engaged time with children. Table 3Predictor Children's Learning Behavior Children's Emotional Distress Changes in Indoor Engaged Time Changes in Outdoor Engaged Time Anxiety/Withdrawal Fear Acting Out COVID-19-related Trauma β SE β SE Β SE β SE β SE β SE β SE Child's age in months .010 .037 ‒.041 .038 ‒.130*** .038 ‒.022 .039 .038 .040 ‒.075 .039 ‒.095* .039 Child's sex .047 .035 ‒.006 .036 .015 .036 .010 .037 .051 .037 ‒.092* .040 .008 .041 Primary caregivers’ marital status .051 .036 .047 .036 .009 .036 .056 .037 ‒.028 .037 .004 .038 .042 .038 Family size .026 .041 .048 .042 ‒.046 .042 .003 .042 ‒.025 .043 .087* .042 .059 .042 Parental education .079 .045 ‒.056 .046 .046 .046 .081 .047 .012 .048 .062 .046 ‒.148*** .046 Parental occupation .006 .044 ‒.052 .044 ‒.006 .044 ‒.071 .044 ‒.014 .045 ‒.008 .045 .006 .045 Annual household income in 2019 .078 .048 ‒.049 .047 ‒.028 .048 ‒.003 .047 ‒.028 .051 .049 .048 .018 .048 Changes in monthly household income during the pandemic .038 .040 ‒.021 .040 ‒.052 .040 ‒.033 .040 ‒.029 .042 ‒.100* .040 ‒.028 .040 Physical health of family during the pandemic .116* .036 ‒.095 .036 ‒.083* .037 .015 .037 .059 .038 .056 .038 .040 .038 Household chaos during the pandemic ‒.203*** .037 .098* .039 .146* .038 .144* .039 ‒.012 .040 ‒.071 .039 ‒.061 .039 Changes in indoor engaged time .164* .036 ‒.120* .038 ‒.021 .038 ‒.023 .038 ‒.020 .039 ‒ ‒ ‒ ‒ Changes in outdoor engaged time ‒.003 .037 .096* .038 ‒.032 .038 ‒.051 .038 ‒.016 .039 ‒ ‒ ‒ ‒ Note. N = 764. Sex: 0 = girl, 1 = boy. Marital status: 0 = couple, 1 = single. * p < 0.05; p < 0.01; * p < 0.001. Fig. 1 The Path Model with a Completely Standardized Solution Predicting Child Outcomes and Changes in Caregivers’ Indoor and Outdoor Engaged Time from the Start of the Pandemic. Note. N = 764. For ease of presentation, predictive paths that were not significant at p < 0.05 are not shown. * p < 0.05, ⁎⁎p < 0.01, ⁎⁎⁎p < 0.001. Fig 1 Finally, the item-by-item analyses of individual indoor activities showed that perceived changes in indoor play time were negatively predicted by changes in monthly household income during the pandemic (β = −.118, p = 0.004). Perceived changes in demonstration time were not associated with any demographic or familial variables. Perceived changes in reading and story time were positively predicted by family size (β = .088, p = 0.03) and negatively predicted by household chaos (β = −.075, p < 0.05). Perceived changes in school-work or learning time were positively predicted by family size (β = .089, p = 0.03) and physical health (β = .092, p = 0.01), and negatively predicted by changes in household income during the pandemic (β = −.115, p = 0.004). Perceived changes in disciplining time were negatively predicted by changes in household income during the pandemic (β = −.133, p = 0.001). The Relation between Caregivers’ Perceived Changes in Engaged Time and Children's Learning Behavior and Emotional Distress (RQ3) As shown in Table 2, Table 3, and Fig. 1, perceived changes in caregivers’ indoor engaged time with preschool-aged children during the pandemic were positively associated with children's learning behavior and negatively associated with their symptoms of anxiety and withdrawal, but perceived changes in caregivers’ outdoor engaged time with children during the pandemic were positively associated with children's symptoms of anxiety and withdrawal. In other words, primary caregivers who perceived increases in the amount of time spent engaging with their children indoors to a greater extent during the pandemic had children with more positive learning behavior and fewer symptoms of anxiety and withdrawal. In contrast, primary caregivers who perceived decreases in the amount of time spent playing with their children outdoors to a greater extent during the pandemic had children with fewer symptoms of anxiety and withdrawal. No relation was found between perceived changes in caregivers’ indoor or outdoor engaged time with children and children's symptoms of fear, acting out, and COVID-19-related trauma. The model explained 12.4% (p < 0.001), 7.3% (p < 0.001), 6.1% (p = 0.001), 4.0% (p = 0.006), and 1.1% (p = 0.163) of the variance in learning behavior, anxiety/withdrawal, fear, acting out, and COVID-19-related trauma of preschool-aged children, respectively. Finally, the item-by-item analyses of individual indoor activities showed that perceived changes in reading and story time were positively associated with children's learning behavior (β = .135, p < 0.001) and negatively associated with their symptoms of anxiety and withdrawal (β = −.133, p < 0.001), fear (β = −.100, p = 0.007) but not acting out or COVID-related trauma. Perceived changes in school-work and learning time were positively associated with children's learning behavior (β = .137, p < 0.001) and negatively associated with their symptoms of anxiety and withdrawal (β = −.102, p = 0.007). Perceived changes in disciplining time were positively associated with children's learning behavior (β = .145, p < 0.001) and negatively associated with their symptoms of anxiety and withdrawal (β = −.090, p = 0.02). Perceived changes in demonstration time were positively associated with children's learning behavior (β = .128, p < 0.001). Finally, perceived changes in indoor play time were not associated with any child outcomes. Discussion The present study set out to examine how primary caregivers in Wuhan perceived the changes in their allocation of engaged time with their preschool-aged children from the start of the COVID-19 pandemic and how these changes were associated with a large pool of family characteristics and the learning behavior and emotional distress of their children. It has yielded important findings about caregivers’ time investments during the pandemic. Caregivers’ Perceived Changes in Engaged Time with Preschool-aged Children during the COVID-19 Pandemic Caregivers usually spend larger amounts of time in helping, supervising, teaching, and taking care of younger children (Milkie et al., 2004). Prior research on sociohistorical crises (Bauer & Sonchak, 2017), including the COVID-19 pandemic (Kreyenfeld & Zinn, 2021; Zhou et al., 2020), has documented that they led to greater changes in caregivers’ engaged time with preschool-aged children than older children. It is thus important to study the changes in caregivers’ allocation of engaged time with preschool-aged children during the pandemic. The finding that primary caregivers in Wuhan generally perceived increases in the amount of time spent engaging with their preschool-aged children indoors from the start of the COVID-19 pandemic is consistent with the study hypothesis. It also adds to evidence from studies in Germany (Kreyenfeld & Zinn, 2021) and the United States (Lee et al., 2021). The finding that the caregivers in this sample generally perceived decreases in their outdoor play time with their preschool-aged children is also consistent with the hypothesis but is contrary to the finding of Lee et al. (2021) that parents in the United States engaged more often in going on walks with their children from the start of the pandemic. These findings may be explained by the containment measures implemented in Wuhan. The city of Wuhan was placed under strict lockdown during the pandemic. Similar to other cities around the world, Wuhan's lockdown involved such containment measures as school closures and work-from-home arrangements. These measures forced all Wuhan children and their caregivers to stay at home for 11 weeks between 23 January and 7 April 2020. Even after the lockdown, preschools in Wuhan, including those in this study, did not reopen until September 2020. Although school closures disrupted education and care for children of various ages, they might have been especially challenging for preschool-aged children, who typically do not have strong self-care abilities and often have trouble concentrating and sustaining attention on task (Zhang, 2021). Unlike primary and secondary schools, moreover, preschools were strictly prohibited by the Ministry of Education of China (2020b) from conducting online teaching during school closures. Consequently, nearly all responsibilities of educating and caring for preschool-aged children shifted to their caregivers (Zhang, 2022). These circumstances may explain why the primary caregivers who participated in this study generally perceived increases in the amount of indoor engaged time spent playing, disciplining, and reading to their preschool-aged children. Finally, Wuhan's lockdown was unique in that almost all residents were prohibited from leaving their residential communities, which may explain why primary caregivers generally perceived decreases in the engaged time spent playing with their preschool-aged children outdoors. Among the six indoor and outdoor activities measured in this study, the perceived decrease in engaged time spent on outdoor play and the perceived increase in engaged time spent on indoor play from the start of the pandemic were relatively large. The finding indicates that the containment measures implemented in Wuhan might have largely led caregivers to play with their preschool-aged children indoors rather than outdoors. Notably, however, this study showed that caregivers’ perceived increases in indoor play time were not associated with any positive child outcomes. Furthermore, although perceived increases in reading and story time and school-work and learning time were associated with children's positive learning behavior and few symptoms of emotional distress (e.g., anxiety and withdrawal and fear), these increases were relatively small. Together, these findings suggest that Wuhan caregivers are in need of assistance in allocating their indoor engaged time with preschool-aged children. I shall discuss this further below. The Role of Family Characteristics in Caregivers’ Perceived Changes in Engaged Time with Preschool-aged Children Consistent with the study's hypotheses, the family's poorer physical health and higher levels of chaos during the pandemic were related to a lower likelihood of perceiving increases in caregivers’ engaged time spent educating their preschool-aged children. Specifically, caregivers living in households with higher levels of chaos during the pandemic were less likely to perceive increases in engaged time spent reading or telling stories to their children, and caregivers living in families with poorer physical health during the pandemic were less likely to perceive increases in engaged time spent helping children with school work or learning tasks. Both of these family circumstances might have created competing demands on caregivers’ time during the pandemic, such as the demands for household organization and healthcare (Cheng et al., 2021). This may have left caregivers with little time for educating their children. Moreover, if a family member was infected with SARS-CoV-2 or had COVID-19-like symptoms, it would have been natural for the entire family to keep their distance from each other and reduce caregivers’ engaged time to avoid infection. These findings are cause for alarm, given that the COVID-19 pandemic has caused poor physical health among many people and pushed many families into chaos and disorganization (Chong et al., 2020; Zhang, 2022). Surprisingly, the observed associations between the other family characteristics and caregivers’ perceived changes in engaged time with preschool-aged children were inconsistent with the hypotheses. First, the hypothesis that a large family size and family income instability during the pandemic would be potential barriers to caregivers spending time engaging with their preschool-aged children was not supported. Instead, these family characteristics were found to be associated with caregivers’ perceived increases in indoor engaged time with their preschool-aged children. Specifically, a greater loss of household income during the pandemic was associated with bigger increases in caregivers’ engaged time spent disciplining the child, playing indoors with the child, and helping the child with school work or learning tasks, and a larger family size was associated with bigger increases in reading and story time and school-work and learning time. It is worth noting that Wuhan's lockdown took place during Chinese New Year, when family members get together to celebrate. This is reflected in the family size observed in this study. Due to the newly implemented two-child policy that replaced the one-child policy in 2016, the vast majority of nuclear families in China typically have a size equal to or fewer than four people (Xinhua News Agency, 2015). The family size of 4.87 indicates that many of the participating children might have been living with relatives beyond their parents and siblings, such as their grandparents, aunts and uncles, at the time of the study. Driven by the Chinese ideology of collectivism, members of an extended family often work together for the benefits of the whole family (Cong & Silverstein, 2012). In particular, intergenerational families with preschool-aged children in China often take it for granted that grandparents should share the responsibility for household chores and childcare (Sun et al., 2019). I thus speculate that grandparents and other relatives in extended families in Wuhan might have been greatly involved in sharing household tasks such as doing chores and grocery shopping, thereby giving primary caregivers more time to engage in educational activities (i.e., reading, telling stories, and helping with school work) with their preschool-aged children. In contrast, primary caregivers who lived in a smaller, nuclear family without any relatives had to manage multiple duties (e.g., chores, jobs, and childcare) during the pandemic, which might have created competing demands on their time (Cheng et al., 2021) and left them with little available time for educating their children. The observed relation between loss of family income and caregivers’ perceived increases in indoor engaged time might be due to the possibility that caregivers who suffered income loss either lost their jobs or worked fewer hours during the pandemic than before it. In the literature, becoming unemployed and working fewer hours have been associated with an increase in caregivers’ time spent engaging with their preschool-aged children (Kendig & Bianchi, 2008; Roeters et al., 2010). Second, the hypothesis that higher levels of parental education would be associated with caregivers’ perceived increases in engaged time spent with their preschool-aged children during the pandemic was not supported. Rather, the results showed that higher levels of parental education were associated with greater decreases in caregivers’ outdoor play time with their preschool-aged children. In a recent study, the participants with higher levels of education were found to perceive a higher risk of COVID-19 (Wong et al., 2020). Intuitively, caregivers who had a higher education level may have been less likely to go outdoors with their preschool-aged children to avoid potential infection risks. Third, marital status, parental occupation, and family income in 2019 were not associated with caregivers’ perceived changes in engaged time with their preschool-aged children. This finding suggests that these family characteristics did not produce inequalities or individual differences in primary caregivers’ allocation of engaged time to their children. Caregivers’ Perceived Changes in Engaged Time in Relation to Preschool-aged Children's Learning Behavior and Emotional Distress Consistent with the hypothesis, I found important benefits of caregivers’ increased indoor engaged time with a preschool-aged child for the child's learning behavior and emotional health during the COVID-19 pandemic. Specifically, greater perceived increases in caregivers’ time spent engaging with their preschool-aged children indoors were associated with higher levels of positive learning behavior and fewer symptoms of anxiety and withdrawal in the children. Spending quality time together at home provides opportunities for communicative exchanges and interpersonal interactions that facilitate positive learning behavior (Hayes et al., 2018). During these exchanges and interactions, caregivers may convey their values or expectations regarding children's learning and reward children's positive learning behavior, and children may model their caregivers’ positive learning behavior (Pomerantz & Grolnick, 2017). The observed benefit of indoor engaged time for children's anxiety and withdrawal might be attributed to the nurturing and supportive context that caregivers’ involvement provides for their children's socioemotional development (Barger et al., 2019). Among the five indoor activities measured in this study, caregivers’ perceived increases in engaged time in indoor play were not associated with any child outcomes. In contrast, perceived increases in engaged time in the other four indoor activities, especially literacy and educational activities, were associated with preschool-aged children's positive learning behavior and few symptoms of emotional distress. For instance, caregivers’ perceived increases in reading and story time were associated with children's positive learning behavior and low levels of anxiety/withdrawal and fear. In prior studies of non-crisis situations, caregivers’ engaged time in literacy and educational activities was found to be beneficial for children's learning behavior (Hsin & Felfe, 2014) and emotional health (Del Bono et al., 2016). The present study adds to the literature by revealing similar relations in a public health crisis. These findings are encouraging, given the documented negative impact of the pandemic on the learning behavior and emotional health of children across a wide age range (Rosen et al., 2020). Inconsistent with the hypothesis, however, greater perceived decreases in caregivers’ outdoor play time with their preschool-aged children during the COVID-19 pandemic were associated with fewer symptoms of anxiety and withdrawal in the children. This finding might be due to children's concerns about possible infection when going outdoors with their caregivers. Preschool-aged children might have acquired knowledge about the coronavirus infection through various channels, such as mass media and caregivers’ and teachers’ instructions. I speculate that these channels might have conveyed to preschool children a message about the risk of catching this coronavirus outdoors, especially considering the fact that Wuhan's lockdown was very restrictive in that almost all people were prohibited from leaving their residential communities (Chong et al., 2020). As such, playing outdoors is likely to have engendered preschool children's insecurity and worries about possible infection, which might explain why more outdoor play time was found to be associated with more symptoms of anxiety and withdrawal in these children. No relations were found between caregivers’ perceived changes in indoor and outdoor engaged time and their preschool-aged children's symptoms of acting out and COVID-19-related trauma. Arguably, these symptoms, especially trauma, are more severe than symptoms of anxiety, withdrawal, and fear. It is likely that quality family time could mitigate the pandemic's impact on mild symptoms of emotional distress but not severe symptoms. Limitations and Future Directions Although this study represents an important endeavor to examine caregivers’ allocation of engaged time to their preschool-aged children during the COVID-19 pandemic, it has several limitations. First, all of the study constructs were reported by children's primary caregivers. This might not only have involved social desirability issues but also might have led to shared method variance, both of which are likely to have biased the results. It would be useful for future studies to use alternative data collection methods (e.g., child assessments, observation) or a multi-informant approach (e.g., involving both primary and secondary caregivers) to replicate the present findings. Second, the measure of changes in caregivers’ engaged time was a newly developed scale. Given the timing of this study and the context of the COVID-19 pandemic, I was not able to collect more validation data. Also, the measure of changes in outdoor engaged time had only one item (i.e., playing outdoors), and its psychometric properties might not be good. However, most children and their caregivers in the city of Wuhan live in condominiums, and the 11-week tough lockdown prohibited them from leaving their residential communities and severely restricted the type of outdoor activities that they were able to engage in during the pandemic. Therefore, it might not be reasonable to measure children's engagement in other outdoor activities such as hiking, playing ball, or riding bikes. Third, the sample was drawn from Wuhan. As discussed above, Wuhan's situation was unique during the COVID-19 pandemic. The extent to which the present results are unique to samples in Wuhan or universal to all samples in China or even around the world is unclear. In particular, considering that Wuhan's lockdown was much tougher than that imposed in most, if not all, other countries and cities around the world (Chong et al., 2020), caution should be applied in generalizing the results, especially those related to caregivers’ outdoor engaged time with children, to different populations. Conditions during the COVID-19 pandemic have not only differed across cities but have also varied among countries and between rural and urban areas. More studies are needed to reveal how caregivers’ allocation of engaged time to their children has been affected by the pandemic and is related to children's learning behavior and emotional health in different societies. Last, this study was cross-sectional in nature and did not allow to examine the prospective or causal relations between the study variables. Moreover, caregivers were asked to retrospectively report the change in the engaged time allocated to their children from the start of the pandemic, which may have introduced confounding factors such as memory distortion. It would be valuable for a longitudinal design to be used in future pandemics to assess the real change before and during the pandemic and the prospective relations between the variables. Theoretical and Practical Implications The findings of this study have important theoretical implications regarding caregivers’ time allocation to their children during a public health crisis. First, this study showed that primary caregivers in Wuhan altered their patterns of time allocation to their preschool-aged children in response to the COVID-19 pandemic and resulting containment measures. This finding joins evidence from studies on COVID-19 in other countries (Kreyenfeld and Zinn, 2021, Lee et al., 2021) and provides strong evidence for the notion of life course theory that sociohistorical events affect family life and the interaction between family members (Bengtson & Allen, 1993; Elder, 1998). More important, the finding that caregivers in Wuhan generally perceived decreases in their outdoor play time with their children is in sharp contrast to Lee et al.’s (2021) finding in the United States. Arguably, the difference might be caused by the differential containment policies implemented in Wuhan versus the United States. Together, these findings highlight the importance of life course theory's principle of time and place (Elder et al., 2003), which emphasizes that the impact of a sociohistorical event on individual and family development varies depending on where the event is experienced. Notably, a place is not merely a location, but ‘space filled up by people, practice, objects, and representations’ (Gieryn, 2000, p. 465). Wuhan, the original epicentre of the COVID-19 pandemic, had many people infected with and dying from COVID-19, experienced shortages of medical staff and equipment, and imposed stringent lockdown measures. I speculate that all of these and other social characteristics are the root of the forces that ultimately led to the observed time allocation patterns of caregivers in Wuhan. Second, the present results showed that family-level characteristics were important sources of individual differences in the perceived change of caregivers’ engaged time with their preschool-aged children during the COVID-19 pandemic. Specifically, a large family size and diminished family income during the pandemic were associated with caregivers’ perceived increases in indoor engaged time with preschool-aged children, and a high level of parental education was associated with caregivers’ perceived decreases in outdoor play time with children. Moreover, higher levels of household chaos and family members’ poorer physical health during the pandemic were related to a lower likelihood of perceiving increases in caregivers’ engaged time in literacy and educational activities. These findings are also consistent with life course theory and suggest that the pandemic differentially affected the time allocation patterns of caregivers from different family backgrounds. Third, this study suggests that the findings of studies on the developmental benefits of caregivers’ indoor engaged time during normal situations can be generalized to the COVID-19 pandemic situation, whilst those related to outdoor engaged time cannot, at least in the context of Wuhan. It is likely that preschool-aged children perceived a high risk of infection when engaging in outdoor activities during the pandemic. If this is the case, time spent engaging in outdoor activities with caregivers is unlikely to have produced positive effects on preschool-aged children's emotional health; it may instead have harmed these children's emotional health. Interestingly, the present findings suggest that a health crisis situation such as the COVID-19 pandemic is potentially beneficial for caregivers’ allocation of engaged time with their preschool-age children. Specifically, the pandemic led caregivers in Wuhan to generally increase the indoor time spent engaging with their preschool-aged children and this additional quality time with caregivers might have mitigated some of the adverse impacts of the pandemic on children's learning behavior and emotional health. The findings resemble those of recent studies on the effects of the Great Recession (Cano, 2019), the East Japan Earthquake (Kaneko & Noguchi, 2020), and the COVID-19 pandemic (Kreyenfeld & Zinn, 2021; Zhou et al., 2020) on caregivers’ time allocation to young children. Together, these findings seem to suggest that some sociohistorical crises can lead to caregivers’ positive time investment in young children's development. Moreover, although it appears to be a misfortune that caregivers in Wuhan had no choice but to follow the lockdown rules and decrease outdoor play time with their children, this decrease might in fact have been beneficial to preschool-aged children's emotional health. Nevertheless, we do not know whether more overwhelming disasters, such as wars, can be equally beneficial for caregivers’ time allocation or not, due to the lack of evidence from previous studies regarding this topic. Finally, it should be noted that even milder sociohistorical crises can have adverse effects on other aspects of caregiver-child interactions, such as increased maternal spanking (Brooks-Gunn et al., 2013) and decreased autonomy support (Bülow et al., 2021). The findings of this study also have important implications for enhancing preschool-aged children's learning behavior, reducing their emotional distress, and helping them to navigate the COVID-19 pandemic. According to the findings, it might be wise for parents and caregivers of preschool-aged children to increase their at-home time spent engaging with children in important activities such as reading, telling stories, and doing school-related work during the pandemic. Preschool children may reap the most benefits and demonstrate the most positive learning behavior and emotional health when the time spent on these activities is increased. Additionally, because children whose caregivers perceived greater decreases in outdoor play time exhibited fewer symptoms of anxiety and withdrawal, it may be valuable for parents and caregivers of preschool-aged children who worry about coronavirus infections when going outdoors to reduce outdoor time. Notably, caregivers who live in smaller families and have little support from extended families and who live in families with higher levels of chaos and poorer physical health may have less time available for engaging with their preschool-aged children during the pandemic, and caregivers with a lower level of education tend to spend more time outdoors with their children. Policymakers and social workers should be made aware that these caregivers need more support to learn how to best allocate their engaged time to improve preschool-aged children's learning behavior and emotional health. Authorship contribution statement Xiao Zhang: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Formal analysis, Writing - original draft, Writing - revision Appendix Table S1 Table S1 Demographics of the Sample Participants Table S1Variable M (SD) % Child characteristics Age (months) 59.07 (12.28) Year of study in preschoolFirst year (junior class)Second year (middle class)Third year (senior class) 40.6%25.8%33.6% Sex (numeric code) Boy (1) 52.7% Girl (0) 47.3% Family characteristics Marital status (numeric code; n = 749) Couple (0)Single (1) 95.2%2.9% Annual income in 2019 (numeric code; n = 575) <10,000 RMB or $1,429 (1) 3.7% 10,000∼19,999 RMB or $2,858 (2) 3.9% 20,000∼29,999 RMB or $4,287 (3) 2.2% 30,000∼39,999 RMB or $ 5,716 (4) 3.3% 40,000∼49,999 RMB or $7,145 (5) 2.7% 50,000∼59,999 RMB or $8,574 (6) 3.8% 60,000∼69.999 RMB or $10,003 (7) 3.0% 70,000∼79,999 RMB or $11,432 (8) 3.1% 80,000∼89,999 RMB or $12,862 (9) 3.4% 90,000∼99,999 RMB or $ 14,290 (10) 5.6% 100,000∼109,999 RMB or $15,719 (11) 8.0% 110,000∼119,999 RMB or $17,148 (12) 2.7% 120,000∼129.999 RMB or $18,577 (13)130,000∼139.999 RMB or $20,006 (14) 5.2%1.3% 140,000∼149,999 RMB or $21,435 (15) 1.8% 150,000∼159,999 RMB or $22,864 (16) 2.5% 160,000∼169.999 RMB or $24,293 (17) 1.2% 170,000∼179,999 RMB or $25,722 (18) 0.8% 180,000∼189,999 RMB or $27,151 (19) 1.2% 190,000∼199,999 RMB or $ 28,580 (20) 4.2% ≥200,000 RMB or $30,009 (21)Parental education (numeric code) 11.4%Mothers (n = 736) Fathers (n = 705) Elementary school or below (1)Middle school (2)High school or vocational school degree (3)Associate degree (4)Bachelor's degree (5) 1.6%13.9%22.6%29.8%26.7% 0.8%11.4%21.5%29.1%26.6% Master's degree (6)Doctoral degree (7) 1.7%0% 3.1%0.1% Parental occupation (numeric code) Mothers (n = 721) Fathers (n = 679) Unemployed, nontechnical worker, farmer (1) 30.0% 6.4% Semitechnical worker, self-employed small business owner: e.g., construction worker (2) 16.1% 25.3% Technical worker or semiprofessional: e.g., driver (3) 30.4% 31.9% Professional or officer: e.g., doctor, teacher, technician (4) 15.8% 15.7% High-level professional or administrator: e.g., manager (5) 2.1% 9.6% Appendix B Supplementary materials Image, application 1 Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.ecresq.2022.03.001. ==== Refs References Barger M.M. Kim E.M. Kuncel N.R. Pomerantz E.M. The relation between parents’ involvement in children’s schooling and children’s adjustment: A meta-analysis Psychological Bulletin 145 9 2019 855 890 10.1037/bul0000201 31305088 Bauer P. Sonchak L. The effect of macroeconomic conditions on parental time with children: Evidence from the American time use survey Review of Economics of Household 15 2017 905 924 10.1007/s11150-017-9368-y Benner A.D. Mistry R.S. Child development during the COVID‐19 pandemic through a life course theory lens Child Development Perspectives 14 4 2020 236 243 10.1111/cdep.12387 33230400 Bengtson V.L. Allen K.R. The life course perspective applied to families over time Boss P.G. Doherty W.J. LaRossa R. Schumm W.R. Steinmetz S.K. Sourcebook of family theories and methods: A contextual approach 1993 Plenum Press 469 504 Bountress K.E. Gilmore A.K. Metzger I.W. Aggen S.H. Tomko R.L. Danielson C.K. Impact of disaster exposure severity: Cascading effects across parental distress, adolescent PTSD symptoms, as well as parent-child conflict and communication Social Science & Medicine 264 2020 113293 10.1016/j.socscimed.2020.113293 Brooks-Gunn J. Schneider W. Waldfogel J. The great recession and the risk for child maltreatment Child Abuse & Neglect 37 2013 721 729 10.1016/j.chiabu.2013.08.004 24045057 Brown S.M. Doom J.R. Lechuga-Peña S. Watamura S.E. Koppels T. Stress and parenting during the global COVID-19 pandemic Child Abuse & Neglect 110 2020 104699 10.1016/j.chiabu.2020.104699 Bülow A. Keijsers L. Boele S. Van Roekel E. Denissen J.J.A. Parenting adolescents in times of a pandemic: Changes in relationship quality, autonomy support, and parental control Developmental Psychology 57 2021 1582 1596 10.1037/dev0001208 34807682 Cai X. The development and pilot testing of an assessment tool of senior preschoolers’ approaches to learning. Unpublished Master's dissertation 2015 Shaanxi Normal University Shaanxi Cano T. Changes in fathers’ and mothers’ time with children: Spain, 2002–2010 European Sociological Review 35 2019 616 636 10.1093/esr/jcz020 Carroll N. Sadowski A. Laila A. Hruska V. Nixon M. Ma D.W.L. Haines J. The impact of COVID-19 on health behavior, stress, financial and food security among middle to high income Canadian families with young children Nutrients 12 2020 2352 10.3390/nu12082352 Cheng Z. Mendolia S. Paloyo A.R. Savage D.A. Tani M. Working parents, financial insecurity, and childcare: Mental health in the time of COVID-19 in the UK Review of Economics of the Household 19 2021 123 144 10.1007/s11150-020-09538-3 33456425 Chong K.C. Cheng W. Zhao S. Ling F. Mohammad K. Wang M. Wang J. Transmissibility of coronavirus disease 2019 (COVID-19) in Chinese cities with different transmission dynamics of imported cases medRxiv 2020 10.1101/2020.03.15.20036541 Cong Z. Silverstein M. Caring for grandchildren and intergenerational support in rural China: A gendered extended family perspective Ageing and Society 32 2012 425 450 10.1017/S0144686X11000420 Craig L. Churchill B. Dual-earner parent couples’ work and care during COVID-19 Gender, Work & Organization 28 2021 66 79 10.1111/gwao.12497 32837023 Del Bono E. Francesconi M. Kelly Y. Sacker A. Early maternal time investment and early child outcomes The Economic Journal 126 2016 F96 F135 10.1111/ecoj.12342 Elder G.H. The life course as developmental theory Child Development 69 1998 1 12 10.1111/j.1467-8624.1998.tb06128.x 9499552 Elder G.H. Johnson M.K. Crosnoe R. The emergence and development of life course theory Mortimer J.T. Shanahan M.J. Handbook of the life course 2003 Kluwer Academic/Plenum Publishers New York, NY 3 19 England P. Strivastava A. Educational differences in US parents’ time spent in child care: The role of culture and cross-spouse influence Social Science Research 42 2013 971 988 10.1016/j.ssresearch.2013.03.003 23721668 Esping-Andersen G. The incomplete revolution: Adapting to women's new roles 2009 Polity Press Cambridge, UK Feehan J. Is COVID-19 the worst pandemic? Maturitas 149 2021 56 58 10.1016/j.maturitas.2021.02.001 33579552 Fiorini M. Keane M.P. How the allocation of children’s time affects cognitive and noncognitive development Journal of Labor Economics 32 2014 787 836 10.1086/677232 Fomby P. Musick K. Mothers’ time, the parenting package, and links to healthy child development Journal of Marriage and Family 80 2018 166 181 10.1111/jomf.12432 31839685 Gieryn T.F. A space for place in sociology Annual Review of Sociology 26 2000 463 496 10.1146/annurev.soc.26.1.463 Green E. Chase R.M. Zayzay J. Finnegan A. Puffer E.S. The impact of the 2014 Ebola virus disease outbreak in Liberia on parent preferences for harsh discipline practices: A quasi-experimental, pre-post design Global Mental Health 5 2018 1 7 10.1017/gmh.2017.24 Hayes N. Berthelsen D.C. Nicholson J.M. Walker S. Trajectories of parental involvement in home learning activities across the early years: Associations with socio-demographic characteristics and children’s learning outcomes Early Child Development and Care 188 10 2018 1405 1418 10.1080/03004430.2016.1262362 Hsin A. Felfe C. When does time matter? Maternal employment, children’s time with parents, and child development Demography 51 2014 1867 1894 10.1007/s13524-014-0334-5 25280840 Hubei Health Commission The situation of the COVID-19 pandemic in Hubei Province on 31 May, 2020 2020 Accessed on 20 October, 2021 Available at: http://www.xinhuanet.com/politics/2020-06/01/c_1126058098.htm Hyson M. Enthusiastic and engaged learners: Approaches to learning in the early childhood classroom 2008 Teachers College Press New York, NY Johns Hopkins University COVID-19 dashboard 2021 Accessed on 4 August, 2021 Available at: https://coronavirus.jhu.edu/map.html Kaneko S. Noguchi H. Impacts of natural disaster on changes in parental and children's time allocation: Evidence from the Great East Japan Earthquake 2020 Accessed on 12 December, 2021 Available at: https://www.waseda.jp/fpse/winpec/assets/uploads/2020/06/E2006.pdf Kendig S.M. Bianchi S.M. Single, cohabitating, and married mothers’ time with children Journal of Marriage and Family 70 2008 122 1240 10.1111/j.1741-3737.2008.00562.x Ko C.H. Yen C.F. Yen J.Y. Yang M.J. Psychosocial impact among the public of the severe acute respiratory syndrome epidemic in Taiwan Psychiatry and Clinical Neurosciences 60 2006 397 403 10.1111/j.1440-1819.2006.01522.x 16884438 Kreyenfeld M. Zinn S. Coronavirus and care: How the coronavirus crisis affected fathers’ involvement in Germany Demographic Research 44 2021 99 124 10.4054/DemRes.2021.44.4 Kupferschmidt K. Cohen J. Can China’s COVID-19 strategy work elsewhere? Science 267 6492 2020 1061 1062 10.1126/science.367.6482.1061 Lee S.J. Ward K.P. Chang O.D. Downing K.D. Parenting activities and the transition to home-based education during the COVID-19 pandemic Children and Youth Services Review 122 2021 105885 10.1016/j.childyouth.2020.105585 Li-Grining C.P. Votruba-Drzal E. Maldonado-Carreño C. Haas K. Children’s early approaches to learning and academic trajectories through fifth grade Developmental Psychology 46 2010 1062 1077 10.1037/a0020066 20822223 Liu T. Zhang X. Zhao K. Chan W.L. Teacher-child relationship quality and Chinese toddlers’ developmental functioning: A cross-lagged modelling approach Children and Youth Services Review 116 2020 105192 10.1016/j.childyouth.2020.105192 Matheny Jr A.P. Wachs T.D. Ludwig J.L. Phillips K. Bringing order out of chaos: Psychometric characteristics of the confusion, hubbub, and order scale Journal of Applied Developmental Psychology 16 3 1995 429 444 10.1016/0193-3973(95) 90028-4 Milkie M.A. Mattingly M.J. Nomaguchi K.M. Bianchi S.M. Robinson J.P. The time squeeze: Parental statuses and feelings about time with children Journal of Marriage and Family 66 3 2004 739 761 10.1111/j.0022-2445.2004.00050.x Milkie M.A. Nomaguchi K.M. Denny K.E. Does the amount of time mothers spend with children or adolescents matter? Journal of Marriage and Family 77 2015 355 372 10.1111/jomf.12170 Ministry of Education of China The 2019 statistical bulletin on China's national educational development 2020 Accessed on 20 October, 2021 Available at: http://www.moe.gov.cn/jyb_sjzl/sjzl_fztjgb/202005/t20200520_456751.html Ministry of Education of China Suspension of classes without suspending learning’ does not simply mean taking on-line courses 2020 Accessed on 20 October, 2021 Available at: http://www.moe.gov.cn/jyb_xwfb/s5147/202002/t20200212_420201.html Monahan E.K. Income instability and child maltreatment: Exploring associations and mechanisms Children and Youth Services Review 108 2020 104596 10.1016/j.childyouth.2019.104596 Pomerantz E.M. Grolnick W.S. The role of parenting in children's motivation and competence: What underlies facilitative parenting? Elliot A.J. Dweck C.S. Yeager D.S. Handbook of competence and motivation: Theory and application 2017 The Guilford Press New York, NY 566 585 Prime H. Wade M. Browne D.T. Risk and resilience in family well-being during the COVID-19 pandemic American Psychologist 75 5 2020 631 643 10.1037/amp0000660 32437181 Roeters A. van der Lippe T. Kluwer E.S. Work characteristics and parent-child relationship quality: The mediating role of temporal involvement Journal of Marriage and Family 72 2010 1317 1328 10.1111/j.1741-3737.2010.00767.x Rosen, Z., Weinberger-Litman, S. L., Rosenzweig, C., Rosmarin, D. H., Muennig, P., Carmody, E. R., . . . Litman, L. (2020). Anxiety and distress among the first community quarantined in the US due to COVID-19: Psychological implications for the unfolding crisis. doi: 10.31234/osf.io/7eq8c Russell B.S. Hutchison M. Tambling R. Tomkunas A.J. Horton A.L. Initial challenges of caregiving during COVID-19: Caregiver burden, mental health, and the parent–child relationship Child Psychiatry and Human Development 51 2020 671 682 10.1007/s10578-020-01037-x 32749568 Saylor C.F. Swenson C.C. Stokes Reynolds S. Taylor M. The Pediatric Emotional Distress Scale: A brief screening measure for young children exposed to traumatic events Journal of Clinical Child Psychology 28 1 1999 70 81 10.1207/s15374424jccp2801_6 10070608 Sprang G. Silman M. Posttraumatic stress disorder in parents and youth after health-related disasters Disaster Medicine and Public Health Preparedness 7 2013 105 110 10.1017/dmp.2013.22 24618142 Singletary B. Schmeer K. Time Spent by Mothers and Others Scale. Unpublished research instrument 2020 Ohio State University Sun A. Zhang C. Hu X. Boys, girls, and grandparents: The impact of the sex of preschool-aged children on family living arrangements and maternal labor supply Demography 56 2019 813 833 10.1007/s13524-019-00783-5 31087284 UNICEF-IRC. (2014). Children of the recession: The impact of the economic crisis on child well-being in rich countries. Innocenti Report Card, No. 12. New York, NY: United Nations. United Nations (2020). Policy brief: The impact of COVID-19-19 on children. Available at: https://unsdg.un.org/resources/policy-brief-impact-COVID-19-19-children Wong C.L. Chen J. Chow K.M. Knowledge, attitudes and practices towards COVID-19 amongst ethnic minorities in Hong Kong International Journal of Environmental Research and Public Health 17 2020 7878 10.3390/ijerph17217878 Wu M. Xu W. Yao Y. Zhang L. Guo L. Fan J. Chen J. Mental health status of students’ parents during COVID-19 pandemic and its influence factors General Psychiatry 33 2020 e100250 10.1136/gpsych-2020-100250 Wuhan Bureau of Statistics Wuhan Statistical Yearbook: 2020 2020 China Statistics Press Beijing, China Xinhua News Agency Top legislature amends law to allow all couples to have two children 2015 Accessed on 27 December, 2015 Available at: http://www.xinhuanet.com//politics/2015-12/27/c_1117591640.htm Zhang X. Parent-child and teacher-child relationships in Chinese preschoolers: The moderating role of preschool experiences and the mediating role of social competence Early Childhood Research Quarterly 26 2011 192 204 10.1016/j.ecresq.2010.09.001 Zhang X. Barriers and benefits of primary caregivers’ involvement in children’s education during COVID-19 school closures International Journal of Disaster Risk Reduction 66 2021 102570 10.1016/j.ijdrr.2021.102570 Zhang X. Household chaos and caregivers’ and young children’s mental health during the COVID-19 pandemic: A mediation model Journal of Child and Family Studies 2022 10.1007/s10826-022-02283-4 In press Zhang X. Hu B.Y. Ren L. Huo S. Wang M. Young Chinese children’s academic skill development: Identifying child-, family-, and school-level factors New Directions for Child and Adolescent Development 163 2019 9 37 10.1002/cad.20271 Zhou M. Hertog E. Kolpashnikova K. Kan M. Gender inequalities: Changes in income, time use and well-being before and during the UK COVID-19 lockdown Socarxiv 2020 10.31235/osf.io/u8ytc
PMC009xxxxxx/PMC9005340.txt	==== Front High Educ (Dordr) High Educ (Dordr) Higher Education 0018-1560 1573-174X Springer Netherlands Dordrecht 35431322 830 10.1007/s10734-022-00830-y Article Assessing knowledge of and attitudes towards plagiarism and ability to recognize plagiaristic writing among university students in Rwanda Clarke Olivia oclarke@ughe.org 1 Chan Wai Yin Debbie 2 Bukuru Saddam 3 Logan Jenae 45 Wong Rex 36 1 grid.507436.3 0000 0004 8340 5635 Educational Development and Quality Center, University of Global Health Equity, Butaro, Rwanda 2 grid.38142.3c 000000041936754X Harvard University, Cambridge, MA USA 3 grid.507436.3 0000 0004 8340 5635 Bill and Joyce Cummings Institute of Global Health, University of Global Health Equity, Butaro, Rwanda 4 grid.507436.3 0000 0004 8340 5635 Executive Education, University of Global Health Equity, Butaro, Rwanda 5 grid.417182.9 0000 0004 5899 4861 Partners In Health, Boston, MA USA 6 grid.47100.32 0000000419368710 School of Public Health, Yale University, New Haven, CT USA 13 4 2022 117 14 2 2022 © The Author(s), under exclusive licence to Springer Nature B.V. 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Plagiarism is a serious type of scholastic misconduct. In Rwanda, no research has been conducted to assess university students’ attitudes and knowledge of plagiarism and if they have the skills to avoid plagiarizing. This study was conducted to assess knowledge of and attitudes towards plagiarism, as well as ability to recognize plagiaristic writing, among university students in Rwanda. An online questionnaire containing 10 knowledge questions, 10 attitude statements, and 5 writing cases with excerpts to test identification of plagiarism was administered between February and April 2021. Out of the 330 university students from 40 universities who completed the survey, 75.8% had a high knowledge level (score ≥ 80%), but only 11.6% had a high score in recognizing plagiaristic writing (score ≥ 80%). There was no statistically significant association between knowledge level and ability to recognize plagiaristic writing (P = 0.109). Lower odds were found in both diploma/certificate and bachelor students of having high knowledge as well as of having high ability to recognize plagiaristic writing than in master’s students. Although respondents generally disapproved of plagiarism, approximately half of the respondents indicated that sometimes plagiarism is unavoidable, and self-plagiarism should not be punished in the same way as plagiarism of others’ work. Inter-collegial collaboration on effective plagiarism policies and training programs is needed. Keywords Plagiarism Rwanda Academic integrity Higher education university of global health equity ==== Body pmcIntroduction Plagiarism is a serious type of scholastic misconduct (Anderson & Steneck, 2011; Ewing et al., 2016; Gómez et al., 2013; Owens & White, 2013) and is considered one of the most severe breaches of academic integrity. Academic integrity is the commitment to honesty, trust, fairness, respect, and responsibility as the foundational cornerstones of academia (Keohane, 1999). With plagiarism defined as stealing and passing off the ideas of another, or from an existing source, as one’s own without crediting the source (Merriam-Webster n.d), plagiarism is arguably one of the most pressing concerns facing higher education institutions today. With advances in information technology, it has become increasingly easy to access information, articles, and other students’ assignments on the internet, presenting students with ample opportunities to plagiarize the work of others (Honing & Bedi, 2012; Owens & White, 2013; Ramzan et al, 2012; Snow, 2006). This has been of particular importance in the higher education space with the rapid growth of online education due to the COVID-19 pandemic and the closure of many higher education institutions all over the world. It has been suggested that the switch to online learning has resulted in an increase in plagiarism. One study suggested that there has been a significant rise in “contract cheating” during the COVID-19 pandemic, whereby students pay for online tutors to take examinations or write assignments for them (Hill et al. 2021). As well as easy access to information online, a lack of support for students’ writing development and the constraints on faculty teaching capacity in challenging circumstances has resulted in plagiarism becoming more prevalent in recent times (Bailey, 2020). However, more broadly, studies have found that academic dishonesty of varying degrees among students in higher education institution settings has generally always been a common issue worldwide (Amos, 2014; Owens & White, 2013). In a study in the UK surveying 49 higher education institutions, 91% of the institutions reported mild to moderate levels of plagiarism among their students, and 2% reported severe (Parry & Houghton, 1996). In a study conducted in 2010 in the USA, more than 50% of students surveyed reported having cheated or plagiarized at some point in their academic careers (Owunwanne et al. 2010). Similarly, in Taiwan, the prevalence of plagiarism was 66.1% (Lin & Wen, 2007). The Sustainable Development Goal (SDG) 4, Quality Education, aims to promote quality education and lifelong learning opportunities, especially in developing countries (UN, 2021). Targets from SDG 4 such as increasing access to tertiary education and ensuring that young people and adults are equipped with employable skills are directly linked to the quality of education provided at university level. Given that the long-term personal and professional success of both universities and their students goes hand-in-hand with academic honesty and integrity (Cavico & Mujtaba, 2009), studying and addressing issues of plagiarism in higher education institutions in developing countries is vital to achieve the SDG 4 targets. For example, when examining African research studies, Rohwer et al (2018) identified that out of 495 biomedical research journal articles studied, 313 had evidence of plagiarism. Quality of academic work and research is severely compromised by plagiarism. In the higher education institution setting, a study conducted by Iloh et al. in 2018 found that among Nigerian university students, although knowledge of plagiarism was high, attitudes towards plagiarism were poor, resulting in a high level of plagiarism among students. In Rwanda, although there are no studies measuring prevalence or attitudes towards plagiarism, plagiarism is thought to be widespread, and there is a general lack of training for students in this area (Rugira 2015). When it comes to what students may consider plagiarism, one study examining students’ perceptions of plagiarism in the UK found that there was a significant lack of knowledge around whether taking someone else’s ideas, colluding on a paper, or downloading material from the internet were considered plagiarism (Dawson & Overfield, 2006). Further to this, some studies have found that students did not see any harm in plagiarizing, particularly with the internet providing free access to information (Evering & Moorman, 2012), and that this failure to recognize plagiarism or that it constitutes academic dishonesty was more prevalent in lower-income countries (Iloh et al, 2018; Ramzan et al, 2012). Students who received a lower grade point average (GPA) were more likely to plagiarize than those who were more academically successful (Honing & Bedi, 2012), and one study, conducted by Gomez et al. in 2013, found that students were more likely to plagiarize when the assignment contributes less to their overall final grade, suggesting an awareness of the consequences of plagiarism for higher stake assignments. As pointed out by different studies, variations in the attitude or the understanding of what constitutes plagiarism exist due to different cultures (Lin & Wen, 2007; Magnus et al., 2002). It is therefore important to study plagiarism and the extent to which it occurs in local contexts. Higher education institutions around the world are undertaking efforts to educate students by offering them guidance and training to explain the types and consequences of plagiarism, as well as how to avoid it. Different approaches are used to deliver this education and include classroom teaching, online modules, writing exercises, quizzes, and anti-plagiarism software (Belter & Du Pre, 2009; Dunn et al, 2013; Elander et al, 2010; Owens & White, 2013). Combining these approaches—for example, complementing classroom sessions with writing exercises and the use of plagiarism detection software—is a promising way to effectively reduce instances of plagiarism (Owens & White, 2013). However, evidence also suggests that even when plagiarism-detection tools were available, it is still necessary to train students on proper referencing, plagiarism, and academic integrity (Warn, 2006). This prevents unintentional plagiarism occurring due to inadequate knowledge of these, which is a major cause of plagiarism (Belter & du Pre, 2009). Training on such topics also instills high standards of academic integrity and honesty (Babalola, 2012). However, before instituting such training, it is important to understand the students’ ability to recognize plagiaristic writing in order to create appropriate training that is contextually relevant. As mentioned before, although there are some studies on plagiarism focused on the African continent, there has been little research into this area in Rwanda specifically. It has been recognized that there is a lack of training for students in this area and the discussion around plagiarism has reached national attention through mainstream media publications (Rugira, 2015). However, students’ attitudes, knowledge, and abilities to recognize plagiaristic writing have never been formally assessed. Furthermore, without a clearer understanding of what students in Rwanda understand as areas of plagiarism, it is impossible to design effective education programs to reduce plagiarism across the country. This study is the first of its type in Rwanda and aimed to assess the attitudes, knowledge, and ability among students in Rwanda to recognize plagiaristic writing. Assessing the attitudes towards and knowledge of plagiarism provides important baseline information about how plagiarism is perceived by students and whether they are aware of what constitutes plagiarism. Examining ability will determine whether students can apply knowledge in practical situations to identify whether plagiarism is occurring. Assessing the association between knowledge level and application will identify potential gaps and assist in proposing intervention(s) to reduce plagiarism incidence among university students in Rwanda. This will provide important baseline information from which future studies can be conducted. Materials and methods Design The study utilized an online questionnaire, conducted between February and April 2021, to assess knowledge of and attitudes towards plagiarism among university students in Rwanda, as well as their ability to recognize plagiarism in writing. Sample and sampling method Students who were enrolled in a degree program in any university in Rwanda were the study population. The higher education institutions in Rwanda enroll over 44,000 students every year (US Embassy 2019), with 70% in public universities. English is the official academic language used in all higher education institutions in Rwanda. A non-probability sampling method was used by sharing the survey link to different university faculty and students through personal and professional contacts. Deans and principals of universities were contacted by the study team via a letter delivered via email containing information about the study and a link to the survey. University leadership were requested to circulate the survey link to faculty and students, who in turn were encouraged to further share the link to other university students in Rwanda. Data collection tool and methods A questionnaire in English was designed for this study, with some questions adapted from a previously published study (Lindahl & Grace, 2018). The questionnaire was piloted at one university. The final questionnaire consisted of five main sections:Part 1 collected basic demographic information including age, sex, program degree, and university. Part 2 contained 10 “yes” and “no” questions related to knowledge of plagiarism. Part 3 presented five writing samples alongside excerpts from source materials. Respondents were asked to identify if the writing was plagiarized or not as well as to provide the reason. This section was designed to assess the respondent’s ability to identify plagiaristic writings. Part 4 had 10 Likert-scale statements related to the respondent’s attitude towards plagiarism. Each statement has five options: “Strongly agree,” “Agree,” “Neither agree nor disagree,” “Disagree,” and “Strongly disagree.” Part 5 had one open-ended question to ask respondents to provide comments. All students fulfilling the selection criteria were invited to complete the anonymous online questionnaire. No identifying information was collected. All relevant information pertaining to the study and the protection of data was provided to participants on the first page of the online questionnaire. A statement of consent was also presented to participants on the first page of the online questionnaire to serve as a proxy for written consent. This study was approved by the Institutional Review Board of the University of Global Health Equity. Measures The four key measures of this study were as follows:Knowledge level on plagiarism. The knowledge score was calculated as the percentage of correct answers. The knowledge level was defined as high if the score was 80% or above and as low to moderate if the score was below 80% (Koo, Poh and Ruzita, 2015). Attitude towards plagiarism. Percentages of respondents strongly agreed/agreed and strongly disagreed/disagreed on each of the 10 attitude statements. Level of ability to recognize plagiaristic writing. The score was calculated as the percentage of correctly identified plagiaristic writings in the case studies. The ability level was defined as high if the score was 80% or above and as low to moderate if the score was below 80% (Koo, Poh and Ruzita, 2015). Themes emerging from the open-ended question. Data management and analysis Data collected from the online survey were downloaded to Microsoft Excel and cleaned before being uploaded to SPSS for analysis. Descriptive statistics were used to summarize the demographic data as well as the results of the key measures. Fisher’s exact tests were used to examine association between demographics, knowledge level, and level to detect plagiaristic writings. Binomial tests were used to detect if there was a significant difference in the percentage distribution of respondents between strongly agree/agree and strongly disagree/disagree of each of the 10 attitude statements. Content analysis was performed on the comments provided by the respondents through the open-ended question. All tests were performed by using SPSS v. 26, with P-value set at 0.05. Results Demographics A total of 330 university students from 40 universities completed the online survey, with 222 (70%) of them attending private institutes and 94 (30%) attending public university. The mean age was 25.2 (SD 4.6), and the majority of students (n = 193, 59%) were in a bachelor’s degree program (Table 1).Table 1 Summary of the demographic information, knowledge, and ability to recognize plagiaristic writing N (%) Sample 330 Age (year) Mean (SD) 25.2 (± 4.6) University Number participated 24 Private 222 (70.3%) Public 94 (29.7%) Degree program Diploma or certificate 75 (23.0%) Bachelor 193 (59.2%) Graduate degree 58 (17.8%) Knowledge score Mean (SD) 83.1% (± 16.3%) Below 80% (low to moderate) 79 (24.2%) 80% or more (high) 248 (75.8%) Recognizing plagiarism score Mean (SD) 49.0% (± 20.0%) Below 80% (low to moderate) 291 (88.4%) 80% or more (high) 38 (11.6%) Knowledge on plagiarism and ability to recognize plagiaristic writing The mean knowledge score was 83.1% (SD ± 16.3%). Out of the sample of 330, 248 (75.8%) had a high knowledge level (score ≥ 80%). On average, 49% (SD ± 20.0%) of the plagiarism was recognized, ranging from 0 to 100%, with 38 (11.6%) having scored 80% or above (Table 1). Out of the 10 knowledge questions on plagiarism, there were only three in which less than 80% of students answered correctly. They were (1) question 5, expressing well-known common knowledge in your own words without a citation (60.2% correct); (2) question 4, copying a text from a paper you have already written previously without citation (75.2% correct); and (3) question 7, hiring others to write some parts of a paper you will turn in as your own (78.4% correct) (Table 2).Table 2 Results of knowledge on plagiarism Correct n (%) 1. Summarizing the ideas of another author in your own words with a citation 284 (88.2%) 2. Downloading a paper from the internet and presenting it as your own work 306 (95.3%) 3. Copying an excerpt from a paper, using quotation marks, adding a proper citation for the original text, and including a full reference at the end of the paper 275 (85.9%) 4. Copying a text from a paper you have already written previously without citation 240 (75.2%) 5. Expressing well-known common knowledge in your own words without a citation 192 (60.2%) 6. Copying text directly from another paper with quotation marks only 270 (84.4%) 7. Hiring others to write some parts of a paper you will turn in as your own 250 (78.4%) 8. Using pictures from the internet without citing the source 281 (87.5%) 9. Using information from a website that is publicly available without referencing 284 (88.5%) 10. Translating information from a paper in a foreign language without referencing 279 (87.5%) Mean (SD) 83.1% (± 16.3%) Below 80% (low to moderate) 79 (24.2%) 80% or more (high) 248 (75.8%) Attitudes towards plagiarism There was no statistical significance detected between the percentages of respondents who agreed/strongly agreed and those who disagreed/strongly disagreed on three statements: (1) 44% agreed/strongly agreed that “sometimes plagiarism is unavoidable” (P = 0.052); (2) 51.2% agreed/strongly agreed that “self-plagiarism should not be punished the same way as plagiarism of others” (P = 0.697); and (3) 46.5% agreed/strongly agreed that “I should not change the words of an author who wrote something better than I could write it” (P = 0.222) (Table 3).Table 3 Summary of percentage distribution on attitude statements N (%) Agree Disagree P-value Sometimes plagiarism is unavoidable 144 (44.4%) 180 (55.6%) 0.052 All forms of plagiarism are unacceptable 207 (63.3%) 120 (36.7%) < 0.001 For writers whose first language is not English, it is okay to copy parts of a paper written in English 67 (20.8%) 255 (79.2%) < 0.001 Self-plagiarism should not be punished the same way as plagiarism of others 166 (51.2%) 158 (48.8%) 0.697 It is not plagiarism if my colleague/classmate gives me permission to copy from his/her paper 62 (19.1%) 262 (80.9%) < 0.001 If I found out that my colleague/classmate is plagiarizing, I would report the offense to a faculty member/supervisor 192 (59.4%) 131 (40.6%) 0.001 Plagiarizing is as wrong as cheating on an examination 277 (85.8%) 46 (14.2%) < 0.001 Plagiarism should not be penalized after an assignment grade is finalized and released 100 (31.3%) 220 (68.8%) < 0.001 Plagiarism is acceptable if only a small amount is copied 104 (32.1%) 220 (67.9%) < 0.001 I should not change the words of an author who wrote something better than I could write it 151 (46.5%) 174 (53.5%) 0.222 Significantly more respondents agreed/strongly agreed on the following statements: (1) “All forms of plagiarism are unacceptable” (63.3%, P < 0.001); (2) “If I found out that my colleague/classmate is plagiarizing, I would report the offense to a faculty member/supervisor” (59.4%, P = 0.001); and (3) “Plagiarizing is as wrong as cheating on an examination” (85.8%, P < 0.001) (Table 3). Significantly more respondents disagreed/strongly disagreed on the following statements: (1) “For writers whose first language is not English, it is okay to copy parts of a paper written in English” (79.2% P < 0.001); (2) “It is not plagiarism if my colleague/classmate gives me permission to copy from his/her paper” (80.9%, P < 0.001); (3) “Plagiarism should not be penalized after an assignment grade is finalized and released” (68.8%, P < 0.001); and (4) “Plagiarism is acceptable if only a small amount is copied” (67.9%, P < 0.001) (Table 3). Factors associated with knowledge on plagiarism and ability to recognize plagiaristic writing Level of knowledge or ability to recognize plagiaristic writings was not significantly associated with private or public universities or year in program. The degree program respondents were enrolled in was associated with the knowledge level (P = 0.002), and the ability to recognize plagiaristic writing (P < 0.001) (Table 4).Table 4 Association between demographic, knowledge level, and ability level to recognize plagiaristic writing Knowledge level Recognizing ability High (≥ 80%) Low to moderate (< 80%) P-value High (≥ 80%) Low to moderate (< 80%) P-value University Private 168 (70%) 54 (71.1%) 0.887 29 (80.6%) 192 (68.8%) 0.177 Public 72 (30%) 22 (28.9%) 7 (19.4%) 87 (31.2%) Degree program Diploma or certificate 55 (22.4%) 20 (25.3%) < 0.001* 8 (21.1%) 67 (23.3%) < 0.001* Bachelor 136 (55.3%) 56 (70.9%) 9 (23.7%) 183 (63.8%) Master or higher 55 (22.4%) 3 (3.8%) 21 (55.2%) 37 (12.9%) Year in program Final year 163 (65.7%) 49 (62%) 0.589 30 (78.9%) 185 (63.6%) 0.070 Not final year 85 (34.3%) 30 (38%) 8 (21.1%) 106 (36.4%) Knowledge level High (≥ 80%) NA NA NA 33 (86.8%) 215 (74.7%) 0.109 Low to moderate (< 80%) NA NA 5 (13.2%) 73 (25.3%) Significant at P = 0.05. The odds of having a high knowledge level was significantly lower for diploma/certificate students (OR = 0.15, P = 0.003, 95% CI = 0.042, 0.534) and for bachelor students (OR = 0.132, P = 0.001, 95% CI = 0.040, 0.441), compared to master students. The odds of having a high ability to recognize plagiaristic writing was significantly lower for diploma/certificate students (OR = 0.21, P < 0.001, 95% CI = 0.085, 0.522) and for bachelor students (OR = 0.087, P < 0.001, 95% CI = 0.037, 0.204), compared to master students (Table 5).Table 5 Odds ratios of having high knowledge level and high ability to recognize plagiaristic writing among students from different degree programs High knowledge level (≥ 80%) OR P-value (90% CI) High recognizing ability (≥ 80%) OR P-value (90% CI) Diploma or certificate 55 (22.4%) 0.15 0.003 (0.042, 0.534) 8 (21.1%) 0.21 < 0.001 (0.085, 0.522)* Bachelor 136 (55.3%) 0.132 0.001 (0.040, 0.441)* 9 (23.7%) 0.087 < 0.001 (0.037, 0.204)* Master or higher 55 (22.4%) Ref NA 21 (55.2%) Ref NA *Significant at P = 0.05. Statistical association was not detected between the knowledge level and ability to recognize plagiaristic writing (P = 0.109) (Table 4). These results are consistent with the open-ended comments discussed below. Summary of results from open-ended questions Four general themes were identified through the analysis of the open-ended question:Not understanding self-plagiarism is a form of plagiarism One common comment from the open-ended question was related to the lack of clear understanding of self-plagiarism, as illustrated by the following excerpts:"I did not know that self-plagiarism exists, my understanding of plagiarism is using/presenting someone’s work as yours. Wish to know more about it." (Public university, 28 years old, Diploma/certificate student) "Self-plagiarism shouldn’t be taken as a big deal. That’s my work, I copied mine, not someone’s else’s." (Private university, 23, Bachelor) 2. Acknowledging plagiarism was a serious problem Eighteen of the 56 comments from the open-ended question indicated that plagiarism was a serious problem and was very prevalent. Two mentioned that even their lecturers committed plagiarism."Sometimes the lecturers make students plagiarize because they also copy most things on the internet." (Public university, 25, Bachelor) "I need to change [my] program because people are [plagiarizing] every day." (unknown, 22, Bachelor) "Thanks for thinking about this pandemic in academic affairs of nowadays." (Private university, 29, Diploma/certificate) 3. Seeing plagiarism as an academic dishonest act, but committed plagiarism regardless, for various reasons Some students indicated they committed plagiarism because they want to achieve certain academic performance."[students] have to do that because they have to get some marks in the courses, students will do everything in power to have the marks needed." (Private university, 25, Bachelor) Although most respondents indicated plagiarism was not acceptable and should be taken seriously, some also indicated that plagiarism was interpreted differently at different institutions."Plagiarism is culture specific. Depending on the context i.e. school system one went through, there interpretation of plagiarism may be different." (Private university, 32, Master) 4. Ability to recognize plagiaristic writing was low despite understanding what plagiarism means Among the 56 comments received through the open-ended question, 26 indicated that they were uncertain of the answers of the questions related to the knowledge or ability to recognize plagiaristic writings."It was a bit challenging." (Private university, 23, Bachelor) "All of the samples were not plagiarized but they were not also all well summarized." (unknown, 24, Master) Discussion The study results showed that university students in Rwanda overall have a high level of knowledge of plagiarism, with over 75% of participants having a high knowledge level and an average knowledge score of 83.1%. With regard to both knowledge of plagiarism and ability to recognize it, master degree level students were much more likely to have higher skills in both these areas as compared to diploma or bachelor degree level students. This however was not associated with whether students were enrolled in a public or private university or their year of study. Similar results were found in previous studies. In 2007, Lin and Wen found that first year university students were more likely to cheat academically, compared to more senior students. The prevalence also significantly decreased from high school to college (Davis et al., 2002; Jendreck, 1989; McCabe et al., 2001). It is likely that the longer students stay in school, the more exposure they have in writing and reading within the discipline-specific literacy practices required of them, resulting in better understanding and avoidance of plagiarism. These findings are consistent with our study results. Similar to other studies conducted in other low- and middle-income countries, there were significant gaps in understanding what constitutes plagiarism (Amos, 2014; Evering & Moorman, 2012; Honing & Bedi, 2012; Iloh et al, 2018). About 40% of participants did not know that expressing well-known common knowledge without a source is not considered plagiarism, and about 25% did not know that self-plagiarizing is considered plagiarism. The results showed that although the overall knowledge level of plagiarism is high, there is still room for improvement and clarifications. About half of the respondents did not think self-plagiarism was wrong, with more than half of the respondents believing that self-plagiarism should not be punished the same way as plagiarism of others. Self-plagiarism is common in scholarly writing; authors reuse their own texts that have already been published previously as if they were new ideas (Goldblatt, 1984; Roig and Caso 2005; Broome, 2004), and it was among the most common misconceptions in our study. Authors who self-plagiarize do not make any new contributions to the scholarly world (Lowe, 2003). Such intent to deceive the readers is the main reason that self-plagiarism is considered wrong; readers should be informed about this duplication (Goldblatt, 1984). Our results demonstrate there was a general lack of understanding of the difference between citing themselves and reusing one’s own previous work. Our study also found that over 20% of respondents did not know that hiring others to write some parts of their own paper was considered plagiarism. While significantly more respondents (63.3%) indicated that all forms of plagiarism were unacceptable, 44.4% said plagiarism was unavoidable. Some comments from the respondents suggested that even if they knew plagiarism was not acceptable, they still did it. Perhaps relatedly, 31.3% of respondents said plagiarism should not be penalized after an assignment grade is finalized and released. These trends suggest that plagiarism was likely prevalent among the respondents, despite knowing that it is a punishable act. Linked to this, this study also highlighted a huge gap between knowing the principles of plagiarism and applying them. Results showed that students’ overall high knowledge level about plagiarism did not translate to their ability to recognize plagiaristic writings. Only 11.6% of participants had a high level of ability to recognize plagiaristic writings. No association was found between their knowledge level and ability to recognize plagiaristic writings (P = 0.109). Similar results were found in previously published literature in other contexts. In another study conducted to assess whether undergraduate students could determine plagiaristic work, 40 to 50% of students could not correctly identify plagiarized writing samples (Roig, 1997). Other studies have indicated that students who did not receive adequate preparation in research and writing skills were found to be more likely to plagiarize as they did not learn skills such as referencing and paraphrasing (Babalola, 2012; Warn, 2006). This literature is consistent with our results demonstrating that while students might know plagiarism is wrong in most instances, they are simply not able to recognize it or apply key principles to avoid it. Our study is consistent with the literature in that it highlights that greater attention to plagiarism education is needed, with special emphasis on the actual application to recognize plagiaristic writing and skills in referencing and paraphrasing. In addition to this, approaches to plagiarism education must not only focus on these technical aspects of paraphrasing and citing correctly, but must also focus on the purpose of academic writing, which is knowledge dissemination, and the disciplinary norms and values of this. Students must be made aware that they are producers of knowledge through the academic work they produce at university, and therefore should not write to avoid plagiarism, but write to produce new knowledge. Previous studies have found that students plagiarize for multiple reasons. As discussed above, they may not know how to reference or cite, or recognize what plagiarism is, but they also may be driven to plagiarism to achieve better academic scores or because they lack academic writing skills (Anderson & Steneck, 2011; Babalola, 2012; Dawson & Overfield, 2006; Ramzan et al, 2012; Warn, 2006). Our results provided similar insights into the reasons students plagiarize. Specifically, over 46% of respondents indicated that they would not change the words of an author who wrote something better than they could write it. Although the official language in Rwanda is English, the most common language used is Kinyarwanda. Adding to the linguistic complexity is that the government officially changed the medium of education from French to English in 2008, and for many citizens, including parents, teachers, and students, English is not their first language. This lower comfort with the language may help explain the desire to use the words of others who express arguments well in English. In addition to this, uncertainty around how to write academically may have been a major contributing factor in this area. Research conducted by Sheridan (2011) into student diversity and academic literacy found that students who are writing in a second or third language express anxiety over their ability to express themselves and struggle to convey complex ideas in their writing. This is an important consideration in our study, particularly when considered in relation to the results indicating that students undertaking a diploma or bachelor degree were less likely to have a high knowledge level or ability to recognize plagiarism. This is the first time these students are exposed to academic writing at the higher education level, and they may face anxiety or struggles over expressing themselves when writing. Plagiarizing by not changing the words of an author could be seen as an “easy way out” and help assuage anxiety and feelings of helplessness when trying to convey complex ideas in an academic environment. This study has highlighted that students plagiarize because they may lack essential skills and confidence in academic writing. This calls for higher education institutions in Rwanda and across the region to prioritize academic and scientific writing development training for students. This training should not only cover principles of avoiding plagiarism and ensuring academic integrity, for example, how to reference, cite, and paraphrase correctly, as well as what constitutes plagiarism, but it should also cover principles of academic writing. Given that the mode of instruction in Rwandan higher education institutions is English, students should also receive English writing training to build confidence in expressing ideas in a second language. Universities should be cognizant that such training should be integrated when the student first enters higher education at the diploma or bachelor degree level, and early on in their program, as our results showed both the knowledge level and ability to recognize plagiarism were lower among students in diploma/certificate and undergraduate programs, compared to master program students. This foundational training must be combined with practical activities to promote and reinforce learning to application. As reported in this study, simply knowing about plagiarism does not necessarily mean that plagiarism can be correctly identified. Application of knowledge to practical examples will build students’ confidence in knowing how and when plagiarism occurs, and this can be directly applied to their own writing. Providing this comprehensive academic writing and how to avoid plagiarism is important for students to equip them to adhere to standards of academic integrity and produce acceptable academic work while at university, but is of particular importance with regard to authorship and publications, as the implications of plagiarism extend well beyond the time a student spends studying in a higher education institution. Studies have shown that researchers in low- and middle-income countries were underrepresented in first and last authorship positions in peer reviewed publications, even within studies conducted in Africa (Hedt-Gauthier et al, 2019; Mbaye et al, 2019; Schneider & Maleka, 2018). A recent systematic review found that less than 50% of the publications related to studies conducted in Africa had an African first or last author; the underrepresentation was even more significant for publications related to studies conducted in non-Anglophone countries (Mbaye et al, 2019). This underrepresentation is an issue of structural inequity; it is also likely that one contributing factor is low- and middle-income country (LMIC) authors’ limited command of English for non-native English speakers and limited training in scientific writing. By enhancing students’ understanding of plagiarism and improving their scientific writing skills scholars in LMICs may elevate their recognition in the publication space. As such, this study recommends that comprehensive scientific and academic writing training should be instituted in higher education institutions in Rwanda and across similar contexts to promote the importance of academic integrity in scholarly work, improve students’ ability to apply skills to avoid plagiarism, and increase opportunity for publication. In addition to this, higher education institutions should establish policies on plagiarism and enforce strict penalties if plagiarism occurs. Both pedagogical approaches and policy should also be extended to academic faculty, as academic integrity must be enforced at all levels in the academic world. This study presents an opportunity for higher education institutions and other stakeholders in Rwanda, or other similar contexts, to collaboratively establish clear standards, guidelines, and training on plagiarism to increase standards of education, academic writing, and publications across higher education institutions. The project has successfully accomplished the objectives of this study; however, the results must be seen in light of some limitations. Firstly, the majority of our respondents were from private universities, despite most students in Rwanda attending public universities. In addition to the non-probability sampling method, the results therefore may not be completely representative of the higher education landscape in Rwanda. Further to this, we did not examine the role of gender in this study. The extent to which gender differences play a role in whether students plagiarize has been inconclusive in previous literature. While some studies found that more males than females cheated or found it acceptable to cheat (Smyth and Davis, 2004; Brown and Choong, 2005; Lin & Wen, 2007), others found no significant difference (Roig & Caso, 2005). As the goals of our study were to assess the overall level of understanding and ability to identify plagiarism among higher education students, given that this has not been studied in Rwanda before, we did not include gender as a variable. Incorporating gender analysis in future studies will enhance the assessment of plagiarism in Rwanda and provide useful insight into the demographics of plagiarism. Lastly, in our study, we did not identify in what language the respondents received their secondary education. Future studies to examine knowledge of and attitudes towards plagiarism in relation to language may provide more insight into whether speaking English as a second or third language has an impact on plagiarism. Added to this, measuring the skills of students in English and academic writing may shed light on whether lower skills in English and academic writing is associated with higher levels of plagiarism, which is what previous research in other contexts has suggested. Future studies to include faculty in the study would provide more insights into the culture of plagiarism at higher education institutions. Further to this, future studies could also examine skill levels of students in relation to academic writing and whether they are able to implement strategies to avoid plagiarism. This would provide further insight into the discussion around the linkages between knowledge of and attitudes towards plagiarism and whether students are able to implement strategies to avoid plagiarism. However, despite the limitations, this study provided important baseline information for future studies and initiatives. Conclusion This is the first nationwide study of its kind in Rwanda. Not only will the study results provide the baseline data to inform program design, the study effort itself also presented an opportunity for inter-collegial collaboration to design an appropriate plagiarism standard and training programs. With e-learning becoming more commonly used as technology advances, the development and utilization of an online platform to promote best practices in scholarly writing should be considered. Acknowledgements The authors would like to thank all who have participated in this study. Author contribution All authors contributed to the study conception and design. Material preparation and data collection tools were prepared by Olivia Clarke, Debbie Chan, and Jenae Logan. Data collection and analysis were performed by Rex Wong and Saddam Bukuru. The first draft of the manuscript was written by Olivia Clarke and Rex Wong, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding This research was supported by the University of Global Health Equity. Data availability The data that support the findings of this study are openly available in Mendeley Data at http://doi.org/10.17632/46v7bhb7sr.1. Code availability N/A. Declarations Conflict of interest The authors declare no competing interests. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Amos KA The ethics of scholarly publishing: Exploring differences in plagiarism and duplicate publication across nations Journal of the Medical Library Association 2014 102 2 87 10.3163/1536-5050.102.2.005 24860263 Anderson M Steneck N The problem of plagiarism Urologic Oncology: Seminars and Original Investigations 2011 29 91 94 10.1016/j.urolonc.2010.09.013 Babalola YT Awareness and incidence of plagiarism among undergraduates in a Nigerian private university African Journal of Library, Archives and Information Science 2012 22 1 53 Bailey, J. (2020). 5 Ways COVID-19 is changing academic plagiarism. Plagiarism Today. Retrieved October 1 2021 from https://www.plagiarismtoday.com/2020/10/26/5-ways-covid-19-is-changing-academic-plagiarism/ Belter RW Du Pre A A strategy to reduce plagiarism in an undergraduate course Teaching of Psychology 2009 36 257 261 10.1080/00986280903173165 Broome ME Self-plagiarism: Oxymoron, fair use, or scientific misconduct? Nursing Outlook 2004 52 6 273 274 10.1016/j.outlook.2004.10.001 15614263 Cavico, F. J., and Mujtaba, B. G. (2009). Making the case for the creation of an academic honesty and integrity culture in higher education: Reflections and suggestions for reducing the rise in student cheating. American Journal of Business Education, 2(5), 75–88. 10.19030/ajbe.v2i5.4072 Davis SF Grover CA Becker AH McGregor LN Academic dishonesty: Prevalence, determinants, techniques, and punishments Handbook for Teaching Introductory Psychology with an Emphasis on Assessment 2002 3 153 157 Dawson MM Overfield JA Plagiarism: Do students know what it is? Bioscience Education 2006 8 1 1 15 10.3108/beej.8.1 Dunn DS Saville BK Baker SC Marek P Evidence-based teaching: Tools and techniques that promote learning in the psychology classroom Australian Journal of Psychology 2013 65 1 5 13 10.1111/ajpy.12004 Elander J Pittam G Lusher J Fox P Payne N Evaluation of an intervention to help students avoid unintentional plagiarism by improving their authorial identity Assessment and Evaluation in Higher Education 2010 35 157 171 10.1080/02602930802687745 Evering LC Moorman G Rethinking plagiarism in the digital age Journal of Adolescent and Adult Literacy 2012 56 1 35 44 10.1002/JAAL.00100 Ewing H Anast A Roehling T Addressing plagiarism in online programmes at a health sciences university: A case study Assessment & Evaluation in Higher Education 2016 41 4 575 585 10.1080/02602938.2015.1033612 Goldblatt D Self-plagiarism Journal of Aesthetics and Art Criticism 1984 43 71 78 10.2307/430193 Gómez J Salazar I Vargas P Dishonest behavior and plagiarism by university students: An application to management studies Procedia - Social and Behavioral Sciences 2013 83 766 770 10.1016/j.sbspro.2013.06.144 Hedt-Gauthier, B. L., Jeufack, H. M., Neufeld, N. H., Alem, A., Sauer, S., Odhiambo, J., and Volmink, J. (2019). Stuck in the middle: A systematic review of authorship in collaborative health research in Africa, 2014–2016. BMJ Global Health, 4(5). 10.1136/bmjgh-2019-001853 Hill, G., Mason, J., and Dunn, A. (2021). Contract cheating: An increasing challenge for global academic community arising from COVID-19. Research and Practice in Technology Enhanced Learning 16(24). 10.1186/s41039-021-00166-8 Honing B Bedi A The fox in the hen house: A critical examination of plagiarism among members of the Academy of Management Academy of Management, Learning and Education 2012 11 1 101 123 10.5465/amle.2010.0084 Iloh GUP Amadi AN Chukwuonye ME Godswill-Uko EU Plagiarism in a resource-constrained context: A cross-sectional study of post-graduate medical college trainees and fellows in a tertiary health institution in South East Nigeria Archives of Medicine and Health Sciences 2018 6 2 270 10.4103/amhs.amhs_103_18 Jendreck MP Faculty reactions to academic dishonesty Journal of College Student Development 1989 30 401 406 Keohane, N. (1999). The fundamental values of academic integrity. The Center for Academic Integrity, Duke University, 1–12. Retrieved October 1 2021 from https://academicintegrity.org/images/pdfs/20019_ICAI-Fundamental-Values_R12.pdf Koo, H. C., Poh, B. K., and Ruzita, A. T. (2015). Assessment of knowledge, attitude and practice towards whole grains among children aged 10 and 11 years in Kuala Lumpur. International Journal of Food Science, Nutrition and Dietetics, 4(1), 171–177. 10.19070/2326-3350-1500032 Lin CHS Wen LYM Academic dishonesty in higher education—A nationwide study in Taiwan Higher Education 2007 54 1 85 97 10.1007/s10734-006-9047-z Lindahl JF Grace D Students’ and supervisors’ knowledge and attitudes regarding plagiarism and referencing Research Integrity and Peer Review 2018 3 1 10 10.1186/s41073-018-0054-2 30386644 Lowe N Publication ethics: Copyright and self-plagiarism Journal of Obstetric, Gynecologic and Neonatal Nursing 2003 32 145 146 10.1111/j.1552-6909.2003.tb00137.x Magnus JR Polterovich VM Danilov DL Savvateev AV Tolerance of cheating: An analysis across countries Journal of Economic Education 2002 33 2 125 135 10.1080/00220480209596462 Mbaye R Gebeyehu R Hossmann S Mbarga N Bih-Neh E Eteki L Thelma O Oyerinde A Kiti G Mburu Y Haberer J Siedner M Okeke I Who is telling the story? A systematic review of authorship for infectious disease research conducted in Africa, 1980–2016 BMJ Global Health 2019 4 5 e001855 10.1136/bmjgh-2019-001855 McCabe DL Trevino LK Butterfield KD Cheating in academic institutions: A decade of research Ethics and Behaviors 2001 11 3 219 232 10.1207/S15327019EB1103_2 Merriam-Webster. (n.d.). Plagiarize. In Merriam-Webster.com dictionary. Retrieved October 1, 2021, from https://www.merriam-webster.com/dictionary/plagiarize Owens C White FA A 5-year systematic strategy to reduce plagiarism among first-year psychology university students Australian Journal of Psychology 2013 65 1 14 21 10.1111/ajpy.12005 Owunwanne, D., N. Rustagi, and Dada, R. (2010). Students’ perceptions of cheating and plagiarism in higher institutions. Journal of College Teaching and Learning 7(11), 59–68. http://www.cluteinstitute.com/ojs/index.php/TLC/article/view/253/243. Parry G Houghton D Plagiarism in UK universities Education and the Law 1996 8 3 201 215 10.1080/0953996960080303 Ramzan M Asif Munir M Siddique N Asif M Awareness about plagiarism amongst university students in Pakistan Higher Education 2012 64 73 84 10.1007/s10734-011-9481-4 Rohwer A Wager E Young T Garner P Plagiarism in research: A survey of African medical journals British Medical Journal Open 2018 8 11 e024777 10.1136/bmjopen-2018-024777 Roig M Can undergraduate students determine whether text has been plagiarized? The Psychological Record 1997 47 1 113 122 10.1007/BF03395215 Roig M Caso M Lying and cheating: Fraudulent excuse making, cheating, and plagiarism The Journal of Psychology 2005 139 6 485 494 10.3200/JRLP.139.6.485-494 16419439 Roig, M. (2006). Avoiding plagiarism, self-plagiarism, and other questionable writing practices: A guide to ethical writing. Retrieved September 22 2018 from https://ori.hhs.gov/sites/default/files/plagiarism.pdf Rugira, L. (June 08, 2015). University students are plagiarising? Say it ain’t so. The NewTimes. Retrieved May 20 2017 from http://www.newtimes.co.rw/section/read/189538/. Schneider H Maleka N Patterns of authorship on community health workers in low-and-middle-income countries: An analysis of publications (2012–2016) BMJ Global Health 2018 3 3 e000797 10.1136/bmjgh-2018-000797 Sheridan V A holistic approach to international students, institutional habitus and academic literacies in an Irish third level institution Higher Education. 2011 62 129 140 10.1007/s10734-010-9370-2 Snow, E. (2006). Teaching students about plagiarism: An internet solution to an internet problem. Innovate: Journal of Online Education, 2(5). Retrieved September 22 2018 from http://www.innovateonline.info/index.php?view=article&id=306. United Nations Department of Economic and Social Affairs (n.d). The 17 Goals. Retrieved September 15 2021 from https://sdgs.un.org/goals. US Embassy (n.d). Rwanda Education System. Retrieved March 10 2020 from https://rw.usembassy.gov/education-culture/rwanda-education-system/ Warn J Plagiarism software: No magic bullet Higher Education Research and Development 2006 25 2 195 208 10.1080/07294360600610438
PMC009xxxxxx/PMC9005341.txt	==== Front GeoJournal GeoJournal Geojournal 0343-2521 1572-9893 Springer Netherlands Dordrecht 35431409 10633 10.1007/s10708-022-10633-4 Article The tide of tiger poaching in India is rising! An investigation of the intertwined facts with a focus on conservation http://orcid.org/0000-0002-6587-699X Nittu George http://orcid.org/0000-0002-2306-1821 Shameer Thekke Thumbath Nishanthini Nanjanad Kannan http://orcid.org/0000-0003-2226-2012 Sanil Raveendranathanpillai sanilravi@live.in grid.413002.4 0000 0001 2179 5111 Molecular Biodiversity Lab, Department of Zoology & Wildlife Biology, Government Arts College, Udhagamandalam-643002, The Nilgiris, Tamil Nadu India 13 4 2022 114 11 3 2022 © The Author(s), under exclusive licence to Springer Nature B.V. 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Poaching and illegal trafficking are major threats to biodiversity, especially when endangered felids are concerned. Tigers are iconic animals, and there is huge demand for their body parts both in the national and international illegal markets. India forms the largest tiger conservation unit in the world and poaching is at its peak even though there are stringent laws and strict enforcement. In the present study, we analytically estimated the tiger seizure cases in India from 2001–2021 using newspaper archives as the main source of data. The data was geo-referenced to understand the details of seizure, demand, and locality. We statistically correlated the seizure rate with the density of tigers, tiger reserves, and various other socio-economic factors. Our result shows that skin, claws, bones, and teeth have more demand, with nails and teeth being the most preferred in local markets. The bones, flesh, and other parts were mostly seized in the border states of the north and eastern states. The intensity of seizures is very high in the states of Maharashtra, Karnataka, Tamil Nadu, and Assam. From our analysis, we predict four trade routes for the export of the seized parts: the Nepal-Bhutan border, Assam border, the Brahmaputra, and the Mumbai port. This corresponds to the five tiger conservation blocks in India, and we observed the seizure rate is high near the Western Ghats region, which has not yet been noticed. Apart from the seizure, we are unconcerned with the seizure's origin or the local trading routes. The study demonstrates the importance of identifying the source population using DNA methods and carefully enforcing the rules in area of poaching. We assert that current approaches are incapable of resolving the issue and that a more precise and effective forensic procedure capable of resolving the issue at the minute local level is critical for precisely tracing trade channels. Supplementary Information The online version contains supplementary material available at 10.1007/s10708-022-10633-4. Keywords Wildlife seizure Tiger dominated landscapes Trade routes Skin and claws Panthera tigris ==== Body pmcIntroduction The illegal trafficking of commercially valuable body parts of endangered animals has a well-developed international network. Such practises are of great concern in biodiversity-rich countries like India, because they negatively impact conservation efforts (Tensen, 2016). Despite strict laws and enforcement, enigmatic species such as the Bengal tiger (Panthera tigris) face severe threats from trafficking (Nittu et al., 2021). The species has fewer than 4,981 individuals worldwide, with India accounting for a sizable portion of the population (Jhala et al., 2021). The species is well-known, and body parts are in high demand on the illegal international market. The tiger modulates the food web and extemporises natural resources as the top predator of the Indian forests (Mohan et al., 2021). The Indian government is taking every possible step to increase the population in natural habitats and enforce stringent laws to prevent poaching. India, China, Russia, and other tiger-ranging countries jointly launched the Global Tiger Recovery Programme to increase tiger populations in their natural habitats (Joshi et al., 2016). As a result, tiger-roaming countries enacted strict legislation to protect the species from extinction. Even though strict rules are in place, illegal trafficking in the body parts of this elusive ionic species is common in illegal markets. In the last two decades, the global wild tiger population has shrunk by half (Morgan et al., 2021). Because of the increased demand for body parts, they are highly vulnerable to poaching (Nowell, 2010; Stoner et al., 2016). The various enforcement agencies seized 2,359 tiger parts globally between 2000 and 2018 (Morgan et al., 2021; Wong & Krishnaswamy, 2019). China is the largest consumer of this endangered wild felid's body parts and byproducts (Coals et al., 2020; Jiao et al., 2021). The illegal products from other tiger-ranging countries make their way to China via an international trading network. Trafficked tiger products are primarily used as decorative materials or as an ingredient in Chinese traditional medicines (Shepherd et al., 2020; van Uhm, 2020; Wong, 2019, 2020) and the production of wine (Coals et al., 2020; Wong, 2016). Skin, claws, and teeth are part of fashion items, home decor, jewellery, holy materials, and the symbol of bravery. Flesh, blood, bones, fats, and genitalia are all used as ingredients in traditional medicines, tonics, and treatments for various ailments (Ellis, 2013; Gratwicke et al., 2008; Karmacharya et al., 2018; Mainka & Mills, 1995; Moyle, 2009; Nowell, 2000; Wong, 2016). When tiger bone gets boiled, it transforms into an adhesive known as tiger bone glue, an ingredient of traditional medicines in many countries (Davis et al., 2020; Drury, 2011; Nowell, 2010; Stoner et al., 2016). The intensity of tiger part smuggling is proportional to the market demand and people's incorrect perceptions. According to reports, the tiger is being bred in captivity in China to meet public demand (Coals et al., 2020; Wang et al., 2019). The handling and trade of tiger parts became illegal in India with the implementation of the Wildlife Protection Act in 1972, followed by the documentation of cases in the 1980s. Southern Indian states recorded the maximum tiger smuggling cases as they have a viable tiger population. Unlawful trafficking grew into an extensive network that included Bengal, Madhya Pradesh, Andhra Pradesh, Uttar Pradesh, Bihar, Maharashtra, and Karnataka (Ellis, 2013; Palita, 2007). The dealers take advantage of the tribal people's economic conditions and hunt the tiger (Ellis, 2013). Many tribal hamlets are within reserves, and tribes have extensive knowledge of animal movement and handling. Because strict enforcement may be challenging to implement in a scheduled tribal hamlet, poachers take advantage of this. The Indian scenario is unclear regarding the severity of tiger poaching and the origins of trade routes. The majority of the literature on tiger trade and poaching implements the India-China-Nepal trade route and the intensity of seizures in states bordering China and Nepal (Karmacharya et al., 2018; Sharma et al., 2014). They also forecast a trade route through Ladakh, Tibet, Laos, and Bhutan (Wong, 2016), with China as the primary consumer. Domestic demand or the starting point of the trade route was not the focus of any research, and studies mainly focused on recognising and improving tiger habitats and populations. Objectives It is unknown from which part of India the poaching of tiger parts constantly occurs, but the Maharashtra-Madhya Pradesh belt may be the major poaching site (Gopal et al., 2010; Wright, 2010). Though strict enforcement can control poaching to some extent, the aspects of poaching motivation by locals need to be thoroughly investigated. There are limitations to conducting such a study because protected area managers, locals, and tribal dwellers in the forests may not cooperate. The seizure data records may not be an accurate figure, but they are a random sample of the actual figure. As a result, the issue can be addressed on a broad scale, assuming the state as a whole and the various sociological factors. The current study aims to assess the intensity and prevalence of poaching based on secondary seizure data. The study intends to investigate the impact of demographic, economic, and sociological factors on poaching intensity in a state. We also attempted to extrapolate ecological factors related to tiger demography to predict potential tiger poaching areas. Review of literature Zoogeographers such as Newbigin (1913) drew attention to animal geographies, which led to the development of cultural animal geographies that placed a priority on space and spatial distributions (Bennett, 1960). Later on, animal geographies faded away and resurfaced with the work of Anderson (1997), Wolch and Emel (1995), Elder et al. (1998), and Wilbert (2000). Many animal geographers advanced arguments demonstrating how animals and the networks in which they are entangled leave imprints on specific locations, regions, and landscapes over time, sparking research on animals and place (Anderson, 1995; Davies, 2000; Emel et al., 2002; Gruffudd, 2000; Ufkes, 1995; Yarwood & Evans, 2000). The essays by Emel and Urbanik (2010) drew attention to the physical and conceptual dimensions of human-animal relationships. Urbanik's (2012) concepts attracted widespread notice for their numerous connections between geography and human-animal relations. She emphasised the importance of establishing techniques, pointing out that doing so will enable us to get closer to animals. This was followed by a body of writing (see Buller, 2014) arguing that "animals are no longer the sole domain of the sciences." Hovorka (2017) emphasised the global diversity of human perspectives, placements, and interactions with animals, as well as the global diversity of animal circumstances, experiences, and lives. Animal research in geography could benefit from "further expansion into human–environment geographies, physical geographies, and geomatics in order to leverage hybridity as a concept and practise" (Hovorka, 2017). Hovorka (2019) states that "examining the breadth and complexity of these power relations is critical given that we live in a multispecies world and continue to seek ways to de-center "the human" in “theory and practise." There are thirteen tiger range countries in the world: India, China, Indonesia, Myanmar, Bangladesh, Vietnam, Russia, Malaysia, Thailand, Nepal, Cambodia, Bhutan, and Lao PDR (Kumar, 2021). However, India, Western China, Southern Asia, and some regions of Russia currently hold the majority of the world's tiger population, with India holding the majority of the wild tiger population (Karanth et al., 2017; Knoka et al., 2018; Kumar et al., 2019). According to the most recent census, India has 2,967 tigers, up from 1706 in 2010. (Jhala et al, 2020; Jhala et al., 2021). India, as a primary tiger habitat, also has the highest number of tiger seizures worldwide (Kumar et al., 2019; Wong & Krishnaswamy, 2019). The plight of Indian tigers began in the late nineteenth century with the widespread use of modern weapons (Sharma et al., 2014). Despite strict legislation such as the WPA 1972 and its 2006 tiger amendment, poaching incidents occur in India (Bijoy, 2011). Between 1994 and 2003, the Environmental Investigation Agency (EIA) reported 684 cases of tiger poaching and seizure (Banks & Newman, 2004). According to WPSI's online reports (http://www.wpsi-india.org/), India lost 1,110 tigers between 1994 and 2016. According to another report by Wong and Krishnasamy (2019), 369 tiger seizures occurred in India between 2000 and 2018, which is more than any other tiger range country. Tiger skin, teeth, claws, meat, genitals, fat, and even bones are poached for traditional medicines (Ellis, 2013; Gratwicke et al., 2008; Kitpipit et al., 2012; Mainka & Mills, 1995; Moyle, 2009). Body parts such as eyeballs, tails, and whiskers are also used in medicine (Mills & Jackson, 1994). The majority of tiger part trafficking is done to meet Chinese demand (Gill, 2014). According to Paudel et al. (2020), the main illegal trade route for tiger body parts is across the Indo-Chinese border. When tiger parts were scarce, leopards replaced the scarcity in the market (Niraj et al., 2012). According to Sethi et al. (2019), between 2012 and 2016, 650 leopards were poached in India, putting both endangered felids in grave danger. Many online reports speculated that tiger poaching peaked (up to 151%) during the COVID-19 lockdowns (see the information in the Times of India, the Print, etc.). There haven't been any studies that have looked at how tiger poaching affects the environment and how it affects the people who do it. Methodology Data collection of the tiger confiscations. Because seizures are often talked about in print and online media, newspapers are a great source of information, and archives are easy to find. We used the archives of major English and regional language newspapers and search engines like Google and Bing. We gathered this data over the last 20 years (2001–2021). Before 2001, information was scarce and was challenging to obtain. We also gathered information from the WPSI's website (wpsi-india.org), which contains a consolidated report of Indian law enforcement agencies. To learn more about law enforcement, we refered the Wildlife Crime Control Bureau website (wccb.gov.in). After a thorough examination, we ignored repeated data of the same incidents from different newspapers to avoid repetitions. We consolidated every piece of data into an excel sheet with the details of date, place, and state, seizure details, and convicts (if interested, readers can obtain it by sending an email to the corresponding author). As the data given on the WPSI website differs from our data, we ignored the data prior to 2010 in our primary statistical analysis. The WPSI data is a year-to-year figure, and we cannot use it to analyse state-wise trends, geo-referencing, or other seizure details. We converted the year-wise WPSI data (2010–2019) to a logarithmic value and compared it (Chi-square analysis) with the logarithm of the data we obtained to rule out the trend difference. Collection of supporting data. We got reports of the seizure case from states such as Andhra Pradesh, Arunachal Pradesh, Assam, Chhattisgarh, Delhi, Goa, Haryana, Jammu & Kashmir, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Nagaland, Odisha, Tamil Nadu, Telangana, Uttar Pradesh, Uttarakhand and West Bengal. So, we further collected the data related to only the particular states. The secondary data we collected includes the following: total number of tiger reserves in India (ntca.gov.in), mortality rate across states (ntca.gov.in), percentage of forest cover per state (fsi.nic.in), details of tiger population and density (Jhala et al., 2020), total number of urban areas per state (mohua.gov.in), total area of tiger reserves (wiienvis.nic.in), total number of village areas per state (censusindia.gov.in), population status of each state as per 2011 census (censusindia.gov.in), literacy rate per state (mospi.nic.in), GSDP (Gross State Domestic Productivity) of each state (mospi.nic.in/data), and year wise GDP (Gross Domestic Productivity) of India (2001–2020) (mospi.nic.in). We provided the consolidated information as supplementary material ESM1 for reference. Tiger increment rate and the Seizure rate analysis. We calculated the total tiger increment in every state using the four NTCA tiger census reports (2006, 2010, 2014, and 2018). Similarly, we calculated the seizure rate as the ratio of the number of seizures reported to the number of years. We compared the annual seizure rate with the state-wise tiger mortality, annual tiger increment, number of tiger reserves, and tiger density by creating line and bar graphs (see supplementary data ESM2). Statistical Analysis. We analysed the data using the seizure as a sample of original poaching to determine the state-by-state prevalence and year-by-year trends. Using the ggplot 2 (Valero-Mora, 2010) and hrbrthemes (github.com/hrbrmstr/hrbrthemes) packages, we examined the relationship between seizure rate and various ecological and socioeconomic variables in R studio (RStudio Team 2020). Data geo referencing and mapping. We obtained the latitude and longitude (geographical position in degrees and minutes) of the seizure using Google Maps for the data geo-referencing. We classified the primary data based on the seized parts, which included skin, bones (except the skull), claws (including nails), carcass (seizure of a dead specimen from a person/group), kill (a case in which a person/group killed the tiger with the intent of poaching), organs (any parts other than the skin and bones), teeth (including all types of teeth), skull (except bones), and whiskers. We categorised and plotted the data on the map of India using QGIS (Mileu & Queiros, 2018) to determine which tiger parts are most poached and their regional preference. We then graded the states based on the intensity of poaching and created a heat map of India based on each state's annual rate of poaching. Results The major seizure reports were during transit or attempts to hand over or sell to a third party. Only 13 of the total cases report the domestic sale of tiger parts. The majority of domestic cases involve the sale of paws, claws, and teeth. There have been reports of people attempting to sell tiger skin in the Kolkata market, and they have also tried to sell fake tiger skin and claws for a high price. The authorities seized the tiger paw, claw, and teeth from the perpetrator at the time of the fake skin seizure. The study shows that in the Indian market, most demand is for the claws and teeth, with less demand for the other parts. In most cases where the seizure occurred during handover, the convicts were primarily young, but their ages ranged from 24 to 60 years. The agencies typically apprehend tribes who have tiger parts in their possession. The punishment details from the Indian court are available on the Wildlife Crime Control Bureau website. The general information says that only 14 people were punished for seven years of imprisonment, amongst which only two were punished for possessing tiger body parts and the others for hunting. Our survey clearly shows there is a discrepancy among the data regarding tiger seizures and related offenses. When looking at individual cases from the newspaper archives, it appears that the articles (tiger parts) are even available in jewelry stores. According to the report, regular customers want the tiger nails and teeth fixed in their ornaments. Such incidents are more common in the Karnataka cities of Bangalore and Chamarajanagar. According to the Chamarajanagar report, people in this region have superstitious beliefs that if they hold tiger parts like a claw, it will bring them wealth and protect them from disease. The data also shows that most seizures occurred in private vehicles, with only a few occurring on public transportation. It is clear that once it reaches the edge of the forest, there is a network that distributes the items via multiple links, making it impossible for law enforcement to determine the end and beginning of the chain. However, the overall report indicates that the sample came from tiger reserves in India, and the poachers were mostly local forest dwellers. The intensity of poaching. The intensity of seizures is very high in Maharashtra, Karnataka, Tamil Nadu, and Assam (Fig. 1), indicating trade in this region or a poaching site. When analysing the points where the maximum seizure occurred, it is near a reserve, a country's border, or an important international port. The intensity of seizure activity increased from Tamil Nadu through Karnataka-Maharashtra, an indication of the trade route. The former two states have a good number of tiger populations, which can be considered the origin of poaching. A similar track and trend can be seen from Odisha through West Bengal to Assam. Mumbai may be receiving tiger parts either from the Western Ghats or from the central India. Similarly, Assam can be another exit port to meet the requirements of China through the Tibetan or other border regions. We do not get ample data to study the demand, either in the internal domestic market or in the international market.Fig. 1 Cumulative poaching intensity in Indian states More demanded tiger parts. The major seizure is the tiger skin when assessing India (Fig. 2) as a whole, followed by the other body parts like claws, bones, and teeth. The bone seizure is mainly happening in the north-eastern Indian border regions. Skin, claws, and teeth are the most frequently confiscated items in peninsular India, indicating a high demand for these parts. The figure also clearly shows that the seizure of tiger parts is closer to the tiger conservation blocks, demonstrating the origin of the seized parts. Our data depicts Maharashtra (particularly Mumbai) may be the primary destination for poached tiger parts. It is also clear that bone is a significant tiger part in high demand along the Nepal-China border. The Shivalik hills may be the source of tiger parts that have been taken to the borders of Nepal and Ladakh.Fig. 2 Geo-referenced tiger body part seizure from India Analysis of Socio-economic and the demographic attributes. We used tiger seizures as the function of poaching, and the correlation (r) shows that poaching correlates (P0.05) more with tiger-related attributes than with human socio-economic parameters (Fig. 3). The values show the number of tiger reserves, tiger population density, tiger mortality, and forest cover, which are all major factors correlated with poaching. As a result, it stands to reason that a state with these factors will have more poaching cases. Poaching is unaffected by factors such as low literacy rates, the number of villages in a state, urbanised areas, average gross state domestic product (GSDP), and even population density. Our analysis correlating the tiger loss and gain concerning poaching clearly shows a similar trend (Fig. 4). Hence, in regions where the tiger population is increasing, the intensity of poaching also increases. Similarly, tiger poaching shows an opposite trend of increment in some regions, where tiger loss is experienced. Gross domestic product (GDP) is a commonly used measure of the economic activity of a country. For the past 20 years, the GDP showed a significantly correlated trend with seizure data (Fig. 5).Fig. 3 Heat map showing the correlation between various attributes and seizure rate (A: Human population density, B: GSDP, C: Number of villages, D: Urban area, E: Percentage of literacy, F: Forest cover, Tiger picture: Tiger increment rate per annum, G: Tiger mortality, H: Tiger density, I: Number of tiger reserves, J: Seizure rate per annum, r: Pearson's correlation coefficient). Fig. 4 State wise annual seizure rate (y axis range 0.01-1) and tiger increment rate (y' axis range -10 to 25) Fig. 5 Correlation between GDP and seizure rate. GDP is significantly correlated with seizure rate (r = 0.8), the dotted lines show the trend line Discussion. Market demand for seized products. Tigers are constantly being poached for their body parts, and their demand in illegal markets is increasing. Wong and Krishnasamy (2019) reported that in India, tiger skin, bones, and claws were the most commonly seized items between 2000 and 2018. According to the current results, skin is the most traded or demanded part, followed by claws, bones, and teeth, corroborating previous observations. Other parts, such as paws, skulls, teeth, and whiskers, have a minor need in illegal markets and are in high demand in the Indian domestic market. Tiger skin is used in the decoration and manufacture of luxury furniture. Tiger bones are used in Asian traditional medicine to treat rheumatism (Ellis, 2013; Hitchens & Blakeslee, 2020; Moreto & Lemieux, 2015; Nowell, 2000; Still, 2003). Tiger bones are also used as bone wine in traditional Chinese medicine (Coals et al., 2020; Gratwicke et al., 2008). Tiger bone wine is available in the markets of China and Vietnam under brand names (Coals et al., 2020). Paws are used in black magic in some parts of India (Sethi et al., 2019). Tiger claws are primarily used in the creation of lockets, sculptures, religious items, and as medicine in traditional practise (Coals et al., 2020; Sharma et al., 2019). Because of the high demand for tiger claws, fake claws are manufactured for illegal marketing (Sharma et al., 2016), demonstrating the intensity of the demand for tiger claws. Tiger parts, such as skin, bones, claws, whiskers, canines, penises, and so on, are in demand in the market. Skidmore (2021), says that the demand for poached parts is different because people use different things for different purposes. Seizure trends. The seizure rate was showing a trend of increasing year-wise in India, which was correlated with the increased tiger density in various landscapes. It can also be argued that the increased seizure corresponds to the strict implementation of the Indian laws. We show that the seizure is more near the forested regions in India, where similar observations were reported in previous studies by Sharma et al. (2014). In the period of COVID-19 lockdown, there was a sharp increase in the number of seizures. Hence, it can be assumed that the major tiger habits in India are facing a huge tiger loss, which remains unaccounted. The efforts of the NTCA to broaden the number of tiger habitats in view of increasing the number of tigers are yielding good results (Jhala et al., 2020). Most of the seizures are based on information given to enforcement, and the convicts arrested are mostly tribes. The data clearly demonstrates that most seizures occur near the forest ranges and the arrested ones are tribes, when the tiger parts once handed over to the network are rarely seized. Other than the regions near reserves, the seizure mainly occurs near places of export. As per the analysis from 2000 to 2018 (Wong & Krishnasamy 2019), the highest rate of arrest for tiger smuggling is reported in India. From our observations, we predict four routes for the export of the seized parts: the Nepal border, Assam border, the Brahmaputra, and the Mumbai port. In most of the seizures, the end and start may only be traced, while the middle part of the network remains untraced. Once seized, the enforcement agencies do not make enough effort to find the track or origin of the sample. Trade routes. Land ways are majorly used in trafficking from India to China, with several routes passing through the states of Jammu (Upadhyay et al., 2007; Wong, 2016) and Kashmir (Upadhyay et al., 2007). Our analysis also shows that most of the tiger parts seizures occur during the time of transport via land roads. Smugglers may choose from a variety of nearby trade routes to avoid detection at border checkpoints. Most likely, traders in India will use road transportation to move tiger parts from one location to another. They may use water as a mode of transportation to transport the parts to other countries, as it is less dangerous than roads. Water, land, and air routes are used by the smugglers to trade the tiger parts to China (Moyle, 2009). The report of Oswell (2010) shows that tiger parts for China are easily traded through boats or roads. The Brahmaputra river may be the best way for traders from Assam and other nearby states to move their goods around. The river has a connection with the border of Tibet, an autonomous region of China. From China, tiger parts will flow to the neighbouring countries like Taiwan, Japan, and South Korea. Moreover, China acts as an intermediary distributor; they sell the items at a high rate to the neighbouring countries. Wong (2016) reported that smuggling routes pass through Ladakh between India and Tibet, which provides an opportunity for the Indian poachers to transport the tiger skin to the dealers in Shiquanhe (Western Tibet), which later reaches Lhasa and other regions of China through retailers. According to Paudel et al. (2020), tiger reserves such as Corbett, Udwan, and Katernighat are close to Nepal's western border, and the regions of both India's and Nepal's western and northern borders offer more direct trade routes and borders between India and China. Our analysis is similar to this report in that seizure cases occurred in the Sivalik range of India, which is close to the Corbett, Pilibhit, and Dudhwa tiger reserves and has a direct trade route to Nepal. Another possible trade route is through the Ganga, Yamuna, and Brahmaputra rivers, which flow directly to Tibet and from there through land route to China. Western Ghats un-noticed. There is a lot of trafficking going on in the Western Ghats and the tiger parts are being sent to the Mumbai port area. The Western Ghats being detected with high tiger confiscations, despite the fact that it is home to a large number of tiger reserves. The tiger protected areas of Western Ghats spotted in the tiger seizure analysis map are the Anamalai Tiger Reserve, Parambikulam Tiger Reserve, Mudhumalai Tiger Reserve, Nagarahole Tiger Reserve, Sathyamangalam Tiger Reserve, and Bandipur Tiger Reserve regions. However, no specific reports of tiger seizures were observed from the Nilgiri Hills. Another cluster of crimes was spotted in central India. More than 15 tiger reserves are in the central part of India, where Tadoba-Andhari, Melghat, Pench, Kanha, Bandhavarh, and Indravati are major tiger habitats. The states that cover the parts of the Brahmaputra River such as West Bengal, Assam, and Arunachal Pradesh, and the nearby areas with about nine tiger reserves is a major tiger smuggling site. Among them, Assam holds greater number of tiger seizure cases. According to the report by Wong and Krishnasamy (2019), the major affected tiger habitats are Bandipur, Nagarhole, Aanaimalai Tiger Reserve, Silent Valley, Kanha Tiger Reserve, Dongargarh-Dhaara, Malwenda forest complex, Bhander, Sundarban Park, Pilibhit Tiger Reserve, and Valmiki Tiger Reserve. We also found the similar spotting places for tiger seizures on the map and are located near the major Tiger Reserves. This also provides strong evidence that the majority of crimes occur near tiger habitats, despite the fact that Indian laws are strict. Future needs and conclusion. Furthermore, the primary question for forest officials and wildlife forensic experts is where the seized tiger parts came from. Is it from the same state or imported from another country? There are currently no precise molecular techniques for determining the origin of seized tiger parts. Despite the fact that there are studies that suggest microsatellites and SNP-based profiles to identify the origin of seized samples (Kolipakam et al., 2019; Natesh et al., 2017), no work has yet demonstrated a proper molecular marker to identify the reserve or meta-populations. Furthermore, it is highly unlikely that law enforcement will ever attempt to trace the geographical origin of the seized material. We say that it is very important to develop a better and more precise molecular forensic technique for the identification of tiger parts that have been seized. This helps to provide more strong and authentic evidence in court to punish the person who poached the tiger. This necessitates the collaboration of various researchers from across the country in order to share and analyse data. Identification of high-risk poaching areas can only help to reduce the illegal trade by increasing vigilance in these areas. This will aid in determining the circumstances that justify the tribal and locals' poaching the tiger. The situation can be handled by addressing the local poachers' economic and educational status. India is a country that spends a significant amount of public money on tiger conservation; a portion of this money could be diverted to address this serious issue. We forget about the future threat of poaching because of the satisfactory increase in the number of individuals recorded in the most recent tiger census. According to our findings, authorities should be more careful about protecting tigers from poaching and illegal trafficking. If that doesn't happen, many Indian tiger reserves will be like Sariska and Panna, which both had a lot of tigers killed. Supplementary Information ESM 1 (PDF 27 kb) ESM 2 (PDF 214 kb) Acknowledgements We acknowledge Dr. Shekhar Kumar Niraj IFS and Mr. K.K. Kausal IFS for the suggestions and the media desk people for their support in searching data. We also acknowledge Idea Wild, Colorado, USA for supporting this study by providing equipment’s to the second author. Funding The study is self-funded by the authors. Data availability All the data used in the present study is properly acknowledged or provided as supplementary materials. Code availability Not applicable. Declarations Conflicts of interest/Competing interests The authors declare that there is not financial and non-financial conflict of address between authors. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Anderson, K. (1995). Culture and nature at the Adelaide Zoo: At the frontiers of ‘human’ geography. Transactions of the Institute of British Geographers, 20 (3), 275–294. 10.2307/622652 Anderson K A walk on the wild side: A critical geography of domestication Progress in Human Geography 1997 21 463 485 10.1191/030913297673999021 Banks, D., & Newman, J. (2004). The Tiger skin trail. Environmental Investigation Agency, London, UK. https://eia-international.org/wp-content/uploads/TheTigerSkinTrail-Low-Res.pdf. Bennett CF Cultural animal geography: An inviting eld of research Professional Geographer 1960 12 5 12 14 10.1111/j.0033-0124.1960.125_12.x Bijoy, C.R. (2011). The great Indian tiger show. Economic and Political Weekly, 46(4), 36–41. https://www.jstor.org/stable/27918063 Buller H Animal geographies I Progress in Human Geography 2014 38 2 308 318 10.1177/0309132513479295 Coals P Moorhouse TP D’Cruze NC Macdonald DW Loveridge AJ Preferences for lion and tiger bone wines amongst the urban public in China and Vietnam Journal for Nature Conservation 2020 57 125874 10.1016/j.jnc.2020.125874 Davies G Philo C Wilbert C Virtual animals in electronic zoos: The changing geographies of animal capture and display Animal spaces, beastly places: New geographies of human-animal relations 2000 Routledge 243 246 Davis EO Willemsen M Dang V O’Connor D Glikman JA An updated analysis of the consumption of tiger products in urban Vietnam Global Ecology and Conservation 2020 22 e00960 10.1016/j.gecco.2020.e00960 Drury, R. (2011). Hungry for success: urban consumer demand for wild animal products in Vietnam. Conservation and Society, 9(3):247–257. https://www.jstor.org/stable/26393047 Elder G Wolch J Emel J Wolch J Emel J Le pratique sauvage: Race, place, and the human animaldivide Animal geographies: Place, politics, and identity in the nature-culture borderlands 1998 Verso 72 90 Ellis R Tiger bone & rhino horn: The destruction of wildlife for traditional Chinese medicine 2013 Island Press Emel J and Urbanik J (2010) Animal geographies: Exploring the spaces and places of human-animal encounters.In: DeMelloM(ed.) Teaching the Animal: Human AnimalStudies Across Disciplines. New York: Lantern Press, 202–217. Gill, A.K. (2014). Analyzing the Pathway to Improve Tiger Conservation in India. Madras School of Economics, Chennai, India. Gopal R Qureshi Q Bhardwaj M Singh RJ Jhala YV Evaluating the status of the endangered tiger Panthera tigris and its prey in Panna Tiger Reserve, Madhya Pradesh India. Oryx 2010 44 3 383 389 10.1017/S0030605310000529 Gratwicke B Mills J Dutton A Gabriel G Long B Seidensticker J Wright B You W Zhang L Attitudes toward consumption and conservation of tigers in China PloS one 2008 3 7 e2544 10.1371/journal.pone.0002544 18596926 Gruffudd P Philo C Wilbert C Biological cultivation: Lubetkin’s modernism at London Zoo in the 1930s Animal spaces, beastly places: New geographies of human-animal relations 2000 Routledge 222 242 Hitchens RT Blakeslee AM Trends in illegal wildlife trade: Analyzing personal baggage seizure data in the Pacific Northwest PloS one 2020 15 6 e0234197 10.1371/journal.pone.0234197 32520961 Hovorka AJ Animal geographies I: Globalizing and decolonizing Progress in Human Geography 2017 41 3 382 394 10.1177/0309132516646291 Hovorka AJ Animal geographies III: Species relations of power Progress in Human Geography 2019 43 4 749 757 10.1177/0309132518775837 Jhala, Y. V., Qureshi, Q., & Nayak, A. (Eds) (2020). The status of tigers, co-predators and prey in India 2018. National Tiger Conservation Authority, Government of India, New Delhi and Wildlife Institute of India Dehradun. ISBN 81–85496–50–1. Jhala Y Gopal R Mathur V Ghosh P Negi HS Narain S Yadav SP Malik A Garawad R Qureshi Q Recovery of tigers in India: Critical introspection and potential lessons People and Nature 2021 3 2 281 293 10.1002/pan3.10177 Jiao Y Yeophantong P Lee TM Strengthening International Legal Cooperation to Combat the Illegal Wildlife Trade Between Southeast Asia and China Frontiers in Ecology and Evolution 2021 9 105 10.3389/fevo.2021.645427 Jody Emel,1 Chris Wilbert, and Jennifer Wolch (2002) Animal Geographies Society & Animals 10:4 Koninklijke Brill NV, Leiden, Joshi AR Dinerstein E Wikramanayake E Anderson ML Olson D Jones BS Seidensticker J Lumpkin S Hansen MC Sizer NC Davis CL Tracking changes and preventing loss in critical tiger habitat Science Advances 2016 2 4 e1501675 10.1126/sciadv.1501675 27051881 Karanth, K.U., Nichols, J.D., Goodrich, J.M., Reddy, G.V., Mathur, V.B., Wibisono, H.T., Sunarto, S., Pattanavibool, A., & Gumal, M.T. (2017). Role of monitoring in global tiger conservation. In Methods For Monitoring Tiger And Prey Populations (pp. 1–13). Springer, Singapore. 10.1007/978-981-10-5436-5_1 Karmacharya D Sherchan AM Dulal S Manandhar P Manandhar S Joshi J Bhattarai S Bhatta TR Awasthi N Sharma AN Bista M Species, sex and geo-location identification of seized tiger (Panthera tigris tigris) parts in Nepal—A molecular forensic approach PloS one 2018 13 8 e0201639 10.1371/journal.pone.0201639 30138352 Kitpipit T Tobe SS Kitchener AC Gill P Linacre A The development and validation of a single SNaPshot multiplex for tiger species and subspecies identification—Implications for forensic purposes Forensic Science International: Genetics 2012 6 2 250 257 10.1016/j.fsigen.2011.06.001 21723800 Knoka AM Sawosz E Chwalibog A Reminder about tigers: Current status and conservation International Journal of Avian & Wildlife Biology 2018 3 2 98 99 Kolipakam V Singh S Pant B Qureshi Q Jhala YV Genetic structure of tigers (Panthera tigris tigris) in India and its implications for conservation Global Ecology and Conservation 2019 20 e00710 10.1016/j.gecco.2019.e00710 Kumar, A. (2021). Conservation Status of Bengal Tiger (Panthera tigris tigris)-A Review. Journal of Scientific Research, 65(2), 1–5. https://www.bhu.ac.in/research_pub/jsr/Volumes/JSR_65_02_2021/1.pdf Kumar U Awasthi N Qureshi Q Jhala Y Do conservation strategies that increase tiger populations have consequences for other wild carnivores like leopards? Scientific Reports 2019 9 1 1 8 10.1038/s41598-019-51213-w 30626917 Mainka, S. A., & Mills, J. A. (1995). Wildlife and traditional Chinese medicine: supply and demand for wildlife species. Journal of zoo and wildlife medicine, 193–200. Mileu, N., & Queiros, M. (2018). Development of a QGIS plugin to dasymetric mapping. In Free and open source software for geospatial (FOSS4G) conference proceedings, 18(1):9. Mills JA Jackson P Killed for a cure: A review of the worldwide trade in tiger bone 1994 Traffic International Mohan, G., Yogesh, J., Nittu G., Shameer, T. T., Backer, S. J., Nandhini, S., Ramakrishan, B., Jyothi, M., & Sanil, R. (2021). Factors influencing survival of tiger and leopard in the high-altitude ecosystem of the Nilgiris, India. Zoology and Ecology, 32(2), 116–133. 10.35513/21658005.2021.2.6. Moreto WD Lemieux AM From CRAVED to CAPTURED: Introducing a product-based framework to examine illegal wildlife markets European Journal on Criminal Policy and Research 2015 21 3 303 320 10.1007/s10610-014-9268-0 Morgan KI Ewart KM Nguyen TQ Sitam FT Ouitavon K Lightson AL Kotze A McEwing R Avoiding common numts to provide reliable species identification for tiger parts Forensic Science International: Reports 2021 3 100166 10.1016/j.fsir.2020.100166 Moyle B The black market in China for tiger products Global Crime 2009 10 1–2 124 143 10.1080/17440570902783921 Natesh M Atla G Nigam P Jhala YV Zachariah A Borthakur U Ramakrishnan U Conservation priorities for endangered Indian tigers through a genomic lens Scientific Reports 2017 7 1 1 11 10.1038/s41598-017-09748-3 28127051 Newbigin, M. I. (1913). Animal geography: The faunas of the natural regions of the globe.Oxford: Clarendon. Niraj SK Krausman PR Dayal V A stakeholder perspective into wildlife policy in India The Journal of Wildlife Management 2012 76 1 10 18 10.1002/jwmg.263 Nittu, G., Bhavana, P. M., Shameer, T. T., Ramakrishnan, B., Archana, R., Kaushal, K. K., Khedkar, G. D., Mohan, G., Jyothi, M. & Sanil, R. (2021). Simple Nested Allele-Specific approach with penultimate mismatch for precise species and sex identification of tiger and leopard. Molecular Biology Reports, 48(2), 1667–1676. 10.1007/s11033-021-06139-w. Nowell, K. (2000). Far from a cure: the tiger trade revisited. TRAFFIC International, Cambridge. https://www.traffic.org/site/assets/files/4014/far_from_a_cure.pdf Nowell, K. (2010). Tiger farms and pharmacies: the central importance of China’s trade policy for tiger conservation. In: Tigers of the World. William Andrew Publishing, 463–475. 10.1016/B978-0-8155-1570-8.00038-4 Oswell AH The big cat trade in Myanmar and Thailand 2010 Petaling Jaya, Selangor, Malaysia TRAFFIC Southeast Asia Palita, S. K. (2007). Royal Bengal Indian Tiger: Past, Present and Future–An Analysis in Orissa Context. Paudel PK Acharya KP Baral HS Heinen JT Jnawali SR Trends, patterns, and networks of illicit wildlife trade in Nepal: A national synthesis Conservation Science and Practice 2020 2 9 e247 10.1111/csp2.247 RStudio Team (2020). RStudio: Integrated Development for R. RStudio, PBC, Boston, MA URL http://www.rstudio.com/ (accessed 15 July 2021). Sethi, S., Goyal, S.P., & Choudhary, A.N. (2019). 12 Poaching, Illegal Wildlife Trade, and Bushmeat Hunting in India and South Asia. International wildlife management: Conservation challenges in a changing world, p.157. Sharma CP Sharma S Sharma V Singh R Rapid and non-destructive identification of claws using ATR-FTIR spectroscopy–A novel approach in wildlife forensics Science & Justice 2019 59 6 622 629 10.1016/j.scijus.2019.08.002 31606099 Sharma K Wright B Joseph T Desai N Tiger poaching and trafficking in India: Estimating rates of occurrence and detection over four decades Biological Conservation 2014 179 33 39 10.1016/j.biocon.2014.08.016 Sharma V Sharma CP Kumar VP Goyal SP Pioneer identification of fake tiger claws using morphometric and DNA-based analysis in wildlife forensics in India Forensic Science International 2016 266 226 233 10.1016/j.forsciint.2016.05.024 27322503 Shepherd CR Kufnerová J Cajthaml T Frouzová J Gomez L Bear trade in the Czech Republic: An analysis of legal and illegal international trade from 2005 to 2020 European Journal of Wildlife Research 2020 66 6 1 10 10.1007/s10344-020-01425-7 Skidmore A Using crime script analysis to elucidate the details of Amur tiger poaching in the Russian Far East Crime Science 2021 10 1 1 25 10.1186/s40163-021-00150-z Still Use of animal products in traditional Chinese medicine: Environmental impact and health hazards Complementary Therapies in Medicine 2003 11 2 118 122 10.1016/S0965-2299(03)00055-4 12801499 Stoner, S., Krishnasamy, K., Wittmann, T., Delean, S., & Cassey, P. (2016). Reduced to skin and bones re-examined: Full analysis. An analysis of Tiger seizures from 13 range countries from 2000-2015. TRAFFIC, Selangor, Malaysia. https://www.traffic.org/site/assets/files/2350/reduced-to-skin-and-bones-re-examined-full-analysis.pdf Tensen L Under what circumstances can wildlife farming benefit species conservation? Global Ecology and Conservation 2016 6 286 298 10.1016/j.gecco.2016.03.007 Ufkes FM Lean and mean: U.S. meat-packing in an era of agro-industrial restructuring Environment and Planning d: Society and Space 1995 13 683 706 10.1068/d130683 Upadhyay, S. K., Ali, S., & Sharma, K. B. (2007). Tiger Poaching and Trade in Asia: An Overview. In B.N. Pandey & G.K. Kulkarni (Eds) Biodiversity And Environment (pp 161–170), New Delhi, APH Publishing Corporation. Urbanik J (2012) Placing Animals: An Introduction to the Geography of Human-Animal Relations. Lanham, MD:Rowman and Littlefield. Valero-Mora PM ggplot2: Elegant graphics for data analysis Journal of Statistical Software 2010 35 1 1 3 21603108 van Uhm, D. (2020). Wildlife trafficking and criminogenic asymmetries in a globalised world. In: Routledge International Handbook of Green Criminology, Routledge, 529–542. Wang W Yang L Wronski T Chen S Hu Y Huang S Captive breeding of wildlife resources—China's revised supply-side approach to conservation Wildlife Society Bulletin 2019 43 3 425 435 10.1002/wsb.988 32327862 Wilbert, C. (2000). Anti-this-against-that: Resistances along a human non-human axis.In J. Sharp et al., (Eds.), Entanglements of power (pp. 238–255). London: Routledge. Wolch, J., & Emel, J. (1995). Bringing the animals back in. Environment and Planning D: Society and Space 13: 632–636. Wong RW The organization of the illegal tiger parts trade in China British Journal of Criminology 2016 56 5 995 1013 10.1093/bjc/azv080 Wong, R. W. (2019). China and the illegal wildlife trade. In: The Illegal Wildlife Trade in China. Palgrave Macmillan, Cham 13–38. 10.1007/978-3-030-13666-6_2 Wong RW The Illegal Wildlife Trade in China: Understanding the Distribution Networks. 2020 10.1007/s11417-020-09335-y Wong RW Shadow operations in wildlife trade under China’s Belt and Road Initiative China Information 2021 35 2 201 218 10.1177/0920203X20948680 Wong, R. & Krishnasamy, K. (2019). Skin and Bones Unresolved: An Analysis of Tiger Seizures from 2000–2018. TRAFFIC, Selangor, Malaysia. https://www.traffic.org/site/assets/files/12344/skin_and_bones_unresolved-web-1.pdf Wright, B. (2010). Will the tiger survive in India?. In Tigers of the World (pp. 87–100). William Andrew Publishing. Yarwood R Evans E Philo C Wilbert C Taking stock of farm animals and rurality Animal spaces, beastly places: New geographies of human-animalrelations 2000 Routledge 98 114
PMC009xxxxxx/PMC9005342.txt	==== Front Bull Malays Math Sci Soc Bull Malays Math Sci Soc Bulletin of the Malaysian Mathematical Sciences Society 0126-6705 2180-4206 Springer Nature Singapore Singapore 35431363 1287 10.1007/s40840-022-01287-z Article Regressive Class Modelling for Predicting Trajectories of COVID-19 Fatalities Using Statistical and Machine Learning Models Chowdhury Rafiqul I. rachowdhury@upei.ca 1 http://orcid.org/0000-0001-7802-2910 Hasan M. Tariqul thasan@unb.ca 2 Sneddon Gary 3 1 grid.139596.1 0000 0001 2167 8433 School of Mathematical and Computational Sciences, University of Prince Edward Island, Charlottetown, PEI C1A 4P3 Canada 2 grid.266820.8 0000 0004 0402 6152 Department of Mathematics and Statistics, University of New Brunswick, Fredericton, NB E3B 5A3 Canada 3 grid.260303.4 0000 0001 2186 9504 Department of Mathematics and Statistics, Mount Saint Vincent University, Halifax, NS B3M 2J6 Canada Communicated by Shahariar Huda. 13 4 2022 2022 45 Suppl 1 235250 10 1 2022 17 3 2022 20 3 2022 © The Author(s), under exclusive licence to Malaysian Mathematical Sciences Society and Penerbit Universiti Sains Malaysia 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. The COVID-19 (SARS-CoV-2 virus) pandemic has led to a substantial loss of human life worldwide by providing an unparalleled challenge to the public health system. The economic, psychological, and social disarray generated by the COVID-19 pandemic is devastating. Public health experts and epidemiologists worldwide are struggling to formulate policies on how to control this pandemic as there is no effective vaccine or treatment available which provide long-term immunity against different variants of COVID-19 and to eradicate this virus completely. As the new cases and fatalities are recorded daily or weekly, the responses are likely to be repeated or longitudinally correlated. Thus, studying the impact of available covariates and new cases on deaths from COVID-19 repeatedly would provide significant insights into this pandemic’s dynamics. For a better understanding of the dynamics of spread, in this paper, we study the impact of various risk factors on the new cases and deaths over time. To do that, we propose a marginal-conditional based joint modelling approach to predict trajectories, which is crucial to the health policy planners for taking necessary measures. The conditional model is a natural choice to study the underlying property of dependence in consecutive new cases and deaths. Using this model, one can examine the relationship between outcomes and predictors, and it is possible to calculate risks of the sequence of events repeatedly. The advantage of repeated measures is that one can see how individual responses change over time. The predictive accuracy of the proposed model is also compared with various machine learning techniques. The machine learning algorithms used in this paper are extended to accommodate repeated responses. The performance of the proposed model is illustrated using COVID-19 data collected from the Texas Health and Human Services. Keywords SARS-CoV-2 virus Repeated measures Model accuracy Deep learning techniques Joint modelling Mathematics Subject Classification 62H20 62F03 62F10 62N03 issue-copyright-statement© The Author(s), under exclusive licence to Malaysian Mathematical Sciences Society and Penerbit Universiti Sains Malaysia 2022 ==== Body pmcIntroduction Countries around the world are overwhelmed by the outbreak of the COVID-19 pandemic, as the SARS-CoV-2 virus (COVID-19) has been transmitted across the globe at an alarming rate. As of December 31, 2021, according to World health organization (WHO) [1], a total confirmed new cases and deaths are over 281 millions and 5.4 millions, respectively. Due to its astonishing nature of transmission, public health experts and epidemiologists are struggling to formulate policies on how to control and end this pandemic. Health policy planners from all the countries are trying to flatten the curves for both the new cases and deaths of COVID-19. However, public health and government officials are witnessing a recent new spike of cases in countries where the spread of the COVID-19 was once assumed to be under control. Thus, the initial flattened curves of COVID-19 do not mean the crisis is over. Health officials introduced various restrictions on people’s mobility and socialization to reduce the spread and hence death. Consequently, the socio-economic, psychological, and other impacts of COVID-19 on people have already started to become visible continuously [1, 2]. However, as there has been more than one wave of the virus already, there could be another increase or spike in cases after the waves of infections stabilized. Besides, this pandemic curve can have multiple severe surges or spikes in new cases before it stabilizes with more effective doses of vaccine. Lifting of existing restrictions to control this pandemic too quickly could send countries around the world ”back into confinement” once a consecutive wave hits. To identify the COVID-19 transmission patterns, spread and development, various statistical modelling techniques were employed by health researchers around the world since the beginning of 2020 [3–9]. For example, to predict COVID-19-related new cases and deaths, autoregressive integrated moving average (ARIMA)-based modelling techniques were incorporated by researchers in the literature ([3–7]). Shastri et al [8] proposed recurrent neural network (RNN)-based deep learning techniques to forecast COVID-19 new cases and death. To predict the dynamic deportment of COVID-19, Da Silva et. al. [9] have exploited artificial intelligence (AI)-based modelling techniques such as Bayesian regression neural network, cubist regression, k-nearest neighbours, quantile random forest, and support vector regression. To predict the transmission of COVID-19, machine learning algorithm has also been employed in the literature [10]. Our research is motivated by COVID-19 data collected from the Texas Health and Human Services website (https://dshs.texas.gov/coronavirus///additionaldata.aspx). In this paper, we have considered the daily new cases and deaths for 254 counties in Texas between March 4, 2020, and July 25, 2020, along with the population size for each county. As some counties have no daily deaths for several days, we summed weekly deaths and new cases for our analysis. Thus, for modelling, we have used any death or no death during a week for a total of 20 consecutive weeks. As COVID-19 is causing excess fatalities all over the world, predicting fatality trajectories has gained compelling interest among public health researchers. In this paper, we are interested in predicting the trajectory risks of weekly death and the impact of available risk factors on the trajectory. As the repeated weekly deaths in a specific county are likely to be correlated, we propose a joint modelling-based regressive class approach to estimate the trajectory risks [11, 12]. Studying the impact of available covariates and new cases on deaths from COVID-19 longitudinally would provide significant insights into this pandemic’s dynamics. To better understand the dynamics of spread, we need to study the impact of risk factors and new cases on deaths over time, first conditionally and then jointly, which is crucial to the health policy planners for taking necessary measures. The conditional model is a natural choice to study the underlying property of dependence in consecutive weeks [13]. Using this model, one can examine the relationship between outcomes and predictors, and it is possible to calculate risks of the sequence of events longitudinally [14]. Also, we must understand how the transitions between weeks occur and how the covariates influence these transitions for better prediction. The advantage of repeated measures is that one can see how individual responses change over time [15]. Lindsey and Lamber [15] suggested that this must generally be conditioned on the previous history of a subject. Lee and Nelder [16] concluded that the conditional models are of fundamental interest, and marginal predictions can be made from conditional models. Sometimes the transition probabilities may depend on different models, each model representing a transition from one stage to another. Lindsey and Lamber [15] examined some important theoretical aspects concerning the use of marginal models (e.g. GEE) and demonstrated various limitations. Another alternative is the subject-specific models considering the random effects by allowing random effect terms in the linear predictor [17]. However, the proposed method provides a more general and flexible set up for addressing the risk prediction of repeated categorical outcomes emerging from longitudinal studies. The regressive models at the subsequent follow-ups provide the estimates of the parameters of the conditional models. We organize the rest of the paper as follows. In Sect. 2, we present proposed joint modelling approach for the repeated measurements to predict fatality trajectories along with various machine learning techniques. Estimation techniques of the model parameters are presented in Sect. 3. The results of data analysis are presented in Sect. 4. Finally, in Sect. 5, some concluding remarks are presented. Predicting Fatality Trajectories for COVID-19 In this section, we present a proposed marginal-conditional based joint modelling approach for predicting trajectories of repeated COVID-19 deaths. In addition, we also incorporate three machine learning algorithms neural network (NN), support vector machine (SVM) and random forest (RF), and compared their performance with the proposed method. The chosen machine learning algorithms have widespread use in various fields. However, these machine learning algorithms are developed for the classification of categorical outcomes measured cross-sectionally or data collected from a single time point. These models also allow the prediction of the probability of the outcome categories. But the problem at our hands is longitudinal, and we are interested in predicting trajectory risk. In other words, we want to predict the joint probability of a sequence of events that occurs longitudinally. Besides, we need to assess the covariate impact on the series of events. We extended these machine learning algorithms using the proposed marginal conditional approach to obtain the marginal and conditional probability and, ultimately, the joint probability to predict the trajectory risks. In other words, we extend these machine learning models for repeated measures data to predict the trajectory risks based on the covariates. First we present proposed joint modelling approach to predict fatalities trajectories for COVID-19 in Sect. 2.1. We then present various machine learning algorithms in Sect. 2.2. Joint Modelling Approach Let Yi1,Yi2,...,YiJi represent the occurrence of COVID-19-related death in the i-th county of Texas, USA at j-th week (i=1,2,...,nandj=1,2,...,Ji). Note that the data set we used in this paper has n=254 and Ji=J=20. We also assume Yij=s follows a binomial distribution where (s=0,1) with the category 0 denotes no COVID-19 related death and 1 denotes COVID-19 related death during a week for the ith county in Texas, USA. To be specific, in Fig. 1, we present the trajectory path using three consecutive weeks for the ith county.Fig. 1 Trajectory path for three consecutive weeks for the ith county The marginal trajectory probability of the response at the first week Yi1 for the ith county can be calculated using the following formula:1 Pis.y1(zi1)=P(Yi1=s∣zi1)=e(zij′βy1)s1+e(zij′βy1),s,=0,1. where βy1=[β0,β1,...,βp]′ is a 1×(p+1) regression coefficients corresponding to the covariate vector z=[1,z1,...,zp]′=[1,x1,...,xp]′. Similarly, the conditional trajectory probability of the response at the second week Yi2 given the first week Yi1 can be calculated using the first-order regressive logistic models as:2 Pis.y2∣y1(zi2)=P(Yi2=s∣Yi1=yi1,zij)=e(zij′βy2.y1)s1+e(zij′βy2.y1),s,yi1=0,1, where βy2.y1=βy20,βy21,...,βy2p,βy2y1′ is 1×(p+2) vector of regression coefficients corresponding to the covariate vector z=[1,z1,...,zp,zp+1]′=[1,x1,...,xp,y1]′. Thus, the general form of the first-order regressive logistic models for j consecutive outcomes Yi1,⋯,Yij can be shown to be:3 Pis.yj∣yj-1(zij)=P(Yij=s∣Yi(j-1)=yi(j-1),zij)=e(zij′βyj.yj-1)s1+e(zij′βyj.yj-1),s,yij-1=0,1, where βyj.yj-1=βyj0,βyj1,...,βyjp,,βyjyj-1′ is a 1×(p+2) vector of regression coefficients corresponding to the covariate vector z=1,z1,...,zp,zp+1′=1,x1,...,xp,yj-1′. Consequently, the joint probability mass function of Yi1,Yi2,...,YiJ with covariate vector X=x can be expressed as:4 P(Yi1=yi1,Yi2=yi2,...,YiJ=yiJ∣X=x)=P(Yi1=yi1∣X=x)×P(Yi2=yi2∣Yi1=yi1;X=x)×...×P(YiJ=yiJ∣Yi(j-1)=yi(j-1);X=x)=Pis.yi1(x)×Pis.yi2∣yi1(x)×...×Pis.yiJ∣yi(J-1)(x),s,yi1,...,yij-1=0,1, where X′=[1,x1,...,xp] is vector of covariates. It should be noted that X=x can be time dependent. Now using the fitted marginal and conditional models, we can estimate the trajectory risks as follows:5 P^(Yi1=yi1,Yi2=yi2,...,YiJ=yiJ∣X=x)=P^(Yi1=yi1∣X=x)×P^(Yi2=yi2∣Yi1=yi1;X=x)×...×P^(YiJ=yiJ∣Yi(j-1)=yi(j-1);X=x)=P^is.yi1(x)×P^is.yi2∣yi1(x)×...×P^is.yiJ∣yi(J-1)(x),s,yi1,...,yij-1=0,1, In the next subsection, we discuss various machine learning techniques to predict fatalities of COVID-19. Machine Learning Approaches In this section, we present three machine learning algorithms: neural network (NN), support vector machine (SVM) and random forest (RF), which will be used to predict trajectories. Although there are other machine learning approaches available, we considered these three commonly used techniques for illustration and comparisons. It may be noted that these machine learning algorithms are developed for cross-sectional data or data from a single time point. However, using the proposed approach, we extended these algorithms for trajectory risk prediction for longitudinal events. Neural Networks (NN) Neural networks are often used as a building block for deep learning due to the excellent predictive performance. The NN algorithm is tuned by constructing many layers and neurons, then trimming off the unnecessary neurons and layers using regularization through cross-validation. A neural network does not assess associations between predictors and the response as its objective is to predict outcome class. Also, multicollinearity does not create many problems in the neural network. More details of the theory of a neural network can be found in Ripley [18]. Support Vector Machine (SVM) A popular classifier, support vector machine (SVM), was originally introduced by Boser et. al. [19]. It has been updated by Cortes and Vapnik [20]. The main idea is to separate the space of the feature variables using hyperplanes so that the response classes become as distinct as possible. Also, SVM is not affected by the presence of multicollinearity. This model often provides excellent predictive performance and becomes intuitive, visualizing the support vectors in the lower dimension. This algorithm is computationally demanding, and much attention is required to tune the model parameters. Random Forests (RF) A random forest is an aggregated collection of classifiers of classification trees (an ensemble). This prediction model is compelling and provides one of the top prediction performances. The strength of the ensemble depends on the depths of the constituent classification trees and diversity between the trees. The RF needs to tuned using the number of trees to grow in a forest. The random selection of independent variables keeps the correlation between predictors to the minimal. When the objective is to make better predictions, multicollinearity does not affect the results. However, the algorithm may appear computationally demanding for high-dimensional data. Estimation of Model Parameters In this section, we develop estimation techniques for estimating model parameters to predict trajectories using the proposed approach. To do that, we obtain the log-likelihood function of the joint mass function in (4) which can be expressed as:6 l=∑i=1n∑j=1JlnP(Yi1=yi1,Yi2=yi2,⋯,YiJ=yiJ∣x)=∑i=1n∑j=1J[lnP(Yi1=yi1∣x)+lnP(Yi2=yi2∣yi1;x)+⋯+lnP(YiJ=s∣Yi(j-1)=yi(j-1);x)]=∑i=1n∑j=1J[lnPis.yi1(xij)+lnPis.yi2∣yi1(x)+⋯+lnPis.yiJ∣yi(j-1)(x)] From Eq. (6), it is clear that the joint mass function boils down to univariate mass function for any first-order regressive model. In other words, all the first-order regressive models are equivalent to that of the marginal model and follow the same estimation procedure. Then, the likelihood function of the first-order regressive model for sample of n independent observations can be written as:ln[Pis.yij∣yi(j-1)(z)]=∑i=1n[Yis.yjzij′βyj.yj-1-log[1+e(zij′βyj.yj-1)]] where z=[1,z1,z2,...,xz,y(j-1)]′=[1,x1,x2,...,xp,y(j-1)]′. Differentiating the log-likelihood with respect to the parameters and equating the derivatives to zero, we obtain the equations whose solutions give the following maximum likelihood estimates for (p+2) parameters.∂ln[Pis.yij∣yi(j-1)(z)]∂βjq=∑i=1n[Yis.yj-Pis.yij∣yi(j-1)(z)]zq,q=0,1,...,p,(p+1). Thus, the observed information matrix can be obtained using the second derivatives as follows:∂2ln[Pis.yij∣yi(j-1)(z)]∂βjq∂βjq′=∑i=1n[Pis.yij∣yi(j-1)(z){1-Pis.yij∣yi(j-1)(z)}]zq′zq, and∂2ln[Pis.yij∣yi(j-1)(z)]∂βjq∂βjq′=∑i=1n[Pis.yij∣yi(j-1)(z)Pis.yij∣yi(j-1)(z)]zq′zq, whereq,q′=0,1,...,p,(p+1). The observed information matrix I(β) is the (p+2)×(p+2) matrix where elements are the negative of the second derivatives. It can be shown that the asymptotic covariance matrix is [I(β)]-1. Consequently, we can use a Newton–Raphson iterative logarithm to obtain the estimated regressive logistic coefficients. To do that, we first arrange the data into a new structure. Then, all these models can be estimated in a usual manner with an appropriate data structure and using Python, R, or other software capable of fitting logistic regression. Then, we can exploit the parallel programming to analyse all the subsets of data using multiple cores in a single computer or using several CPU in a distributed system. Significance of the Joint Model The significance of the joint model based on Eq. (4) can be tested using a likelihood ratio test between the joint constant only model (reduced) and joint full model (full) as follows:7 -2lnLReduced(β^Reduced)-lnLFull(β^Full) which is distributed asymptotically as χ2 with [(p+1)+{(p+2)×(j-1)}]-j degrees of freedom. It is noteworthy to point out that β^Reduced includes all the regression parameters from the constant only joint model and β^Full includes all the parameters from the full joint model. Predictive Models and Joint Probabilities The predicted joint probability that a subject with covariate vector (X∗=x∗) for a trajectory as shown in Fig. 1 can be predicted using Eq. (5) and using the predicted marginal first-order conditional probabilities from the fitted marginal and all first-order regressive logistic models. Based on Eq. (5), the predicted joint probabilities for Yi1=yi1 and Yi2=yi2 are:P^yi1,yi2(x)=P^(Yi1=yi1,Yi2=yi2∣x)=P^(Yi1=yi1∣X=x)×P^(Yi2=yi2∣Yi1=yi1;X=x)=P^is.yi1(x)×P^is.yi2∣yi1(x). Now, the predicted marginal probabilities P^(Yi1=1∣X=x) and P^(Yi1=0∣X=x) can be estimated from the fitted marginal model. The first-order conditional probabilities P^(Yi2=s∣Yi1=yi1;X=x), s,y1=0,1, can be estimated from the fitted first-order regressive model using covariate vector Z′=[x∗,yi1] and X∗=x∗. Then, we can plug-in the predicted marginal and conditional probabilities in Eq. (5) to obtain the joint probabilities for events. Analysis of Texas COVID-19 Data Texas COVID-19 data used in this paper were downloaded from the Texas Health and Human Services website (https://dshs.texas.gov/coronavirus///additionaldata.aspx). It provides daily new cases and deaths for 254 counties in Texas. We used data from March 4, 2020, to July 25, 2020. Also, the population size for each county is considered as covariate, which ranges from 92 to 4978845 for different counties. As some counties do not have any daily fatality for many days, we summed weekly deaths and new cases for our analysis. For modelling purposes, we have converted death as binary variable with any death as 1 or no death as 0, as the goal of public health experts and scientists are to minimize the death as 0. In our analysis, we have considered death as repeated binary outcome variable which repeats over the 20 weeks and are defined as Yi1,⋯,Yi20, respectively. We are interested in predicting the trajectory risks of death over the weeks and the impact of available risk factors on the trajectory. Due to the small sample size, i.e. only 254 counties in Texas, we divide the data into training (90%) and test (10%) sets based on the first-week samples to determine over(under)fitting and to assess the generalization ability of the models used. We used R and parallel programming package to fit all the models simultaneously. Table 1 displays the distribution of weekly death.Table 1 Distribution of deaths for consecutive weeks Fatality Weeks 1 2 3 4 5 6 7 8 9 10 No deaths 221 119 99 106 94 96 84 41 26 20 Any deaths 33 135 155 148 160 158 170 213 228 234 Total 254 254 254 254 254 254 254 254 254 254 Fatality Weeks 11 12 13 14 15 16 17 18 19 20 No deaths 249 228 207 209 211 214 205 213 191 175 Any deaths 5 26 47 45 43 40 49 41 63 79 Total 254 254 254 254 254 254 254 254 254 254 Table 2 Training and test data accuracy for different models Weeks Models Proposed model NN SVM RF Train Test Train Test Train Test Train Test 1 0.925 0.964 0.938 0.964 0.925 0.964 0.983 1.000 2 0.841 0.964 0.845 0.964 0.181 0.143 0.904 0.933 3 0.801 0.750 0.841 0.821 0.195 0.357 0.904 1.000 4 0.805 0.857 0.801 0.857 0.199 0.143 0.921 0.867 5 0.832 0.786 0.832 0.786 0.177 0.143 0.891 0.733 6 0.827 0.821 0.832 0.821 0.208 0.179 0.891 0.800 7 0.796 0.857 0.819 0.786 0.204 0.143 0.904 0.867 8 0.889 0.929 0.889 0.929 0.128 0.179 0.950 0.933 9 0.903 0.964 0.907 0.964 0.102 0.107 0.975 1.000 10 0.916 1.000 0.934 0.036 0.088 1.000 0.975 1.000 11 0.987 1.000 0.982 0.964 0.982 0.964 0.992 1.000 12 0.925 0.893 0.898 0.893 0.925 0.857 0.962 1.000 13 0.885 0.964 0.885 0.964 0.881 0.964 0.925 1.000 14 0.889 0.929 0.903 0.929 0.894 0.964 0.933 0.933 15 0.916 0.786 0.920 0.786 0.920 0.786 0.921 0.933 16 0.889 0.857 0.850 0.786 0.894 0.857 0.933 0.933 17 0.863 0.821 0.867 0.821 0.867 0.786 0.925 1.000 18 0.898 0.857 0.889 0.857 0.903 0.857 0.925 0.933 19 0.845 0.821 0.850 0.786 0.150 0.179 0.904 1.000 20 0.854 0.821 0.845 0.821 0.841 0.750 0.887 0.867 The prediction accuracy of the proposed joint model along with the three machine learning techniques is presented in Table 2. In general, the weekly accuracy of all three models such as proposed regressive model, neural network (NN), random forest (RF) is relatively very high compared to support vector machine (SVM) technique. The accuracy of SVM is low for most of the weeks. The same is true for both the training and test data. Also, we observed overfitting a few models. Regarding the model performance, comparing the accuracy of all these three models, it is found that the random forest model showed the highest accuracy, followed by the proposed regressive model and neural network, respectively. The fitted coefficients for marginal and proposed regressive models are shown in Table 3. The population size is positive and significantly associated with the outcomes for all the marginal and first-order regressive models with few exceptions. A similar pattern is observed between daily new cases and death. Daily new counts of cases for all the first-order regressive models are positive and significantly associated with the outcomes except for some. While we assessed the effect of covariates using proposed statistical learning models as shown in Table 3, we could not do so using machine learning models as those have interpretation problems. But we used those as one of our objectives is to predict fatality.Table 3 Parameter estimates, standard errors (SE) and P-values using proposed marginal conditional modelling approach Covariates Weeks β^ SE p-value Weeks β^ SE p-value Constant Week 1 – 3.14 0.34 0.00 Week 11 – 20.61 3961.56 1.00 New cases – 1.11 3.11 0.72 Population 0.00 0.00 0.00 0.00 0.00 0.00 Deaths 15.43 3961.56 1.00 Constant Week 2 – 2.36 0.34 0.00 Week 12 – 2.99 0.32 0.00 New cases 1.36 0.51 0.01 0.77 1.15 0.51 Population 0.00 0.00 0.00 0.00 0.00 0.00 Deaths 0.10 1.13 0.93 0.21 2.84 0.94 Constant Week 3 – 1.62 0.29 0.00 Week 13 – 2.57 0.28 0.00 New cases 1.15 0.40 0.00 1.77 0.55 0.00 Population 0.00 0.00 0.00 0.00 0.00 0.01 Deaths 0.80 0.43 0.07 0.45 0.74 0.54 Constant Week 4 – 2.05 0.33 0.00 Week 14 – 2.84 0.31 0.00 New cases 1.52 0.40 0.00 1.09 0.56 0.05 Population 0.00 0.00 0.00 0.00 0.00 0.00 Deaths 0.29 0.41 0.49 0.92 0.55 0.10 Constant Week 5 – 2.22 0.36 0.00 Week 15 – 3.12 0.35 0.00 New cases 1.54 0.44 0.00 1.63 0.69 0.02 Population 0.00 0.00 0.00 0.00 0.00 0.08 Deaths 1.19 0.43 0.01 1.82 0.61 0.00 Constant Week 6 – 1.93 0.33 0.00 Week 16 – 2.93 0.32 0.00 New cases 0.72 0.40 0.07 2.03 0.54 0.00 Population 0.00 0.00 0.00 0.00 0.00 0.11 Deaths 1.06 0.41 0.01 1.11 0.60 0.06 Constant Week 7 – 1.21 0.28 0.00 Week 17 – 2.58 0.28 0.00 New cases 1.31 0.39 0.00 1.00 0.54 0.07 Population 0.00 0.00 0.08 0.00 0.00 0.01 Deaths 1.28 0.39 0.00 1.71 0.52 0.00 Constant Week 8 – 1.15 0.39 0.00 Week 18 – 2.77 0.30 0.00 New cases 2.04 0.55 0.00 0.73 0.58 0.21 Population 0.00 0.00 0.01 0.00 0.00 0.00 Deaths 0.57 0.56 0.31 0.78 0.55 0.16 Constant Week 9 – 0.68 0.44 0.13 Week 19 – 2.34 0.27 0.00 New cases 1.69 0.61 0.01 0.47 0.63 0.46 Population 0.00 0.00 0.15 0.00 0.00 0.00 Deaths 1.40 0.62 0.02 0.83 0.55 0.13 Constant Week 10 – 0.61 0.55 0.27 Week 20 – 2.45 0.29 0.00 New cases 1.70 0.66 0.01 1.03 0.44 0.02 Population 0.00 0.00 0.42 0.00 0.00 0.00 Deaths 1.76 0.63 0.00 1.25 0.46 0.01 The predicted trajectory of conditional probabilities for Harris county is displayed in Fig. 2. The population of this county is 4978845. We presented the predicted trajectories using all four models. Both the proposed regressive and random forest models predicted the highest risks of deaths over time and these two lines overlapped. The dashed line is from the proposed model and the dashed-dotted line from the random forest models. The neural network (dotted line) closely follows these trajectories. The trajectory (the solid bottom line) predicted using the support vector machine showed the worst performance. This county has the highest population size and observed death in all the 20 consecutive weeks. However, there were no new cases during week 11 for this county. The predicted trajectories using the proposed regressive model and random forest are similar and predicted the highest risk, which is expected. However, the support vector machine algorithm predicted much lower risks and downward trends up to the 10th week. This may be due to no new cases during the 11th week. The corresponding trajectories of joint probabilities are shown in Fig. 3. After the pre-processing, the data for two counties (Anderson and Archer) used for model fitting are shown in Table 4. In Fig. 4, we present the trajectory of conditional probabilities for Andrewes county using all the models. The population size of this county is much smaller (22269) compared to Harris county. Also, this county did not observe death during the majority of the weeks. It is clear from the figure that the probability of death is lower than 0.50 or close to zero when no new cases are observed during a week. This flattening trend of lowering death risk is clearly visible from the trajectories of joint probabilities shown in Fig. 5. In general, the predicted paths from all the models are relatively close except for support vector machine.Fig. 2 Trajectory of conditional probabilities for Harris county using four models Fig. 3 Trajectory of joint probabilities for Harris county using four models Fig. 4 Trajectory of conditional probabilities for Andrews county using four models Fig. 5 Trajectory of joint probabilities for Andrews county using four models Conclusions This paper proposed a marginal-conditional modelling approach to predict trajectory risks of fatalities from COVID-19 longitudinally using daily new cases and other available covariates from 254 counties from Texas in the USA. We assess the impact of weekly new cases and the county population size on the trajectories. Also, we extended some commonly used machine learning algorithms that use for the classification of categorical outcomes from cross-sectional data for the trajectory risk predictions from repeated measures data. It is evident from our study that controlling the occurrence of new cases reduces the risk of death. Also, the trajectory risk of death is much higher for the densely populated county. Relevant authorities of all countries are trying hard to reduce the new cases of COVID-19 towards zero. Only by ending the pandemic everywhere, we can achieve this goal.Table 4 Pre-processed data for model fitting from Anderson and Archer counties County Nd Nc Population County Nd Nc Population Week Anderson 0 0 58199 Archer 0 0 1948 1 Anderson 1 0 58199 Archer 0 0 1948 2 Anderson 1 1 58199 Archer 1 0 1948 3 Anderson 1 1 58199 Archer 1 0 1948 4 Anderson 1 1 58199 Archer 0 0 1948 5 Anderson 1 1 58199 Archer 1 0 1948 6 Anderson 1 1 58199 Archer 0 0 1948 7 Anderson 1 1 58199 Archer 0 1 1948 8 Anderson 1 1 58199 Archer 0 0 1948 9 Anderson 1 1 58199 Archer 1 0 1948 10 Anderson 0 0 58199 Archer 0 0 1948 11 Anderson 0 0 58199 Archer 0 0 1948 12 Anderson 0 0 58199 Archer 0 0 1948 13 Anderson 0 0 58199 Archer 0 0 1948 14 Anderson 0 0 58199 Archer 0 0 1948 15 Anderson 0 0 58199 Archer 0 0 1948 16 Anderson 0 0 58199 Archer 0 0 1948 17 Anderson 0 1 58199 Archer 0 0 1948 18 Anderson 1 0 58199 Archer 0 0 1948 19 Anderson 0 1 58199 Archer 0 0 1948 20 Nd: weekly death from COVID 19: 0 = No death and 1 = 1 or more deaths Nc: Number of weekly new cases Governments around the world are implementing various strict restrictions, including travel bands, social distancing, and unnecessary movements, besides travel restriction and quarantine of both suspected individuals and subjects who have had close contacts with suspected cases. The aim is that reducing new infections towards zero will reduce the spread of this disease hence death. While people around the world overwhelmingly support pandemic-related restrictions, but many communities became impatient for these restrictions and want those removed. A positive association of population size with death from our studies may reiterate the need for those restrictions for highly populated places to reduce the new cases and hence an increasing number of fatalities. We believe that analysis with the availability of more detailed data along with related risk factors may provide an in-depth understanding regarding the dynamics. Acknowledgements This research was supported by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC). Declarations Conflict of interest The authors declare that they do not have any conflict of interest to declare. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. WHO Coronavirus Disease (COVID-19) Dashboard, Map Data, 2020. Available from: https://covid19.who.int/ 2. Carpenter TE O’Brien JM Hagerman AD Epidemic and economic impacts ofdelayeddetectionoffoot-and-mouthdisease: acasestudyofasimulatedoutbreakinCalifornia J. Vet. Diagn. Invest. 2010 23 26 33 10.1177/104063871102300104 3. Sahai AK Rath N Sood V Singh MP ARIMA modelling & forecasting of COVID-19 in top five affected countries Diabetes Metab. Syndr.: Clin. Res. Rev. 2020 14 5 1419 1427 10.1016/j.dsx.2020.07.042 4. Dal Molin Ribeiro MH da Silva RG Mariani VC dos Santos Coelho L Short-term forecasting COVID-19 cumulative confirmed cases: perspectives for Brazil Chaos Solitons Fractals 2020 10.1016/j.chaos.2020.109853 5. Vena, A., Giacobbe, D. R., Di Biagio, A., Mikulska, M., Taramasso, L., De Maria, A., et. al.: Clinical characteristics, management and in-hospital mortality of patients with coronavirus disease 2019 in Genoa, Italy, Clinical Microbiology and Infection, Available online 15 August 2020, doi: 10.1016/j.cmi.2020.07.049 6. Ceylan Z Estimation of COVID-19 prevalence in Italy, Spain, and France Sci. Total Environ. 2020 729 10 10.1016/j.scitotenv.2020.138817 7. Yang Q Wang J Ma H Wang X Research on COVID-19 based on ARIMA modelΔ-Taking Hubei, China as an example to see the epidemic in Italy J. Infect. Public Health 2020 13 10 1415 1418 10.1016/j.jiph.2020.06.019 32624404 8. Shastri S Singh K Kumar S Kour P Mansotra V Time series forecasting of Covid-19 using deep learning models: India-USA comparative case study, Chaos Solitons Fractals 2020 10.1016/j.chaos.2020.110227 9. da Silva RG Dal Molin Ribeiro MH Mariani VC dos Santos Coelho L Forecasting Brazilian and American COVID-19 cases based on artificial intelligence coupled with climatic exogenous variables, Chaos Solitons Fractals 2020 10.1016/j.chaos.2020.110027 10. Tuli S Shikhar T Rakesh T Rakesh SS Gill Predicting the growth and trend of COVID-19 pandemic using machine learning and cloud computing Internet Things 2020 10.1016/j.iot.2020.100222 11. Islam MA Chowdhury RI Prediction of disease status: a regressive model approach for repeated measures Stat. Methodol. 2010 7 520 540 10.1016/j.stamet.2010.03.001 12. Chowdhury RI Islam MA Regressive models for risk prediction for repeated multinomial outcomes: an illustration using Health and Retirement Study (HRS) data Biometric. J. 2020 62 898 915 10.1002/bimj.201800101 13. Islam MA Chowdhury RI Huda S Markov Models with Covariate Dependence for Repeated Measures 2009 New York Nova Science 14. Islam MA Chowdhury RI Singh KP A Markov model for analyzing polytomous outcome data Pak. J. Stat. Op. Res. 2012 8 593 603 10.18187/pjsor.v8i3.530 15. Lindsey JK Lamber P P, On the appropriateness of marginal models for repeated measurements in clinical trials Stat. Med. 1998 17 447 469 10.1002/(SICI)1097-0258(19980228)17:4<447::AID-SIM752>3.0.CO;2-G 9496722 16. Lee Y Nelder JA Conditional and marginal models: another view Stat. Sci. 2004 19 219 238 10.1214/088342304000000305 17. Breslow NE Clayton DG Approximate inference in generalized linear mixed models J. Am. Stat. Assoc. 1993 88 9 25 18. Ripley Brian D Pattern Recognition and Neural Networks 1996 Cambridge Cambridge University Press 19. Boser, B.E., Guyon, I.M. and Vapnik, V.N.: A training algorithm for optimal margin classifiers. 5th Annual ACM Workshop on COLT. Pittsburgh, PA, ACM Press, pp. 144-152 (1992) 20. Cortes C Vapnik V Support-vector networks Machine Learn. 1995 20 273 297 10.1007/BF00994018
PMC009xxxxxx/PMC9005343.txt	==== Front Theor Appl Climatol Theor Appl Climatol Theoretical and Applied Climatology 0177-798X 1434-4483 Springer Vienna Vienna 35431378 4041 10.1007/s00704-022-04041-4 Original Paper Climate change perception in Romania http://orcid.org/0000-0001-6412-1918 Cheval Sorin sorin.cheval@meteoromania.ro 1 Bulai Ana 2 Croitoru Adina-Eliza 34 Dorondel Ștefan 5 Micu Dana 1 Mihăilă Dumitru 6 Sfîcă Lucian 7 Tișcovschi Adrian 8 1 grid.425939.0 0000 0004 0495 5672 Department of Climatology, National Meteorological Administration, 97 Sos. București-Ploiești, 01686 Bucharest, Romania 2 AB European Research Group, 1H, Rubinului Street, 077025 Bragadiru, Ilfov, Romania 3 grid.7399.4 0000 0004 1937 1397 Department of Physical and Technical Geography, Faculty of Geography, Babeş-Bolyai University, 5-7 Clinicilor Street, 400006 Cluj-Napoca, Romania 4 grid.7399.4 0000 0004 1937 1397 Faculty of Geography, Babeş-Bolyai University, Research Centre for Sustainable Development, 5-7 Clinicilor Str, 400006 Cluj-Napoca, Romania 5 Francisc I, Rainer Institute of Anthropology, Bucharest, 13 Calea 13 Septembrie, 050731 Bucharest, Romania 6 grid.12056.30 0000 0001 2163 6372 Department of Geography, Faculty of History and Geography, Ștefan Cel Mare University of Suceava, 13 Universităţii Str, 720229 Suceava, Romania 7 grid.8168.7 0000000419371784 Department of Geography, Faculty of Geography and Geology, Alexandru Ioan Cuza University of Iași, 20A Carol I Blvd, 700506 Iași, Romania 8 grid.5100.4 0000 0001 2322 497X Faculty of Geography, University of Bucharest, 1 Nicolae Bălcescu Blvd, 010041 Bucharest, Romania 13 4 2022 2022 149 1-2 253272 10 9 2021 30 3 2022 © The Author(s), under exclusive licence to Springer-Verlag GmbH Austria, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. In the last decades, anthropogenic drivers have significantly influenced the natural climate variability of Earth’s atmosphere. Climate change has become a subject of major interest for different levels of our society, such as national governments, businesses, local administration, or citizens. While national and local policies propose mitigation and adaptation strategies for different sectors, public perception is a key component of any implementation plan. This study investigates the CC perception in Romania, based on a national-scale online survey performed in the spring of 2020, aiming to outline the prominence of environmental and CC issues, level of information and interest, perceived causes, changes perceived in meteorological phenomena at the regional scale, perceived impacts, and the psychological representation of the CC. The study investigates single causal factors of perception. We found that particularly (i) the regional differences on climate change intensity strongly bias the perception of CC causes; (ii) age is very likely to influence the acceptance of CC, the importance of environmental issues, and the levels of information and interest; while (iii) age, gender, and place of residence (rural–urban) are very likely to control the changes perceived in the occurrence of various meteorological phenomena, and their impact. This research is the first statistically relevant analysis (± 4%, statistical significance) developed at national and regional scales and the only study of climate change perception performed during the COVID-19 pandemic in Romania. Its results may represent the baseline for more in-depth research. Supplementary Information The online version contains supplementary material available at 10.1007/s00704-022-04041-4. Keywords Climate change Climate change perception Romania issue-copyright-statement© Springer-Verlag GmbH Austria, part of Springer Nature 2022 ==== Body pmcIntroduction Climate change (CC) is one of the major concerns of the worldwide scientific community, stakeholders, and general public. At the beginning of the 2010s, the EU citizens considered CC as the second most serious problem facing the world. They identified the economic benefits of tackling the associated phenomena and adaptation challenges (European Commission 2014). The most recent Eurobarometer on Climate Change Perception (CCP) within the EU showed that 93% of the European citizens consider it a severe challenge in the near future (European Commission 2014). The tremendous interest is justified by a plethora of clear evidence, such as measurements, observations, and impacts, but it is also extensively promoted by scientific publications and media coverage. For example, IPCC (2021) has very soundly documented the climate system’s unequivocal warming, induced mainly by human activity, with unprecedented impacts on the environment and society. Media communication focuses on different perspectives of CC and contributes to shaping people's attitudes toward it (Pasquaré and Oppizzi 2012; Lorenzoni and Whitmarsh 2014; Metag et al. 2017). Mainstream movies and documentaries can increase anxiety or motivation to act (Lowe et al. 2006), and participatory workshops may change the public perception (Fernández-Llamazares et al. 2015). The individual perception of media messages is likely to be among the front line drivers of group behavior. Personal perception plays the ultimate role in developing and implementing adaptive responses (Wolf and Moser, 2011) and in the people’s engagement to minimize the CC impacts (Reser and Bradley 2020). Understanding the causes and especially, the impacts may support adaptation and mitigation measures, so numerous studies on various aspects of climate perception have been performed in the recent decades (Capstick et al. 2015). Stehr and Von Storch (1995) introduced the concept of social construct of climate, directly linked to the public perception, as a sustainable instrument for smooth adaptation of society and CC policies. According to external and internal factors, people and communities may perceive climate variability and change very differently (e.g., Schnegg et al. 2021; Lee et al. 2020; Ruiz et al. 2020; Magistro and Roncoli 2001). For instance, those groups who rely more on exploiting natural resources such as fishermen, farmers, or woodcutters experience CC more acutely than those who live in the cities and rely indirectly on natural resources (e.g., Bacha et al. 2021; Sullivan and White, 2020). In this regard, Romania has the highest ratio of the population working in agriculture (21.2%) among the European countries (World Bank, 2021). Some studies investigated how factors like age, gender, education, income, political views, and occupation can be associated with the perception and willingness to act related to CC issues (Kabir et al. 2016; Luo and Zhao 2019; Pitpitunge 2013). Ruiz et al. (2020) separate direct factors which influence CCP, such as the common principles and ideals, and the physical experience of weather within a community, from indirect factors, namely the level of development of a community and the spread of CC information. Rühlemann and Jordan (2021) emphasized the need for inclusive climate risk and development strategies and the importance of the relationships between organizations responsible for CC adaptation and vulnerability reduction and at-risk populations. These factors seem to contribute to either inaction or effective action to climate risk. Based on a global scale analysis, Hansen et al. (2012) assumed that a person old enough to remember the climate of 1951–1980 should perceive CC signals, mainly regarding the summer months. Still, significant variations can occur among individuals, groups and geographical regions, pledging to perform local, national, and regional studies. People living in different regions of the planet perceive, experience, and adapt differently to CC. Thus, the local knowledge of crop-climate linkage shapes the Indian farmers’ CCP (Vedwan 2006; Puri 2015). Some other studies focused on the perception of extreme environments such as the Arctic, high mountainous, dry-places, low-lying islands in the South-Pacific, high Tibet, or villages from the Lower Danube Valley (Baer and Singer, 2018; Byg and Salick 2009). Many studies on CCP focus on the local level (i.e., cities or villages in different regions). At the same time, the national scale has been less approached, probably due to logistic difficulties and the different sizes of the countries. For example, the Germans’ risk perception of hazards is more frequent within the CC context (i.e., heatwaves, storms, and floods) (Frondel et al. 2017). In Hungary, CC is perceived as an ongoing but geographically remote phenomenon (Jankó et al. 2018). Only a small percentage of people hold a high level of CCP and intend to reduce the carbon lifestyle in Lebanon (Hussein et al. 2019). Comprehensive insights into the public perception of CC across France, Germany, Norway, and the UK were provided by the European Perceptions of Climate Change and Energy Preferences (EPCC) Project supported through the JPI-Programme, based on a survey conducted in 2016 (Steentjes et al. 2017). One of the main conclusions of the EPCC Project is that CC and environmental issues may be shadowed in public priorities by more urgent threats like immigration, unemployment, and economic situation. Another analysis conducted at the European scale revealed that findings cannot always be generalized, and the national context is an essential factor that shapes the CCP (Poortinga et al. 2019). Previous studies tackled the Romanian public’s perception of different environmental issues, such as natural hazards (Cheval 2003; Andrei et al. 2020), CC impact on forest (Blujdea 2005; Cosofret et al. 2014), flood risk (Armaş and Avram 2009), and earthquakes (Armaş 2006) or the CC issue at city scale (e.g., Bere-Semeredi and Bere-Semeredi 2020). Besides, the Romanians’ perception of the CC issue was investigated in continental-scale studies (e.g., Poortinga et al. 2019; Hagen et al. 2015) or systematically reviewed (e.g., Capstick et al. 2015). This research is the first study focused on the CCP performed over a statistically relevant sample at Romania’s national and regional scale. This investigation aims to provide the first overall representation of the CCP in Romania, based on broad topics and national coverage approach, but detailed at the regional level. Besides, this work aims at providing a consistent framework for further, more in-depth research, development of CC policies, and effective management of risks associated or exacerbated by CC. Data and methods Romania—geographical context, relevant facts, and figures By its geographical position within the continent (Fig. 1), Romania extends over a region of interferen European Commission 2021 ce among five major pressure centers acting across Europe (the Azores, East-European and Scandinavian Highs, and the Mediterranean and Icelandic Lows). The interactions between their influence and the underlying topography play a key role in defining the regional climate conditions across the country, which belong to four climate groups according to the Köppen-Geiger climate classification system (dry—Bsk, temperate—Cfa, Cfb, continental—Dfa, Dfb, and polar—ET) (Kottek et al. 2006), leading to a consistent temperature range between summer and winter, and moderate precipitation amounts. The average climate of the country is characterized by a mean annual air temperature varying from below 0 °C in high mountains to more than 11 °C in the South and Southwestern lowland regions, and annual precipitation amount of 400–500 mm in the Southeastern lowlands to more than 1000 mm in the high altitudes of the Western Carpathians (Sandu et al. 2008).Fig. 1 Study area: A Geographic location of Romania in Europe. B Development regions (NUTS 2), topography, and other relevant geographic features of Romania The Carpathian Mountains, located in the central part of the country, increase the diversity of climatic conditions in the country. Thus, they induce a decrease of temperature and increase the precipitation amount with the altitude, prevent or block the air masses advection, impose climatic asymmetries across the country, and different climatic patterns in regions located inside and outside the range. Thus, Eastern and Southern Romania are subject to more severe cold waves generated by Eastern, or Northern Europe originated air masses. In contrast, Western Romania, sheltered from this type of advection, remains warmer (Apostol and Sfîcă 2013). The regional differences in the climate variability and trends observed for air temperature (Marin et al. 2014), precipitation amount (Croitoru et al. 2018), aridity (Cheval et al. 2017) or heat waves (Sfîcă et al. 2017; Croitoru et al. 2018) can lead to differences on how CC is perceived in different areas. For instance, the increasing frequency of the heat waves in the Southern regions or the occurrence of high amounts of precipitation in the mountains could build the local population's CCP. Survey delivery This study relies on the results provided by a cross-national survey experiment conducted between April 30th and May 16th 2020, corresponding also with COVID-19 lockdown in Romania, to detect the level of public perception and views on CC across Romania. The survey used a structured questionnaire with 21 items in Romanian, including 15 opinion questions (Q) and six identification items (I) (Supplementary material 1). It was applied via an internet-based form to 2180 respondents from different social and demographic groups across the eight Development Regions (DRs) of Romania: North-West (NW), Center, North-East (NE), South-East (SE), South-Muntenia, Bucharest-Ilfov, South-West Oltenia (SW Oltenia), and West (Fig. 1). The regions correspond to the NUTS 2 level, and they are the basic territorial entities for applying regional policies in the EU member states (EUROSTAT 2020). The sample was weighted according to age, gender, residence (urban vs. rural), and level of education according to the statistical structure of the population of each DR (National Institute for Statistics 2020), resulting in a total sample of 835 respondents (N = 835), statistically relevant for the Romania population (± 4%) (Supplementary material 2). The methodology of applying the questionnaire is associated with some representativeness features of the study, as follows: (i) the study considered the population with internet access and who frequently uses social media for communication; (ii) even after weighting, the sample records a bias for young, urban people with a high level of education population; data collection was performed by employing online relational groups, which alters the random nature of the selection; (iii) the sampled population comprises interdisciplinary character, with a majority of respondents belonging to the academia (29%), and ongoing higher education (students 25%); (iv) it was assumed that all terms are understood equally by all respondents, an assumption that was not subsequently qualitatively validated; and (v) the margin of error for the weighted sample (Cochran method) is about ± 4% at a 95% confidence interval (p < 0.05) (Cochran 1977). All the figures are rounded to the nearest integer, considering the limitations associated with the representativeness of the study. The answers were ranked in five Likert-scale classes (Likert 1932; Findlater et al. 2019). The Relative Importance or Relative Influence Index (RII) was computed as weighted averages of the percentage allocated to each class to bring to a standard measure the CCP over the territorial entities analyzed (DRs). Results and discussions The results referring to different aspects related to Romanians’ CCP are presented in the following sections, with the associated questions: (1) Prominence of environmental and CC issues (Q1–Q5); (2) Level of information and interest for CC issues (Q6–Q8); (3) Causes of CC (Q9); (4) Changes perceived in meteorological phenomena at regional scale (Q10–Q12); (5) CC impacts (Q13-Q14); and (6) Psychological representation of CC (Q15). Prominence associated with environmental and CC issues Is CC real? Climate variability is a common feature over the Earth’s geological history. The increased frequency and impact of different extreme events specific over the last and estimated for the future decades are currently attributed to the ongoing changes in the atmospheric system (Brown 2020; Perkins et al. 2012). While most scientific literature and mainstream mass media agree that CC is a fact, its denial is not uncommon (Gross 2018; Medimorec and Pennycook 2015). In Romania, most respondents say that CC can affect the region where they live (92%), but 7% of the respondents declare they do not know, and 1% are pessimistic about this issue (Q1). Almost all the interviewed people believe that CC is a reality (95%), either as it is presented (60%), exaggerated for unmentioned reasons (17%), or used for the benefit of some interest groups (18%). At the same time, only a very negligible share of respondents considers it as a made-up theory (2%) (Q2) (Fig. 2).Fig. 2 Share of respondents who perceive CC as a reality, either explicit or distorted, or a made-up theory Complete or partial denial of the CC records the highest values among older people (about 6%), who consider CC an invented topic for the interest of a limited group of influence, whereas 39% accept CC as a reality, although presented exaggeratedly. The belief that the topic is used to benefit some groups of interest prevails among the category of 30–44 years old (33% of the age interval, representing 2% of the entire sample). How important are environmental and CC issues for Romanians? This query aims to reveal the respondents’ perception regarding the importance of the environmental and CC issues, both for the country and for the region of residence. It was addressed through Q3–Q5. Although most respondents do not hold specialized environmental education (99%), the vast majority (91%) considers environmental issues important or very important topics for Romania. In comparison, CC is perceived as an essential or very important national or regional issue by only about 80% of the people (Fig. 3). The outputs show that Romanians perceive general environmental problems as more important than the CC for the country, which can be justified by the urgency and immediate impact on their well-being or that the people subsume the CC issues to environmental ones. Moreover, the proportion of Romanians considering the CC as the single most serious problem facing the world is one of the lowest in Europe (7%) (European Commission 2021).Fig. 3 Level of perceived importance, interest, and information for environmental and CC issues Age seems to be a very influential factor that shapes this way of thinking. For example, 87% of the persons above 65 years old believe that environmental issues are very important, whereas only 55% of the people between 18 and 29 share the same belief. The reason could derive from the young people’s priorities, more focused on professional and short-term goals, and less related to environmental issues. Regional differences regarding the overall importance for the country refer only to quantitative characteristics (i.e., environmental issues are very important for 87% of the respondents from NE DR, 90% from SW Oltenia DR, 67% from NW DR, and 74% from the Centre DR). In contrast, the opinion that environmental issues are very important is qualitatively dominant in all the DRs (Fig. 4).Fig. 4 The relative importance of the environmental issues Overall, most people (79%) answer that CC is an important or very important issue for the region where they live, whereas only 4% consider the topic is rather not important at a regional scale (Fig. 5). There are people (17%) who cannot decide if CC is important or unimportant for their region. The Carpathian chain triggers a clear limit between two distinct views about this issue. Thus, the RII have higher values in the DRs outside the Carpathians. About 90% of the people living in the extra-Carpathian regions perceive the CC as an important or very important topic. Only 66% of the people living in the intra-Carpathians regions value CC as an important or very important topic important regional issue (Fig. 6).Fig. 5 The relative importance of the CC issues for Romania Fig. 6 The relative importance of the CC issues for the region of residence People’s perception is inherently socially, economically, and culturally influenced and may be different from the perception of a person living a few hundred kilometers away in a different ecological setting. They judge according to their direct living experience of climate. The personal experience with varying patterns of climate and CC identified at a regional scale may explain such a clear distinction between the two areas (Barnes et al. 2013; Welch-Devine et al. 2020). The intra-Carpathian regions are less exposed to extreme weather conditions, such as heatwaves, cold waves, extreme precipitation, blizzards, and drought events in the present climate, and less susceptible to immediate impacts of CC than extra-Carpathians regions. Thus, they perceive less the danger induced by extreme weather events (Sandu et al. 2008; Croitoru et al. 2018). Level of information and interest for CC issues Specific questions (Q6–Q8) addressed the respondents’ opinions about other people’s knowledge, their own level of information/knowledge, and their interest in the CC and its impact. The outcome shows a perception structure relevant for marginal problems, when the level of declared interest for a topic is considerably much higher than the level of information, assumed as an indicator for ongoing action and potential implication (Sherif and Hovland 1961; Van der Linden 2015). Thus, 82% of the respondents declare that they are interested or very interested in the CC and its impact. Still, only 43% of the total sample consider themselves well or very well informed. A consistent majority (64%) appreciates that the other Romanians than themselves are rather poorly or very poorly informed about this topic (Fig. 7). The perception structure revealed by this study may explain the very low level of personal implication in actions to fight against CC recorded in Romania at present, which is the lowest in the EU (European Union 2021). Fig. 7 Level of perceived relative information of Romanians about the CC issues Regional differences are revealed for the level of both information (either own or others) and interest. For example, the respondents from the NE (RII = 2.22) and SW (RII = 1.63) regions have the lowest trust in the Romanians’ level of information, while the people from the SE (RII = 2.55) and Bucharest-Ilfov (RII = 2.76) regions are the most confident in others’ level of knowledge about CC issues. On the contrary, most respondents from the SW DR (RII = 4.06; 81% from the total sample) declare that they are well or very well informed about CC, whereas people from NW (RII = 3.00), NE (RII = 3.22), and West (RII = 3.23) regions consider themselves rather moderately informed (Fig. 8). The perceived level of personal information has the highest value in the rural areas (54%), mainly at the population between 30 and 65 years old, with an under average education level and low income, suggesting a strong agricultural dimension of the CCP issues in Romania. The urban respondents above 45 years old with higher education level are the category stating the uppermost interest in this topic, and consider that the others are not sufficiently informed. This may explain the regional differences, but a more detailed investigation is needed at a regional and local scale.Fig. 8 Level of perceived relative own information of the respondents about the CC issues Inhabitants of all DRs declare a relatively high to a very high interest for the CC issues, with the highest values in SW (RII = 4.77), Bucharest-Ilfov (RII = 4.48), and NE (RII = 4.44) regions (Fig. 9). However, there is an important gap between the declared self-information (i.e., relatively low) and interest (i.e., relatively high) in all the DRs (Fig. 10). It is possible that people either underestimate their level of information, overestimate their level of interest, or even both causes are valid simultaneously.Fig. 9 Level of perceived relative interest of the respondents about the CC issues Fig. 10 Differences between declared high and very high levels of self-information and own interest for CC issues Causes of CC The perception of the possible causes of CC at a regional scale was examined based on a predefined set of variables, leaving open the possibility to propose additional triggering factors. Multiple choices were permitted. Extensive deforestation was considered to generate the CC by about 77% of the respondents, whereas the overall anthropogenic activities and industry were mentioned as influencing factors in more than 60% of the answers each (Fig. 11). The national administrations of the industrialized countries or urbanization were also considered responsible by 41–46% of the people, while natural causes or agriculture recorded lower percentages (30% and respectively 23%). About 4% of the respondents assume “Other causes” may be important, but they are all associated with the predefined variables (i.e., pollution, radioactivity, and traffic are industrial factors). “Divinity” is perceived as responsible for the CC by 3% of the investigated population. This low share may be influenced by the structure of the respondents, with the majority belonging to academia and ongoing higher education.Fig. 11 Perception of causes of CC (All people: all anthropogenic activities; GovInd: Governments of Industrialized Countries) “Extensive deforestation” was more frequently mentioned as a factor generating CC by the people between 45 and 64 years old (82%) living in rural areas, with high school as the highest level of education (81%). Women see “extensive deforestation” more frequently responsible for CC than men (85% vs. 65%). The highest share of people incriminating the “extensive deforestation” as a CC trigger lives in the southern DRs (SW Oltenia, Bucharest-Ilfov, and SE regions), where the coverage of forest land is lower than in the other areas. One can assume that the impact of deforestation news is stronger in areas where woodland spots are already sparser (Fig. 11). Overall, this response may be an output of the public discussion on deforestation in Romanian mass media during the last 10 years. Also, this role attributed to deforestation as a CC driver is a consequence of the respondents’ overall poor understanding of the CC complexity. Age and gender biased the perception of the triggering factors of CC. About 55% of the young respondents and 62% of the women in Romania consider that “all people” are responsible for it, which is considerably higher compared to the elderly (41%) or men (44%). Industrial development was identified as a triggering factor by 61% of 18–29-year-old people and 28% of the 65 + people. The last one is the generation who witnessed the industrial boom of the 1970s and society’s intense focus on the important role and performance of the Romanian industry developed during the 1980s. In contrast, environmental protection and CC were minor issues on the public agenda. On the other hand, the actions of the governments of the industrialized countries are considered an important CC factor mainly by older people: 47% of the people within the category 65 + compared to 26% from the people in the age category 18–29. Both “urbanization” and “natural causes” are more frequently identified as causes for CC by the people living in cities. Also, almost 20% of the urban dwellers believe that intensive agriculture has an important influence on CC, compared to only 11% of the respondents living in rural areas. Perception of changes in the local occurrence of dangerous meteorological phenomena The survey also investigated the public perception related to the dangerous meteorological phenomena (severe weather conditions) most often associated with CC in Romania. The results revealed how the Romanians perceive the local changes of temperature, precipitation, and several meteorological phenomena proposed in the survey based on existing reports on observed changes: heat and cold waves (Piticar et al. 2017; Croitoru et al. 2016, 2018), rainfall events (Busuioc et al. 2017), snow cover (Micu et al. 2015; Birsan and Dumitrescu 2014), drought, storms, and new or rare events, such as tornadoes (Andrei et al. 2020). Individuals’ conceivable timescales for visioning and concrete engagement extend to about two decades into the future (Lorenzoni and Hulme 2009), while the perception of the observed changes in temperature and precipitation focused on the 15–20 years before the survey. About 85% of the people believe that the regional climate is warmer and drier than 15–20 years before, and 7–8% declare that they perceive no change or the climate is colder and wetter (Fig. 12). The bias related to age, gender, region or other factors is negligible. The perception is in perfect agreement with the observed temperature variability, but the perceived decrease of precipitation is not supported by scientific evidence (Dumitrescu et al. 2015; Croitoru et al. 2016, 2018). The increasing rain intensity trends associated with warming processes (Busuioc et al. 2017) trigger longer intervals with no or little precipitation amounts and less water available in some periods, influencing the public perception of the drier climate over Romania.Fig. 12 Perception of changes in temperature and precipitation over the last decades The main issue reported as a CC marker within the local environment is snow cover, perceived as decreasing over the last decades by 75% of the respondents (Fig. 13). The decreasing precipitation (i.e., droughts) and increasing heatwaves frequency are also associated with CC by most people (almost 70%). More often, storms and the increasing frequency of phenomena less common or new in Romania, such as tornadoes, are noticeably related with CCs by 22–23% of the respondents, while other severe weather conditions are less present in the answers, i.e., more frequent cold waves (15%), more frequent heavy rainfall events (3–4%), and other phenomena (less than 1% each).Fig. 13 Perception of changes occurring in the recent decades at local scale The perceived changes in general climate at the local scale varies significantly (p < 0.05) with age, type of settlement, geographical location (DR), and gender. The heatwaves and drought events were perceived as more frequent nowadays than over the last decades by the majority of two age groups: 65 + population (77%) and those of 30–44 (52%). The young generation (18–29 years old) considered more frequent heavy rainfalls as the most important change in weather conditions. Both categories agree with scientific findings (Bojariu et al. 2015; Croitoru et al. 2018). Urban residents associate rainfall events with CC three times rarer than the rural population. Still, heatwaves are indicated 30% more often, in close relation to the events impacting the most both urban and rural population (Herbel et al. 2018; Ichim and Sfîcă 2020). At the regional level, one can notice that the respondents from the Bucharest-Ilfov DR provided multiple answers and put the new or unusual phenomena in relation to CC much more often than the others. Almost all respondents from Bucharest (95%) consider that the heatwaves are more frequent than in the recent past, with this share being well above the country average (67%). The respondents of the SW Oltenia DR are the least sensitive to CC issues relative to the meteorological phenomena listed in the survey. Other relevant differences between DRs are listed in Table 1.Table 1 Dominant changes associated to CC identified by respondents Development region (DR) Perceived CC Bucharest-Ilfov New phenomena (i.e. tornado); More frequent storms; More frequent heatwaves; More frequent droughts NW Decreasing snow cover NE More frequent heatwaves South Muntenia More frequent heatwaves; Decreasing snow cover SE Oltenia More frequent heatwaves; More frequent droughts West More frequent storms; More frequent cold waves; More frequent droughts Centre More frequent heavy rainfall events The gender proportion is significantly unbalanced, as the women respondents more often associate the phenomena proposed to CC than the men sample (+ 8%). CC impacts Perception of CC impacts at regional scale Most people (92%) say that CC may impact the region where they live, and only 1% of the subjects deny the possibility of such an impact (Q1), which is in very good agreement with the acknowledgement of CC as a reality by 95% of the respondents (Q2). The difference from the results revealed by the 2014 Eurobarometer on CC (European Commission 2014), suggesting a low level of awareness of Romanians on CC topics, may be induced by the sample structure of this study. However, this topic requires more refined and consistent investigations. The difference between young people and the older generation perception is negligible in this case (87% vs. 83%). The perception of environmental issues suggests that the young generation is more interested in CC than in the general environmental agenda. Still, the clear distinction between CC and general environmental issues within the public perception should be investigated in further studies. A high percentage of the respondents (85%) associate the potential impact of CC with the importance of environmental issues since they attribute a high value to both issues (Q3, Q4). Land degradation, biodiversity loss, water resources decline, and river and lake levels drop are the issues most frequently perceived as specific local CC impacts from the 12 predefined options proposed in the questionnaire (Q13) (Fig. 14). Each category mentioned above is present in more than 50% of answers, with a maximum of 75% for land degradation.Fig. 14 Perception of changes of the frequency of occurrence of specific CC impacts in the recent decades in Romania Generally, the respondents’ proposed impacts have been well incriminated as potentially caused by CC (i.e., around 30–35% of answers indicated “more frequent floods,” “deforestation,” “more frequent landslides, epidemics material damages, or pest episodes”). The low percentage of answers indicating “sea level rise” as a CC impact can be explained primarily by the physical distance between respondents’ residence location and the coastal region or by the actual increase in sea level is not perceived as being dangerous for the moment. However, this general perception level is also in line with some previously developed studies (Medvedev et al. 2016), showing that this process is expected to exert a limited impact even in the worst-case CC scenarios within a low-tide water body of the Black Sea. The high percentage of respondents indicating 'more epidemics' as a consequence of CC could be biased by the period when the questionnaire was applied, which overlapped the beginning of the COVID19 breakdown period in Romania. In many cases, regional differences have essential values: SE vs. West regions regarding the land degradation, or NE vs. West regions regarding the biodiversity decline (Table 2). The respondents in the SE region indicate biodiversity loss as a main effect of CC in relation to the loss of diversity in marine life or along the Danube and its delta. The perception of CC impacts is unbalanced by gender, residence, and age. Thus, women and rural residents mentioned more categories of impacts than men and urban respondents. The young generation (18–29) is dominantly concerned about 'deforestation', while biodiversity loss and reduction of water resources are significantly more frequently selected as possible CC impacts by people between 30 and 44 years old (70%) or by rural residents (73%) and higher educated people (71%). The perception of the sea-level rise as a CC impact dramatically diminishes with age (11% of the 18–29 years old group and 1% of the 65 + group).Table 2 Changes in the frequency of occurrence of specific CC impacts in the recent decades at the local scale, as perceived in each DR The reduction of water resources is an issue with a high level of visibility within the 30–44 age group (70%). Similarly, the same impact is significantly lower in the view of younger people (49%). The SW Oltenia DR stands out for the highest share of 97% of the total population mentioned in the survey. This share could be explained in relation to the underlying semi-arid climatic conditions and with a high frequency of drought phenomenon in this region (Sandu et al. 2010). Conversely, the lowest share was recorded in the NE DR (40%). People living in rural areas indicated this CC effect more frequently than those living in urban areas (73% vs. 55%). Women from urban areas are more sensitive to this issue than men (68% vs. 58%). “Decreasing water level in rivers and lakes” showed a similar level of concern in the population sample regardless of age. This issue has significantly higher visibility than the average in the NW (65%), West (62%), and SW Oltenia (62%) DRs. The invasion of pests is a category of CC effects attributed by most respondents aged 30–44 years (54%), which shows relatively similar shares in all regions of Romania. The respondents from rural areas (44%) and a high share of people with an education level below average (49%) indicated a more significant sensitivity to the same topic. “Material damage” induced by CC is a consequence mainly tackled by adult respondents aged between 30 and 44 years (44% mentions), as well as by people with an education level below average (48%), within a higher share of responses in the NW (42%), SW Oltenia (42%), and Bucharest-Ilfov (40%) regions. The “increasing number of epidemics” has been attributed to CC by all age groups (somewhat similar shares of responses). Regionally, this consequence received higher-than-average visibility in the NW (53%) and South-Oltenia (49%) DRs. People from rural areas with below-average education levels are significantly more likely in this attribution than the rest of the population. The landslides show a similar concern in all categories of the population. Unexpectedly, a higher-than-average percentage of respondents mentioned this effect in the Bucharest-Ilfov DR (52%), which is not prone to these geomorphic processes. People in this highly urbanized area also indicate “land degradation”—a problem that apparently should be more important for the rural citizens—as a very important output of CC. These results suggest that the response of this group of population is not driven by their direct experience on the topic but shaped by their effort to gather information on CC issues. Perception of CC impacts on the personal life Almost half of the respondents believe that CC will impact their personal lives negatively (23%) or even profoundly negative (25%). In contrast, only a small share of answers indicated positive (5%) or profoundly positive impacts (2%) in this respect (Fig. 15).Fig. 15 Perception of CC impact on personal life at country scale Respondents living in the countryside show higher concern about the potential impact of CC on their lives (62% of them) compared to the urban residents (37% of the sample). This perception may be generated by the dominant agriculture-based economy in the rural areas, which is directly influenced and more vulnerable to CC and associated extreme events. For instance, farmers in the SW Oltenia DR may perceive CC more than other Romanians due to its direct influence on their crops and living and socio-economic conditions. Or, the land degradation of the lowlands in Southern Romania may trigger a more acute perception of CC than for people who do not directly experience such rapid environmental changes (Stringer and Harris 2014). However, the SW Oltenia and South Muntenia DRs, where rural areas are quite extended, reported the lowest concern regarding the CC impact on personal life (Fig. 16), despite the differences noted in the level of knowledge and interest.Fig. 16 Perception of CC impact on personal life in each DR In contrast, the people from NW DR perceived that the CC impact could be high or very high, although they expressed very low interest in the topic (Figs. 8 and 9). This contradiction suggests that various factors influence the CC perception, and the result is not always predictable, which demonstrates the need for more detailed perception studies. The way people evaluate their own capacity to adapt to CC is an essential element in shaping their perception. For instance, even if they experience CC as a very important issue, the people in SW consider they can adapt to it relatively easy in the future. Therefore, the impact of CC on their life should be minimized. Psychological representation of CC The “free association” of terms by writing without censorship was applied to investigate the content of individual consciousness and gain insight into how people represent CC (Joffe and Elsey 2014; Vulchanova et al. 2019). This method indicates the overall picture of CC designed by the subjects, with no particular focus on causes or impact. The initial list was expert-based filtered and validated as helpful for this study considering the following criteria: (a) similarity between the meaning of the words in the climate context (e.g., cold and very cold; dry and drought; starvation and famine); (b) typos; (c) the meaning of the terms in the CC context (e.g., “I don’t know” or too general terms, such as “impact” or “scale” were eliminated). In the second step, the valid terms were grouped into four categories, namely (1) Societal and Economic Issues, (2) Psychological Factors, (3) Environmental elements, and (4) Climate and Weather (Table 2). Most respondents (77.8%) associated CC with “environmental elements” or “climate and weather” terms; about 20% of the respondents indicated either “societal and economic issues” or “psychological factors” as the first related to CC, and a negligible share of answers was not usable (i.e., 2%). “Famine and poverty” (2.5%), “recklessness, carelessness, and irresponsibility” (1.4%), “pollution and anthropogenic activities” (7.4%), “drought and aridity” (14.9%) are the syntagms most frequently mentioned from each of the four main categories. In comparison, “drought and aridity” and “warming and heatwaves” (13.1%) had the highest frequency among all terms (Table 3).Table 3 Categories and terms indicated by using the “free association” with CC. The top three words of each category are in bold (with the frequency of occurrence from the total number of answers mentioned in brackets) Societal and economic Issues (11.6%) Psychological factors (8.6%) Environmental elements (27.9%) Climate and weather (49.9%) Not valid (2.0%) • Famine and poverty (2.5%); •Danger (General, Crisis, Self-destruction, Problems, Destruction, Evil, Death) (2.4%); • Agriculture and food (1.8%); • Health, hygiene, and disease (including epidemic); • Instability and chaos; Economy, consume, damages and costs; • Industry; Education and information; • Energy; Technology and science; • Financial incentives and penalties; • Administration; • Mass media • Recklessness, carelessness, and irresponsibility (1.4%); • Humankind (1.1%); • Responsibility, activism, NGO (0.9%); • Psychological impact (fear, sadness, hate, unsafety, concern, curiosity) (0.9%); • Changes and instability; • Ignorance and indifference; • Adaptation; • Uncertainties and unpredictable; • Life (in general) and life quality; • God and religion; • Experiments, • Manipulation; • Future generations; • Overpopulation; • Unavoidable; • Exaggeration or unbelievable; • Fast, short term; • Reality; • Injustice; • Holiday; • Marvellous; • Pollution and anthropogenic activities (7.4%); • Glacier melting (general), including polar ice caps and mountain (4.3%); • Nature, environment and ecosystems, Biodiversity (Fauna and vegetation) (4.2%); • Deforestation; • Desertification; • Derived hazards (landslides, land degradation, floods); • Snowpack (missing or reduced); • Physical characteristics of the environment (albedo, atmospheric chemical composition, inversion of Poles, air pressure, solar radiation, relief, soil, water, thermohaline, beach, dust); • Sea level rise; Forest fire; • Waste; • Habitat loss; • Resources; • Cities; • Earth; • Recycling; • Miscellaneous (Earthquake); • Protection; • Globalization; • Fossil fuels • Drought and aridity (14.9%); • Warming and heat waves (13.1%); • Atmospheric hazards (extremes, risks, deviations, disasters) (7.0%); • Changes in seasons / Continentalism; • Storms and hurricanes; • Atmospheric chemistry, gases, radiation (CO2, O3 and ozone layer, UV); • Air temperature; • Greenhouse effect and GHG emissions; • Tornadoes; • Cold; • Rainfall; • Wind; • Precipitation (in general); • Sun; • Meteorology and weather; • Hail; • Blizzard; • Thunderstorms • Not usable answers; • I don't know; • Impacts; • Not a topic of interest; • Past; • Regional scale Conclusions This study provides a general view regarding the perception of CC in Romania captured in the spring of 2020, including regional differences, and consideration about age, gender, living environment, and type and level of education which are likely to bias the outputs. To the best of our knowledge, this is the first study addressing the national level, but considering the perception at the regional level, too and associated with sociological and definite statistical relevance. Overall, the perception of CC in Romania is dominated by the following characteristics:CC is considered a fact, even though media and different groups of interests can overrate it. Both environmental issues and CC are considered important or very important topics of the current societal agenda, and environmental education does not bias the perception. Most people declared a high to very high interest in CC issues; however, less than a half considered they are well or very well informed, whereas the other people are generally seen as poorly or very poorly informed about CC. Most people from rural areas, with an under average education level and low income, considered themselves well or very well informed. Most urban respondents with higher education declared a very high interest in CC issues and thought the others were not sufficiently informed. Extensive deforestation, overall anthropogenic activities and industry, governments of industrialized countries, and urbanization are the most influencing factors that control the CC. This is explicable by the magnitude of public debate on legal and especially illegal deforestation in Romania over the last 2 years. TV shows, public debates, and social media accusations to private forest enterprises gained visibility in the public space and made deforestation a critical political issue. The regional climate is perceived as being warmer and drier than 15–20 years before. Decreasing snow cover depth and precipitation amount (i.e., droughts) and increasing frequency of heatwaves are the primary CC markers, which are in agreement with the scientific findings. At the individual level, the CC is likely to have a negative impact rather than positive consequences. CC is most frequently associated with “environmental elements” or “climate and weather” terms, while “societal and economic issues” or “psychological factors” are also present in the “free association” exercise. “Drought and aridity” and “warming and heatwaves” have the highest occurrence of terms related to CC. Frequently, the responses are different from region to region of Romania due to the specific climate conditions in the country regions or the local importance of various environmental issues. Geography (emphasized by analysis considered at DR level), age, gender, and living environment (i.e., urban or rural) often bias the CCP. For example, the older generation tends to focus more on the influence of the establishment and is less convinced about the industry’s impact. Women and young people tend to blame anthropogenic activities more often, while urban citizens are more likely to indicate “urbanization,” “natural causes,” and “agriculture” as important CC generators. In summary, (i) the regional characteristics (DRs) have the most robust bias on the perception of CC causes, changes in the frequency of various meteorological phenomena, and impact; (ii) the age strongly influence mainly the acceptance of the existence of the CC, the importance of environmental issues, the level of information and interest, the changes in the occurrence frequency of various meteorological phenomena, and their impact; (iii) the gender is essential for perceiving the changes in the frequency of various meteorological phenomena, and their impact; while (iv) the residence habitat (rural vs urban) strongly influences the level of information and interest for CC issues, the changes perceived in the frequency of various meteorological dangerous phenomena, and their impact. Further studies must address this exploratory investigation of the CCP in Romania, as well as some limitations. For example, the complete statistical representativeness for the country scale of the research is biased by several methodological issues, such as the “internet only”-based approach, the confidence intervals of the sampling, or insufficient coverage of the entire population. The high complexity of factors influencing the CC perception requires more in-depth investigations, while this study has provided the general framework. However, this study proposes consistent follow-ups and in-depth examination of CCP in Romania. It could aim to a better spatial resolution (e.g., local or regional scale), more diverse influences (e.g., economic factors, religion, or occupation), or context (e.g., national or international policy, or natural disasters). Besides, this study is a genuine argument for implementing systematic reviews of the CCP at a national scale to support the development of CC policies well aligned to real societal needs. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 25 KB) Supplementary file2 (DOCX 14 KB) Acknowledgements The authors kindly acknowledge the anonymous reviewer for the valuable comments and suggestions that helped improve this paper's quality. Special acknowledgements are for Dr. Cristina Florina Vîrtan (Roşca) for technical mapping support. Author contribution Conceptualization: Sorin Cheval, Ana Bulai, Adina-Eliza Croitoru, Ștefan Dorondel, Lucian Sfîcă, Adrian Tișcovschi; methodology: Sorin Cheval, Ana Bulai Adina-Eliza Croitoru, Lucian Sfîcă, Adrian Tișcovschi; formal analysis and investigation: Sorin Cheval Ana Bulai, Adina-Eliza Croitoru, Ștefan Dorondel, Dumitru Mihăilă, Lucian Sfîcă, Adrian Tișcovschi; writing—original draft preparation Sorin Cheval, Ana Bulai, Adina-Eliza Croitoru, Ștefan Dorondel, Dumitru Mihăilă, Lucian Sfîcă; writing—review and editing: Sorin Cheval, Ana Bulai, Adina-Eliza Croitoru, Ștefan Dorondel, Dana Micu, Lucian Sfîcă; supervision: Sorin Cheval, Ana Bulai, Adina-Eliza Croitoru, Dana Micu, Lucian Sfîcă. Funding Stefan Dorondel’s work in this paper is supported by the Grant PN-III-P4-ID-PCE-2020–1238 of the UEFISCDI, The Romanian Ministry of Research and Digitalisation. Adina-Eliza Croitoru’s work for this research was partially conducted under the framework of the experimental demonstration project Redefining the agro-suitability zones for maize and winter wheat crops toward a smart climate change-oriented agriculture in Romania (www.agroclim.ro) funded by the Executive Unit for the Financing of Higher Education, Research, Development and Innovation in Romania (UEFISCDI), grant number PN-III-P2-2.1-PED-2019–2310. Data availability The datasets generated during and/or analyzed during the current study are not publicly available due to GDPR regulations but are available from the corresponding author on reasonable request. Code availability No code was used in this research. Declarations Ethics approval The authors confirm that the manuscript upholds the integrity of the scientific records. Consent to participate All authors have given the consent to participate in this research. Consent for publication All authors have read and agreed to the published version of the manuscript. Conflict of interest The authors have no conflicts of interest. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Andrei S Andrei MD Hustiu M Cheval S Antonescu B Tornadoes in Romania—from forecasting and warning to understanding the public's response and expectations Atmosphere 2020 11 9 966 10.3390/atmos11090966 Apostol L Sfîcă L Thermal differentiations induced by the Carpathian Mountains on Romanian territory Carpathian J Earth Environ Sci 2013 8 215 221 Armaș I (2006) Earthquake risk perception in Bucharest, Romania. Risk Analysis, 26(5):1223-1234. 10.1111/j.1539-6924.2006.00810.x Armaş I Avram E Perception of flood risk in Danube Delta Rom Nat Hazards 2009 50 2 269 287 10.1007/s11069-008-9337-0 Bacha MS Muhammas M Kilic Z Nafees M The dynamics of public perceptions and climate change in Swat Valley, Khyber Pakhtunkhwa Pak Sustain 2021 13 4464 10.3390/su13084464 Baer H, Singer M (2018) The anthropology of climate change: an integrated critical perspective (2nd ed.). Routledge. 266p, 10.4324/9781315818702 Barnes J Dove MR Lahsen M Mathews A McElwee P McIntosh R Moore F O'Reilly J Orlove B Puri R Weiss H Yarger K Contribution of anthropology to the study of climate change Nat Clim Chang 2013 3 514 544 10.1038/nclimate1775 Bennett NJ Using perceptions as evidence to improve conservation and environmental management Conserv Biol 2016 30 3 582 592 10.1111/cobi.12681 26801337 Bere-Semeredi I, Bere-Semeredi A-A., 2020. Perception, knowledge, attitude and behavior toward climate change—A survey among citizens in Timisoara, Romania. In: Prostean G., LaviosVillahoz J., Brancu L., Bakacsi G. (eds) Innovation in sustainable management and entrepreneurship. SIM 2019. Springer Proceedings in Business and Economics. Springer, Cham. 10.1007/978-3-030-44711-3_15. Birsan M-V, Dumitrescu A (2014) Snow variability in Romania in connection to large scale atmospheric circulation. International Journal of Climatology. 34. 10.1002/joc.3671 Blujdea V (2005) Percepţia silvicultorilor privind impactul schimbărilor climatice asupra pădurilor. Analele ICAS, I (48), 151–160. Available from: http://www.editurasilvica.ro/analeleicas/48/1/blujdea.pdf. Accessed 18 Aug 2021 (in Romanian) Bojariu R, Birsan MV, Cica R, Velea L, Burcea S, Dumitrescu A, Dascalu SI, Gothard M, Dobrinescu A, Carbunaru F, Marin L, 2015 Schimbările climatice - de la bazele fizice la riscuri și adaptare. Editura Printech, București, 200. Brown SJ Future changes in heatwave severity, duration and frequency due to climate change for the most populous cities Weather Clim Extremes 2020 30 100278 10.1016/j.wace.2020.100278 Busuioc A Baciu M Breza T Dumitrescu A Stoica C Baghina N Changes in intensity of high temporal resolution precipitation extremes in Romania: implications for Clausius-Clapeyron scaling Clim Res 2017 72 239 249 10.3354/cr01469 Byg A, Salick J (2009) Local perspectives on a global phenomenon—climate change in Eastern Tibetan villages. Global Environmental Change, 19(2), 156–166. 10.1016/j.gloenvcha.2009.01.010 Campbell-Lendrum D Bertollini R Science, media and public perception: implications for climate and health policies Bull World Health Organ 2010 88 242 243 10.2471/BLT.10.077362 20431780 Capstick S Whitmarsh L Poortinga W Pidgeon N Upham P International trends in public perceptions of over the past quarter century Wiley Interdiscip Reviews Clim Change 2015 6 1 35 61 10.1002/wcc.321 Cheval S Percepţia hazardelor natural. Rezultatele unui sondaj de opinie desfăşurat în România Riscuri și Catastrofe 2003 II December 2002 49 60 Cheval S, Dumitrescu A, Birsan M-V (2017) Variability of the aridity in the South-Eastern Europe over 1961–2050. Catena, 151: 74-86. 10.1016/j.catena.2016.11.029 Cochran WG Sampling Techniques 1977 New York J. Wiley and Sons Inc Cosofret C, Duduman C, Mutu M, Palaghianu C, Bouriaud L (2018) Exploring media influence in determining forest engineers' perceptions on climate change. Bulletin of the Transilvania University of Brasov. 11(60):17-30 Croitoru AE Piticar A Burada DC Changes in precipitation extremes in Romania Quatern Int 2016 415 325 335 10.1016/j.quaint.2015.07.028 Croitoru AE Sfîcă L Roșca C-F Tudose T Horvath C Ionuț M Ciupertea A-F Scripcă S Harpa G Extreme temperature and precipitation events in Romania 2018 Bucharest Publishing House of the Romanian Academy 360p Dumitrescu A, Bojariu R, Birsan MV, Marin L, Manea A (2015) Recent climatic changes in Romania from observational data (1961–2013). Theor Appl Climatol 122, 111–119. 10.1007/s00704-014-1290-0 European Commission (2014) Special Eurobarometer 372: Social Climate (v1.00). [Data set]. European Commission, Directorate-General for Communication. http://data.europa.eu/88u/dataset/S1007_75_4_EBS372 European Union (2021) Special Eurobarometer 513: Climate Change, March-April 2021. https://doi.org/10.2834/437. Available from https://europa.eu/eurobarometer/surveys/detail/2273. Accessed 29 Jun 2021 EUROSTAT (2020) Statistical regions in the European Union and partner countries. Nuts and statistical regions 2021. Luxembourg: Publications Office of the European Union. 188 p 10.2785/850262 Vulchanova M, Vulchanov V, Fritz I, Milburn EA (2019) Language and perception: introduction to the Special Issue "Speakers and Listeners in the Visual World". Journal of Cultural Cognitive Science, 3, 103–112. 10.1007/s41809-019-00047-z Fernández-Llamazares Á Méndez-López ME Díaz-Reviriego I McBride MF Pyhälä A Rosell-Melé A Reyes-García V Links between media communication and local perceptions of climate change in an indigenous society Clim Change 2015 131 2 307 320 10.1007/s10584-015-1381-7 26166919 Findlater K Kandlikar M Satterfield T Donner S Weather and climate variability may be poor proxies for climate change in farmer risk perceptions Weather, Climate, and Society 2019 11 4 697 711 10.1175/WCAS-D-19-0040.1 Frondel M Simora M Sommer S Risk perception of climate change: empirical evidence for Germany Ecol Econ 2017 137 173 183 10.1016/j.ecolecon.2017.02.019 Gross L Confronting climate change in the age of denial PLoS Biol 2018 16 10 10 13 10.1371/journal.pbio.3000033 Hansen J, Sato M, Ruedy R (2012) Perception of climate change. Proceedings of the National Academy of Sciences of the United States of America, 109 (37) E2415-E2423. 10.1073/pnas.1205276109 Herbel I Croitoru AE Rus A Rosca CF Harpa GV Ciupertea A-F Rus I The impact of heat waves on surface urban heat island and local economy in Cluj-Napoca city Romania Theoretical and Appl Climatol 2018 133 24 681 695 10.1007/s00704-017-2196-4 Hussein A Hussein T Moussaid G Moussaid G Climate change perception in Lebanon: An Exploratory Study Int J Market Stud 2019 11 4 53 10.5539/ijms.v11n4p53 Ichim P Sfîcă L The influence of urban climate on bioclimatic conditions in the city of Iași Romania Sustain 2020 12 22 9652 10.3390/su12229652 IPCC, Climate Change (2021) The physical science basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M. I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J. B.R. Matthews, T. K. Maycock, T. Waterfield, O. Yelekçi, R. Yu and B. Zhou (eds.)]. Cambridge University Press. In Press. Jankó F Bertalan L Hoschek M Komornoki K Németh N Papp-Vancsó J Perception, understanding, and action: attitudes of climate change in the Hungarian population Hungarian Geographical Bulletin 2018 67 2 159 171 10.15201/hungeobull.67.2.4 Joffe H Elsey J Free association in psychology and the grid elaboration method Rev Gen Psychol 2014 18 173 185 10.1037/gpr0000014 Kabir MI Rahman MB Smith W Lusha MAF Azim S Milton AH Knowledge and perception about climate change and human health: findings from a baseline survey among vulnerable communities in Bangladesh BMC Public Health 2016 16 1 1 10 10.1186/s12889-016-2930-3 26728978 Kottek M Grieser J Beck C Rudolf B Rubel F World map of the Köppen-Geiger climate classification updated Meteorol Z 2006 15 259 263 10.1127/0941-2948/2006/0130 Lee K Gjersoe N O'Neill S Barnett J Youth perceptions of climate change: a narrative synthesis Wires Clim Change 2020 11 3 e641 10.1002/wcc.641 Likert R A Technique for the Measurement of Attitudes Archives of Psychology 1932 140 1 55 Lorenzoni I Hulme M Believing is seeing: Laypeople's views of future socio-economic and climate change in England and in Italy Public Underst Sci 2009 18 4 383 400 10.1177/0963662508089540 Lorenzoni I Whitmarsh L Climate change and perceptions, behaviors, and communication research after the IPCC 5th Assessment Report - a WIREs Editorial Wiley Interdisciplinary Reviews Clim Change 2014 5 6 703 708 10.1002/wcc.319 Lowe T Brown K Dessai S De França Doria M Haynes K Vincent K Does tomorrow ever come? Disaster narrative and public perceptions of climate change Public Underst Sci 2006 15 4 435 457 10.1177/0963662506063796 Luo Y Zhao J Motivated attention in climate change perception and action Front Psychol 2019 10 1541 10.3389/fpsyg.2019.01541 31379643 Magistro J Roncoli C Anthropological perspectives and policy implications of climate change research Climate Res 2001 19 2 91 96 10.3354/cr019091 Marin L Birsan M-V Bojariu R Dumitrescu A Micu D Manea A An overview of annual climatic changes in Romania: Trends in air temperature, precipitation, sunshine hours, cloud cover, relative humidity and wind speed during the 1961–2013 period Carpathian J Earth Environ Sci 2014 9 253 258 Medimorec S Pennycook G The language of denial: text analysis reveals differences in language use between climate change proponents and skeptics Clim Change 2015 133 4 597 605 10.1007/s10584-015-1475-2 Medvedev IP, Rabinovich AB, Kulikov EA (2016) Tides in three enclosed basins: The Baltic, Black, and Caspian Seas. Front Marine Sci, 46(3) 10.3389/fmars.2016.00046 Metag J Füchslin T Schäfer MS Global warming’s five Germanys: a typology of Germans’ views on climate change and patterns of media use and information Public Underst Sci 2017 26 4 434 451 10.1177/0963662515592558 26142148 Micu DM, Dumitrescu A, Cheval S, Birsan M-V (2015) Climate of the Romanian Carpathians. Variability and change. Springer Atmospheric Sciences, Springer International Publishing, 213 p. 10.1007/978-3-319-02886-6 National Institute for Statistics (2020) POP107D - LEGALLY RESIDENT POPULATION, by age group and ages, sex, counties and localities at January 1st. Available from http://statistici.insse.ro:8077/tempo-online/#/pages/tables/insse-table [Accessed April 24th 2020] OECD (2019) Measuring distance to the SDG targets 2019: an assessment of where OECD Countries stand. In Measuring Distance to the SDG Targets 2019. Organization for economic co-operation and development, Paris, France. 10.1787/e0f4d2ac-en. Available from http://www1.oecd.org/std/OECD-Measuring-Distance-to-SDG-Targets.pdf [Accessed August 18th 2021]. Pasquaré FA Oppizzi P How does the media affect public perception of climate change and geohazards? An Italian case study Global Planet Change 2012 90–91 152 157 10.1016/j.gloplacha.2011.05.010 Perkins SE Alexander LV Nairn JR Increasing frequency, intensity and duration of observed global heatwaves and warm spells Geophys Res Lett 2012 39 20 1 5 10.1029/2012GL053361 Piticar A Croitoru AE Ciupertea F-A Harpa G-V Recent changes in heat waves and cold waves detected based on excess heat factor and excess cold factor in Romania Int J Climatol 2017 38 4 1777 1793 10.1002/joc.5295 Pitpitunge AD (2013) Students' perceptions about climate change. Asian Journal of Biology Education, 7, 2–11. Available from http://aabe.sakura.ne.jp/Journal/Papers/Vol7/02 Pitpitunge.pdf [Accessed August 17th 2021]. Poortinga W Whitmarsh L Steg L Böhm G Fisher S Climate change perceptions and their individual-level determinants: a cross-European analysis Glob Environ Chang 2019 55 25 35 10.1016/j.gloenvcha.2019.01.007 Puri R (2015) The uniqueness of the every day: Herders and invasive species in India. In: J. Barnes and M.R. Dove, eds. Climate cultures: Anthropological perspectives to climate change, New Haven: Yale University Press. 249–272. 10.12987/yale/9780300198812.003.0011 Reser JP Bradley GL The nature, significance, and influence of perceived personal experience of climate change Wiley Interdisciplinary Reviews: Climate Change 2020 11 5 1 28 10.1002/wcc.668 Roncoli C, Crane T, Orlove B (2009) Fielding climate change in cultural anthropology. In: S.A. Crate and M. Nuttall, eds., Anthropology and Climate Change: From Encounters to Actions. 1 ed., New York: Routledge. 87–115. 10.4324/9781315434773 Ruiz I Faria SH Neumann MB Climate change perception: driving forces and their interactions Environ Sci Policy 2020 108 112 120 10.1016/j.envsci.2020.03.020 Rühlemann A, Jordan JC (2021) Risk perception and culture: implications for vulnerability and adaptation to climate change. Disasters 45(2), 424–452. 10.1111/disa.12429 Sandu I Mateescu E Vătămanu V Schimbări climatice în România şi efectele asupra agriculturii 2010 Craiova SITECH Publishing House 406 Sandu I, Pescaru VI, Poiană I, Geicu A, Cândea I, Țâștea D. (eds.) (2008) Clima României. Administrația Națională de Meteorologie. București: Editura Academiei Române, 365 p. (in Romanian) Schnegg M O'Brian CI Sievert IJ It's our fault: a global comparison of different ways of explaining climate change Hum Ecol 2021 49 327 339 10.1007/s10745-021-00229-w Sfîcă L Croitoru AE Iordache I Ciupertea AF Synoptic conditions generating heatwaves and warm spells in Romania Atmosphere 2017 8 3 1 22 10.3390/atmos8030050 Sherif M, Hovland CI (1961) Social judgment: Assimilation and contrast effects in communication and attitude change. Yale Studies in Attitude and Communication, New Haven: Yale University Press, 218 p. Steentjes K, Pidgeon N, Poortinga W, Corner A, Arnold A, Böhm G, Mays C, Poumadère M, Ruddat M, Scheer D, Sonnberger M, Tvinnereim E, 2017. European perceptions of climate change: Topline findings of a survey conducted in four European countries in 2016. Cardiff: Cardiff University. 72 p. Available from https://orca.cardiff.ac.uk/98660/7/EPCC.pdf [Accessed August 16th 2021] Stehr N Von Storch H The social construct of climate and climate change Climate Res 1995 5 2 99 105 10.3354/cr005099 Stringer LC Harris A Land degradation in Dolj County, Southern Romania: environmental changes, impacts and responses Land Degrad Dev 2014 25 1 17 28 10.1002/ldr.2260 Sullivan A White DD An assessment of public perceptions of risk in three Western U.S. cities Weather Clim Soc 2020 11 2 449 463 10.1175/WCAS-D-18-0068.1 Van der Linden S The social-psychological determinants of climate change risk perceptions: Towards a comprehensive model J Environ Psychol 2015 41 112 124 10.1016/j.jenvp.2014.11.012 Vedwan N Culture, climate and the environment: local knowledge and perception of climate change among apple growers in Northwestern India J Ecol Anthropol 2006 10 1 4 18 10.5038/2162-4593.10.1.1 Welch-Devine M, Sourdril A, Burke BJ. (eds) (2020) Changing climate, changing worlds: local knowledge and the challenges of social and ecological change. Zurich: Springer Cham. 266 p. 10.1007/978-3-030-37312-2 Wolf J Moser SC Individual understandings, perceptions, and engagement with climate change: Insights from in-depth studies across the world Wiley Interdisciplinary Reviews Clim Change 2011 2 4 547 569 10.1002/wcc.120 World Bank (2021) World Development Indicators. Available from http://data.worldbank.org/data-catalog/world-development-indicators. Accessed on 20 Jun 2021.
PMC009xxxxxx/PMC9005344.txt	==== Front Artif Intell Rev Artif Intell Rev Artificial Intelligence Review 0269-2821 1573-7462 Springer Netherlands Dordrecht 35431395 10188 10.1007/s10462-022-10188-3 Article Deep Neuro-Fuzzy System application trends, challenges, and future perspectives: a systematic survey http://orcid.org/0000-0003-4162-7910 Talpur Noureen noureen.baghrani@gmail.com 1 Abdulkadir Said Jadid 1 Alhussian Hitham 1 Hasan Mohd Hilmi 1 Aziz Norshakirah 1 Bamhdi Alwi 2 1 grid.444487.f 0000 0004 0634 0540 Centre for Research in Data Science (CeRDaS), Computer Information Science Department, Universiti Teknologi PETRONAS, Seri Iskandar, Perak, Malaysia 2 grid.412832.e 0000 0000 9137 6644 Department of Computer Sciences, College of Computing, Al Qunfudhah, Umm Al-Qura University, Makkah, Saudi Arabia 13 4 2022 2023 56 2 865913 5 4 2022 © The Author(s), under exclusive licence to Springer Nature B.V. 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Deep neural networks (DNN) have remarkably progressed in applications involving large and complex datasets but have been criticized as a black-box. This downside has recently become a motivation for the research community to pursue the ideas of hybrid approaches, resulting in novel hybrid systems classified as deep neuro-fuzzy systems (DNFS). Studies regarding the implementation of DNFS have rapidly increased in the domains of computing, healthcare, transportation, and finance with high interpretability and reasonable accuracy. However, relatively few survey studies have been found in the literature to provide a comprehensive insight into this domain. Therefore, this study aims to perform a systematic review to evaluate the current progress, trends, arising issues, research gaps, challenges, and future scope related to DNFS studies. A study mapping process was prepared to guide a systematic search for publications related to DNFS published between 2015 and 2020 using five established scientific directories. As a result, a total of 105 studies were identified and critically analyzed to address research questions with the objectives: (i) to understand the concept of DNFS; (ii) to find out DNFS optimization methods; (iii) to visualize the intensity of work carried out in DNFS domain; and (iv) to highlight DNFS application subjects and domains. We believe that this study provides up-to-date guidance for future research in the DNFS domain, allowing for more effective advancement in techniques and processes. The analysis made in this review proves that DNFS-based research is actively growing with a substantial implementation and application scope in the future. Keywords Big data Classification systems Deep neuro-fuzzy systems Deep neural network Fuzzy systems Optimization methods Universiti Teknologi Petronas (MY)015MA0-013 Talpur Noureen issue-copyright-statement© Springer Nature B.V. 2023 ==== Body pmcIntroduction Over the past decades, significant progress in the field of artificial intelligence (AI), machine learning, and deep learning has been made with several real-world problems having been successfully solved. The success in these fields has resulted in the emergence of various methods including fuzzy logic, swarm intelligence, genetic programming, and hybrid approaches, such as neuro-fuzzy and genetic fuzzy systems, all of which have contributed to the design and analysis of complex intelligent systems. Among these methods, deep learning techniques such as deep neural networks (DNN) have made major advances in solving problems that have resisted the best attempts of the AI community for many years. The term “deep” is used because the depth of the network is greater than that of conventional neural networks, which are often referred to as shallow networks (Paul and Singh 2015). Conventional neural networks are limited in their ability to process natural data in their raw form. For decades, the construction of pattern-recognition or machine-learning systems requires careful engineering and considerable domain expertise when designing a feature extractor that transforms the raw data into a suitable internal representation or feature vector from which the learning system (often a classifier) can detect or classify patterns in the input (Ashraf et al. 2020). Compared to typical neural networks with a single hidden layer, a DNN applies representation learning that allows a machine to be fed with raw data and automatically discover the representations needed for detection or classification using multiple hidden layers (LeCun et al. 2015). Hence, a DNN has turned out to be very good at discovering intricate structures in high-dimensional data, and thus is applicable to many domains of science, business, and engineering. Although a DNN is an effective approach for handling big data problems, the superior accuracy of the model, however, comes at the cost of high complexity. It is therefore essential to note a few points before employing this type of network to solve certain problems. Because a DNN uses more than one hidden layer, it can provide a deeper analytical model; however, each added layer adds computational complexity (Sharma 2019). Further, such networks are inspired by a traditional neural network that utilizes the gradient descent optimization approach for network training. Hence, the DNN frequently encounters the problem of being stuck in the local minima. In addition to these challenges, as the major disadvantage of a DNN, the model is often criticized as being non-transparent and its predictions are not traceable by humans owing to its black-box nature (Buhrmester et al. 2019). It is challenging to trust the findings generated by such deep networks. Hence, there is always the possibility of a communication gap occurring between analysts and DNNs (Bonanno et al. 2017; Hayashi 2020). This downside more often limits the usability of such networks in the majority of real-world problems, where verification of the predicted results is a major concern. To cope with these problems, few studies from the literature (Aviles et al. 2016; El Hatri and Boumhidi 2018; Zhang et al. 2020a, b, c) have combined a DNN with fuzzy systems to produce a novel deep neuro-fuzzy system (DNFS). Fuzzy systems are structures established on fuzzy techniques oriented toward information processing, and are mainly used for implementation in systems where the use of classical binary logic is impossible or difficult. Their main characteristic involves a symbolic knowledge representation in the form of fuzzy conditional IF–THEN rules (Czabanski et al. 2017). Therefore, the novel hybridization of a DNN and fuzzy systems has demonstrated an effective way to reduce uncertainty using fuzzy rules. As an emerging hybrid approach, the use of DNFS has gained enormous popularity among research communities during the past 5 to 6 years in the field of AI. Hence, positive growth in the implementation of this model can be seen in distributed systems, cloud computing, healthcare, and various other areas. However, to the best of the authors’ knowledge, no systematic review has been conducted with the sole focus on highlighting the current progress in the domain of DNFS with detailed facts and figures. This study, therefore, presents a systematic literature survey of the research work published between the year 2015 and 2020 with the following major contributions: First, a comprehensive methodology has been designed to perform an in-depth search in a systematical way by following a revised study mapping process comprised of seven phases (shown in Fig. 1).Fig. 1 Revised study mapping process This paper contributes to deliver the basic concept of DNFS and highlights some of the open questions covering different variations of structural designs that have been introduced in the literature with a combination of deep neural networks and fuzzy systems. The study also covers the optimization methods and techniques that have been widely used to train and optimize the parameters of DNFS. In addition, this paper presents information regarding the intensity of the research conducted in this discipline by performing extensive searches in different scientific databases. One of the research questions included in this paper intends to highlight the applications of DNFS, which is one of the main focuses of this study. Finally, this survey highlights the research gaps, issues, and challenges that require further attention from researchers. It provides a comprehensive body of knowledge and delivers the current status of this particular field, while suggesting some potential future directions. The remainder of this paper proceeds as follows. Section 2 highlights related research based on the available survey studies from the literature, whereas Sect. 3 presents the methodology designed to conduct this systematic review. Section 4 answers the research questions set out in our systematic survey by analyzing the synthesized results of identified publications from available sources in the literature. The identified issues, gaps, challenges, and future areas of study are discussed in Sect. 5. Finally, the conclusions of this study are presented in Sect. 6. Related work From the past 5 to 6 years, successful attempts to hybridize deep learning and fuzzy systems have attracted researchers to implement such a method in various real-world applications. An extensive literature have been published to date, focusing on experimenting with the model in new domains where it has not been implemented in the past. However, considering that DNFS is a novel approach, at present, very few survey studies have been carried out delivering the overall insight regarding this domain. Therefore, the focus of this section is to highlight the survey studies that are conducted under the domain of DNFS. These survey studies were carefully selected and studied to create a general idea of the present state of DNFS research. The survey performed by Dorzhigulov and James (2020) mainly focuses on neuro-fuzzy and similar machine learning models from the perspective of functionality and architectures. In their study, the authors presented an overview of fuzzy systems and described the stunning journey related to the hybridization of fuzzy systems and neural networks. Moreover, deep learning methods such as a DNN can be integrated with fuzzy systems to introduce automated optimization of neural architectures. Therefore, this study described the DNFS architectures, including an adaptive neuro-fuzzy inference system (ANFIS), fuzzy neural networks, fuzzy trees, and overviews of neural architectures that use some fuzzy elements, such as radial basis function networks (RBFN) and a fuzzy adaptive resonant theory map (ARTMAP). The literature has shown significant interest in the domain of control systems and classification using neuro-fuzzy systems. However, most of the neuro-fuzzy systems presented in the literature are software-based solutions that provide improved training algorithms or mathematical and architectural modifications of the model. However, neuro-fuzzy systems still face challenges of slow training when dealing with big data, which affects its overall performance. There have been limited studies implementing neuro-fuzzy systems as dedicated high-performance hardware, including (Jhang et al. 2018; Khati et al. 2019; Mata-Carballeira et al. 2019), that have proposed the use of field-programmable gate array (FPGA) devices. This hardware solution tends to be more efficient and faster, but with a trade-off in flexibility. Recently, only one study (Marlen and Dorzhigulov 2018) can be seen using memristive crossbar arrays with a fuzzy membership function that acts as a resistor, capacitor, and inductor. Hence, the study of Dorzhigulov and James (2020) suggests that in the near future, hardware solutions should be used to improve the performance and speed of these hybrid approaches. In the same vein, another recent and interesting study (Das et al. 2020) found in the literature explored the different ways in which deep learning is improved with fuzzy logic systems along with the utilization of the model in various real-life applications. It can be seen that using fuzzy theory along with deep learning can improve the performance of models in which the data are noisy, heterogeneous, incomplete, or vague. However, a problem of computational complexity may occur when utilizing fuzzy systems. The availability of software platforms such as the Compute Unified Device Architecture (CUDA) of Nvidia, the Radeon Open Compute (ROCm) ecosystem released by Advanced Micro Devices, Inc. (AMD), and the Math Kernel Library (MKL) by Intel further accelerate the deep learning processes. The computation of fuzzy parameters is time-consuming using the presently available architectures, despite the models providing resistance to noise and searches over a wider space. Alternatively, fuzzy logic can be used alongside standard deep learning models to process the input or output. Models can make use of fuzzified inputs coupled with standard deep learning models such as deep belief network (DBN) or convolutional neural networks (CNN). This allows leveraging software platforms to accelerate DNN training using fuzzy systems. This study further suggests exploring better ways to improve the performance of fuzzy deep learning models in the future. Taking a deeper look into deep learning-based neuro-fuzzy systems, an excessive and appealing approach can be found in (Singh and Lone 2020). This study develops the basics for readers from fuzzy sets to the concepts of fuzzy rules and reasoning to understand membership functions with the help of real-world scenarios and case studies in simple mathematics. Furthermore, it describes the working style of Mamdani fuzzy inference systems, Takagi–Sugeno–Kang (TSK) fuzzy inference systems, and Tsukamoto fuzzy inference systems, along with explanations of how these three models vary from each other. A CNN, natural language processing (NLP), and recurrent neural networks (RNN) were implemented in the subject area of computer vision and time-series prediction. Different variations of the architectures have been defined with the integration of fuzzy systems and deep learning. Insight into these hybrid approaches as intelligent systems in the modern world are provided. In addition, it simplifies the implementation of fuzzy logic, neural networks, DNFS, and related concepts using Python, which encourages readers to experiment with these machine learning and deep learning methods. This study not only builds the fundamentals but also encourages newcomers in the field of AI to implement these methods in their respective research areas. The fourth and last survey study was conducted by de Campos Souza (2020). This study aims to describe the proposed methodologies and existing and improved techniques, including the implementation of neuro-fuzzy in applications such as pattern classification, time-series prediction, fault detection, and various other approaches developed since the year 2000. Moreover, the author provided a well-defined model architecture, describing its problem-solving abilities, mechanisms, training algorithms, and different ways to extract information through fuzzy rules. The major emphasis of the study was to specifically survey neuro-fuzzy systems from the literature that provide supervised learning. The central focus of this survey is to gain in-depth knowledge of the neuro-fuzzy systems without the implementation of deep learning. However, a small section of the study focuses on the training of neuro-fuzzy models using deep learning methods to perform the tasks of data classification (Deng et al. 2017), traffic incident detection (El Hatri and Boumhidi 2018), and sentiment analysis (Nguyen et al. 2018). The survey also mentioned a few studies implementing a semi-supervised DNFS for image classification (Xiaowei Gu and Angelov 2018) and remote sensing scene classification tasks (Gu et al. 2018). As future work, this study not only suggests exploring dynamic hybrid model architectures, it also advises on addressing new learning algorithms. From this section, it is clear that currently there are only four survey studies present in the literature that discuss novel DNFS approaches. It is important to note that these studies did not focus on DNFS only. The main motivation of these survey studies is to build general concepts about DNFS models along with various similar techniques, such as neuro-fuzzy systems without a deep learning approach. Although these studies comprehend helpful knowledge on understanding the basic concept of a hybrid model, they do not explore the trends and developments regarding DNFS. Based on these observations, Table 1 provides a summarized comparative analysis of the above-mentioned four survey studies.Table 1 Comparative analysis of the review studies presented in the literature Study Advantages Limitations Findings Dorzhigulov and James (2020) Emphasis on building basic understanding about fuzzy inference Well explained different architectures of neuro-fuzzy systems Does not cover application areas Scope in the domain of deep neuro-fuzzy is missing Less discussion on combining deep learning methods with neuro-fuzzy systems The study is mainly a survey, rather than a systematic literature review showing the trend and progress of DNFS The optimization methods of DNFS are not highlighted The study suggests proposing hardware solutions (FPGA, memristive) for the neuro-fuzzy model to deal with big data efficiently Das et al. (2020) The study explains the architectures of deep neuro-fuzzy architecture in depth The study covers the majority of the application areas The study is mainly a survey, rather than a systematic literature review showing the trend and progress of DNFS The optimization methods of DNFS are not highlighted Combining fuzzy systems with deep learning can increase computational complexity. The study advises using software platforms such as CUDA, ROCm, and MKL to improve the speed of deep learning. However, the improvement based on the performance of the deep learning training method needs to be explored further in the future Singh and Lone (2020) Explains the DNFS architecture with simple mathematics The study provides examples of implementing CNN, NLP, and RNN for the tasks related to computer vision and time series prediction Does not cover application areas Scope in the domain of deep neuro-fuzzy is missing Less discussion on combining deep learning methods with neuro-fuzzy systems The study is mainly a survey, rather than a systematic literature review showing the trend and progress of DNFS The optimization methods of DNFS are not highlighted N/A de Campos Souza (2020) The study provides a review mainly focused on neuro-fuzzy systems, their architectures, training algorithms, and applications areas However, it slightly touched on training neuro-fuzzy using deep learning Covers supervised neuro-fuzzy models The overall study is much appealing providing in-depth knowledge regarding neuro-fuzzy systems Does not explain DNFS architecture Scope in the domain of DNFS is missing The majority of application areas mentioned are on the neuro-fuzzy systems without deep learning The study is mainly a survey, rather than a systematic literature review showing the trend and progress of DNFS The optimization methods of DNFS are not highlighted Future work suggests providing studies more focused on presenting new learning algorithms. Since neuro-fuzzy systems with deep learning are becoming the hot topic, the author also encouraged to come up with a revised version of the study that covers knowledge in this domain from a wider perspective The summary of the comparative analysis presented in Table 1 provides confirmatory evidence of a research gap. The current literature does not offer comprehensive, systematic, and more importantly, quantitative research knowledge for researchers who wish to explore the scope and current research progress on DNFS in the area of AI, particularly in hybrid approaches. Therefore, our systematic survey aims to broaden the knowledge of readers by presenting not only a basic structural understanding, but also aims to deliver the information regarding the optimization methods and application domains through a quantitative analysis, facts, challenges, scope, and future suggestions. To fulfill the above-mentioned objectives, the next section (Sect. 3) provides the detailed methodology that has been followed to construct this systematic survey. Methodology The purpose of this systematic literature survey is to provide a complete list of all possible studies related to DNFS reported in the recent literature between the year 2015 and 2020. Various attractive approaches to conduct the survey studies have been implemented in (Singh and Singh 2020; Yu and Pan 2021; Yu and Sheng 2020). The guidelines followed in this paper are taken from systematic literature reviews published by Baashar et al. (2020), Schön et al. (2017), and Muhammed et al. (2018), whereas, a revised study mapping process illustrated in Fig. 1 is designed for this review paper, which is a combination of preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement (Moher et al. 2009; Safdar et al. 2018); and a systematic mapping process presented by Hussain et al. (2019). It is comprised of seven phases (as shown in Fig. 1), i.e., a preliminary study, formulation of research questions, identification of the search criteria, all papers found in a literature search, the screening process, an eligibility and quality assessment of the selected studies, data extraction and compilation, and the final list of articles included in this systematic study after applying the exclusion criteria. Phase 1: Preliminary study A preliminary study is an initial and significant phase applied when conducting a systematic literature survey. In this phase, we narrow down the search parameters by obtaining the background information for studies related to deep neuro-fuzzy systems. A random search was performed on the common search engines over the Internet using a single domain-specific keyword, such as deep neuro-fuzzy systems. Afterward, we filtered out the search results and focused on those studies that helped us to choose search strings strictly related to the domain of our study. Phase 2: Formulating research questions In this phase, we formulated the key research questions that can direct us throughout the research and writing process. We determined the quality of the research questions based on the elements of constructiveness, focus, and relevance to a specific area or problem. Therefore, an extensive research has been conducted on the development of neuro-fuzzy systems (NFS). However, it is challenging to find sufficient literature guidance regarding novel deep neuro-fuzzy systems (DNFS) that involve deep learning methods such as basic DNNs. Therefore, the primary motivation behind this study is to prepare a systematic review regarding the development of DNFS, their optimization methods, and their application subjects. The following are the fundamental research questions:RQ-1 What are the fundamental concepts related to deep neuro-fuzzy systems? Motivation Answering this question, will set the basics and help in understanding the primary knowledge regarding deep neuro-fuzzy systems. RQ-2 What approaches have been widely employed for the optimization of deep neuro-fuzzy systems? Motivation This question aims to identify optimization techniques involved in the training and learning of deep neuro-fuzzy systems. RQ-3 What is the intensity of publications in the domain of deep neuro-fuzzy systems in terms of year-wise and directory-wise? Motivation In this question, the potential of deep neuro-fuzzy systems is investigated using conference proceedings, journals, articles, and book chapters. RQ-4 What are the most promising and practical application subjects and domain areas where deep neuro-fuzzy systems have been implemented? Motivation This question investigates the studies that highlight the application, particularly in the domain of deep neuro-fuzzy systems in multiple potential subject and domain areas. Phase 3: Identification of search criteria The preliminary search was conducted using the Google Scholar search engine, which helped to formulate the research questions for this study. Next, research questions were used in this survey to identify the relevant keywords and search terms/strings related to the DNFS topic. The identification stage of this study further helped us to explore the topic from a broader perspective while employing specific search criteria. In this survey study, five scientific databases and venues (Table 2) were selected for a more advanced search using finalized keywords (Table 3) from the domain of DNFS. The selected databases were chosen because they offer a wide variety of the most essential and highest-impact journals and conference proceedings. Because each scientific database uses different search features and filters to perform a systematic search, it is important to adjust the search string for every scientific database. Table 2 presents the search criteria used to apply the systematic search for this study using the selected scientific databases and search engines.Table 2 Identified search criteria No Scientific databases and search engine Search scheme Publication type Publication year 1 The ACM Digital Library Title, abstract, and keywords Journals papers, conference proceedings, book chapters, technical reports 2015 to 2020 2 IEEE Xplore Title, abstract, and keywords 2015 to 2020 3 Scopus Title, abstract, and keywords 2015 to 2020 4 ScienceDirect Title, abstract, and keywords 2015 to 2020 5 SpringerLink Title, abstract, and keywords 2015 to 2020 6 Google Scholar (Search Engine) Full Text 2015 to 2020 Table 3 Keywords and search strings Domain focus Keywords Search string Deep neuro-fuzzy systems (DNFS) Deep neuro-fuzzy systems, Deep neuro-fuzzy optimization, Deep neuro-fuzzy applications subjects (“fuzzy systems” OR “fuzzy logic” OR “fuzzy inference systems”) AND (“deep neural network” OR “deep learning” OR “deep networks”) OR (“fuzzy deep” OR “deep fuzzy”) AND (“neural networks” OR “networks” OR “systems” OR “models” OR “approach”) AND (“optimization methods” OR “optimization techniques” OR “optimization”) AND (“applications” OR “utilization” OR “implementation”) OR (“practices”) The search was performed with the aim of obtaining the maximum number of relevant studies published in these databases based on the specific keywords and search strings using the logical operators OR and AND, as presented in Table 3. However, it is important to modify the search strings to meet the unique specifications of each database. Moreover, the search strings highlighted in Table 3 were used to retrieve research papers for RQ1, RQ2, and RQ4. Meanwhile, RQ3 is specifically formulated to investigate the intensity and potential of DNFS based on the publications published between the year 2015 and 2020 using the keywords and search strings shown in Table 3. Phase 4: Screening process (inclusion and exclusion criteria) After completing the database search using the identified keywords and search strings, all retrieved papers were screened based on the inclusion and exclusion criteria used to filter and select only the studies that were relevant to answering the research questions, while excluding studies that were not. Table 4 lists the criteria that were followed throughout the screening process to evaluate each paper and decide whether to include or exclude the paper in the systematic literature survey.Table 4 Inclusion and exclusion criteria for search screening Inclusion criteria Journal papers, conference proceedings, articles, and book chapters relevant to the domain of DNFS Studies with optimization techniques of DNFS Studies with application areas of DNFS in computing, engineering, industry, economics, healthcare, etc Published between the year 2015 until 2020 Studies published in English only Exclusion criteria Publications not relevant to the domain of DNFS Papers less than five pages, tutorial, seminar, interview, blog, or poster Duplicate articles Studies published in languages other than English Phase 5: Eligibility and quality assessment In addition to the inclusion and exclusion criteria, it is critical to assess the eligibility and relevance of the primary studies found in the previous stage. In this survey, three quality assessment scores applied to answer every question were adopted from (Hordri et al. 2017) to examine the eligibility of individual study based on the factors, such as the significance of a particular study, the quality of the results and analysis, and future research guidelines or findings. The scoring procedures are 1 (Yes), 0.5 (Partly), and 0 (No), whereas Table 5 describes the quality assessment scores employed to check the eligibility of each DNFS paper from the records.Table 5 Eligibility criteria in terms of the scores for each paper for a quality assessment Item Criteria Score Description Q1 Are the objectives and goals of the study well-defined? 1 It clearly defines the objectives and goals of the study 0.5 It follows the objectives of the article, but the goals are not clearly explained 0 It does not follow the objectives, nor are the goals of the study well-defined Q2 Are the methodology and experimental process well-explained? 1 It clearly explains the methodology and experimental process performed on DNFS 0.5 It evidently illustrates the methodology but does not provide an explanation of the experimental process for DNFS 0 It does not clearly explain the methodology and experimental process Q3 Are the limitations of the study well-acknowledged? 1 The limitations of the study on DNFS are well-acknowledged 0.5 The limitations of the study on DNFS are stated, but not in detail 0 The limitations of the study are not well-acknowledged Q4 Is there a clear statement of the research findings? 1 Findings are explicit, easy to understand, and in a logical progression. Tables, if present, are explained in the text. Results relate directly to the aims. Sufficient data are presented to support the findings 0.5 Findings were mentioned, but more explanation could be given. Data presented relate directly to results 0 Findings presented randomly, not explained, and do not progress logically from results, and/or the findings are not mentioned or do not relate to the aim of the study Phase 6: Data extraction and compilation After appropriately classifying the studies to be included in the systematic literature survey following the steps in Phases 4 and Phase 5, we performed a data extraction and compilation to examine and compare the relevant studies. Generally, the data extraction and compilation are performed using a variety of available tools and software, including Microsoft Excel (spreadsheets), REDCap, and Google Sheets. We utilized Microsoft Excel to record data from the publications to answer the research questions and achieve the goals of the study. The following information was extracted from each included study: title, abstract, keywords, authors, publication year, scientific databases/venue, publication type (e.g., journal, conference, or book chapter), the technique used (new method, modified, or hybrid approach), and study type (e.g., analysis, survey or a mixture of both). Phase 7: Included (final findings and results) Figure 2 summarizes the entire revised study mapping process according to the PRISMA guidelines using a flow chart. As illustrated in Fig. 2, a total of 252 studies were found using the keywords and search strings (as listed in Table 3) in scientific databases including ACM, IEEE Xplore, Scopus, ScienceDirect, and SpringerLink from the years 2015 to 2020 during the identification phase. The screenings of the 252 collected records were then conducted based on the inclusion and exclusion criteria mentioned in Table 4.Fig. 2 PRISMA flow chart for selection of the studies in the systematic literature survey At this stage, 166 relevant studies were included. The preferences were given to the journal articles, conference proceedings, and book chapters relevant to the domains of DNFS, its optimization methods, and applications in various domains that are written in English language and published between the year 2015 and 2020. A total of 86 studies were excluded at this stage because they were not related to the DNFS domain, written in languages other than English, published prior to the year 2015, or were published as tutorials, short papers, interviews, blogs, or duplicated publications. The 166 relevant studies screened from the previous phase were further cross-checked for eligibility based on the eligibility criteria specified in Table 5 of phase 5 before being included in this systematic literature survey. The eligibility criteria were designed to assess whether the studies have clearly defined objectives, a clearly presented methodology, a clear experimental process, stated limitations, and findings based on scoring outputs of 1 (yes), 0.5 (partly), and 0 (no). As a result of the quality assessment process, a total of 105 studies remained in the records, whereas 61 studies were excluded because they did not meet the eligibility criteria. Therefore, after following the mapping process, the final 105 collected studies were included in the systematic literature survey to answer the RQs of this study and highlight the research gaps, issues, and challenges of this particular domain. Furthermore, Fig. 3 shows the publication type of the 105 included studies based on their publication avenues, such as studies published in journals, conference proceedings, book chapters, and preprint servers for the analysis.Fig. 3 Included studies based on the publication types Analysis and synthesis of data This section answers the research questions in this study and provides a better understanding of the collected data. RQ-1: What are the fundamental concepts related to deep neuro-fuzzy systems? A deep neuro-fuzzy system (DNFS) is an advanced concept of hybridization, where deep learning approaches, such as deep neural networks and fuzzy logic approaches, are combined to solve various real-world complex problems involving high-dimensional data. That said, before going into the details of DNFS models, it is essential to build preliminary knowledge for the readers by providing an overview of deep neural networks and neuro-fuzzy systems, which makes it easier to comprehend the idea behind developing DNFS. Deep neural network (DNN) Deep learning enables multi-layer cognitive models to learn and interpret data with several levels of abstraction, replicating the brain perception and knowledge representation; hence, it is indirectly capable of capturing complex large-scale data structures. In comparison to various existing state-of-the-art methods, the importance of deep learning approaches is growing rapidly owing to their extraordinary performance in several applications, such as visual, audio, social, and medical data (Voulodimos et al. 2018). In general, AI models are trained to perform data processing tasks based on hand-crafted features derived from raw data or features learned from other basic AI models. Using deep learning, computers can automatically learn useful representations and features directly from the raw data, avoiding the challenging step of manually crafting the features (Lundervold and Lundervold 2019). Deep learning techniques have achieved an excellent performance in computer vision, automatic speech recognition, and natural language processing. It is evident from the term “deep learning” that a DNN model involves a greater number of processing layers, instead of fewer layers in a simple neural network recognized as a shallow learning model. The advancement from shallow to deep learning models has increased the possibility of dealing with complex and nonlinear functions (Shrestha and Mahmood 2019). Neuro-Fuzzy Systems (NFS) Several NFS have been presented in the literature. However, the ANFIS model is the most frequently used approach. The concept of ANFIS was presented by Jang in 1993, which is a proficient combination of neural networks and fuzzy logic (Hussain et al. 2015). The key benefit of a neural network is the ability to learn from data. However, such a network is considered a “black-box,” because it does not clarify how the final outcome is achieved. Therefore, with the help of IF–THEN rules in fuzzy systems, one can interpret the results generated by the model (Kruse and Nauck 1998). The following is a presentation of the standard fuzzy rule:1 IFxiAandyisBTHENz=f(x,y) where A and B are fuzzy sets, and z is a polynomial or a constant (Emad Hussen et al. 2020). Hence, a neuro-fuzzy model can incorporate human knowledge and self-learning competencies that can potentially approximate every situation (Mohd Salleh and Hussain 2016). ANFIS has gained popularity among the research community as compared to other variations of fuzzy and neuro-fuzzy systems because it has been successfully implemented in various classification, rule-based process controls, and pattern recognition applications. It also embeds learning mechanisms to adapt and update all adjustable parameters of the model with two-pass learning algorithms, i.e., forward pass and backward pass. This algorithm is a hybrid of gradient descent (GD) and a least squares estimator (LSE), which helps to adjust the antecedent and consequent parameters of the ANFIS model to minimize the error between the actual output and the targeted output (Salleh et al. 2018). Novel hybrid approach of deep neuro-fuzzy system (DNFS) For many years, the scientific community has explored several ways to execute sophisticated algorithms that can learn from data, the main barriers of which were the computational power and limited data. Such attempts and years of development to solve these problems have resulted in an exciting new subfield of machine learning called deep learning (DL). A popular algorithm within DL is a DNN (Bonanno et al. 2017). A DNN attempts to learn multiple levels of abstractions and representations to locate complex relations between data. This evolving subfield has been rising rapidly over the past few years owing to the ever-increasing computational power and unlimited access to data. Although DNNs have shown remarkable progress with respect to feature learning from big data, the model is unable to express the uncertainties with data because of its “black-box” nature (Shwartz-Ziv and Tishby 2017). Despite the usefulness of a DNN, a gap in communication exists between the analysis and DNN. Since the beginning of big data, massive amounts of data have been produced daily around the world within the domain of science and industry. However, when the volume of data increases, the existence of noise and unpredictable uncertainties in large amounts cannot be ignored, which is another critical issue of data ambiguity (Angelov and Gu 2018). This issue again becomes challenging for many autonomous systems because the majority of these machine learning models are designed and trained using labeled data (Bonanno et al. 2017). The drawbacks of such methods have been solved by introducing an additional machine learning process into neural networks, that is, fuzzy inference, to create an explainable rule-based structure known as a neuro-fuzzy system (NFS). The NFS allows experts to generate rule-based structures. Once the rules are generated, it is possible for the experts to bias features generated from the DL by providing feedback to the system. In addition, through these rule-based structures, an analyst can easily understand how a decision has been made by the system (An et al. 2019). Using fuzzy if–then rules, neuro-fuzzy systems such as ANFIS can approximate complex nonlinear problems. These systems can be applied in various applications that involve encoding both objective measurements and subjective information (Bonanno et al. 2017). Therefore, with the emergence of the deep learning concept and a DNN, some researchers have introduced fuzzy inference system elements and modules into such systems to address the possible uncertainties in the raw data, similar to ANFIS. Recently, a few studies have incorporated the concepts of an explainable rule-based structure called fuzzy inference with a DNN as a DNFS to overcome the black-box problem of DNN (An et al. 2019). Figure 4 shows DNFS, combining the advantages of fuzzy systems and a DNN (Aviles et al. 2016).Fig. 4 Representation of DNFS by combining the advantages of fuzzy systems and a DNN DNFS comprises a broad category of hybrid systems that combine the properties of the DNN and fuzzy inference systems in different architectures. The structural designs found in the literature are classified into three categories, i.e., a sequential, parallel, and cooperative DNFS by integrating fuzzy inference systems with a DNN. The following subsection briefly explains the structures of the three categories and their respective examples from the literature. Sequential structural designs of deep neuro-fuzzy system A sequential DNFS is suitable for solving problems involving high linearities, such as time-series data, text documents, sentiment or video classification, and speech recognition. In sequential structural design, data processing in a fuzzy system and a DNN take place one after the other, as presented in Fig. 5a, b. In fuzzy theory, a fuzzy set A in a universe of discourse X is represented by a membership function μA taking the values from the unit interval as μA:X→[0,1]. At this stage, a membership function shows the degree of similarity for a data point within the universe of discourse of x∈X (Yazdanbakhsh and Dick 2020). The approximate reasoning and decision-making ability of fuzzy logic assist the fuzzy system in effectively describing the uncertainty of the real world. It can work with the data having the characteristics of imprecision, ambiguity, and uncertainty (Gallab et al. 2019; Vlamou and Papadopoulos 2019).Fig. 5 Sequential DNFS: a fuzzy systems incorporated with a DNN and b a DNN incorporated with fuzzy systems The process in the first approach of DNFS’ sequential structural designs (Fig. 5a) starts by taking the input features of the data and converting the values into fuzzy sets which are processed by DNN. This means that the inputs enter the fuzzy system to get the fuzzy linguistic values. Afterward, the neural network helps to generate the outputs of the sequential DNFS model. Similarly, the process in Fig. 5(b) works in the opposite way (Vieira et al. 2004). The DNN model assists the fuzzy system in determining the desired parameters when the DNFS model cannot measure the input values directly from the data (Abraham 2001). To develop a better understanding of sequential DNFS, the study of (Sarabakha and Kayacan 2019) is presented in this review paper. This study presents a deep fuzzy neural network (DFNN) proposed by Sarabakha and Kayacan, which uses one antecedent fuzzification layer for learning control of nonlinear systems. As illustrated in Fig. 6, the DFNN neurons are organized in an input layer with ninp neurons, a Gaussian fuzzification layer μx with ngF neurons, hidden layers nhL with (nH + 1) neurons in each layer, and an output layer with yout neurons.Fig. 6 Illustrates an example of the sequential DNFS In a fuzzy set theory, the degree of truth is defined by the membership function which contains the curve. This curve represents every single point in a specified input space. The inputs x1,⋯xninp to the DFNN are fuzzified using three Gaussian membership functions, i.e., cgF,1=-1, cgF,2=0, cgF,3=1, and σgF,1=σgF,1=σgF,1=1, as indicated in Eq. (2). The two parameters of c and σ defines the center/mean and width of the curve/variance. Figure 7 shows the representation of the Gaussian membership function.Fig. 7 Gaussian membership function 2 μgFlxj=e-12xj-cgF,lσgF,l2,j=1,⋯,ninpandl=1,2,3 The fuzzified inputs μgF1x1,μgF2x1,μgF3x1,⋯,μgF1xninp,μgF2xninp,μgF3xninp are forwarded to the first hidden layer of the DFNN through the weights w1 of the network. The DFNN hidden layers are aligned in a fully connected model using network weightswi,i=2,⋯,nhL-1. Finally, the output y1,⋯,yout is calculated using the weights wnhL until reaching the output of the final hidden layer. The weights of network Wi,i=1,⋯,nhL, are restricted by some positive constants ofcW,i,i=1,⋯,nhL, i.e.,3 ‖wim‖∞≤cW,i∀mi=1,⋯,nhL The learning of this model is divided into two stages: offline pre-training and online training. During the offline pre-training process, the classical controller executes a series of trajectories and collects a batch of training samples. The controller based on a DFNN, such as DFNN0, is then pre-trained on the data samples obtained to estimate the inverse dynamics of the system. Because DFNN0 cannot tune the new conditions, online training is conducted at this stage. During this stage, the DFNN continuously monitors and updates the input of the controller to improve the performance. The DFNN adaptive information is generated by expert knowledge encoded in the rule-based method using fuzzy mapping. The approximation of the inverse dynamics of the system is a typical problem of regression; thus, the mean square error was set as a cost function in this study for both offline and online training. Various other examples of similar deep neuro-fuzzy structure designs can be found in the literature (An et al. 2019; Aviles et al. 2016; Dabare et al. 2019; El Hatri and Boumhidi 2018; Korshunova 2018; Liu et al. 2020a, b; Ramasamy and Hameed 2019; Yeganejou and Dick 2018, 2019). Parallel structural designs of deep neuro-fuzzy system In a parallel structural design, data are processed separately from the fuzzy systems and a DNN, and then fused to obtain the final output of the data, as shown in Fig. 8. The parallel structure uses a fuzzy system with a hierarchical DNN that derives information from both fuzzy and neural representations. The knowledge learned from these two respective views is then combined to form the final data representation for the classification (Chen et al. 2019; Deng et al. 2017).Fig. 8 Structural design for parallel or fused DNFS An example of a parallel or fused structure can be seen in the study of Chen et al. (2019), where authors have proposed a hierarchical Pythagorean fuzzy deep neural network (HPFDNN). As illustrated in Fig. 9, the HPFDNN model consists of four phases: fuzzification (Pythagorean fuzzification), neural net, fusion, and learning phases.Fig. 9 Illustrates an example of parallel/fusion DNFS Phase 1 - Pythagorean fuzzification phase of HPFDNN In this phase of the model, inputs are linked to several Gaussian membership functions defined in Eq. (2) and Fig. 7 to determine the degree of membership and to specify the input belonging to a certain fuzzy set. In this phase, if p is input, q is an output, and n is the layer number, pin denotes the input of the i-th neuron of the n-th layer. Similarly, qjn represents the output of the jth neuron in the n-th layer. The output is processed through a Pythagorean fuzzification, which is defined with parameter r to indicate the non-membership function as follows:5 qi(n)=spin=μi2pin-ri2(pin) Phase 2 - Neural net phase of HPFDNN In this phase, the neural net is formed based on the perceptron. In the perceptron, input values are multiplied by the weight of each neuron, and when the degree of the entire input signal exceeds the defined threshold value, a neuron produces the output. This is achieved by computing the sum of the weighted inputs with the threshold neural function (NF) on the sum to produce the output. Subsequently, the DNN converts the inputs into high-level representations by activated neurons, and the neural net phase helps the model acquire neural features. In this phase, a sigmoid activation function (Fig. 10) was used as follows:Fig. 10 Sigmoid activation function 6 ai(n)=11+e-pi(n) where qin=wi(n)ai(n)+b(n). Here, the activated weights and biases are presented as the output of the neural net. Phase 3 - Fusion phase of HPFDNN In the proposed HPFDNN model, the fusion of fuzzy and neural nets is processed to obtain the output using the following operation:7 qi(n)=qf(n-1)+qn(n-1) where qf and qn represent the outputs of both the fuzzy and neural phases. Phase 4 - Learning phase of HPFDNN This final phase of the HPFDNN generates a trained DNN model where the output of every layer is used as the input for the upcoming layer as follows:8 q(n)=wn-1(n)q(n-1)+b(n) where a weight matrix connection is presented by wn-1(n) in layer n with n-1. The bias vector is represented by b(n). In addition, a sigmoid function is used in each hidden layer of this phase to transform the de-normalized output q with real values as follows:9 q(n)=11+e-wn-1(n)q(n-1)+b(n) Cooperative structural designs of deep neuro-fuzzy system In cooperative deep neuro-fuzzy designs, there are two potential models of DNFS, as illustrated in Fig. 11a, b. In Fig. 11a, the fuzzy interface block converts the crisp input into fuzzy values to provide an input vector to a multi-layer neural network in response to linguistic statements. Then, the DNN is trained to generate the required outputs, and defuzzification of the outputs is performed to convert the fuzzy value into a crisp output value. As shown in Fig. 11b, the fuzzy inference mechanism is determined by a multilayered DNN. Fuzzy systems obtain the computational characteristics of learning offered by a DNN, and in return, the DNN receives the interpretation and clarity of the system representation (Phuong and Kreinovich 2001).Fig. 11 Structural design of cooperative DNFS: a fuzzy deep neural network and b deep neuro-fuzzy network A simple example of a cooperative DNFS was proposed by (Yeganejou et al. 2020) using a CNN for feature extraction and by transferring the outputs of the final convolutional layer for fuzzy classification, as depicted in Fig. 12.Fig. 12 Illustrated example of cooperative DNFS With the proposed model, modifications can be made to the CNN based on a dataset or individual needs. In Fig. 12, layers 1–3 are the same as those in the LeNet architecture. Subsequently, the data paths are divided into two parts. In the case of a deep fuzzy structure, the feature maps of layer 3 are extracted and sent for fuzzy clustering. Rocchio’s algorithm was employed to generate the final classification results for the network. The other path leads to a more conventional deep network including subsampling layer 4, a fully connected layer with ReLU neurons, and finally to a fully connected layer with softmax activation functions.10 oi=ewiM→.Ofcl→∑n=1NewnM→.Oflc→ where Oi is the i-th network output, Oflc→ is the output from the former fully connected layer, wiPM→ is the modified value of weights for the i-th softmax neuron, and the number of neurons in the softmax layer are represented by n and wiM→. Here, Oflc→ represents the logit value of the output layer. The outputs of the network are defined as [0, 1], and the summation to 1 for all inputs of the network. The fuzzy classifier unit of the proposed model implies a process for selecting a feature map, which states that either all elements are selected from the feature map, or none of the elements are selected. This process helps to omit redundant feature maps, which leads to a better classification accuracy. Next, a PCA-based dimension reduction idea is employed using GK clustering, in which a matrix inversion requires a step with each chosen feature map providing 196 features of a 14 × 14 image. Fuzzy clustering was executed, and Rocchio’s algorithm (Yeganejou et al. 2020) (presented in Fig. 13) was employed as a classifier.Fig. 13 Flow of fuzzy Rocchio’s algorithm A mini-batching variant of stochastic gradient descent was used as an additional momentum to train the network.11 Δwijt=μδjtpit+mΔwij(t-1) where Δwijt represents the tuning of the ij-th weight after observing the t-th mini-batch of the input patterns. The learning rate is presented as μ, δjt shows the error for neuron j, and m is the constant of the momentum. Subsequently, the former mini-batch of the ij-th weight is tuned using Δwij(t-1) and a cross-entropy loss function LCE is used as follows:12 LCE=-∑n=1Nbnlog(lOn\|I→) where bn states the ground-truth of the class that the current input I→ is n, and l shows the probability that the output On will be projected by the deep network for input I→. In addition, other examples of cooperative-type DNFS have been reported in (Greeshma and Bindu 2017; Nguyen et al. 2018, 2019; Samanta et al. 2019). RQ-2: What approaches have been widely employed for the optimization of deep neuro-fuzzy systems? Optimization plays an extremely important role in discovering the best solution from a set of available options with minimal resources. In the field of computing, engineering, or a simple task of online shopping, to find solutions rationally, optimization methods help to identify the best solution from a wider range of possible options. Similarly, various optimization methods, such as a gradient descent, stochastic gradient descent, and conjugate gradient methods, have been adapted by machine learning and deep learning techniques for parameter optimization. These methods are well known as exact methods. However, metaheuristic algorithms have gained more popularity over exact methods for solving optimization problems owing to the simplicity and robustness of the results generated when implemented in a wide range of fields, including engineering, industry, transport, and even social sciences (Hussain et al. 2019). The exact methods are suitable for delivering optimal solutions for smaller problems by following local search mechanisms, whereas metaheuristic-based methods have shown a significant performance in finding optimal solutions when solving large-scale problems using their ability of a global search (Kolajo et al. 2019). The metaheuristic concept further offers a search mechanism based on a single solution and population-based methods. Population-based (PB) metaheuristics offer a wide range of algorithms, such as evolutionary algorithms (EA) or swarm intelligence (SI)-based metaheuristics. The EAs are composed in a population of individuals, where each individual represents a search point in the space of possible solutions, and are subjected to a collective learning process for transmitting the information to the next generations. In SI, the individual member in a swarm works independently on the basis of their stochastic behavior and observations from the neighborhood or surroundings environment (Kurban et al. 2014). This section aims to provide an insight into the optimization methods used for optimizing the DNFS in the included studies. A careful investigation of the optimization method is presented in Table 6.Table 6 Search results for optimization methods Publishers Type of optimization methods Exact methods PB metaheuristics Hybrid methods Others Not mentioned ACM 3 1 0 3 1 IEEE Xplore 37 3 2 4 8 Scopus 3 0 0 0 1 ScienceDirect 12 1 1 1 3 SpringerLink 21 0 0 0 0 Total 76 5 3 8 13 The deep analysis of the final data revealed that most papers on the DNFS model have employed exact methods for network optimization, as presented in Table 6. Meanwhile, only five studies have employed population-based (PB) metaheuristic approaches to optimize DNFS such as brain storm optimization (BSO) (Ravi 2020), elephant herd optimization (EHO) (Velliangiri and Pandey 2020), genetic algorithms (GA) (Lee 2020), crow search algorithm (CSA) (Chandrasekar 2020), and the Jaya optimization algorithm (JOA) (Siva Raja and Rani 2020). However, in three studies, the optimization was performed by combining one metaheuristic with another metaheuristic algorithm such as the genetic algorithm (GA) with big bang-big crunch (BB-BC) (Chimatapu et al. 2018), biogeography-based optimization (BBO) with hessian-free (HF) (Zheng et al. 2017) and BBO with greedy layer-wise training method (Zheng et al. 2016). Few studies in the records were found to explain the model without mentioning any optimization methods. Figures 14 and 15 provides a clearer image of the distribution and record found in scientific databases for each optimization method presented in Table 6.Fig. 14 Overall distribution of the optimization methods used with DNFS Fig. 15 Distribution of the DNFS optimization methods in scientific databases Based on Fig. 15, IEEE Xplore has the highest number of publications in the optimization of DNFS when using different methods. The publication record found in Scopus indicates the least number of studies from the literature to optimize DNFS, and SpringerLink shows that no studies have been published in the directory using optimization techniques other than exact methods. Along with the type and intensity of each optimization method in scientific databases, Fig. 16 identifies the trend of studies in the DNFS domain employing exact optimization methods.Fig. 16 Trend of exact methods in the DNFS domain Similarly, a deeper analysis was carried out to understand the beginning of the trend for researchers using population-based (PB) metaheuristics for DNFS optimization. This will help to identify the scope of metaheuristics-based methods within this domain. From Fig. 17, it is obvious that the implementation of metaheuristic methods in the DNFS domain first took place in the year 2017, and only eight studies were conducted from 2017 to 2020. Moreover, we can conclude that most studies have employed EA methods compared to SI in optimizing DNFS.Fig. 17 Trend of population-based (PB) metaheuristic methods in the DNFS domain RQ-3: What is the intensity of publications in the domain of deep neuro-fuzzy systems in terms of year-wise and directory-wise? According to the primary data extracted from scientific databases, it can be concluded that the advanced research related to the integration of deep learning techniques such as a fuzzy-based DNN in the form of DNFS initially took place in the year 2015 (Laleye et al. 2015) and has gained the attention of the research community ever since. Since then, the model has been employed to solve problems in various application areas. Although the implementation of such a system is still in its early stages, the rise in the number of publications, as indicated in Fig. 18, cannot be ignored. It is clear from the figure that the research towards the integration of a DNN and fuzzy systems attracted researchers more effectively from 2015 to 2016, and a steady growth in the number of publications had occurred until 2017. After a decrease in 2018, a positive increase in the publications on DNFS can be seen in Fig. 18 for the period of 2019 to 2020, along with application subjects and an analysis of DNFS optimization methods. Hence, by observing the intensity of the publications in Fig. 18, it can be concluded that the novel DNFS has a promising scope at present and as well as in the future.Fig. 18 Publications of DNFS year-wise A total of 252 publications found in the literature were published over the past 6 years (from 2015 to 2020). This does not necessarily imply that all publications in the literature have been found, and there remains more to be explored. However, during the DNFS-related keyword search, most of the databases displayed only related studies on the first three pages. This shows that there is still much to be explored in the domain of DNFS and the intensity of the research conducted in this domain has continued to increase over the past 6 years, as shown in Fig. 18. To be more careful, we exceeded our search to a maximum of five pages for each scientific database. The priority was given to studies appearing on the initial five search result pages from ACM, IEEE Xplore, Scopus, ScienceDirect, and SpringerLink. Figure 19 shows the intensity of publications for DNFS in each scientific database. It was revealed that the most popular avenue for published literature related to the DNFS is IEEE Xplore (also with the highest number of conference proceedings, journals, papers, and book chapters) followed by SpringerLink, ScienceDirect, Scopus, and ACM.Fig. 19 Publications of DNFS directory-wise RQ-4: What are the most promising and practical application subject and domain areas where deep neuro-fuzzy systems have been implemented? Combining fuzzy systems with a DNN enables the development of AI models that are not only accurate in prediction but also inherently interpretable and understandable to humans. Furthermore, the experimental results presented in (Yazdanbakhsh and Dick 2019) show that the DNFS approach is capable of achieving better accuracy than a DNN with the same level of abstraction/depth. Because basic neuro-fuzzy systems have been major research topics for over 27 years, various surveys and systematic review studies can be found in the literature. However, there is a lack of current research on the novel DNFS hybrid technique (Yeganejou and Dick 2018). Hence, researchers have started exploring the potential of this new domain by implementing the model in various applications ranging from the computing domain to the healthcare, manufacturing, and aviation industries. Most breakthroughs regarding the implementation of DNFS in various application subject domains are highlighted in the following subsections. Deep neuro-fuzzy system applications in the subject domains of computing Techniques in the field of AI have made significant contributions to the solution of different real-world problems, including those in the computing domain. Likewise, the novel DNFS has also shown positive potential in solving multiple problems, mainly from this domain. In this subsection, we discuss the maximum number of studies found in records implementing the DNFS in different subject domains of computing, including distributed systems, cloud computing, cybersecurity, Internet marketing, software testing, and the classification of image, speech, text, and video. (i) DNFS application on distributed systems The difficulty of classifying sentiments on Twitter is important for real-world situations such as decision-making and information systems, where customers might obtain relevant information through online reviews. Service ratings can serve as an excellent point for the decision-making process as they provide quick information on the online reviews (Uma 2020). Therefore, an optimization-based fuzzy deep learning classification was proposed in (Uma 2020) and (Bedi and Khurana 2020) for sentiment analysis. The proposed method was developed to solve the misclassification problem in social media reviews. Similarly, taking the advantages of deep learning and fuzzy inference, the authors in (Nguyen et al. 2018) proposed a hybrid fuzzy convolutional neural network (FCNN) with the integration of fuzzy logic and a CNN model for text sentiment classification of Twitter sentiment and movie reviews. The proposed model can resolve ambiguities in data with linguistic labels that are important for emotion detection for sentiment analysis tasks. A comparison of the results between the proposed FCNN and a conventional CNN showed that the proposed FCNN achieves better classification accuracies on emotional data. In another study (Zhou et al. 2014), the authors embedded prior knowledge into the learning structure, making a two-step semi-supervised learning method called fuzzy deep belief networks (FDBN) for sentiment classification. In addition, DNFS has been implemented for congestion control in wireless sensor networks (WSNs) (Monisha and Ranganayaki 2018) and data streaming processing (Mahardhika Pratama et al. 2018). (ii) DNFS application on cloud computing A few years ago, personal computers were not capable of tackling heavy workloads with vast amounts of data for processing. Hence, processing massive data was (and in some regards still is) a challenging problem until cloud computing was introduced to offer services as a solution to the massive data storage. While attempting to reduce the service rates, the most important aspect was the capability to schedule an upscale of cloud system resources for future or on-demand use. To ensure that cloud services are affordable to the customers, the authors in (Chen et al. 2018a, b) proposed a fuzzy deep neural network (FDNN) model to predict the demand for cloud computing resources. This model can assist customers in deciding the number of resources to be reserved for their computing needs, thereby reducing their operational costs. In another study, the same authors developed a hierarchical Pythagorean fuzzy deep neural network (HPFDNN) model by incorporating the properties of Pythagorean fuzzy logic and DNN (Chen et al. 2019). (iii) DNFS application on cybersecurity Modern malware is an alarming threat to both individual and organizational security. Over the last few decades, several malware families have been codified and differentiated based on their behavior and functionality. Most machine learning methods work well with the general benign-malicious classification, but are unable to distinguish new malware among many classes (Shalaginov and Franke 2017). Therefore, the authors Shalaginov and Franke (2017) proposed a novel deep neuro-fuzzy architecture for multi-label malware classification and fuzzy rule extraction. In addition, smart grids (SGs) are critical and intelligent systems. However, to ensure the security, network requires cybersecurity method with advanced mechanisms of intrusion detection and prevention systems (IDPSs). Therefore, to provide the best possible security for SGs, a smart collaborative IDPS was introduced in (Patel et al. 2017) with a fully distributed management system to support and prevent the network from attacks. In (Amosov et al. 2019), a hybrid model of fuzzy logic with convolutional layers was introduced to detect the denial-of-service (DoS) attacks in a highly loaded corporate network by recognizing the abnormal network traffic. (iv) DNFS application on Internet marketing Apart from the sentiment analysis, Internet marketing and online advertising are seen as successful approaches for promotional engagement because of their solid and personalized communication capabilities. As a result, many researchers have shown interest in Internet advertisements, which have become an important source of revenue for online businesses. The click-through rate (CTR) is an effective factor for determining the effect of targeted advertising. Therefore, an FDNN was proposed in (Jiang et al. 2018) to predict the advertising CTR. In addition, fuzzy clustering and deep learning were combined in (Yin et al. 2020) to forecast the sales of the new products. (v) DNFS application in software testing In the field of software engineering, human-based software testing consumes a lot of time and resources. However, software testing is an essential part of the process to validate the performance of a product under different circumstances before being released for consumer usage. Therefore, to save cost and time during the testing process, various software testing tools related to Oracle have been reported in the literature. Among such studies, in (Monsefi et al. 2019), the authors implemented a novel deep neuro-fuzzy approach for software testing on Oracle. The proposed approach was validated using four different applications and produced better accuracy in detecting the errors with correct data. By contrast, in the study (Liu et al. 2019), a deep-learning and fuzzy oversampling-based model called DeepBalance was used for software vulnerability detection. (vi) DNFS application on image, speech, and text classification Earlier, deep learning was successfully applied in various tasks such as data, text, and image classification. Likewise, the majority of the articles in the literature on DNFS applications can be seen handling these challenges by combining fuzzy systems and a DNN. The author Korshunova (2018) proposed a convolutional fuzzy neural network approach with the help of convolutional, pooling, fully connected, and fuzzy self-organization layers for the classification of real-world objects and image scenes. The approach combines the advantages of a CNN and fuzzy logic to tackle ambiguity in the interpretation of the input sequence. Inspired by the current advancements of CNNs, the study of Greeshma and Bindu (2017) developed a fuzzy deep learning algorithm for single-image super-resolution. This novel approach uses a fuzzy rule layer along with a deep network to recreate a high-resolution image. In addition, the authors of (Deng et al. 2017) presented an FDNN for classification tasks, such as natural scene image categorization and stock trend prediction. Similar studies for image classification can be found in (Guan et al. 2020; Kunchala et al. 2020; Liu et al. 2020a, b; Liu et al. 2020a, b; Manchanda et al. 2020; Tianyu and Xu 2020; Yeganejou and Dick 2018, 2019; Yeganejou et al. 2020; Zhang et al. 2020a, b, c). In addition to image classification, DNFS has been successfully implemented to perform speech classification. A speech enhancement framework using a fuzzy deep belief network (FDBN) was reported in (Samui et al. 2019). In this model, the network is implemented to perform pre-training with the help of multiple FDBNs to enhance the stability and speed of feature learning. Meanwhile, Xu and Xiao (2018) suggested a theoretical method using a combination of fuzzy optimization with deep learning to cope with the fuzziness of emotions. Since data are growing at an exponential rate, it is essential to summarize a text document in order to understand the key elements of the document. Many studies have been conducted on summarization methods, and most of them are extractive summarizers. Henceforth, the research work by Chopade and Narvekar (2017) presented a hybrid approach of DNN and fuzzy logic systems. In this study, a restricted Boltzmann machine (RBM) is proposed with a DNN and fuzzy rule based on the phrases to extract the features using a sentence matrix. (vii) DNFS application on video classification and robotics The study of Nguyen et al. (2019) presented a novel convolutional neuro-fuzzy network, which incorporated a CNN into the fuzzy logic domain to derive high-level features of emotions from the text, audio, and image data. Alternatively, the authors in (Cunha Sergio and Lee 2020) proposed a novel hybrid DNN with ANFIS to interpret the emotions of a video from its visual features and a deep long short-term memory recurrent neural network to produce the related audio signals with an equal emotional impression. Likewise, in another study (Savchenko et al. 2018), the authors implemented a fuzzy analysis with a CNN in still-to-video recognition. Moreover, the model has also been implemented in human action recognition and robotics (Bendre et al. 2020; Chen et al. 2020a, b; Liao et al. 2020; Mohmed et al. 2020; Wu et al. 2020). Figure 20 shows the intensity of publications for various DNFS applications in the computing domain. Based on Fig. 20, most DNFS applications are focused on image, speech, and text classifications, followed by video classification and robotics, distributed systems, cybersecurity, cloud computing software testing, and Internet marketing.Fig. 20 Intensity of DNFS-related publications for application subjects in the computing domain Deep neuro-fuzzy system applications in the subject domains of healthcare Similar to the computing domain, DNFS has been successfully implemented in the healthcare sector to perform robotic surgeries and predict various diseases. A deep neuro-fuzzy approach was implemented in (Aviles et al. 2016) to estimate the interaction forces while performing a Robotic surgery. A fuzzy hybridized FCNN model was used in (Ramasamy and Hameed 2019) to classify healthcare data. In the study (Davoodi and Moradi 2018), a modern fuzzy deep model was proposed for intensive care units (ICUs) for mortality prediction. For this, a deep framework was developed based on the layered structure of the fuzzy rule base, which can address big data problems. In (Park et al. 2016), the researchers proposed fuzzy deep learning (FDL), which is a specific estimation method for intra-and inter-fractional variations in many patients. The proposed FDL was built by breathing clustering, a prediction of precise movements, and decreasing the computational cost. Sharma et al. (2020) employed a DNFS as a decision-making system to predict the risk and severity of diseases. The study presented a hybrid diagnosis strategy (HDS) using fuzzy inference and a DNN to detect COVID-19 patients. Moreover, the novel hybrid approach to DNFS has been implemented for tumor and cancer detection and segmentation (Banerjee et al. 2020; Lima et al. 2020; Mudiyanselage et al. 2020; Özyurt et al. 2019; Pitchai et al. 2020; Rahouma et al. 2019; Sengan et al. 2020; Shen et al. 2020; Yang et al. 2020; Zhang et al. 2020a, b, c). Deep neuro-fuzzy system applications in the subject domains of finance and economics After the field of financial engineering expanded over the last few years from financial signal analysis to financial prediction methods, this field has become the most important topic among the academic communities and the financial world. Several hybrid intelligent financial prediction systems incorporating neural networks, fuzzy logic, and genetic algorithms have been proposed over the past 20 years. Likewise, to make a worldwide financial prediction, Lee (2020) introduced a chaotic type-2 transient-fuzzy deep neuro-oscillatory network (CT2TFDNN) with retrograde signaling. Other studies (Chandrasekar 2020; Chen et al. 2020a, b; Wang 2020; Xiao 2020) have implemented the DNFS method in Bitcoin price prediction, stock index prediction, and e-commerce platforms. Deep neuro-fuzzy system applications in the subject domains of traffic flow and incident prediction In the era of AI today, the use of intelligent software for travel assistance is growing rapidly. An intelligent transportation system (ITS) is an advanced transport management system that incorporates electronic information, AI, global positioning system (GPS) tracking, communications engineering, and other techniques. Generally, traffic flow data has the limitation of complexity and noise interaction. However, compared with the previous deterministic explanation, fuzzy theory is capable of generalizing the original data more logically (An et al. 2019). Hence, a fuzzy-based convolutional neural network (F-CNN) approach was implemented in (An et al. 2019) for predicting traffic flow. This approach uses a fuzzy inference system (FIS) to produce uncertain knowledge about traffic incidents. In addition, a CNN training algorithm is used to learn the characteristics of internal traffic data, traffic accident information, and external information, thus forming an F-CNN prediction model to predict the traffic flow. A few similar studies have employed DNFS for traffic flow prediction and incident detection in (Chen et al. 2018a, b; El Hatri and Boumhidi 2018; Sumit and Akhter 2019; Usman et al. 2020). At the same time, the authors in (Chai et al. 2020; Ivanov et al. 2019) proposed the same model for monitoring unmanned surface vehicles (USVs) and hypersonic vehicles. Deep neuro-fuzzy system applications in the subject domain of the manufacturing industry In the welding industry, methods and techniques must consider trends of robotic usage and a large multi-structural architecture to meet the criteria of current development projects in the market. In addition, the technologies in modern manufacturing have led to new developments in welding techniques. Drawing inspiration from the AI technique, the study in (Kesse et al. 2020) proposed the implementation of an AI-based tungsten inert gas (TIG) algorithm for welding to identify the control parameters and predict the optimal welding bead width using fuzzy deep learning. Similarly, for industrial accidents, it is important to prevent and control industrial accidents with an early warning. The existing approaches are time-consuming, unreliable, and incompetent in coping with uncertainty. Therefore, an FDNN was implemented in (Gobinath and Madheswaran 2020; Lin et al. 2020; Yun et al. 2020; Zhang et al. 2020a, b, c; Zheng et al. 2017) to diagnose the faults in machines, provide a forecast, and alert managers for possible industrial accidents in advance. In addition, the study of (Remya and Sasikala 2019) used hybridization of the back-propagation in deep learning and fuzzy logic decision tree for rubberized coir fiber classification. Deep neuro-fuzzy system applications in the subject domain of the aviation industry In the aviation sector, passenger profiling plays a vital role in maintaining commercial airline security. However, the conventional methods have become inefficient in handling the rapidly increasing amounts of electronic records. Hence, the researchers in (Zheng et al. 2016) proposed a deep neuro-fuzzy approach with the integration of ordinary Pythagorean-type fuzzy sets and a deep Boltzmann machine (DBM) as a Pythagorean fuzzy deep Boltzmann machine (PFDBM). This study further proposed a hybrid learning algorithm combining a biogeography-based optimization (BBO) metaheuristics algorithm to improve the exploration search and a gradient-based method to enhance the exploitation search. The simulation results performed on the Air China datasets indicate that the proposed solution offers a high classification accuracy with a great learning ability. In addition, many pattern-analysis tasks can be solved using this approach. Deep neuro-fuzzy system applications in the subject domains of energy and load forecasting The incorporation of smart meters in energy management systems has made it easier for electrical companies to access the electricity usage data of their customers. However, extracting and analyzing enormous amounts of data is challenging for these companies. Therefore, researchers have started to utilize various AI techniques to analyze data retrieved from smart meters (Javaid et al. 2019). The efforts have been made for the energy management of the residential buildings in (Javaid et al. 2019). This study focused on an efficient load and cost optimization by proposing the use of DNFS for solving uncertain behaviors of consumers with large amounts of data. The final finding of the study confirms the robustness of the proposed model in terms of cost optimization and energy efficiency. Apart from this, one more study was found in the literature that combines fuzzy and deep learning methods to predict the hourly load of the next 7 days. The proposed technique showed a superior performance compared to the traditional load forecasting schemes (Sideratos et al. 2020). Figure 21 summarizes and visualizes the DNFS research applications in various domains, together with the intensity of publications for each application subject domain. By contrast, Fig. 22 visualizes the distribution of records found in each application domain for the papers included in this systematic literature survey.Fig. 21 The intensity of publications in different application domains of DNFS Fig. 22 Distribution of records found in each application domain Discussion As a comprehensive analysis of the studies found through a designed study mapping process, this systematic literature survey reports a total of 105 relevant studies addressing the research questions. This section has been thoroughly organized into two subsections. The first subsection (Sect. 5.1) highlights the research gaps, issues, and challenges found while answering the four research questions. In addition, a few recommendations are presented to facilitate the researchers in finding potential directions for future work. Meanwhile, the limitations of this systematic literature survey are presented in second subsection (Sect. 5.2). Research gap, challenges, and future recommendation for RQ1, RQ2, RQ3, and RQ4 Research Question (RQ1) Based on the literature published during the past 5 to 6 years, several variations of DNFS have been proposed since its emergence, and the successful implementation of this emerging model is growing rapidly in a variety of application domains. Because adding a fuzzy layer into a DNN is extremely flexible, it is possible to include it anywhere in the network architecture, depending on the desired behavior of the fuzzy layer. Hence, this study also covered three different structural designs that have been developed in the formation of deep neuro-fuzzy-based models, such as sequential structural designs, parallel structural designs, and cooperative structural designs. Since these novel deep neuro-fuzzy systems are a type of deep network with a hybrid of fuzzy rules, membership degrees along with DNN parameters such as the learning rate, number of layers, number of nodes per layer, a huge number of weights, activation functions, and an optimizer. Therefore, despite various choices of structural designs, answering the first research question of this study revealed that most of the examples found during the period of 6 years have used sequential structural designs to develop DNFS as compared to parallel and cooperative structure designs to keep the model simple. However, sequential structures are designed to be linear and are considered as slow models compared with the other two structural designs. This behavior of linearity becomes a challenge in implementing DNFS in the big data paradigm owing to its deep architecture. Whereas a parallel and cooperative structural design could be more successful in solving complex real-world problems involving large-scale data because of their flexibility in learning and interpretability. Hence, in the future, research on the development of DNFS could be further oriented toward efficient hybridization of fuzzy or neuro-fuzzy systems with a DNN to create parallel and cooperative models. Moreover, because these models have been proposed to tackle big data problems, the computational complexity increases when dealing with huge and complex data owing to the deep architecture. In addition to suggested structural changes, few studies from the literature have recommended introducing hardware solutions such as the use of powerful GPUs, an FPGA, and memristors in the future to overcome the problem of computational complexity. In the same context of big data, it is often challenging to model suitable techniques and methods to deal with streaming data continuously generated by different sources at high speeds. In the past, various AI-based decision-making systems have been presented (Almuammar and Fasli 2019; Lobo et al. 2018; Mahardhika Pratama and Wang 2019; Ullah et al. 2019). However, when it comes to DNFS, only a single study (M. Pratama et al. 2020) has been found in developing this model for the continuous learning of non-stationary data streams. In this study, a deep evolving fuzzy neural network (DEVFNN) with an elastic structure is introduced. This approach helps to make dynamic modifications in fuzzy rules and the depth of the network structure. In the future, more research work should be directed toward constructing DNFS models that can analyze streaming data dynamically to generate instant and reliable outcomes. Research Question (RQ2) The second research question covered the methods employed to optimize the parameters of the DNFS. Based on the data presented in Table 6 and Fig. 14, it is obvious that most of the studies have used the exact methods such as a gradient descent (GD) algorithm, which is iterative and prone to being stuck in the local minima. Therefore, when facing large-scale data, deep neuro-fuzzy models often deal with slow convergence and poor outcomes (Das et al. 2020). This problem ultimately affects the accuracy of the model during the classification tasks. As a result, several modern optimization techniques under “metaheuristics” have been introduced and implemented in the literature to efficiently optimize machine learning models. However, very limited efforts have been made in the literature from 2017 to 2020 to solve this problem of the local minima with the help of metaheuristic techniques for DNFS models. Moreover, based on our findings for this research question, the majority of studies have used evolutionary-based metaheuristic optimization approaches, whereas only two studies have adopted the swarm intelligence approach. However, according to the research work presented in (Kurban et al. 2014), swarm-based algorithms are generally more accurate and reliable than evolutionary algorithms. In contrast, an analysis based on the study in (Janga Reddy and Nagesh Kumar 2020) states that evolutionary algorithms outperform swarm-based algorithms in terms of finding a near-optimal solution within a reasonable computational time. In addition, we cannot ignore the concept of the “No Free Lunch theorem”, which states that no single metaheuristic is better than another metaheuristic algorithm for solving all real-world problems. Therefore, it is challenging to generalize a particular metaheuristic optimization algorithm that can be used to solve classification, time-series, computer vision, natural language processing, and other tasks. To date, several attempts have been made to introduce new metaheuristic techniques in the literature. Among these methods, some of the popular algorithms are cuckoo search (CS) (Gandomi et al. 2013), bat algorithm (BA) (Yang and He 2013), grey wolf optimizer (GWO) (Mirjalili et al. 2014), animal migration optimization (AMO) (Li et al. 2014), whale optimization algorithm (WOA) (Mirjalili and Lewis 2016), emperor penguins colony (EPC) (Harifi et al. 2019), mayfly algorithm (MA) (Zervoudakis and Tsafarakis 2020), and equilibrium optimizer (EO) (Faramarzi et al. 2020). With respect to the issues identified in the above statement, research on optimizing DNFS with metaheuristic-based algorithms still needs significant work in the future. Researchers may in the future investigate and explore the newly introduced metaheuristic optimization methods mentioned above to compare and further improve the performance of DNFS. Research Question (RQ3) The intensity of publications in the domain of DNFS was carefully examined and addressed in this research question. Based on our extensive search from the online databases, it can be concluded that the research on the development of DNFS first took place in 2015. According to Fig. 18, the domain of DNFS was not explored much during the first four years (2015–2018) of its development. This might have caused difficulties and challenges for researchers to understand this new concept of integrating deep learning with fuzzy systems. Nevertheless, a major rise in the publication of DNFS can be seen from the years 2019 to 2020. The DNFS model has started gaining attention among research communities while successfully solving various problems in the subject areas of computing, engineering, and industry. Since this domain is still a new area of interest, the majority of the research work has been conducted by implementing the model in different application domains without any major efforts on improving its performance by reducing the computational complexity. Hence, future research should be geared towards exploring efficient ways to improve the training mechanism, hyperparameter tuning (e.g., learning rate and number of hidden layers), fuzzy knowledge base, and structural modifications of the DNFS model. Research Question (RQ4) With this research question, this study tried to cover the majority of the applications in the domain of DNFS. Since the model has been recently introduced, there were minimal studies found in the literature that have used DNFS for different application subjects. Figure 21 of Sect. 4.4 shows that the use of DNFS models in the computing area is the leading trend, followed by healthcare, traffic management systems, and manufacturing industry when compared to other applications. Most studies in the computing domain have used the DNFS model for image, speech, and text classification. Subsequently, video classification and robotics, as well as distributed systems, are trending ahead of other fields such as cybersecurity, cloud computing, software testing, and Internet marketing, as shown in Fig. 20. However, relatively little research using DNFS has been identified in the aviation industry, finance and economics, and energy management. Therefore, while answering this research question, it can be concluded that there is a vast scope related to the future implementation of the model in technologies under the fourth industrial revolution (4IR), such as AI, blockchain, virtual and augmented reality, cybersecurity, biotechnology, the Internet of Things, digital signal processing, robotics, manufacturing industry, and renewable energy. Limitations of this systematic literature survey A critical analysis of the records found in the literature revealed that only four survey studies have been published on DNFS. However, our study is the first initiative in this domain to present a systematic literature survey. The motivation behind conducting this systematic literature survey was to investigate and identify detailed statistics and figures by performing an in-depth analysis of the records obtained from the literature, and to cover all papers related to the area. However, this systematic literature review has a few limitations. For instance, during the screening procedure, only studies published with detailed knowledge and written in English were included in this systematic literature survey. As a result, there may be some short papers or publications that are published in other languages, which may have made a positive contribution in this domain but were not analyzed. Furthermore, to maintain the quality and reliability of this systematic literature survey, we had to exclude research papers that claimed their method was a combination of fuzzy systems and deep learning but had poorly defined methodologies. We are aware that these filters might have affected the final findings of the included studies. Nonetheless, the decision to exclude the aforementioned papers was not taken lightly and was conducted based on the inclusion and exclusion criteria (see Table 4), as well as the eligibility criteria (see Table 5). Thus, the goal of this systematic literature review was to identify and highlight the majority of work published in the DNFS domain. Conclusion This systematic literature survey aims to capture state-of-the-art research in the novel domain of DNFS by following the guidelines of well-written systematic reviews from the literature. As a result, a revised study mapping process comprising seven phases was introduced in this study. Four research questions were designed to lay the foundation of this study and help extract meaningful information from the database to draw a comprehensive picture of the current state of research related to DNFS. A total of 252 studies were retrieved during the first step of the primary search using the selected keywords and search strings. It became obvious that DNFS-based systems are relatively new, with only a few relevant papers found in the literature during the in-depth analysis of the identified publications. However, a total of 105 studies were found during the quality assessment process, which provides an answer to the research questions of this systematic literature survey. Moreover, the well-defined answers to the research questions helped to identify the research gaps, issues, and challenges of this particular domain. In addition, this study addressed possible future directions, including potential structural designs (e.g., parallel and cooperative architectures) to further strengthen the outcomes for solving the classification and prediction-related problems. This study also suggested the implementation of modern optimization methods such as metaheuristic techniques to optimize DNFS in the future. This study also suggests to review the performance of the model by improving the training mechanism, hyperparameter tuning (e.g., learning rate and number of hidden layers), and fuzzy knowledge base. Along with that, this study also discovered and recommended potential application areas where the DNFS has not yet been deployed, such as virtual and augmented reality, business, education, robotics, manufacturing, renewable energy, and engineering. Recommendations were made to address the limitations found in the literature to help both researchers and practitioners interested in this particular domain. Therefore, this comprehensive systematic literature survey aims not only to provide researchers with the maximum information about DNFS in a single paper but also offer a platform for researchers who wish to commence their research and explore the potential of DNFS for future work. In the final analysis and conclusion, this study discussed the limitations of the systematic literature survey that affected the final results of the included studies, such as the fact that few studies were excluded owing to poorly defined methodologies, short papers, and research published in languages other than English. Abbreviations ABC Artificial bee colony AI Artificial intelligence AMD Advanced micro devices AMO Animal migration optimization ANFIS Adaptive neuro-fuzzy inference system ARTMAP Adaptive resonant theory map BA Bat algorithm BB-BC Big bang-big crunch BBO Biogeography-based optimization BBO-GLW Biogeography-based optimization—greedy layer wise BBO-HF Biogeography-based optimization—hessian-free BSO Brain storm optimization CNN Convolutional neural networks CS Cuckoo search CSA Crow search algorithm CT2TFDNN Chaotic type-2 transient-fuzzy deep neuro-oscillatory network CTR Click-through rate CUDA Compute unified device architecture DBM Deep Boltzmann machine DBN Deep belief networks DDAE Deep denoising autoencoder DEVFNN Deep evolving fuzzy neural network DFNN Deep Fuzzy Neural Network DL Deep Learning DNFS Deep neuro-fuzzy systems DNN Deep neural network DoS Denial-of-service EA Evolutionary algorithms EBSE Evidence-based software engineering EHO Elephant herd optimization EO Equilibrium optimizer EPC Emperor penguins colony FA Firefly algorithm FCNN Fuzzy Convolutional Neural Networks FDBN Fuzzy deep belief networks FDL Fuzzy deep learning FDNN Fuzzy deep neural network FPGA Field programmable gate arrays FQL Fuzzy Q-learning GA Genetic algorithms GA-BBBC Genetic algorithm—big bang big crunch GD Gradient descent GPS Global positioning system GWO Grey Wolf optimizer HDS Hybrid diagnosis strategy HF Hessian-free HPFDNN Hierarchical Pythagorean fuzzy deep neural network ICU Intensive Care Units IDPSs Intrusion detection and prevention system IIFDL Intra- And Inter-Fraction Fuzzy Deep Learning ITS Intelligent Transportation System JOA Jaya optimization algorithm LSE Least squares estimator MA Mayfly algorithm MKL Math Kernel library NF Neural function NFS Neuro-Fuzzy Systems NLP Natural language processing PB Population based PFDBM Pythagorean fuzzy deep Boltzmann machine PRISMA Preferred reporting items for systematic reviews and meta-analyses RBFN Radial basis function networks RBM Restricted Boltzmann machine RNN Recurrent neural networks ROCm Radeon Open Compute SGs Smart grids SI Swarm intelligence TIG Tungsten inert gas TSK Takagi–Sugeno–Kang UAV Multiple unmanned aerial vehicle USVs Unmanned surface vehicles WOA Whale optimization algorithm WSN Wireless sensor network Acknowledgements Research reported in this publication was supported by Fundamental Research Grant Project (FRGS) from the Ministry of Education Malaysia (FRGS/1/2018/ICT02/UTP/03/1) under UTP grant number 015MA0-013. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Abraham A (2001) Neuro fuzzy systems: state-of-the-art modeling techniques. In: Paper presented at the connectionist models of neurons, learning processes, and artificial intelligence, Berlin, Heidelberg Almuammar M, Fasli M (2019) Deep learning for non-stationary multivariate time series forecasting. In: 2019 IEEE international conference on big data (Big Data), pp 2097–2106. 10.1109/BigData47090.2019.9006192 Amosov OS, Ivanov YS, Amosova SG (2019) Recognition of abnormal traffic using deep neural networks and fuzzy logic. In: 2019 international multi-conference on industrial engineering and modern technologies (FarEastCon), pp 01–05. 10.1109/FarEastCon.2019.8934327 An J Fu L Hu M Chen W Zhan J A novel fuzzy-based convolutional neural network method to traffic flow prediction with uncertain traffic accident information IEEE Access 2019 7 20708 20722 10.1109/ACCESS.2019.2896913 Angelov PP Gu X Deep rule-based classifier with human-level performance and characteristics Inf Sci 2018 463–464 196 213 10.1016/j.ins.2018.06.048 Ashraf S Aslam Z Saleem S Omer Ali S Aamer M Multi-biometric sustainable approach for human appellative CRPASE 2020 6 146 152 Aviles AI, Alsaleh SM, Montseny E, Sobrevilla P, Casals A (2016) A deep-neuro-fuzzy approach for estimating the interaction forces in robotic surgery. In: 2016 IEEE international conference on fuzzy systems (FUZZ-IEEE), pp 1113–1119. 10.1109/FUZZ-IEEE.2016.7737812 Baashar Y Alhussian H Patel A Alkawsi G Alzahrani AI Alfarraj O Hayder G Customer relationship management systems (CRMS) in the healthcare environment: a systematic literature review Comput Stand Interfaces 2020 71 103442 10.1016/j.csi.2020.103442 34170994 Banerjee S Singh SK Chakraborty A Das A Bag R Melanoma diagnosis using deep learning and fuzzy logic Diagnostics 2020 10 8 577 10.3390/diagnostics10080577 32784837 Bedi P Khurana P Singh P Panigrahi B Suryadevara N Sharma S Singh A Sentiment analysis using fuzzy-deep learning Proceedings of ICETIT 2019. Lecture notes in electrical engineering 2020 Cham Springer Bendre N Ebadi N Prevost J Najafirad P Human action performance using deep neuro-fuzzy recurrent attention model IEEE Access 2020 10.1109/ACCESS.2020.2982364 Bonanno D Nock K Smith L Elmore P Petry F An approach to explainable deep learning using fuzzy inference 2017 Washington SPIE Buhrmester V, Münch D, Arens M (2019) Analysis of explainers of black box deep neural networks for computer vision: a survey. arXiv:1911.12116 Chai R Tsourdos A Savvaris A Xia Y Chai S Real-time reentry trajectory planning of hypersonic vehicles: a two-step strategy incorporating fuzzy multiobjective transcription and deep neural network IEEE Trans Ind Electron 2020 67 8 6904 6915 10.1109/TIE.2019.2939934 Chandrasekar R Fuzzy crow search algorithm-based deep LSTM for bitcoin prediction Int J Distrib Syst Technol 2020 11 4 53 71 10.4018/IJDST.2020100104 Chen D, Zhang X, Wang L, Han Z (2018a) Prediction of cloud resources demand based on fuzzy deep neural network. In: 2018a IEEE global communications conference (GLOBECOM), pp 1–5. 10.1109/GLOCOM.2018a.8647765 Chen D Zhang X Wang LL Han Z Prediction of cloud resources demand based on hierarchical pythagorean fuzzy deep neural network IEEE Trans Serv Comput 2019 10.1109/TSC.2019.2906901 Chen L Su W Wu M Pedrycz W Hirota K A Fuzzy deep neural network with sparse autoencoder for emotional intention understanding in human-robot interaction IEEE Trans Fuzzy Syst 2020 28 7 1252 1264 10.1109/TFUZZ.2020.2966167 Chen W An J Renfa L Fu L Xie G Bhuiyan M Li K A novel fuzzy deep-learning approach to traffic flow prediction with uncertain spatial–temporal data features Future Gener Comput Syst 2018 10.1016/j.future.2018.06.021 Chen X, Rajan D, Quek C (2020b) A deep hybrid fuzzy neural hammerstein-wiener network for stock price prediction. In: International conference on artificial intelligence in information and communication (ICAIIC). pp 288–293 Chimatapu R, Hagras H, Starkey A, Owusu G (2018) Interval type-2 fuzzy logic based stacked autoencoder deep neural network for generating explainable AI models in workforce optimization. In: 2018 IEEE international conference on fuzzy systems (FUZZ-IEEE), 2018, pp 1–8. 10.1109/FUZZ-IEEE.2018.8491679 Chopade HA, Narvekar M (2017). Hybrid auto text summarization using deep neural network and fuzzy logic system. In: 2017 international conference on inventive computing and informatics (ICICI), 2017, pp 52–56. 10.1109/ICICI.2017.8365192 Cunha Sergio G Lee M Emotional video to audio transformation using deep recurrent neural networks and a neuro-fuzzy system Math Probl Eng 2020 2020 8478527 10.1155/2020/8478527 Czabanski R Jezewski M Leski J Prokopowicz P Czerniak J Mikołajewski D Apiecionek Ł Ślȩzak D Introduction to fuzzy systems Theory and applications of ordered fuzzy numbers: a tribute to Professor Witold Kosiński 2017 Cham Springer International Publishing 23 43 Dabare R Wong KW Shiratuddin MF Koutsakis P Gedeon T Wong K Lee M Fuzzy deep neural network for classification of overlapped data Neural information processing. ICONIP 2019. Lecture notes in computer science 2019 Cham Springer Das R Sen S Maulik U A Survey on Fuzzy Deep Neural Networks ACM Comput Surv 2020 53 1 25 10.1145/3369798 Davoodi R Moradi MH Mortality prediction in intensive care units (ICUs) using a deep rule-based fuzzy classifier J Biomed Inform 2018 79 48 59 10.1016/j.jbi.2018.02.008 29471111 de Campos Souza PV Fuzzy neural networks and neuro-fuzzy networks: a review the main techniques and applications used in the literature Appl Soft Comput 2020 92 106275 10.1016/j.asoc.2020.106275 Deng Y Ren Z Kong Y Bao F Dai Q A hierarchical fused fuzzy deep neural network for data classification IEEE Trans Fuzzy Syst 2017 25 4 1006 1012 10.1109/TFUZZ.2016.2574915 Dorzhigulov A James AP James AP Deep neuro-fuzzy architectures Deep learning classifiers with memristive networks: theory and applications 2020 Cham Springer International Publishing 195 213 El Hatri C Boumhidi J Fuzzy deep learning based urban traffic incident detection Cogn Syst Res 2018 50 206 213 10.1016/j.cogsys.2017.12.002 Emad Hussen S Shahzad A Zeeshan A Durr M Fuzzy based multi-line power outage control system J Crit Rev 2020 8 2 1421 Faramarzi A Heidarinejad M Stephens B Mirjalili S Equilibrium optimizer: a novel optimization algorithm Knowl-Based Syst 2020 191 105190 10.1016/j.knosys.2019.105190 Gallab M Bouloiz H Alaoui YL Tkiouat M Risk assessment of maintenance activities using fuzzy logic Procedia Comput Sci 2019 148 226 235 10.1016/j.procs.2019.01.065 Gandomi AH Yang X-S Alavi AH Cuckoo search algorithm: a metaheuristic approach to solve structural optimization problems Eng Comput 2013 29 1 17 35 10.1007/s00366-011-0241-y Gobinath S Madheswaran M Deep perceptron neural network with fuzzy PID controller for speed control and stability analysis of BLDC motor Soft Comput 2020 24 13 10161 10180 10.1007/s00500-019-04532-z Greeshma MS, Bindu VR (2017) Single image super resolution using fuzzy deep convolutional networks. In: 2017 international conference on technological advancements in power and energy (TAP Energy). 10.1109/TAPENERGY.2017.8397224 Gu X Angelov PP Semi-supervised deep rule-based approach for image classification Appl Soft Comput 2018 68 53 68 10.1016/j.asoc.2018.03.032 Gu X Angelov PP Zhang C Atkinson PM A massively parallel deep rule-based ensemble classifier for remote sensing scenes IEEE Geosci Remote Sens Lett 2018 15 3 345 349 10.1109/LGRS.2017.2787421 Guan C Wang S Liew AW Lip image segmentation based on a fuzzy convolutional neural network IEEE Trans Fuzzy Syst 2020 28 7 1242 1251 10.1109/TFUZZ.2019.2957708 Harifi S Khalilian M Mohammadzadeh J Ebrahimnejad S Emperor penguins colony: a new metaheuristic algorithm for optimization Evol Intell 2019 12 2 211 226 10.1007/s12065-019-00212-x Hayashi Y Holzinger A Goebel R Mengel M Müller H Black box nature of deep learning for digital pathology: beyond quantitative to qualitative algorithmic performances Artificial intelligence and machine learning for digital pathology: state-of-the-art and future challenges 2020 Cham Springer International Publishing 95 101 Hordri N Samar A Yuhaniz S Shamsuddin S A systematic literature review on features of deep learning in big data analytics Int J Adv Soft Comput Appl 2017 9 1 32 49 Hussain K, Leman A, Salleh M (2015) Optimization of ANFIS using mine blast algorithm for predicting strength of malaysian small medium enterprises. In: 2015 12th international conference on fuzzy systems and knowledge discovery (FSKD), 2015, pp 118–123. 10.1109/FSKD.2015.7381926 Hussain K Mohd Salleh MN Cheng S Shi Y Metaheuristic research: a comprehensive survey Artif Intell Rev 2019 52 4 2191 2233 10.1007/s10462-017-9605-z Ivanov YS, Zhiganov SV, Ivanova TI (2019) Intelligent deep neuro-fuzzy system recognition of abnormal situations for unmanned surface vehicles. In: 2019 international multi-conference on industrial engineering and modern technologies (FarEastCon), pp 1–6. 10.1109/FarEastCon.2019.8934353 Janga Reddy M Nagesh Kumar D Evolutionary algorithms, swarm intelligence methods, and their applications in water resources engineering: a state-of-the-art review H2Open J 2020 3 1 135 188 10.2166/h2oj.2020.128 Javaid S, Abdullah M, Javaid N, Saba T, Ahmed J, Sattar N (2019) Towards buildings energy management: using seasonal schedules under time of use pricing tariff via deep neuro-fuzzy optimizer. In: 2019 15th international wireless communications & mobile computing conference (IWCMC), pp 1594–1599. 10.1109/IWCMC.2019.8766673 Jhang J-Y Tang K-H Huang C-K Lin C-J Young K-Y FPGA implementation of a functional neuro-fuzzy network for nonlinear system control Electronics 2018 10.3390/electronics7080145 Jiang Z Gao S Li M An improved advertising CTR prediction approach based on the fuzzy deep neural network PLoS ONE 2018 13 5 e0190831 10.1371/journal.pone.0190831 29727443 Uma KK Meenakshisundaram K Optimization based fuzzy deep learning classification for sentiment analysis Int J Sci Technol Res 2020 9 3 7 Kesse M Buah E Handroos H Ayetor G Development of an artificial intelligence powered TIG welding algorithm for the prediction of bead geometry for TIG welding processes using hybrid deep learning Metals 2020 10 451 10.3390/met10040451 Khati H, Mellah R, Talem H (2019). Neuro-fuzzy control of a position-position teleoperation system using FPGA. In: 2019 24th international conference on methods and models in automation and robotics (MMAR), pp 64–69. 10.1109/MMAR.2019.8864681 Kolajo T Daramola O Adebiyi A Big data stream analysis: a systematic literature review J Big Data 2019 6 1 1 30 10.1186/s40537-019-0210-7 Korshunova KP (2018) A convolutional fuzzy neural network for image classification. In: 2018 3rd Russian-Pacific conference on computer technology and applications (RPC), pp 1–4. 10.1109/RPC.2018.8482211 Kruse R Nauck D Kaynak O Zadeh LA Türkşen B Rudas IJ Neuro-fuzzy systems Computational intelligence: soft computing and fuzzy-neuro integration with applications. NATO ASI series (Series F: Computer and Systems Sciences) 1998 Berlin Heidelberg Springer Kunchala A, Kumar DA, Venkatanarayana M (2020) Transfer learning based fuzzy deep neural networks for leaves detection from digital images. In: 2020 international conference for emerging technology (INCET), pp 1–5. 10.1109/INCET49848.2020.9153971 Kurban T Civicioglu P Kurban R Besdok E Comparison of evolutionary and swarm based computational techniques for multilevel color image thresholding Appl Soft Comput 2014 23 128 143 10.1016/j.asoc.2014.05.037 Laleye FAA, Ezin EC, Motamed C (2015) Adaptive decision-level fusion for Fongbe phoneme classification using fuzzy logic and Deep Belief Networks. In: 2015 12th international conference on informatics in control, automation and robotics (ICINCO), pp 15–24 LeCun Y Bengio Y Hinton G Deep learning Nature 2015 521 7553 436 444 10.1038/nature14539 26017442 Lee RST Chaotic type-2 transient-fuzzy deep neuro-oscillatory network (CT2TFDNN) for worldwide financial prediction IEEE Trans Fuzzy Syst 2020 28 4 731 745 10.1109/TFUZZ.2019.2914642 Li X Zhang J Yin M Animal migration optimization: an optimization algorithm inspired by animal migration behavior Neural Comput Appl 2014 24 7 1867 1877 10.1007/s00521-013-1433-8 Liao P Xu M Yang C A fuzzy ensemble method with deep learning for multi-robot system IEEE Access 2020 8 220352 220363 10.1109/ACCESS.2020.3042439 Lima S, Terán L, Portmann E (2020) A proposal for an explainable fuzzy-based deep learning system for skin cancer prediction. In: 2020 seventh international conference on eDemocracy & eGovernment (ICEDEG), pp 29–35. 10.1109/ICEDEG48599.2020.9096799 Lin L Li M Ma L Nazari M Mahdavi S Yunianta A Using fuzzy uncertainty quantization and hybrid RNN-LSTM deep learning model for wind turbine power IEEE Trans Ind Appl 2020 10.1109/TIA.2020.2999436 Liu H Zhu T Shang F Liu Y Lv D Yang S Deep fuzzy graph convolutional networks for PolSAR imagery pixelwise classification IEEE J Sel Top Appl Earth Obs Remote Sens 2020 14 504 514 10.1109/JSTARS.2020.3041534 Liu M Zhou Z Shang P Xu D Fuzzified image enhancement for deep learning in iris recognition IEEE Trans Fuzzy Syst 2020 28 1 92 99 10.1109/TFUZZ.2019.2912576 Liu S Lin G Han Q-L Wen S Zhang J Xiang Y DeepBalance: deep-learning and fuzzy oversampling for vulnerability detection IEEE Trans Fuzzy Syst 2019 28 7 1329 1343 Lobo JL, Del Ser J, Laña I, Bilbao MN. Kasabov N (2018) Drift detection over non-stationary data streams using evolving spiking neural networks. In: Intelligent distributed computing XII. IDC 2018 studies in computational intelligence, Vol 798. Springer, Cham Lundervold AS Lundervold A An overview of deep learning in medical imaging focusing on MRI Z Med Phys 2019 29 2 102 127 10.1016/j.zemedi.2018.11.002 30553609 Manchanda M, Gambhir D, Singh SK (2020). An improved multifocus image fusion algorithm using deep learning and adaptive fuzzy filter. In: 2020 international conference on contemporary computing and applications (IC3A), pp. 70–75. 10.1109/IC3A48958.2020.233272 Marlen A, Dorzhigulov A (2018) Fuzzy membership function implementation with memristor. arXiv:1805.06698 Mata-Carballeira Ó Gutiérrez-Zaballa J del Campo I Martínez V An FPGA-based neuro-fuzzy sensor for personalized driving assistance Sensors 2019 19 18 4011 10.3390/s19184011 31533318 Mirjalili S Lewis A The whale optimization algorithm Adv Eng Softw 2016 95 51 67 10.1016/j.advengsoft.2016.01.008 Mirjalili S Mirjalili SM Lewis A Grey Wolf optimizer Adv Eng Softw 2014 69 46 61 10.1016/j.advengsoft.2013.12.007 Moher D Liberati A Tetzlaff J Altman DG Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement Anna Intern Med 2009 6 7 e1000097 Mohmed G, Lotfi A, Pourabdollah A (2020) Convolutional neural network classifier with fuzzy feature representation for human activity modelling. In: 2020 IEEE international conference on fuzzy systems (FUZZ-IEEE), pp 1–7. 10.1109/FUZZ48607.2020.9177851 Monisha V, Ranganayaki T (2018). Congestion avoidance aware using modified weighted fairness guaranteed DRED-FDNNPID congestion control for MWSN. In: 2018 tenth international conference on advanced computing (ICoAC), pp 133–137. 10.1109/ICoAC44903.2018.8939080 Monsefi AK Zakeri B Samsam S Khashehchi M Grandinetti L Mirtaheri S Shahbazian R Performing software test oracle based on deep neural network with fuzzy inference system High-performance computing and big data analysis. TopHPC 2019. Communications in computer and information science 2019 Cham Springer Mudiyanselage TKB Xiao X Zhang Y Pan Y Deep fuzzy neural networks for biomarker selection for accurate cancer detection IEEE Trans Fuzzy Syst 2020 28 12 3219 3228 10.1109/TFUZZ.2019.2958295 Muhammed MT Obidallah WJ Bijan R Applying deep learning techniques for big data analytics: a systematic literature review Arch Inf Sci Tech 2018 1 1 20 41 10.36959/863/756 Nguyen T-L Kavuri S Lee M A fuzzy convolutional neural network for text sentiment analysis J Intell Fuzzy Syst 2018 35 6 6025 6034 10.3233/JIFS-169843 Nguyen T-L Kavuri S Lee M A multimodal convolutional neuro-fuzzy network for emotion understanding of movie clips Neural Netw 2019 118 208 219 10.1016/j.neunet.2019.06.010 31299625 Özyurt F Sert E Avci E Dogantekin E Brain tumor detection based on Convolutional Neural Network with neutrosophic expert maximum fuzzy sure entropy Measurement 2019 147 106830 10.1016/j.measurement.2019.07.058 Park S Lee SJ Weiss E Motai Y Intra- and inter-fractional variation prediction of lung tumors using fuzzy deep learning IEEE J Transl Eng Health Med 2016 4 1 12 10.1109/JTEHM.2016.2516005 Patel A Alhussian H Pedersen JM Bounabat B Júnior JC Katsikas S A nifty collaborative intrusion detection and prevention architecture for Smart Grid ecosystems Comput Secur 2017 64 92 109 10.1016/j.cose.2016.07.002 Paul S, Singh L (2015) A review on advances in deep learning. In: Paper presented at the 2015 IEEE workshop on computational intelligence: theories, applications and future directions (WCI), pp 1–6. 10.1109/WCI.2015.7495514 Phuong NH Kreinovich V Fuzzy logic and its applications in medicine Int J Med Inform 2001 62 2 165 173 10.1016/S1386-5056(01)00160-5 11470619 Pitchai R Supraja P Victoria AH Madhavi M Brain tumor segmentation using deep learning and fuzzy k-means clustering for magnetic resonance images Neural Process Lett 2020 53 4 2519 2532 10.1007/s11063-020-10326-4 Pratama M Pedrycz W Webb G An incremental construction of deep neuro fuzzy system for continual learning of non-stationary data streams IEEE Trans Fuzzy Syst 2018 28 7 1315 1328 10.1109/TFUZZ.2019.2939993 Pratama M Wang D Deep stacked stochastic configuration networks for lifelong learning of non-stationary data streams Inf Sci 2019 495 150 174 10.1016/j.ins.2019.04.055 Rahouma KH Aly RHM Hamed HFA Brain cancer diagnosis and prediction based on neural gas network and adaptive neuro fuzzy Procedia Comput Sci 2019 163 518 526 10.1016/j.procs.2019.12.134 Ramasamy B Hameed AZ Classification of healthcare data using hybridised fuzzy and convolutional neural network Healthc Technol Lett 2019 6 3 59 63 10.1049/htl.2018.5046 31341629 Ravi C Abraham A Cherukuri A Melin P Gandhi N Image classification using deep learning and fuzzy systems Intelligent systems design and applications. ISDA 2018. 2018 Advances in intelligent systems and computing 2020 Cham Springer Remya S Sasikala R Classification of rubberized coir fibres using deep learning-based neural fuzzy decision tree approach Soft Comput 2019 23 18 8471 8485 10.1007/s00500-019-03961-0 Safdar S Zafar S Zafar N Khan NF Machine learning based decision support systems (DSS) for heart disease diagnosis: a review Artif Intell Rev 2018 50 4 597 623 10.1007/s10462-017-9552-8 Salleh M Hussain K A review of training methods of ANFIS for applications in business and economics Int J u- and e- Serv Sci Technol 2016 9 165 172 10.14257/ijunesst.2016.9.7.17 Salleh M Talpur N HussainTalpur K A modified neuro-fuzzy system using metaheuristic approaches for data classification Artif Intell 2018 10.5772/intechopen.75575 Samanta S Pratama M Sundaram S A novel Spatio-Temporal Fuzzy Inference System (SPATFIS) and its stability analysis Inf Sci 2019 505 84 99 10.1016/j.ins.2019.07.056 Samui S Chakrabarti I Ghosh SK Time–frequency masking based supervised speech enhancement framework using fuzzy deep belief network Appl Soft Comput 2019 74 583 602 10.1016/j.asoc.2018.10.031 Sarabakha A Kayacan E Online deep fuzzy learning for control of nonlinear systems using expert knowledge IEEE Trans Fuzzy Syst 2019 28 7 1492 1503 10.1109/TFUZZ.2019.2936787 Savchenko AV Belova NS Savchenko LV Fuzzy analysis and deep convolution neural networks in still-to-video recognition Opt Mem Neural Netw 2018 27 1 23 31 10.3103/S1060992X18010058 Schön E-M Thomaschewski J Escalona MJ Agile requirements engineering: a systematic literature review Comput Stand Interfaces 2017 49 79 91 10.1016/j.csi.2016.08.011 Sengan S Priya V Syed Musthafa A Ravi L Palani S Subramaniyaswamy V A fuzzy based high-resolution multi-view deep CNN for breast cancer diagnosis through SVM classifier on visual analysis J Intell Fuzzy Syst 2020 39 8573 8586 10.3233/JIFS-189174 Shalaginov A, Franke K (2017) A deep neuro-fuzzy method for multi-label malware classification and fuzzy rules extraction. In: 2017 IEEE symposium series on computational intelligence (SSCI), pp. 1–8. 10.1109/SSCI.2017.8280788 Sharma D Singh Aujla G Bajaj R Deep neuro-fuzzy approach for risk and severity prediction using recommendation systems in connected health care Trans Emerg Telecommun Technol 2020 32 e4159 10.1002/ett.4159 Sharma O (2019) Deep challenges associated with deep learning. In: 2019 international conference on machine learning, big data, cloud and parallel computing (COMITCon), pp 72–75. 10.1109/COMITCon.2019.8862453 Shen T Wang J Gou C Wang FY Hierarchical fused model with deep learning and type-2 fuzzy learning for breast cancer diagnosis IEEE Trans Fuzzy Syst 2020 28 12 3204 3218 10.1109/TFUZZ.2020.3013681 Shrestha A Mahmood A Review of deep learning algorithms and architectures IEEE Access 2019 7 53040 53065 10.1109/ACCESS.2019.2912200 Shwartz-Ziv R, Tishby N (2017) Opening the black box of deep neural networks via information. https://arxiv.org/abs/1703.00810 Sideratos G Ikonomopoulos A Hatziargyriou ND A novel fuzzy-based ensemble model for load forecasting using hybrid deep neural networks Electr Power Syst Res 2020 178 106025 10.1016/j.epsr.2019.106025 Singh H Lone YA Deep neuro-fuzzy systems with python 2020 Berkeley Apress Singh S Singh S Systematic review of spell-checkers for highly inflectional languages Artif Intell Rev 2020 53 6 4051 4092 10.1007/s10462-019-09787-4 Siva Raja PM Rani AV Brain tumor classification using a hybrid deep autoencoder with Bayesian fuzzy clustering-based segmentation approach Biocybern Biomed Eng 2020 40 1 440 453 10.1016/j.bbe.2020.01.006 Sumit SH Akhter S C-means clustering and deep-neuro-fuzzy classification for road weight measurement in traffic management system Soft Comput 2019 23 12 4329 4340 10.1007/s00500-018-3086-0 Tianyu Z, Xu J (2020) Hyperspectral remote sensing image segmentation based on the fuzzy deep convolutional neural network. In: 2020 13th international congress on image and signal processing, biomedical engineering and informatics (CISP-BMEI), pp 181–186. 10.1109/CISP-BMEI51763.2020.9263563 Ullah A Muhammad K Haq IU Baik SW Action recognition using optimized deep autoencoder and CNN for surveillance data streams of non-stationary environments Future Gener Comput Syst 2019 96 386 397 10.1016/j.future.2019.01.029 Usman M Carie A Marapelli B Bedru HD Biswas K A Human-in-the-loop probabilistic CNN-fuzzy logic framework for accident prediction in vehicular networks IEEE Sens J 2020 21 14 15496 15503 10.1109/JSEN.2020.3023661 Velliangiri S Pandey HM Fuzzy-Taylor-elephant herd optimization inspired Deep Belief Network for DDoS attack detection and comparison with state-of-the-arts algorithms Future Gener Comput Syst 2020 110 80 90 10.1016/j.future.2020.03.049 Vieira JMN, Morgado DF, Mota A (2004) Neuro-fuzzy systems: a survey. In 5th WSEAS NNA international conference on neural networks and applications, Udine, Italia, pp 87–92 Vlamou E Papadopoulos B Fuzzy logic systems and medical applications AIMS Neurosci 2019 6 4 266 272 10.3934/Neuroscience.2019.4.266 32341982 Voulodimos A Doulamis N Doulamis A Protopapadakis E Deep learning for computer vision: a brief review Comput Intell Neurosci 2018 2018 7068349 10.1155/2018/7068349 29487619 Wang LX Fast Training algorithms for deep convolutional fuzzy systems with application to stock index prediction IEEE Trans Fuzzy Syst 2020 28 7 1301 1314 10.1109/TFUZZ.2019.2930488 Wu M Su W Chen L Pedrycz W Hirota K Two-stage fuzzy fusion based-convolution neural network for dynamic emotion recognition IEEE Trans Affect Comput 2020 10.1109/TAFFC.2020.2966440 Xiao P (2020) Information management of e-commerce platform based on neural networks and fuzzy deep learning models. In: 2020 International conference on smart electronics and communication (ICOSEC), pp 532–535. 10.1109/ICOSEC49089.2020.9215235 Xu J-C Xiao N-F Qiao F Patnaik S Wang J Speech emotion recognition based on deep learning and fuzzy optimization Recent developments in mechatronics and intelligent robotics. ICMIR 2017. Advances in Intelligent Systems and Computing 2018 Cham Springer Yang CH Moi SH Hou MF Chuang LY Lin YD Applications of deep learning and fuzzy systems to detect cancer mortality in next-generation genomic data IEEE Trans Fuzzy Syst 2020 10.1109/TFUZZ.2020.3028909 Yang X-S He X Bat algorithm: literature review and applications Int J Bio-Inspired Comput 2013 5 3 141 149 10.1504/IJBIC.2013.055093 Yazdanbakhsh O, Dick S (2019) A deep neuro-fuzzy network for image classification. arXiv:2001.01686 Yazdanbakhsh O, Dick S (2020). A deep neuro-fuzzy network for image classification. arXiv:2001.01686, abs/2001.01686 Yeganejou M, Dick S (2018) Classification via deep fuzzy c-means clustering. In: 2018 IEEE international conference on fuzzy systems (FUZZ-IEEE), pp 1–6. 10.1109/FUZZ-IEEE.2018.8491461 Yeganejou M, Dick S (2019).Improved deep fuzzy clustering for accurate and interpretable classifiers. In: 2019 IEEE international conference on fuzzy systems (FUZZ-IEEE), pp 1–7. 10.1109/FUZZ-IEEE.2019.8858809 Yeganejou M Dick S Miller J Interpretable deep convolutional fuzzy classifier IEEE Trans Fuzzy Syst 2020 28 7 1407 1419 10.1109/TFUZZ.2019.2946520 Yin P Dou G Lin X Liu L A hybrid method for forecasting new product sales based on fuzzy clustering and deep learning Kybernetes 2020 49 12 3099 3118 10.1108/K-10-2019-0688 Yu D Pan T Tracing the main path of interdisciplinary research considering citation preference: a case from blockchain domain J Informet 2021 15 2 101136 10.1016/j.joi.2021.101136 Yu D Sheng L Knowledge diffusion paths of blockchain domain: the main path analysis Scientometrics 2020 125 1 471 497 10.1007/s11192-020-03650-y Yun SH Koo YD Na MG Collapse moment estimation for wall-thinned pipe bends and elbows using deep fuzzy neural networks Nucl Eng Technol 2020 52 11 2678 2685 10.1016/j.net.2020.05.006 Zervoudakis K Tsafarakis S A mayfly optimization algorithm Comput Ind Eng 2020 145 106559 10.1016/j.cie.2020.106559 Zhang L Zhu Y Shi X Li X A situation assessment method with an improved fuzzy deep neural network for multiple UAVs Information 2020 11 4 194 10.3390/info11040194 Zhang S Sun Z Wang M Long J Bai Y Li C Deep fuzzy echo state networks for machinery fault diagnosis IEEE Trans Fuzzy Syst 2020 28 7 1205 1218 10.1109/TFUZZ.2019.2914617 Zhang Y, Wu J, Jiang B, Ji D, Chen Y, Wu EX, Tang X (2020c) Deep learning and unsupervised fuzzy c-means based level-set segmentation for liver tumor. In: IEEE 17th international symposium on biomedical imaging (ISBI), pp 1193–1196 Zheng Y-J Chen S-Y Xue Y Xue J-Y A pythagorean-type fuzzy deep denoising autoencoder for industrial accident early warning IEEE Trans Fuzzy Syst 2017 25 6 1561 1575 10.1109/TFUZZ.2017.2738605 Zheng Y-J Sheng W-G Sun X-M Chen S-Y Airline passenger profiling based on fuzzy deep machine learning IEEE Trans Neural Netw Learn Syst 2016 28 12 2911 2923 10.1109/TNNLS.2016.2609437 28114082 Zhou S Chen Q Wang X Fuzzy deep belief networks for semi-supervised sentiment classification Neurocomputing 2014 131 312 322 10.1016/j.neucom.2013.10.011
PMC009xxxxxx/PMC9005353.txt	==== Front J Affect Disord J Affect Disord Journal of Affective Disorders 0165-0327 1573-2517 Elsevier/North-Holland Biomedical Press S0165-0327(22)00374-3 10.1016/j.jad.2022.04.036 Article Social support coping styles and psychological distress during the COVID-19 pandemic: The moderating role of sex McLean Caitlin L. ab⁎ Chu Gage M. ab Karnaze Melissa M. cd Bloss Cinnamon S. cd Lang Ariel J. ebc a VA San Diego Healthcare System, United States of America b University of California San Diego (UCSD), United States of America c Herbert Wertheim School of Public Health and Human Longevity Science, UCSD, United States of America d Center for Empathy and Technology, T. Denny Sanford Institute for Empathy and Compassion, UCSD, United States of America e VA San Diego Center of Excellence for Stress and Mental Health (CESAMH), United States of America ⁎ Corresponding author at: 3350 La Jolla Village Dr., San Diego, CA 92161, United States of America. 13 4 2022 1 7 2022 13 4 2022 308 106110 22 9 2021 11 3 2022 9 4 2022 Published by Elsevier B.V. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background The coronavirus disease 2019 (COVID-19) pandemic led to the onset and exacerbation of mental health problems, such as stress, anxiety, and depression; yet stay-at-home-orders affected individuals' ability to make use of social support as a coping skill in managing distress. We aimed to evaluate how social support (emotional and instrumental) and biological sex were associated with stress, anxiety, and depression early in the COVID-19 pandemic. Methods Participants (n = 7256) had an average age of 50.13 years (SD = 16.75) and 51.6% were male. Using a cross-sequential design, seven cohorts of individuals completed baseline (T1) and one-month follow-up (T2) questionnaires online from March to July of 2020. We used a series of hierarchical regressions to identify types of social support (Brief-COPE, T1) associated with stress (Perceived Stress Scale-10, T1 and T2), anxiety and depression (Patient Health Questionnaire-4, T2). Results Greater emotional support was associated with less perceived stress, anxiety and depression (all ps < 0.001), whereas greater instrumental support predicted increased distress (all ps < 0.036) on all four outcomes. Moderation analyses revealed that greater emotional social support was associated with lower perceived stress at T1 for both women and men, with a stronger association for women relative to men. For women, greater emotional social support predicted lower anxiety. Limitations Self-selection may have introduced bias and participant self-report on brief measures may not have fully captured coping and distress. Conclusions Interventions enhancing emotional social support strategies, which appear especially important for women, might help manage enduring stressors such as the COVID-19 pandemic. Keywords COVID-19 Social support Coping Stress Sex Gender ==== Body pmc1 Introduction The emotionally charged nature of the coronavirus disease 2019 (COVID-19) pandemic has led to the onset and exacerbation of mental health problems (Torales et al., 2020). Social support is a particularly effective coping strategy during emotionally charged events (e.g., buffering against psychological distress following 9/11 and improving psychological adjustment to chronic health threats; see Taylor, 2011 for a review). Broadly, social support refers to the provision of resources from meaningful groups of people around an individual, such as family members, friends, neighbors, and colleagues (Cohen and Syme, 1985). The literature further divides social support into instrumental support, which involves attempts to address the source of distress through tangible assistance such as providing information, services, or goods; and emotional support, which entails attempts to manage the stressor through emotional comfort provided through warmth, validation, or positive regard (Taylor, 2011). This distinction proves useful, as support-seeking styles appear to have different outcomes and interact uniquely with each situation. For example, emotional support was associated with fewer depression symptoms and higher quality of life in women with ovarian cancer, whereas instrumental support was not (Hill, 2016). The present study examines whether a similar pattern is present in the context of a different health threat, the COVID-19 pandemic. Environmental stressors and available resources also inform the way individuals respond to transitory conditions. Variables such as sex,1 age, race and ethnicity have been associated with increased distress during the pandemic (Breslau et al., 2021; Czeisler et al., 2020; McGinty et al., 2020). Sex, for example, is a particularly important determinant of health; women reported poorer health-related quality of life than men prior to the pandemic (Centers for Disease Control and Prevention, 2013). Women have borne a disproportionate stress burden in the context of the pandemic: in comparison to men, more women have lost their jobs, and those who remain employed are more often employed in positions that require increased exposure to infection (Carli, 2020). Women also have experienced an increase in childcare and other responsibilities as schools closed (Carli, 2020). Further, early data suggest that “lockdown” policies may have left women vulnerable to abuse in the wake of social and physical isolation, with increases in domestic violence following stay-at-home orders (Boserup et al., 2020). “Social distancing” policies may have led to increased social isolation, increasing the need for social support interventions as we continue to navigate COVID-19-related restrictions. Due to the uncertainty of the pandemic and increased social isolation, our first aim was to examine the role of the use of social support in psychological distress (perceived stress, anxiety, and depression). We hypothesized that increased use of emotional support, but not instrumental support, would be predictive of better psychological outcomes. Because the pandemic exacerbated structural and institutional limitations on the coping resources available to women, our secondary aim was to examine sex as a moderator in the relationship between use of social support and psychological outcomes. We hypothesized the effect of social support would vary by sex, with women benefitting more from emotional support compared to men. 2 Method 2.1 Participants Subjects were 7680 adults in the U.S. recruited via Amazon Mechanical Turk (MTurk) and Qualtrics Online Panels, online crowdsourcing platforms for social science survey completion. Eligibility criteria for the 15–20 minute research study titled “Thoughts and Feelings about COVID-19” were: being 18 years of age or older, fluency in English, and residing in the U.S. In addition, the MTurk questionnaire was only available to those who had a 95% or higher approval rate on MTurk. Cases were removed for concerns about data quality (4 for age > 91 years, 313 did not pass the Qualtrics quality check). Demographics for the final sample (n = 7256) are presented in Table 1 .Table 1 Participant characteristics. Table 1 Time 1 (n = 7256) Time 2 (n = 3461) Test of difference Measure n (%) n (%) X2 Age in years, M ± SD, t 50.13 ± 16.75 49.63 ± 15.98 2.45⁎⁎ Sex 16.43⁎⁎⁎ Male 3745 (51.6) 1705 (49.3) Female 3506 (48.3) 1755 (50.7) Other 5 (0.1) 1 (0.0) Ethnicity 0.37 Latinx/Hispanic 655 (9.0) 305 (8.8) Race 10.75 White/Caucasian 5924 (81.6) 2800 (80.9) Black/African American 610 (8.4) 302 (8.7) Asian 405 (5.6) 208 (6.0) Native Hawaiian, Pacific Islander 10 (0.1) 7 (0.2) American Indian, Alaska Native 57 (0.8) 22 (0.6) Mixed race/other 250 (3.4) 122 (3.5) Sexual orientation 6.25 Heterosexual 6578 (90.7) 3154 (91.1) Gay/lesbian/bisexual 554 (7.6) 250 (7.2) Asexual 47 (0.6) 26 (0.8) Other 12 (0.2) 5 (0.1) Marital status 12.34⁎ Single 2031 (28.0) 1009 (29.2) Married/domestic partnership 4276 (58.9) 2017 (58.3) Divorced/separated 697 (9.6) 342 (9.9) Widowed 239 (3.3) 93 (2.7) Annual household income 15.03 $0–$25,000 776 (10.7) 375 (10.8) $25,000–$50,000 1570 (21.6) 775 (22.4) $50,000–$100,000 2666 (36.7) 1280 (37.0) $100,000–$150,000 1198 (16.5) 581 (16.8) $150,000+ 789 (10.9) 359 (10.4) Education 10.60 No high school degree 54 (0.7) 19 (0.5) High school degree 1630 (22.5) 788 (22.8) 2-year degree 1063 (14.3) 536 (15.5) 4-year degree 2970 (40.9) 1400 (40.5) Master's degree 1228 (16.9) 570 (16.5) Professional degree or PhD 298 (4.1) 148 (4.3) Emotional coping, M ± SD 4.70 ± 1.82 – Instrumental coping, M ± SD 4.20 ± 1.72 – Perceived stress, M ± SD, t 15.01 ± 8.09 14.28 ± 8.23 4.51⁎⁎⁎ Anxiety, M ± SD 0.97 ± 1.45 Depression, M ± SD 0.84 ± 1.38 Note. Data are given as a number (valid percentage), except where indicated otherwise. ⁎ p < .05. ⁎⁎ p < .01. ⁎⁎⁎ p < .001. 2.2 Procedure This study was approved by the University of California San Diego Institutional Review Board (Protocol #20042949). Using a cross-sequential design, seven cohorts of individuals completed a baseline (T1) questionnaire between March and June 2020. Those who entered their email address at the end of the questionnaire were sent, one month later, an email invitation to complete the follow-up (T2) questionnaire with a three-day window for completion. From the original sample, 5684 completed all T1 questions and 3461 participants completed T2. There was 47.7% retention of the original sample and 60.9% of those who completed T1, which is comparable to other studies conducted during the early pandemic (e.g., 49.1% for Hoffart et al., 2022; 56.9% for Matthes et al., 2021). These data were collected as part of a larger battery that included measures related to COVID-19 impacts, public health attitudes, psychosocial functioning, and health outcomes (Karnaze et al., 2022). 2.3 Measures Demographic questions included age, sex, race, ethnicity, sexual orientation, marital status, educational status, and annual income. Sex was dichotomously coded into male (0) and female (1), as only 5 (0.1%) participants indicated anything other than male or female. To examine any associations between race-ethnicity and outcomes, five categories were created: Latinx/Hispanic, White-non-Latinx, Black-non-Latinx, Asian-non-Latinx-non-Black, and Other-non-Latinx (White-non-Latinx was the reference group) following the analytic guidelines of the National Center for Health Statistics (National Center for Health Statistics, 2018). Use of emotional and instrumental social support were measured using items from the abbreviated COPE Inventory (Brief-COPE; Carver, 1997). The Brief-COPE measures efforts to minimize distress related to stressful life experiences across 14 two-item subscales. The two of focus in the present study are the use of emotional social support (emotional coping subscale) and instrumental social support (instrumental coping subscale). Items were rated on a four-point Likert-scale and summed with higher scores indicating more effective coping (range = 2–8). The two subscales have adequate psychmetric properties (Carver, 1997). Coping was measured at T1 with Cronbach's alphas (α) of 0.88 for use of emotional support and 0.85 for instrumental support. Perceived stress was measured using the Perceived Stress Scale-10 (PSS; Cohen et al., 1983), which measures the degree to which life circumstances are appraised as stressful during the past month. The measure consists of 10 items that are scored on a five-point Likert-scale and summed for a total score (range = 0–40). Higher scores indicate more perceived stress. The PSS has good psychometric properties (Cohen et al., 1983). PSS was measured at both T1 (α = 0.90) and T2 (α = 0.91). Anxiety and depression were measured using the Patient Health Questionnaire-4 (Kroenke et al., 2009). The questionnaire consists of a 2-item anxiety scale and a 2-item depression scale that measure core symptoms and signs of anxiety and depression during the past two weeks. Items are scored on a four-point Likert-type scale and summed (range = 0–6), with higher scores indicating more symptoms. The anxiety and depression scales have good psychometric properties (Kroenke et al., 2009). Anxiety (α = 0.89) and depression (α = 0.90) were measured at T2. 2.4 Data analysis Preliminary analyses consisted of testing group differences between participants who completed T2 and those who did not, using t-tests and chi-square tests (see Table 1). Main analyses consisted of hierarchical regressions to identify types of social support (emotional and instrumental; T1) associated with PSS (T1, T2), anxiety (T2) and depression (T2) with sex as a moderator. In these regressions, four blocks of variables were sequentially entered: the first block included covariates (age, Latinx, Black-non-Latinx, Asian-non-Latinx-non-Black, Other-non-Latinx, and cohort administration; cohort was controlled for due to the rapid changes and adaptation during the pandemic found by Daly and Robinson, 2021); the second block included sex; the third block included use of emotional and instrumental support; the fourth block included the interactions between sex and each factor of social support. All continuous predictor variables were mean-centered to test for interaction effects. Multicollinearity analysis revealed that neither sex nor the social support variables had correlations greater than 0.7 and that no tolerance values were below 0.2 indicating no collinearity (Hair et al., 2014). Statistical analyses were conducted using Statistical Package for the Social Sciences (SPSS) 27.0 software. 3 Results Table 2 shows results of four hierarchical logistic regressions. Across all outcomes, types of social support used explained 1–2% of the variance. Both emotional (all ps < 0.001) and instrumental (all ps < 0.036) social support contributed significantly to all four models. Higher emotional support scores predicted lower scores for all four distress outcomes, whereas increased instrumental support predicted greater distress, while holding all else constant.Table 2 Hierarchical regression models of the moderating effect of sex on the relationships between types of social support and perceived stress, anxiety, and depression. Table 2 Stress T1 (n = 6020) Stress T2 (n = 3358) Anxiety (n = 3345) Depression (n = 3345) β, t β, t β, t β, t First block: covariates Age −0.29, −22.15⁎⁎⁎ −0.26, −14.79⁎⁎⁎ −0.10, −5.76 −0.01, −5.53⁎⁎⁎ Latinxa 0.04, 3.33⁎⁎ 0.06, 3.42⁎⁎ 0.05, 2.59⁎ 0.05, 3.03⁎⁎ Black non-Latinxa −0.02, −1.90 −0.03, −1.83 0.00, 0.21 0.02, 0.85 Asian non-Latinx-Blacka 0.02, 1.23 0.02, 1.33 −0.04, −2.33⁎ −0.02, −1.15 Other non-Latinxa −0.01, −0.80 0.01, 0.30 0.01, 0.44 −0.00, −0.15 Cohort −0.07, −5.29⁎⁎⁎ −0.01, −0.50 0.07, 4.15⁎⁎⁎ 0.06, 3.26⁎⁎ R2 0.101 0.078 0.018 0.016 F 112.26⁎⁎⁎ 47.24⁎⁎⁎ 10.26⁎⁎⁎ 9.13⁎⁎⁎ Second block: sex Sex 0.12, 10.12⁎⁎⁎ 0.12, 7.30⁎⁎⁎ 0.13, 7.87⁎⁎⁎ 0.05, 2.83⁎⁎ Δ R2 since first block 0.015 0.014 0.018 0.002 F 112.49⁎⁎⁎ 48.74⁎⁎⁎ 17.81⁎⁎⁎ 8.99⁎⁎⁎ Third block: social support Emotional coping −0.17, −9.10⁎⁎⁎ −0.21, −8.16⁎⁎⁎ −0.14, −5.37⁎⁎⁎ −0.18, −6.64⁎⁎⁎ Instrumental coping 0.20, 10.52⁎⁎⁎ 0.16, 6.09⁎⁎⁎ 0.15, 5.58⁎⁎⁎ 0.10, 3.66⁎⁎⁎ Δ R2 since second block 0.016 0.018 0.010 0.014 F 101.79⁎⁎⁎ 46.04⁎⁎⁎ 17.80⁎⁎⁎ 12.44⁎⁎⁎ Fourth block: interactions Sexemotional coping −0.06, −2.44⁎ −0.03, −0.94 −0.14, −3.79⁎⁎⁎ −0.06, −1.56 Sexinstrumental coping 0.02, 0.64 0.01, 0.26 0.07, 1.92 0.01, 0.12 Δ R2 since third block 0.001 0.000 0.005 0.001 F 84.23⁎⁎⁎ 37.79⁎⁎⁎ 16.12⁎⁎⁎ 10.65⁎⁎⁎ Note. T1 = time 1, T2 = time 2. a Reference group = non-Hispanic White. ⁎ p < .05. ⁎⁎ p < .01. ⁎⁎⁎ p < .001. Moderation analyses showed two significant interaction effects: for PSS-T1 and anxiety the effects of emotional support differed by sex. Simple slopes indicated that, for women, greater emotional support was associated with decreased PSS-T1 (b = −0.95, t = −7.77, p < .001) and anxiety (b = −0.19, t = −5.95, p < .001). However, for men, more emotional support was associated with decreased PSS-T1 (b = −0.55, t = −4.67, p < .001), but not anxiety (p = .312). 4 Discussion The present research sought to examine emotional and instrumental social support and their associations with psychological outcomes during the COVID-19 pandemic. Using data from a large sample of U.S. adults, we found that both types of social support were significantly associated with perceived stress, anxiety and depression symptoms. Overall, greater use of emotional support was consistently linked to better outcomes, whereas using instrumental support was linked to worse outcomes. For both women and men, emotional support was associated with fewer anxiety symptoms and was associated with less stress at time 1, with a stronger association for women relative to men. This supports prior findings documenting the importance of emotional social support when managing enduring stressors that cannot be quickly and, often, individually resolved (e.g., Compas et al., 2001; Snow-Turek et al., 1996). Instrumental support seeking may increase awareness related to the inability to control situational factors related to the pandemic and in turn increase anxiety symptoms, or result from situational factors that are difficult to control. Similarly, instrumental social support may function as a form of emotional avoidance, subsequently leading to heightened anxiety in women (Panayiotou et al., 2017). For example, those with heightened anxiety surrounding COVID-19 may engage in behaviors to reduce fears through heightened vigilance (e.g., protecting themselves and loved ones from exposure), which then leads to more awareness of possible threat and increased anxiety. Surprisingly, moderation effects were apparent for perceived stress at time 1 but not time 2, perhaps due to attrition of men and individuals with slightly elevated perceived stress scores. These findings generally align with studies on a global scale on coping styles and mental health outcomes during the COVID-19 pandemic. In a large cross-sectional Australian sample, the use of instrumental social support was associated with higher anxiety though neither emotional nor instrumental support were associated with depression or stress (Gurvich et al., 2021). Whereas family support (broadly, a form of social support) was found to be a protective factor against poor mental health in college students in China (Huang et al., 2021). Yet, family support was not parsed into emotional and informational support. A study that sampled 100 people living in lockdown in Saudi Arabia used a four-factor model of coping, which separated items from emotional and instrumental support into problem-focused coping and positive coping factors (Agha, 2021). However, no significant associations were found between these factors and stress, anxiety, and depression. Instead, active avoidance and religious/denial factors were associated with increases in these mental health outcomes. While different coping strategies were evaluated, hindering direct comparisons, this research highlights the need for global studies to better understand if these constructs hold in non–Western, Educated, Industrialized, Rich, and Democratic (WEIRD) contexts. These findings should be interpreted in consideration of study limitations and strengths. Data were collected via an online survey and may not generalize to the U.S. general population, although data collection was through large U.S. national sampling. The most impacted and distressed individuals (e.g., those ill, healthcare workers) may have prioritized immediate concerns as opposed to completing a survey. Illness and increased distress may also account for individuals who did not complete all of time 1 and time 2. While use of both types of social support were significant predictors of all outcomes, they explained minimal model variance (1–2%) after accounting for covariates. Albeit small, this finding generally aligns with the relative importance of various forms of coping on mitigating psychological outcomes, such as when managing ovarian cancer (Hill, 2016) and during the COVID-19 pandemic in a sample of Polish students (Rogowska et al., 2020). This finding may also be explained by the current study sampling the general population, where the pandemic demonstrated a wide range of impact. Additionally, the brevity of measures (which were selected to prioritize a range of measures, length of administration, and funding in the early months of the pandemic) may not have fully captured the breadth of social support, and anxiety and depression symptoms, although all are well-established measures. Moreover, we evaluated social support as a specific type of coping that was likely affected by stay-at-home orders and social distancing restrictions and did not include other potentially relevant coping strategies. Lastly, the moderation analyses must be interpreted in the context of sex and not gender, as the item asked about sex and did not explicitly include non-binary response options. Due to the use of social support being linked to gender socialization (Reevy and Maslach, 2001) and that COVID-19 pandemic has negatively impacted perceived social support and mental health disproportionally in gender minority populations (Moore et al., 2021), future research would benefit from evaluating gender. Despite these limitations, the sample was from a large U.S. convenience sample and conducted using a cohort design where data were collected from March to July 2020, and account for each two-week period during that time, a period in which the nature of the pandemic was rapidly changing. Generally, emotional social support played a protective role for psychological outcomes, while instrumental social support was associated with more intense perceived stress, anxiety and depression symptoms during the COVID-19 pandemic. These findings, albeit small, have important implications for clinical practice and psychoeducation to help manage uncontrollable and enduring periods of distress. As coping skills deficits are amenable to treatment (Folkman and Moskowitz, 2004), future interventions should emphasize enhancing the use of emotional social support, which may include seeking out and strengthening relationships with others who offer empathy, reassurance, and compassion. While not tested, the offering of emotional support to others might foster mutually beneficial relationships. Using emotional support appears especially important for women to help manage distress during a public health threat characterized by uncertainty. Additionally, as informational social support might be ineffective against worsening stress, anxiety, and depression, health professionals might recommend alternatives. As rates of infections decrease and restrictions are lifted, there remains a strong likelihood that negative mental health impacts are likely to remain for select populations. We must consider these findings, which implicate promoting more effective social support, to help inform mental health interventions in our efforts to attenuate deleterious mental health impacts of the COVID-19 pandemic and future public health threats. Funding CLM is supported by the 10.13039/100000738 VA Office of Academic Affiliations Advanced Fellowship in Women's Health. This work was supported by the Center for Empathy and Technology within the T. Denny Sanford Institute for Empathy and Compassion at the University of California San Diego. This work was also supported by NIH/10.13039/100006108 NCATS Colorado CTSA Grant Number UL1 TR002535. Its contents are the authors' sole responsibility and do not necessarily represent the views of the U.S. Government or NIH. CRediT authorship contribution statement Study concept and design: CLM, GMC, AJL. Acquisition, analysis, or interpretation of data: CLM, MMK, CSB, AJL. Drafting of the manuscript: CLM, GMC. Critical revision of the manuscript for important intellectual content: All authors. Conflict of interest None of the authors report any conflicts of interest with this work. Acknowledgements The authors acknowledge Cynthia Triplett, MPH, MA for her assistance with survey data collection and management. 1 We wish to acknowledge the important distinction between sex and gender and lack of precision caused by interchangeability in earlier literature. Because the current study collected data on sex, we use sex here. ==== Refs References Agha S. Mental well-being and association of the four factors coping structure model: a perspective of people living in lockdown during COVID-19 Ethics Med Public Health. 16 2021 100605 10.1016/j.jemep.2020.100605 Boserup B. McKenney M. Elkbuli A. Alarming trends in US domestic violence during the COVID-19 pandemic Am. J. Emerg. Med. 38 2020 2753 2755 10.1016/j.ajem.2020.04.077 32402499 Breslau J. Finucane M.L. Locker A.R. Baird M.D. Roth E.A. Collins R.L. A longitudinal study of psychological distress in the United States before and during the COVID-19 pandemic Prev. Med. 143 2021 106362 10.1016/j.ypmed.2020.106362 Carli L.L. Women, gender equality and COVID-19 Gend Manag. 35 2020 647 655 10.1108/GM-07-2020-0236 Carver C.S. You want to measure coping but your protocol’s too long: consider the brief cope Int. J. Behav. Med. 4 1997 92 10.1207/s15327558ijbm0401_6 16250744 Centers for Disease Control and Prevention CDC health disparities and inequalities report—United States, 2013 62 2013 MMWR https://www.cdc.gov/mmwr/pdf/other/su6203.pdf Cohen S. Syme S.L. Social Support and Health 1985 Academic Press Cohen S. Kamarck T. Mermelstein R. A global measure of perceived stress J. Health Soc. Behav. 24 1983 385 396 10.2307/2136404 6668417 Compas B.E. Connor-Smith J.K. Saltzman H. Thomsen A.H. Wadsworth M.E. Coping with stress during childhood and adolescence: problems, progress, and potential in theory and research Psychol. Bull. 127 2001 87 127 10.1037/0033-2909.127.1.87 11271757 Czeisler M.É. Lane R.I. Petrosky E. Mental health, substance use, and suicidal ideation during the COVID-19 pandemic—United States, June 24–30, 2020 MMWR 69 2020 1049 1057 10.15585/mmwr.mm6932a1 32790653 Daly M. Robinson E. Psychological distress and adaptation to the COVID-19 crisis in the United States J. Psychiatr. Res. 136 2021 603 609 10.1016/j.jpsychires.2020.10.035 33138985 Folkman S. Moskowitz J.T. Coping: pitfalls and promise Annu. Rev. Psychol. 55 2004 745 774 10.1146/annurev.psych.55.090902.141456 14744233 Gurvich C. Thomas N. Thomas E.H. Hudaib A.R. Sood L. Fabiatos K. Sutton K. Isaacs A. Arunogiri S. Sharp G. Kulkarni J. Coping styles and mental health in response to societal changes during the COVID-19 pandemic Int. J. Soc. Psychiatry 67 2021 540 549 10.1177/0020764020961790 33016171 Hair J.F. Jr. Sarstedt M. Hopkins L. Kuppelwieser V.G. Partial least squares structural equation modeling (PLS-SEM): An emerging tool in business research Eur. Bus. Rev. 26 2014 106 121 10.1108/EBR-10-2013-0128 Hill E.M. Quality of life and mental health among women with ovarian cancer: examining the role of emotional and instrumental social support seeking Psychol. Health Med. 21 2016 551 561 10.1080/13548506.2015.1109674 26549407 Hoffart A. Johnson S.U. Ebrahimi O.V. Loneliness during the COVID-19 pandemic: change and predictors of change from strict to discontinued social distancing protocols Anxiety Stress Coping 35 2022 44 57 10.1080/10615806.2021.1958790 34314285 Huang Y. Su X. Si M. The impacts of coping style and perceived social support on the mental health of undergraduate students during the early phases of the COVID-19 pandemic in China: a multicenter survey BMC Psychiatry 21 2021 530 10.1186/s12888-021-03546-y 34706690 Karnaze M.M. Bellettiere J. Bloss C.S. Association of compassion and empathy with prosocial health behaviors and attitudes in a pandemic 2022 Submitted for publication Kroenke K. Spitzer R.L. Williams J.B. Löwe B. An ultra-brief screening scale for anxiety and depression: the PHQ-4 Psychosomatics 50 2009 613 621 10.1176/appi.psy.50.6.613 19996233 Matthes J. Koban K. Neureiter A. Stevic A. Longitudinal relationships among fear of COVID-19, smartphone online self-disclosure, happiness, and psychological well-being: survey study J. Med. Internet Res. 23 2021 e28700 10.2196/28700 McGinty E.E. Presskreischer R. Anderson K.E. Han H. Barry C.L. Psychological distress and COVID-19-related stressors reported in a longitudinal cohort of US adults in April and July 2020 JAMA 324 2020 2555 2557 10.1001/jama.2020.21231 33226420 Moore S.E. Wierenga K.L. Prince D.M. Gillani B. Mintz L.J. Disproportionate impact of the COVID-19 pandemic on perceived social support, mental health and somatic symptoms in sexual and gender minority populations J. Homosex. 68 2021 577 591 10.1080/00918369.2020.1868184 33399504 National Center for Health Statistics National Health and Nutrition Examination Survey: Analytic Guidelines, 2011–2014 and 2015–2016 2018 Centers for Disease Control and Prevention https://wwwn.cdc.gov/nchs/data/nhanes/analyticguidelines/11-16-analytic-guidelines.pdf Panayiotou G. Karekla M. Leonidou C. Coping through avoidance may explain gender disparities in anxiety J. Context. Behav. Sci. 6 2017 215 220 10.1016/j.jcbs.2017.04.005 Reevy G.M. Maslach C. Use of social support: gender and personality differences Sex Roles 44 2001 437 459 10.1023/A:1011930128829 Rogowska A.M. Kuśnierz C. Bokszczanin A. Examining anxiety, life satisfaction, general health, stress and coping styles during COVID-19 pandemic in polish sample of university students Psychol. Res. Behav. Manag. 13 2020 797 811 10.2147/PRBM.S266511 33061695 Snow-Turek A.L. Norris M.P. Tan G. Active and passive coping strategies in chronic pain patients Pain 64 1996 455 462 10.1016/0304-3959(95)00190-5 8783309 Taylor S.E. Social support: a review Friedman H.S. The Oxford Handbook of Health Psychology 2011 Oxford University Press 189 214 Torales J. O’Higgins M. Castaldelli-Maia J.M. Ventriglio A. The outbreak of COVID-19 coronavirus and its impact on global mental health Int. J. Soc. Psychiatry 66 2020 317 320 10.1177/0020764020915212 32233719
PMC009xxxxxx/PMC9005357.txt	==== Front Clin Microbiol Infect Clin Microbiol Infect Clinical Microbiology and Infection 1198-743X 1469-0691 European Society of Clinical Microbiology and Infectious Diseases. Published by Elsevier Ltd. S1198-743X(22)00204-X 10.1016/j.cmi.2022.04.001 Original Article Vertical transmission and humoral immune response following maternal infection with SARS-CoV-2: a prospective multicenter cohort study Massalha Manal 12∗† Yefet Enav 34† Rozenberg Orit 5 Soltsman Sofia 3 Hasanein Jamal 6 Smolkin Tatiana 47 Alter Adi 3 Perlitz Yuri 34 Nachum Zohar 12 1) Department of Obstetrics and Gynecology, Emek Medical Center, Afula, Israel 2) Rappaport Faculty of Medicine, Technion, Haifa, Israel 3) Department of Obstetrics and Gynecology, Baruch Padeh Medical Center, Poriya, affiliated with Azrieli Faculty of Medicine, Bar Ilan University, Israel 4) Azrieli Faculty of Medicine, Bar-Ilan University, Safed, Israel 5) Immunology Laboratory, Emek Medical Center, Afula, Israel 6) Department of Neonatology, Emek Medical Center, Afula, Israel 7) Department of Neonatology, Baruch Padeh Medical Center, Poriya, affiliated with Azrieli Faculty of Medicine, Bar Ilan University, Israel ∗ Corresponding author. Manal Massalha, Department of Obstetrics and Gynecology, Emek Medical Center, Afula, 18101, Israel. † Both authors contributed equally. 13 4 2022 13 4 2022 28 11 2021 27 3 2022 4 4 2022 © 2022 European Society of Clinical Microbiology and Infectious Diseases. Published by Elsevier Ltd. All rights reserved. 2022 European Society of Clinical Microbiology and Infectious Diseases Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Objective To explore maternal humoral immune responses to SARS-CoV-2 infection and the rate of vertical transmission. Methods A prospective cohort study was conducted at two university-affiliated medical centers in Israel. Women positive for SARS-CoV-2 reverse-transcription-polymerase-chain-reaction (RT-PCR) test during pregnancy were enrolled just prior to delivery. Levels of anti-SARS-CoV-2 spike-IgM, spike IgG, and nucleocapsid IgG were tested in maternal and cord blood at delivery, and neonatal nasopharyngeal swabs were subjected to PCR testing. The primary endpoint was the rate of vertical transmission, defined as either positive neonatal IgM or positive neonatal PCR. Results Among 72 women, 36 (50%), 39 (54%) and 30 (42%) were positive for anti-spike-IgM, anti-spike-IgG, and anti-nucleocapsid-IgG, respectively. Among 36 neonates in which nasopharyngeal swabs were taken, one neonate (3%, 95% confidence interval 0.1–15%) had a positive PCR result. IgM was not detected in cord blood. Seven neonates had positive IgG antibodies while their mothers were seronegative for the same IgG. Anti-nucleocapsid-IgG and anti-spike-IgG were detected in 25/30 (83%) and in 33/39 (85%) of neonates of seropositive mothers, respectively. According to the serology test results during delivery with respect to the time of SARS-CoV-2 infection, the highest rate of positive maternal serology tests was 8 to 12 weeks post-infection (89% anti-spike IgG, 78% anti-spike IgM, and 67% anti-nucleocapsid IgG). Thereafter, the rate of positive serology tests declined gradually; at 20 weeks post-infection, only anti-spike IgG was detected in 33 to 50%. Discussion The rate of vertical transmission of SARS-CoV-2 was at least 3% (95% confidence interval 0.1–15%). Vaccination should be considered no later than 3 months post-infection in pregnant women due to a decline in antibody levels. Keywords Antibodies COVID-19 Neonates Pregnancy SARS-CoV-2 Editor: R. Chemaly ==== Body pmcIntroduction The effect of pregnancy on humoral response to SARS-CoV-2 infection as well as the rate of vertical transmission are not fully understood. At the beginning of the current COVID-19 pandemic, evidence pointed to a lack of vertical transmission, as determined by amniocentesis, umbilical cord blood, placenta, neonatal secretion, and breast milk sampling [[1], [2], [3], [4], [5], [6]]. However, recent data, mostly from case reports and case series, demonstrated the presence of SARS-CoV-2 in the placenta [[7], [8], [9]], positive reverse-transcription-polymerase-chain-reaction (RT-PCR) of nasopharyngeal swabs of newborns, and evidence of seropositivity in neonates [[10], [11], [12], [13], [14]]. Evidence for vertical transmission is suggested in either positive neonates for SARS-CoV-2 RT-PCR or the presence of IgM-type antibodies in the newborn since these antibodies do not cross the placenta. The present study explored maternal humoral immune responses to SARS-CoV-2 infection and the rate of vertical transmission. Methods Patient recruitment This prospective multicenter cohort study was conducted between 3 July 2020 and 24 January 2021 at Emek and Baruch-Padeh Medical Centers, two university-affiliated medical centers in north Israel. The study protocol was approved by the Local Institutional Review Boards (60-20-EMC and 90-20-POR). Informed consent was obtained from all individuals who participated in the study. During the study period vaccination was not available in Israel. The study cohort consisted of pregnant women ≥18 years old who had a positive nasopharyngeal swab for SARS-CoV-2, as determined by RT-PCR, during pregnancy. Data collection Women were enrolled at admission to the delivery ward, before delivery, by one of the team investigators. After enrollment, SARS-CoV-2 anti-nucleocapsid-IgG, anti-spike-IgG, and anti-spike-IgM levels in maternal and cord blood were measured near delivery. Nasopharyngeal samples were collected from the neonates in the Department of Neonatology and were subjected to SARS-CoV-2 PCR testing. Participants were excluded from the study if both cord blood serology tests and neonatal RT-PCR could not be obtained due to technical reasons. Determination of SARS-CoV-2 antibody levels Serum was separated from clot and blood cells by centrifugation (1000 × g, 10 min) using gel separator tubes. Samples were either directly tested for SARS-CoV-2 anti-nucleocapsid-IgG antibodies by the Architect i2000 analyzer on the day of sample collection or were separated into a secondary tube and frozen at –200C until the test was performed. After performing the test, samples were frozen at –200C. For determination of SARS-CoV-2 anti-spike (S1/S2) IgG and IgM antibody titers, samples were thawed and mixed by vortex, and then subjected to ready-to-use assays on automated analyzers, as detailed in Supplement 1. Study endpoints The primary endpoint was the rate of vertical transmission, defined as either positive neonatal IgM serology or positive neonatal SARS-COV-2 PCR. Humoral immune response was also evaluated, including the rate of positive mothers for each tested antibody and antibody levels by time between infection and delivery. Correlation between antibody levels and clinical manifestation of COVID-19 was also evaluated as well as demographic and pregnancy characteristics and data regarding fetal malformations. Statistical analysis Sample size was calculated using the binomial proportion test. The rate of vertical transmission was estimated to be 7% when defined by RT-PCR [6]. Assuming that using serology tests increases the rate to 10% versus 0% in noninfected population, 71 women were required (80% power, 5% one-sided alpha). Categorical variables were analyzed using the chi-squared test or Fisher's exact test. The correlation between maternal and neonatal IgG antibody levels was assessed by the Pearson coefficient. The locally scatterplot smoothing nonparametric regression model was utilized to compare the mean drop in antibody levels over time from COVID-19 diagnosis and delivery. Antibodies levels were normalized by dividing each value by the largest value. Anti-spike antibodies were also multiplied by 3 in order to fit the scale. Antibodies levels below the threshold for positive results according to the kit instructions were set to be zero. Statistical analyses were carried out with SAS version 9.4 (SAS Institute, Cary, NC, USA). Significance was set at a p value of <0.05. Data were analyzed by the authors EY and ZN. Results Seventy-nine women were offered participation in the study. All of them agreed to participate. Among them, 72 had available serology test results. Seven women did not have serology information due to technical reasons. Thirty-six neonates did not undergo nasopharyngeal swabbing for SARS-CoV-2 PCR due to parental refusal. One woman did not have enough serum to determine anti-spike IgM levels. Patient characteristics are presented in Table 1 . SARS-CoV-2 antibody profiles of women who had COVID-19 during pregnancy and of their neonates, are presented in Table 2 . Among the 72 women, 36 (50%), 39 (54%), and 30 (42%) women tested positive for anti-spike-IgM, anti-spike-IgG, and anti-nucleocapsid-IgG, respectively. The difference between the rate of positive anti-spike-IgG and anti-nucleocapsid-IgG was statistically significant (p < 0.0001; Table 2).Table 1 Demographic characteristics and pregnancy course of women with COVID-19 during pregnancy (N = 72) Table 1Age 30.2 (4.6) BMI (kg/M2) 26.2 (5.1) Number of children 1.7 (1.5) Place of residency City >20 000 36 (50%) Town≤20 000 30 (42%) Village 6 (8%) Ethnicity Jew 27 (38%) Arab 45 (63%) Gestational week at COVID-19 disease 29.27 (8.94) Trimester at COVID-19 disease 1 5 (7%) 2 21 (29%) 3 46 (64%) Illness duration (days) 8 (12) Symptoms Asymptomatic 14 (19%) Fever 12 (17%) Cough 23 (32%) Dyspnea 20 (28%) Rhinorrhea 9 (13%) Loss of smell sensation 39 (54%) Fatigue 28 (39%) Myalgia 23 (32%) Vomiting 3 (4%) Diarrhea 6 (8%) Headache 17 (24%) Interval between COVID-19 infection diagnosis and delivery (weeks) 9.6 (8.9) Delivery week 38.9 (1.8) Preterm delivery 6 (8%) Gestational hypertension/preeclampsia 4 (6%) GDM 7 (10%) Neonate gender Male 43 (60%) Female 29 (40%) Cesarean delivery 16 (22%) Birth weight 3281 (457) SGA neonate 2 (3%) APGAR at 1 minute 3 (4%) APGAR at 5 minute 1 (1%) Cord pH 7.3 (0.1) BMI, body mass index; COVID, coronavirus disease; GDM, gestational diabetes mellitus; SGA, small for gestational age. Values are presented as mean (standard deviation) or number (percent) missing values: BMI-2, cord pH – 8. Table 2 SARS-COV-2 antibody profile of women with COVID-19 disease during pregnancy and of their neonates Table 2Maternal SARS-CoV-2 nucleocapsid IgG Negative 42 (58%) Positive 30 (42%) Maternal SARS-CoV-2 spike IgG Negative 33 (46%) Positive 39 (54%) Maternal SARS-CoV-2 nucleocapsid and spike IgG Negative 30 (42%) Both positive 27 (38%) One positive 15 (21%) Maternal SARS-CoV-2 spike IgM Negative 35 (49%) Positive 36 (51%) Neonatal SARS-CoV-2 nucleocapsid IgG Negative 42 (58%) Positive 30 (42%) Neonatal SARS-CoV-2 spike IgG Negative 35 (49%) Positive 37 (51%) Neonatal SARS-CoV-2 nucleocapsid and spike IgG Negative 32 (44%) Both positive 27 (38%) One positive 13 (18%) Neonatal SARS-CoV-2 spike IgM Negative 72 (100%) Positive 0 (0%) Neonatal PCR for SARS-CoV-2 Negative 35 (97%) Positive 1 (3%) Values are presented as number (percent). Missing values: Neonatal PCR for SARS-CoV-2 (n = 36); maternal SARS-CoV-2 anti-spike IgM (n = 1). Nasopharyngeal swabs were taken to 36 neonates as mentioned above. Comparison between maternal and pregnancy characteristics of mothers of whom nasopharyngeal swab was taken versus not taken is presented in Supplement 2. Among 36 neonates in which nasopharyngeal swabs were taken, one neonate (3%, 95% confidence interval 0.1–15%) had a positive PCR result. IgM was not detected in cord blood. Seven neonates had positive IgG antibodies while their mothers were seronegative for the same IgG. The rest of the neonates were either seronegative or had the same IgG as their mothers, and therefore whether the IgG transferred from the mothers or was self-produced by the fetus could not be determined. No fetal malformations were detected. Anti-nucleocapsid-IgG and anti-spike-IgG were detected in 25/30 (83%) and in 33/39 (85%) of neonates of seropositive mothers, respectively. Maternal and neonatal IgG antibody levels were positively correlated (Pearson coefficient 0.8, p < 0.001). With regards to the interval between infection and delivery, the highest rate of maternal positive serology tests was when the interval was between 8 to 12 weeks (89% anti-spike IgG, 78% anti-spike IgM, and 67% anti-nucleocapsid IgG). Thereafter, the rate of positive serology tests declined gradually. After 20 weeks, only anti-spike IgG was detected in 33 to 50% (Fig. 1 A).Fig. 1 Maternal humoral immune response following infection with SARS-CoV-2 according to time from infection to delivery. (A) The rate of pregnant women with positive serum anti-spike IgG (IgG-S), anti-spike IgM (IgM-S), and anti-nucleocapsid IgG (IgG-N) antibodies over time (weeks) from infection. (B) Locally scatterplot smoothing smooth curve (smoothing parameter 0.4) of mean antibody levels according to time between COVID-19 disease and delivery (weeks). Antibody levels were normalized by dividing each value with the highest value measured. Anti-spike antibodies were also multiplied by 3 in order to fit the scale. Antibody levels below the seropositivity threshold are shown as zero. Note that the graph represents the rate of antibody level decline and not their actual values. Fig. 1 Maternal anti-spike-IgG responses were the longest compared with anti-nucleocapsid-IgG and anti-spike-IgM responses (Fig. 1B). The rate of women with positive IgG serology was higher if COVID-19 was symptomatic compared to asymptomatic (anti-nucleocapsid-IgG 29 (50%) vs 1 (7%), p = 0.004; anti-spike-IgG 35 (60%) vs 4 (29%), respectively, p = 0.03). Discussion The present study explored the rate of vertical transmission and humoral immune responses following maternal SARS-CoV-2 infection during pregnancy. Evidence of a vertical transmission rate of at least 3% (95% confidence interval 0.1–15%) was observed. In addition, the highest rate of positive serology tests was among women who delivered 8 to 12 weeks postinfection, after which, the rate of positive serology tests declined gradually. Women delivering 20 or more weeks after infection only carried anti-spike IgG antibodies. Data regarding vertical transmission of SARS-CoV-2 are scarce. Most data are based on case reports and small case series. A recent review analyzed 38 studies that assessed COVID-19 and pregnancy. The rate of vertical transmission of SARS-CoV-2 differed by sample source and test type; rates were 2.9%, 7.7%, 2.9%, and 3.7% for neonatal nasopharyngeal swab testing (n = 936), placental sampling (n = 26), cord blood IgM serology (n = 34), and neonatal IgM serology (n = 82), respectively. Amniotic fluid (n = 51) and neonatal urine (n = 17) analyses showed no evidence of vertical transmission [15]. The highest vertical transmission rates (9.7% of n = 31) were observed when testing neonatal fecal/rectal samples [16,17]. In our study, the rate of vertical transmission measured by neonatal nasopharyngeal swab testing was 3%. It should be noted that seven neonates (10%) had positive IgG antibodies with a seronegative mother for the same IgG antibody. This observation may suggest either rapid decline in IgG levels in the mothers or self-produced IgG by the fetus following viral vertical transmission. Approximately 50% of the mothers in our cohort had SARS-CoV-2 antibodies. In a study that examined 392 COVID-19 convalescent subjects, 366 (93.4%) were positive for SARS-CoV-2 IgG antibodies [18]. Time from positive SARS-CoV-2 nasopharyngeal swab correlated with SARS-CoV-2 IgG antibody levels (Pearson r –0.281, p < 0.001), with a 50% decline in antibody levels within 6 months; however, levels were still above the cut-off for positive serology result. Thereafter, antibody levels stabilized and remained similar up to 9 months postinfection. In 15% of participants with tests at two time points (n = 59), SARS-CoV-2 anti-spike antibodies decreased below the positive cut-off [18]. According to our results, during pregnancy, anti-SARS-CoV-2 antibody titers declined more rapidly, with the rate of women with positive anti-spike IgG declining from 89% at 2 to 3 months postinfection to 38% by 5 months postinfection. These data suggest that while immune responses to SARS-CoV-2 infection during pregnancy are similar to those measured in nonpregnant women [18], antibody titer decline more rapidly during pregnancy. One explanation is due to the transition to a Th2 anti-inflammatory environment during pregnancy that may attenuate the immune response [[19], [20], [21]]. Future studies should explore this hypothesis. In a recent study, in the general population, vaccination at least 3 months postinfection has led to 82% protection rate from reinfection. Protection against reinfection was naturally acquired in the first 3 months postinfection [22]. Based on those and our results, we suggest that vaccination should be considered no later than 3 months postinfection in pregnant women due to a decline in antibody levels. The strengths of the study lay in its prospective design and evaluation of several antibodies against SARS-CoV-2. The limitations of the study were the lack of serial maternal serology sampling, insufficient sample size for assessing neonatal complications related to COVID-19, and parental refusal to allow neonatal SARS-CoV-2 nasopharyngeal swabbing. It should be noted that only five neonates were born within 2 weeks of maternal SARS-CoV-2 infection, when serology test sensitivity and specificity might be lower. Taken together, the rate of vertical transmission of SARS-CoV-2 was at least 3% (95% confidence interval 0.1–15%). The highest rate of maternal seropositivity was 8 to12 weeks postinfection. Anti-spike IgG levels remained high for the longest period of time and therefore should be used in serology testing, to avoid false-negative results. We suggest that vaccination should be considered no later than 3 months postinfection in pregnant women due to a decline in antibody levels. Transparency declaration All authors declare no competing interest. The authors have no conflicts of interest to declare. This study was performed in collaboration with the Israeli Ministry of Health. Author contributions MM, EY, OR, SS, JH, TS, AA, YP, and ZN participated in the study design and data collection. EY and ZN analyzed the data and wrote the manuscript. MM, OR, SS, JH, TS, AA, and YP critically reviewed the manuscript. MM and EY contributed equally. Appendix A Supplementary data The following are the Supplementary data to this article:Multimedia component 1 Multimedia component 1 Multimedia component 2 Multimedia component 2 Acknowledgements Special thanks to Mrs. Ibtehal Odeh and Mr. Ariel Abaev for performing the serology tests. Appendix A Supplementary data to this article can be found online at https://doi.org/10.1016/j.cmi.2022.04.001. ==== Refs References 1 Chen H. Guo J. Wang C. Luo F. Yu X. Zhang W. Clinical characteristics and intrauterine vertical transmission potential of COVID-19 infection in nine pregnant women: a retrospective review of medical records Lancet 395 2020 809 815 10.1016/S0140-6736(20)30360-3 32151335 2 Breslin N. Baptiste C. Gyamfi-Bannerman C. Miller R. Martinez R. Bernstein K. Coronavirus disease 2019 infection among asymptomatic and symptomatic pregnant women: two weeks of confirmed presentations to an affiliated pair of New York City hospitals Am J Obstet Gynecol MFM 2 2020 100118 10.1016/j.ajogmf.2020.100118 32292903 3 Chen Y. Peng H. Wang L. Zhao Y. Zeng L. Gao H. Infants born to mothers with a new coronavirus (COVID-19) Front Pediatr 8 2020 104 10.3389/fped.2020.00104 32266184 4 Li N. Han L. Peng M. Lv Y. Ouyang Y. Liu K. Maternal and neonatal outcomes of pregnant women with coronavirus disease 2019 (COVID-19) pneumonia: a case-control study Clin Infect Dis 71 2020 2035 2041 10.1093/cid/ciaa352 32249918 5 Wang L. Shi Y. Xiao T. Fu J. Feng X. Mu D. Chinese expert consensus on the perinatal and neonatal management for the prevention and control of the 2019 novel coronavirus infection (First edition) Ann Transl Med 8 2020 47 10.21037/atm.2020.02.20 32154287 6 Fan C. Lei D. Fang C. Li C. Wang M. Liu Y. Perinatal transmission of 2019 coronavirus disease-associated severe acute respiratory syndrome coronavirus 2: should we worry? Clin Infect Dis 72 2021 862 864 10.1093/cid/ciaa226 32182347 7 Baud D. Greub G. Favre G. Gengler C. Jaton K. Dubruc E. Second-trimester miscarriage in a pregnant woman with SARS-CoV-2 infection JAMA 323 2020 2198 2200 10.1001/jama.2020.7233 32352491 8 Penfield C.A. Brubaker S.G. Limaye M.A. Lighter J. Ratner A.J. Thomas K.M. Detection of severe acute respiratory syndrome coronavirus 2 in placental and fetal membrane samples Am J Obstet Gynecol MFM 2 2020 100133 10.1016/j.ajogmf.2020.100133 32391518 9 Algarroba G.N. Rekawek P. Vahanian S.A. Khullar P. Palaia T. Peltier M.R. Visualization of severe acute respiratory syndrome coronavirus 2 invading the human placenta using electron microscopy Am J Obstet Gynecol 223 2020 275 278 10.1016/j.ajog.2020.08.106 32405074 10 Alzamora M.C. Paredes T. Caceres D. Webb C.M. Valdez L.M. La Rosa M. Severe COVID-19 during pregnancy and possible vertical transmission Am J Perinatol 37 2020 861 865 10.1055/s-0040-1710050 32305046 11 Dong L. Tian J. He S. Zhu C. Wang J. Liu C. Possible vertical transmission of SARS-CoV-2 from an infected mother to her newborn JAMA 323 2020 1846 1848 10.1001/jama.2020.4621 32215581 12 Zamaniyan M. Ebadi A. Aghajanpoor S. Rahmani Z. Haghshenas M. Azizi S. Preterm delivery, maternal death, and vertical transmission in a pregnant woman with COVID-19 infection Prenat Diagn 40 2020 1759 1761 10.1002/pd.5713 32304114 13 Ferrazzi E. Frigerio L. Savasi V. Vergani P. Prefumo F. Barresi S. Vaginal delivery in SARS-CoV-2-infected pregnant women in Northern Italy: a retrospective analysis BJOG 127 2020 1116 1121 10.1111/1471-0528.16278 32339382 14 Zeng H. Xu C. Fan J. Tang Y. Deng Q. Zhang W. Antibodies in infants born to mothers with COVID-19 pneumonia JAMA 323 2020 1848 1849 10.1001/jama.2020.4861 32215589 15 Saadaoui M. Kumar M. Al Khodor S. COVID-19 infection during pregnancy: risk of vertical transmission, fetal, and neonatal outcomes J Pers Med 11 2021 483 10.3390/jpm11060483 34071251 16 Kotlyar A.M. Grechukhina O. Chen A. Popkhadze S. Grimshaw A. Tal O. Vertical transmission of coronavirus disease 2019: a systematic review and meta-analysis Am J Obstet Gynecol 224 2021 35 53 10.3390/jpm11060483 32739398 17 Novazzi F. Cassaniti I. Piralla A. Di Sabatino A. Bruno R. Baldanti F. SARS-CoV-2 positivity in rectal swabs: implication for possible transmission J Glob Antimicrob Resist 22 2020 754 755 10.1016/j.jgar.2020.06.011 32623000 18 Achiron A. Gurevich M. Falb R. Dreyer-Alster S. Sonis P. Mandel M. SARS-CoV-2 antibody dynamics and B-cell memory response over time in COVID-19 convalescent subjects Clin Microbiol Infect 27 2021 1349 10.1016/j.cmi.2021.05.008 e1-1349.e6 33975009 19 Dashraath P. Wong J.L.J. Lim M.X.K. Lim L.M. Li S. Biswas A. Coronavirus disease 2019 (COVID-19) pandemic and pregnancy Am J Obstet Gynecol 222 2020 521 531 10.1016/j.ajog.2020.03.021 32217113 20 Aghaeepour N. Ganio E.A. Mcilwain D. Tsai A.S. Tingle M. Van Gassen S. An immune clock of human pregnancy Sci Immunol 2 2017 eaan2946 10.1016/j.ajog.2020.03.021 21 Enninga E.A. Nevala W.K. Creedon D.J. Markovic S.N. Holtan S.G. Fetal sex-based differences in maternal hormones, angiogenic factors, and immune mediators during pregnancy and the postpartum period Am J Reprod Immunol 73 2015 251 262 10.1111/aji.12303 25091957 22 Hammerman A. Sergienko R. Friger M. Beckenstein T. Peretz A. Netzer D. Effectiveness of the BNT162b2 vaccine after recovery from Covid-19 N Engl J Med 386 2022 1221 1229 10.1056/NEJMoa2119497 35172072
PMC009xxxxxx/PMC9005358.txt	==== Front J Hosp Infect J Hosp Infect The Journal of Hospital Infection 0195-6701 1532-2939 Published by Elsevier Ltd on behalf of The Healthcare Infection Society. S0195-6701(22)00104-9 10.1016/j.jhin.2022.04.002 Article COVID-19 transmission in dental and oral/maxillofacial surgical practice during pandemic: questionnaire survey in 51 university hospitals in Japan Tanaka Hirokazu 1¶ Kurita Hiroshi 1∗¶ Shibuya Yasuyuki 2& Chikazu Daichi 3& Iino Mitsuyoshi 4& Hoshi Kazuto 5& Kobayashi Wataru 6& Yokoo Satoshi 7& Kawano Kenji 8& Mitsudo Kenji 9& Miyazaki Akihiro 10& Ota Yoshihide 11& Kishimoto Hiromitsu 12& Mori Yoshiyuki 13& Yamamoto Tetsuya 14& 1 Department of Dentistry and Oral Surgery, Shinshu University School of Medicine, Matsumoto, Japan 2 Maxillofacial Surgery, Nagoya City University, Graduate School of Medical Sciences, Nagoya, Japan 3 Department of Oral and Maxillofacial Surgery, Tokyo Medical University, Tokyo, Japan 4 Department of Dentistry, Oral and Maxillofacial Plastic and Reconstructive Surgery, Faculty of Medicine, Yamagata University, Yamagata, Japan 5 Department of Oral-Maxillofacial Surgery, Dentistry and Orthodontics, The University of Tokyo, Tokyo, Japan 6 Department of Oral and Maxillofacial Surgery, Hirosaki University School of Medicine, Hirosaki, Japan 7 Department of Oral and Maxillofacial Surgery and Plastic Surgery, Gunma University Graduate School of Medicine, Maebashi, Japan 8 Department of Oral and Maxillofacial Surgery, Faculty of Medicine, Oita University, Yufu, Japan 9 Department of Oral and Maxillofacial Surgery, Yokohama City University Graduate School of Medicine, Yokohama, Kanagawa, Japan 10 Department of Oral Surgery, Sapporo Medical University School of Medicine, Sapporo, Japan 11 Department of Oral and Maxillofacial Surgery, Tokai University School of Medicine, Isehara, Kanagawa, Japan 12 Department of Oral and Maxillofacial Surgery, Hyogo College of Medicine, Nishinomiya, Hyogo, Japan 13 Department of Dentistry, Oral and Maxillofacial Surgery, Jichi Medical University, Shimotsuke, Japan 14 Department of Oral and Maxillofacial Surgery, Kochi Medical School, Kochi University, Nankoku, Kochi, Japan ∗ Corresponding author. 3-1-1 Asahi Matsumoto-city Nagano Japna. Telephone: +81-263-37-2677. ¶ These authors contributed equally to this work. ＆ These authors also contributed equally to this work. 13 4 2022 13 4 2022 18 2 2022 6 4 2022 6 4 2022 © 2022 Published by Elsevier Ltd on behalf of The Healthcare Infection Society. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background The coronavirus disease 2019 (COVID-19) pandemic has become a major public health problem. Dental procedures that generate aerosols are considered to impose a high risk of infection; therefore, dental professionals, such as dentists and dental hygienists, are considered to be at high risk of viral transmission. However, few studies have reported COVID-19 clusters in dental care settings. Aim: This study aimed to investigate whether dental and oral/maxillofacial procedures are associated with the occurrence of COVID-19 clusters and measures taken to prevent nosocomial infection in dental clinics. Methods An online questionnaire survey on clinical activities (administrative control), infection control measures (environmental/engineering control, personal protective equipment [PPE], etc.), and confirmed or probable COVID-19 cases among patients and clinical staff was administered to the faculties of the dental and oral/maxillofacial surgical departments of university hospitals. Findings Fifty-one faculty members completed the questionnaire. All members were engaged in the treatment of dental and oral surgical outpatients and actively implemented standard precautions. Fourteen faculty members treated patients with COVID-19, but no infections transmitted from the patients to the medical staff were observed. In seven facilities, patients were found to have the infection after treatment (medical staff came in close contact), but there was no transmission from patients to medical staff. Four facilities had medical staff with infections, but none of them exhibited disease transmission from staff to patients. Conclusion COVID-19 clusters are unlikely to occur in dental and oral surgical care settings if appropriate protective measures are implemented. Keywords Infection Dental procedure COVID-19 Pandemic PPE Dentist ==== Body pmcIntroduction Coronavirus disease 2019 (COVID-19) is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The COVID-19 outbreak started in December 2019 in the city of Wuhan, Hubei province, China, and has become a major public health problem not only in China but also in other countries worldwide [1], prompting the World Health Organization to declare the COVID-19 outbreak a pandemic on 11th March 2020. The Ministry of Health, Labour, and Welfare of Japan reported that 400,000 people had been infected with SARS-CoV-2 in Japan as of mid-February 2021 [2]. Patients with COVID-19 are the main source of infection. Asymptomatic patients are extremely contagious, with strong infectivity in the average incubation period of 5–6 days. Some studies have reported that this period can be as long as 14 days [3]. The person-to-person transmission routes of COVID-19 are direct transmission (such as coughing, sneezing, and droplet inhalation) and contact transmission (such as contact with oral, nasal, and eye mucosal membranes and through droplets and aerosols) [[4], [5], [6]]. The practice of dentistry and oral/maxillofacial surgery involves the use of rotary dental and surgical instruments such as handpieces, ultrasonic scalers, and air-water syringes. These instruments create a visible spray that may contain droplets of water, saliva, blood, microorganisms, and other debris [7,8]. Hence, dentists, oral/maxillofacial surgeons, clinical staff, and patients have a high potential risk of exposure to and transmission of this virus [9]. However, there is no evidence that aerosols generated during dental and surgical care can lead to the transmission of COVID-19 [10]. To the best of our knowledge, no studies have reported the occurrence of COVID-19 clusters in dental hospitals in Japan. Only small clusters in dental clinics have been reported [11].There is fear that unnecessary refusal of medical treatment or refraining from visiting a doctor or dentist will ultimately have a negative impact on the health of the public. This study aimed to investigate whether COVID-19 clusters actually occurred from dental and oral/maxillofacial procedures and the measures were taken to prevent nosocomial infection. Methods An online questionnaire survey was administered to members of the Japanese Council of Chiefs of Dental and Oral/Maxillofacial Surgical Department of University Hospitals (JCDOUH), a group of dental and oral/maxillofacial surgeons in medical school hospitals in charge of the frontline of medical and dental care. JCDOUH consists of 64 dental and oral/maxillofacial surgical departments at the University School of Medicine in Japan. This survey was conducted using Google form(San Mateo, CA, USA) in February 2021, when vaccination was not available in Japan. JCDOUH contacted 64 departmental heads through email to invite them to participate in the survey. and confirmed or probable COVID-19 among the patients and clinical staff. The first question concerned the number of patients treated at each hospital during the study period. Questions regarding clinical activities related to the implementation of treatment or restrictions on patients and policies for dealing with dental patients with suspected or confirmed COVID-19. Infection control questions were regarding entrance screening and infection control when performing a procedure that produces aerosols. Questions about the dental treatment of patients with confirmed COVID-19 related to treatment history, the procedures performed, and establishment of infection from patient to oral healthcare workers (OHCWs). The questions regarding the exposure status due to close contact with a COVID-19-positive patient concerned the time of contact and the procedure, and transmission from the patient to the OHCWs. There were also questions about the status of COVID-19 among OHCWs during the study period, whether COVID-19-positive staff worked during the period of possible infection, whether transmission from OHCWs to patients could be confirmed, and the route of transmission. Results The staff of 51 (51/64, 79.7%) hospitals answered the questionnaire. More than 70% of hospitals had >2000 new patients from February 2020 to January 2021 (Table I ).Table I Number of patients in each hospital from February 2020 to January 2021 Table INumber of patients 1–500 patients 1 hospital 501–1500 patients 9 hospitals 1501–2000 patients 5 hospitals 2001–2500 patients 11 hospitals 2501–3000 patients 20 hospitals Over 3000 patients 5 hospitals Administrative control Regarding the restriction of treatment and patients, a change in clinical practices, policies, or procedures was reported in 88% (45/51) of hospitals. Restriction of aerosol-generating procedures (e.g., limiting the number of procedures), 31 hospitals; patient (disease) limitations (e.g., urgent, critical, cancer patients), 26 hospitals; prohibition of procedures involving splashes, 21 hospitals; restriction/coordination of patient number, 14 hospitals; and limitation of surgery, three hospitals. Seven hospitals cancelled outpatient care and six hospitals stopped accepting new patients. These limitations persisted for 12 months, with a median duration of 3 months (interquartile range, 2–3 months). Regarding the treatment of patients with confirmed COVID-19, treatment was postponed in 21 hospitals, palliative/emergency care was provided in 19 hospitals, necessary care with infection control measures was provided in five hospitals, and treatment was determined on a case-by-case basis in four hospitals (Table II ).Table II Clinical activities during the period from February 2020 to January 2021 Table IIDid you implement any restriction on treatment and/or patient? Yes 45 hospitals 88.2% No 6 hospitals 11.8% If yes, what type of restriction did you implement? (multiple answers) Restriction of aerosol-generating procedures (e.g., limit the number of procedures) 31 hospitals 60.8% Patient (disease) limitations (e.g., urgent, critical, and cancer patients) 26 hospitals 51.0% Prohibition of procedures involving splashes 21 hospitals 41.2% Restriction/coordination of patient number 14 hospitals 27.5% Limitation of surgery 3 hospitals 5.9% Cancel outpatient care 7 hospitals 13.7% How long (months) were these restrictions implemented? Median time period 3 months Interquartile range 2–3 months Range 1–12 months What was your policy for dealing with dental patients with suspected or confirmed COVID-19? Postpone the treatment 21 hospitals 41.2% Provide palliative/emergency care 19 hospitals 37.3% Provide care with infection control measures 5 hospitals 9.8% Discuss on a case-by-case basis 4 hospitals 7.8% Abbreviation: COVID-19, coronavirus disease 2019 Infection control Regarding administrative control measures in outpatient settings, entrance screening for COVID-19 was performed in 49 (96.1%) hospitals. Visitors were asked for symptoms/signs, close contact exposure, and travel from areas endemic for COVID-19, and body temperature was checked in 49 hospitals. Facemasks were encouraged in 42 hospitals, and hand hygiene was encouraged in 29 hospitals. Regarding the infection control measures during the procedures generating splash and/or aerosol, surgical gloves, face/eye guards, and face masks (surgical mask: 46 hospitals and/or N95 mask: 22 hospitals) were used in hospitals. Surgical gowns were used in 38 (74.5%) hospitals, surgical aprons in 31 (60.1%) hospitals, and surgical caps in 37 (72.5%) hospitals. An extraoral dental suction device was used in 47 hospitals, and cleaning and/or covering of potentially contaminated surfaces were performed in all hospitals (cleaning, 38 hospitals; covering, 12 hospitals). Mouth rinse before oral examination/care was encouraged in 36 hospitals (use of water in 19 hospitals and use of mouthwash in 17 hospitals) (Table III ).Table III Infection control measures Table IIIDid you perform any entrance screening? Yes 49 hospitals 96.1% No 2 hospitals 3.9% If yes, what type of screening did you perform? (multiple answers) Ask for symptoms/signs, close contact exposure, and travel from endemic area of COVID-19 49 hospitals 96.1% Body temperature measurement 49 hospitals 96.1% Wear a mask 42 hospitals 82.4% Hand hygiene compliance 29 hospitals 56.9% PCR test 1 hospital 2.0% What infection control measures were employed during the procedures generating splash and/or aerosol? Use of surgical mask 46 hospitals 90.2% Use of N95 mask 22 hospitals 43.1% Use of surgical gloves 51 hospitals 100.0% Use of face/eye guard 51 hospitals 100.0% Use of surgical gown 38 hospitals 74.5% Use of surgical apron 31 hospitals 60.8% Use of cap 37 hospitals 72.5% Use of extraoral dental suction device 47 hospitals 92.2% Cleaning of possibly contaminated surface 38 hospitals 74.5% Cover possibly contaminated surface 12 hospitals 23.5% Mouth rinse with water 19 hospitals 37.3% Mouth rinse with mouthwash 18 hospitals 35.3% Treatment in negative-pressure room 1 hospital 2.0% Abbreviations: COVID-19, coronavirus disease 2019; PCR, polymerase chain reaction Dental treatment for patients with confirmed COVID-19 Fourteen (27.4%) hospitals had COVID-positive patients for dental/oral surgical treatment. Treatments received by COVID-19-positive patients in different hospitals were as follows: oral care, six hospitals; haemostatic treatment in the oral cavity, three hospitals; tooth extraction, two hospitals; anti-inflammatory medication, one hospital; oral examination, one hospital; denture adjustment, one hospital; and palliative dental treatment, one hospital. No cases of nosocomial viral infection associated with these procedures have been reported (Table IV ).Table IV Dental treatment for patients with confirmed COVID-19 during the period from February 2020 to January 2021 Table IVDid you have any experience of dental/oral surgical treatment for confirmed COVID-19 patients? Yes 14 hospitals 27.5% No 37 hospitals 72.5% If yes, what procedure was performed? (multiple answers) Oral care 6 hospitals 11.8% Hemostatic treatment in the oral cavity 3 hospitals 5.9% Tooth extraction 2 hospitals 3.9% Anti-inflammatory medication 1 hospital 2.0% Oral examination 1 hospital 2.0% Denture adjustment 1 hospital 2.0% Palliative dental treatment 1 hospital 2.0% Is there a confirmed case of viral transmission from patient to dental staff? Yes 0 hospitals No 14 hospitals Abbreviation: COVID-19, coronavirus disease 2019 Experience of treating a patient with close contact exposure to COVID-19-positive patients (Have you ever treated a patient who was later confirmed to be infected?) was asked. Seven hospitals (13.7 %) responded positively to the questionnaire. The dental procedures performed on these patients are listed in Table V . In five of the seven hospitals, dental procedures resulting in droplet formation were performed during the period of possible viral transmission. However, no case of viral infection has been reported in the staff who provided treatment to patients who were later diagnosed with COVID-19.Table V Close contact exposure to COVID-19-positive patients during the period from February 2020 to January 2021 Table VDid you have any experience of treating patients with close-contact exposure to a COVID-19-positive patient? Yes 7 hospitals 13.7% No 44 hospitals 86.3% If yes, when and which procedure was performed at the time of close contact? 7 days before the diagnosis of COVID-19 Extraction of wisdom tooth 1 week before the diagnosis of COVID-19 Extraction of impacted third molar 2 weeks before the diagnosis of COVID-19 Dental scaling 2 days before the diagnosis of COVID-19 Dental scaling 1–2 weeks before the diagnosis of COVID-19 Oral care On the same day of the diagnosis of COVID-19 Oral examination 2 days before the diagnosis of COVID-19 Not described Were there confirmed cases of viral transmission from patient to dental staff? Yes 0 hospitals No 7 hospitals Abbreviation: COVID-19, coronavirus disease 2019 COVID-19 in dental staff Faculties of the four hospitals reported that their dental staff members were infected with SARS-CoV-2. The source of infection in the dental staff members of one hospital was presumed to be outside the hospital, and those of other hospitals were unknown. Although all these staff members provided dental treatment/care during the period of possible viral transmission, none of the patients who received dental treatment or care were diagnosed with COVID-19 (Table VI ).Table VI COVID-19 in dental staff during the period from February 2020 to January 2021 Table VIWere there any staff members who were positive for COVID-19? Yes 4 hospitals 7.8% No 47 hospitals 92.2% If yes, did the infected staff members work during the period of possible viral transmission? Yes 4 hospitals No 0 hospitals Was there a confirmed case of viral transmission from dental staff to patient? Yes 0 hospitals No 4 hospitals Suspected route of infection? Out of hospital 1 hospital Unknown 3 hospitals Abbreviation: COVID-19, coronavirus disease 2019 Discussion Aerosol-generating procedures are routinely performed in dental and oral/maxillofacial surgical practices. As SARS-CoV-2 is found in saliva [12], it is possible that COVID-19 can be transmitted by aerosolised saliva. The risk of nosocomial transmission is high in the dental settings [13]. In particular, patients in the incubation period have an increased risk of transmission among dental practitioners and patients [3,14,15]. Hence, considering the characteristics of dental procedures, such as production of splatters and aerosols and the transmission mode of SARS-CoV-2, it is presumed that dental staff and patients are at a high risk of infection. However, no case of dental and oral/maxillofacial surgical treatment-associated SARS-CoV-2 infection was found in our study to assess the effect of the COVID-19 pandemic in university hospitals, where the staff of nearly all (>70%) hospitals accepted >2000 new dental/oral surgical patients. Few studies have reported cases of COVID-19 transmission in dental setting [15,16]. In Wuhan, 34 cases of SARS-CoV-2 infection among OHCWs were reported during the epidemic period from January to March 2020. Of these, 21 individuals worked in dental outpatient clinics, seven worked in fever outpatient clinics, two worked as clerks, and one worked in the surgical inpatient ward [15]. Dental treatment was not confirmed as the route of infection in the 21 OHCWs working in dental outpatient clinics. It is possible that some of these 21 OHCWs may have been infected because of the spread of COVID-19 in Wuhan, where the infection was prevalent at that time. The Hospital of Stomatology at Wuhan University provided emergency dental treatment during the outbreak. During the period from 23rd January 2020 to 7th April 2020, 320 staff members wearing N95 masks and sophisticated PPE provided dental treatments to 2025 patients; however, none of the staff members were infected with SARS-CoV-2 [15]. Scott et al. reported no nosocomial infections at three dental clinics in New York [16]. This prospective study involved 2810 patients treated over a 6-month period (15th March 2020 to 15th September 2020) in three different dental clinics by two dentists and three hygienists during and shortly after the peak of the pandemic in New York. In addition, Estrich et al. [17] reported that of 2195 dentists in the USA, 20 (0.9%) were already infected with SARS-CoV-2 by June 2020, when the COVID-19 pandemic broke out in the USA. According to an epidemiological investigation, the source of infection in 75% of these dentists was unknown, and that in 15% of dentists was community-acquired. In this study, patients with close contact exposure to COVID-19-positive patients underwent dental treatment in seven of 51 hospitals, and those with SARS-CoV-2 infection received oral treatment in 14 of 51 hospitals; however, no case of SARS-CoV-2 nosocomial infection was reported. In addition, staff members of four hospitals who were later diagnosed with COVID-19 engaged in clinical care during the period of possible transmission; however, no cases of staff-to-patient transmission were confirmed as a result of meticulous follow-up. These results suggest that the occurrence of clusters related to dental treatment in dental-care settings is low. However, the results of this study that the rate of COVID-19 transmission in dental clinics was low should be interpreted with caution. One possible reason for the low transmission rate is that infection control measures, such as the use of PPE, were thoroughly followed in dental and oral/maxillofacial surgical practice during the COVID-19 outbreak. ‘Standard Precautions’ described by the Centers for Disease Control and Prevention in the USA in 1996 [18] indicated avoiding contact with blood, any type of body fluids, secretions, and excretions (excluding sweat) regardless of whether they contain blood, non-intact skin, and mucous membranes. Secondary precautions, ‘Transmission-Based Precautions’, were designed to reduce the risk of pathogen transmission through contact, droplets, and airborne routes. Standard precautions have been reported to be sufficient to prevent the transmission of influenza and rhinoviruses through dental aerosols [19]. Samaranayake and Peiris reported that during the SARS-CoV outbreak in 2003, no dental healthcare workers were infected in clinical settings [20]. However, it has been reported that many health workers who did not wear PPE on a regular basis and failed to use it properly were infected with SARS-CoV [13,18,20]. Regarding COVID-19, Wuhan University Hospital reported that staff members wearing N95 masks and sophisticated PPE who provided dental treatment to 2025 patients did not have COVID-19 [15]. Scott et al. reported no nosocomial infections at three dental clinics that used screening questionnaires in New York [21]. The results of this study revealed that PPE was used in nearly all the hospitals. Level of awareness of standard precautions for droplet infection through aerosols and saliva in daily dental and oral/maxillofacial surgical practice. Importantly, the low transmission rate of COVID-19 is the result of following standard precautions in daily practice. Even in OHCWs for whom COVID-19 was confirmed in this study, it was not considered a nosocomial infection (Table VI). In other words, patient-to-OHCW or OHCW-to-patient transmission of COVID-19 was not confirmed. It could be inferred that there were no SARS-CoV-2 nosocomial infections in any of the hospitals. This may be because the use of PPE, such as surgical masks, face guards, and extraoral dental suctioning, was adequately and appropriately implemented (Table III). Regarding SARS-CoV-2 infection in the dental field, aerosols generated using high-speed handpieces are problematic [22], but it has been reported that aerosol production can be reduced by 90% with the use of extraoral dental suction [8]. In this study, 92.2% of the hospitals were using this system. Another possible reason for the low transmission rate is the entrance screening for COVID-19. In this study, entrance screening was performed in 96% of hospitals. Visitors were asked about symptoms or signs, close contact exposure, and travel from areas endemic for COVID-19; moreover, their body temperature was measured. The use of facemasks was encouraged in 42 hospitals and hand hygiene compliance was encouraged in 29 hospitals. Nsawotebba et al. assessed the effectiveness of thermal screening for the detection of COVID-19 and reported a high specificity of 99.5% [23]. Guan et al. reported that fever with a body temperature above 37.5°C occurs in 88% of COVID-19-positive patients [24]. Peng et al. reported that in addition to measuring body temperature in outpatient settings, medical interviews should be conducted before patients sit on a dental chair to prevent the spread of COVID-19 [6]. Moreover, hospital restrictions on patients and treatment for other reasons may have had an impact. The number of new patients decreased during the COVID-19 pandemic; however, most hospitals accepted >2000 new dental patients. Moreover, 88% (45/51) of the hospitals had restrictions during only one-fourth of the survey period. However, COVID-19 clusters in dentistry were unlikely to occur in this study. Many hospitals that participated in this study restricted the numbers and treatments of the patients and aerosol-generating procedures. These limitations may have affected the results of the present study. A limitation of this study is that it was a retrospective study based on a questionnaire survey, and the patient and procedure restrictions implemented by each hospital were not necessarily based on the same criteria. Conclusions We investigated the prevalence of COVID-19 in dental and oral/maxillofacial surgical care settings using a questionnaire survey and found that if appropriate protective measures are taken, COVID-19 clusters are unlikely to occur in dental and oral/maxillofacial surgical care. Funding statement This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Ethical approval The study protocol was approved by the committee on medical research of Shinshu University (No. 5306). The questionnaires were anonymous. Informed consent was obtained from each participant before the questionnaire was administered. Conflict of interest statement The authors declare that they have no conflicts of interest to disclose. Acknowledgements We would like to thank Editage (www.editage.com) for editing this manuscript for native English language. We would also like to thank the following 51 hospitals for participating in this study (survey-based ordering): Kinki University Hospital, Tokai University Hospital, Juntendo University Hospital, Kobe University Hospital, Kochi University Hospital, Yamagata University Hospital, Yamanashi University Hospital, Oita University Hospital, Hirosaki University Hospital, Nagoya University Hospital, Kurume University Hospital, Toho University Hospital, Chiba University Hospital, Saga University Hospital, Nihon University Hospital, Osaka Medical and Pharmaceutical University, Miyazaki University Hospital, Toyama University Hospital, Sapporo Medical University Hospital, Fukui University Hospital, Hospital of the University of Occupational and Environmental Health, Kagawa University Hospital, Ehime University Hospital, Yokohama City University Hospital, Nagoya City University Hospital, Tottori University Hospital, Asahikawa Medical University Hospital, Kanazawa Medical University Hospital, Tsukuba University Hospital, Wakayama Medical University Hospital, Nara Medical University Hospital, Shiga Medical University Hospital, Teikyo University Hospital, Fukuoka University Hospital, Ryukyu University Hospital, Jichi Medical University Hospital, Kumamoto University Hospital, University Hospital of Kyoto Prefectural University of Medicine, Kansai Medical University Hospital, Keio University Hospital, Shinshu University Hospital, Tokyo Medical University Hospital, Tokyo University Hospital, Yamaguchi University Hospital, Gunma University Hospital, Jikei University Hospital, International University of Health and Welfare Hospital, The Hospital of Hyogo College of Medicine, Fujita Health University Hospital, Kanazawa University Hospital, and Kawasaki Medical School Hospital. ==== Refs References 1 Organization World Health coronavirus disease (COVID-19), https://covid19.who.int/, [accessed March 16, 2020]; n.d. 2 Novel coronavirus (COVID-19), https://www.mhlw.go.jp/stf/seisakunitsuite/bunya/0000164708_00001.html, [accessed February 9, 2021]; n.d. 3 Duruk G. Gümüşboğa Z.Ş. Çolak C. Investigation of Turkish dentists’ clinical attitudes and behaviors towards the COVID-19 pandemic: A survey study Braz Oral Res 34 2020 e054 10.1590/1807-3107BOR-2020.VOL34.0054 32490887 4 Barca I. Cordaro R. Kallaverja E. Ferragina F. Cristofaro M.G. Management in oral and maxillofacial surgery during the COVID-19 pandemic: Our experience Br J Oral Maxillofac Surg 58 2020 687 691 10.1016/j.bjoms.2020.04.025 32386671 5 Xu H. Zhong L. Deng J. Peng J. Dan H. Zeng X. High expression of ACE2 receptor of 2019-nCoV on the epithelial cells of oral mucosa Int J Oral Sci 12 2020 8 10.1038/s41368-020-0074-x 32094336 6 Peng X. Xu X. Li Y. Cheng L. Zhou X. Ren B. Transmission routes of 2019-nCoV and controls in dental practice Int J Oral Sci 12 2020 9 10.1038/s41368-020-0075-9 32127517 7 Centers of Disease Control and Prevention. Interim infection prevention and control guidance for dental settings during the coronavirus Disease 2019 (COVID-19) pandemic. Guid Dent Settings. 2020. 8 Harrel S.K. Molinari J. Aerosols and splatter in dentistry: A brief review of the literature and infection control implications J Am Dent Assoc 135 2004 429 437 10.14219/jada.archive.2004.0207 15127864 9 Elzein R. Bader B. Rammal A. Husseini H. Jassar H. Al-Haidary M. Legal liability facing COVID-19 in dentistry: Between malpractice and preventive recommendations J Forensic Leg Med 78 2021 102123 10.1016/j.jflm.2021.102123 33516144 10 Epstein J.B. Chow K. Mathias R. Dental procedure aerosols and COVID-19 Lancet Infect Dis 21 2021 e73 10.1016/S1473-3099(20)30636-8 32791041 11 Dental clinic clusters in Toyama n.d., https://www.hokkoku.co.jp/articles/tym/399261, [accessed April 2, 2022]. 12 To K.K.W. Tsang O.T.Y. Yip C.C.Y. Chan K.H. Wu T.C. Chan J.M.C. Consistent detection of 2019 novel coronavirus in saliva Clin Infect Dis 71 2020 841 843 10.1093/cid/ciaa149 32047895 13 Shi A.H. Guo W. Chng C.K. Chan B.H. Precautions when providing dental care during coronavirus Disease 2019 (COVID-19) pandemic Ann Acad Med Singapore 49 2020 312 319 10.47102/annals-acadmedsg.2020111 32582908 14 Bescos R. Casas-Agustench P. Belfield L. Brookes Z. Gabaldón T. Coronavirus Disease 2019 (COVID-19): Emerging and future challenges for dental and oral medicine J Dent Res 99 2020 1113 10.1177/0022034520932149 32463715 15 Meng L. Ma B. Cheng Y. Bian Z. Epidemiological investigation of OHCWs with COVID-19 J Dent Res 99 2020 1444 1452 10.1177/0022034520962087 32985329 16 Froum S.H. Froum S.J. Incidence of COVID-19 virus transmission in three dental offices: A 6-month retrospective study Int J Periodontics Restorative Dent 40 2020 853 859 10.11607/prd.5455 33151191 17 Estrich C.G. Mikkelsen M. Morrissey R. Geisinger M.L. Ioannidou E. Vujicic M. Estimating COVID-19 prevalence and infection control practices among US dentists J Am Dent Assoc 151 2020 815 824 10.1016/j.adaj.2020.09.005 33071007 18 Harte J.A. Standard and transmission-based precautions: An update for dentistry J Am Dent Assoc 141 2010 572 581 10.14219/jada.archive.2010.0232 20436107 19 Marui V.C. Souto M.L.S. Rovai E.S. Romito G.A. Chambrone L. Pannuti C.M. Efficacy of preprocedural mouthrinses in the reduction of microorganisms in aerosol: A systematic review J Am Dent Assoc 150 2019 1015 1026 10.1016/j.adaj.2019.06.024 e1 31761015 20 Samaranayake L.P. Peiris M. Severe acute respiratory syndrome and dentistry: A retrospective view J Am Dent Assoc 135 2004 1292 1302 10.14219/jada.archive.2004.0405 15493394 21 Froum S. Froum S. Incidence of COVID-19 Virus Transmission in Three Dental Offices: A 6-Month Retrospective Study Int J Periodontics Restorative Dent 40 2020 853 859 10.11607/prd.5455 33151191 22 Clementini M. Raspini M. Barbato L. Bernardelli F. Braga G. Di Gioia C. Aerosol transmission for SARS-CoV-2 in the dental practice. A review by SIdP Covid-19 task-force Oral Dis 28 Supplement 1 2022 852 857 10.1111/odi.13649 33124127 23 Nsawotebba A. Ibanda I. Ssewanyana I. Ogwok P. Ocen F. Okiira C. Effectiveness of thermal screening in detection of COVID-19 among truck drivers at Mutukula Land Point of Entry, Uganda PLOS ONE 16 2021 e0251150 10.1371/journal.pone.0251150 24 Guan W.J. Ni Z.Y. Hu Y. Liang W.H. Ou C.Q. He J.X. Clinical characteristics of coronavirus disease 2019 in China N Engl J Med 382 2020 1708 1720 10.1056/NEJMoa2002032 32109013
PMC009xxxxxx/PMC9005361.txt	==== Front Womens Health Issues Womens Health Issues Women's Health Issues 1049-3867 1878-4321 Jacobs Institute of Women's Health, George Washington University. Published by Elsevier Inc. S1049-3867(22)00030-5 10.1016/j.whi.2022.04.001 Commentary COVID-19 Vaccine Hesitancy during the Perinatal Period: Understanding Psychological and Cultural Factors to Improve Care and Address Racial/Ethnic Health Inequities Anderson Micheline R. PhD ab∗ Hardy Erica J. MD, MMSc bcd Battle Cynthia L. PhD abe a Department of Psychiatry and Human Behavior, Warren Alpert Medical School of Brown University, Providence, Rhode Island b Women and Infants' Hospital of Rhode Island, Providence, Rhode Island c Departments of Medicine and Obstetrics and Gynecology, Warren Alpert Medical School of Brown University, Providence, Rhode Island d Division of Infectious Diseases, Warren Alpert Medical School of Brown University, Providence, Rhode Island e Butler Hospital, Providence, Rhode Island ∗ Correspondence to: Micheline R. Anderson, PhD, Clinical Psychology Training Program, Warren Alpert Medical School of Brown University, Box G-BH, Providence, RI 02912. Phone: (401) 444-1929. 13 4 2022 July-August 2022 13 4 2022 32 4 317321 20 12 2021 7 4 2022 7 4 2022 © 2022 Jacobs Institute of Women's Health, George Washington University. Published by Elsevier Inc. All rights reserved. 2022 Jacobs Institute of Women's Health, George Washington University Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcFrom the beginning of the COVID-19 pandemic until December 2021, more than 24,400 pregnant people have been hospitalized with COVID-19 in the United States; however, perinatal vaccination against COVID-19 remains disproportionately low, placing pregnant and postpartum individuals at greater risk for morbidity and mortality from COVID-19. Strikingly low rates of vaccination among pregnant individuals from some racial/ethnic groups highlight pre-existing health care disparities and potentially the presence of unique vaccination concerns among some groups. Despite its significance to public health, an evidence-based understanding of how and why pregnant and postpartum individuals decide to accept the COVID-19 vaccine is lacking. Further, COVID-19 vaccine hesitancy may be related to a larger, concerning presence of medical mistrust that has been magnified in an age of misinformation. Accelerating and prioritizing research that can inform targeted and effective campaigns to increase COVID-19 vaccination among perinatal populations is essential. Perinatal COVID-19 Infection Risk and Vaccination It is now known that pregnant and postpartum women with COVID-19 are at increased risk of severe disease, in comparison with women with COVID-19 who are not pregnant. Further, infants of women with COVID-19 infections may experience poor neonatal outcomes (Allotey et al., 2020; Angelidou et al., 2021; Chinn et al., 2021; Metz et al., 2021; Villar et al., 2021; Wei, Bilodeau-Bertrand, Liu, & Auger, 2021; Zambrano et al., 2020). For example, pregnant women with COVID-19 experience almost four times the rate of intensive care unit admission and mechanical ventilation, and almost twice the rate of death as nonpregnant women with COVID-19 (Zambrano et al., 2020). As compared with pregnant women without COVID-19, those with COVID-19 are also significantly more likely to experience preterm birth and the neonatal morbidity that is associated with preterm birth (Allotey et al., 2020; Chinn et al., 2021; Metz et al., 2021; Villar et al., 2021). Mitigation efforts, including the widespread uptake of COVID-19 vaccination, are imperative to prevent loss of life and other negative maternal–infant outcomes. COVID-19 vaccinations have been successful at preventing COVID-19 transmission, hospitalization, severe infection, and COVID-related deaths; moreover, safety data suggest that the vaccine is safe and effective in pregnant and lactating people (Blakeway et al., 2021; Gray et al., 2021; Kachikis et al., 2021; Shimabukuro et al., 2021). Studies evidencing vaccine-triggered immune response among pregnant people and neonatal transfer of antibodies (Gray et al., 2021; Trostle, Aguero-Rosenfeld, Roman, & Jennifer, et al., 2021), alongside data illustrating no increased risks for miscarriage, preterm birth, or stillbirth associated with vaccination (Trostle, Limaye, et al., 2021), have propelled strong recommendations by the American College of Obstetricians and Gynecologists, the Society for Maternal Fetal Medicine, and the Centers for Disease Control and Prevention (CDC) that pregnant individuals and those attempting to become pregnant receive COVID-19 vaccines. However, recent data suggest that pregnant individuals are much more likely to be hesitant about and to refuse the COVID-19 vaccine compared with the rest of the adult population (Murphy et al., 2021). As of December 2021, data from the CDC Vaccine Safety Datalink estimated that only 24% of pregnant individuals in the United States had received at least one dose of a COVID-19 vaccine either during or before pregnancy (in comparison with 84% of U.S. adults; CDC, 2021), with rates varying significantly across ethnic/racial group (Asian, 33%; White, 24%; Hispanic, 22%; Black, 17%). Low maternal COVID-19 vaccine coverage may not resolve after birth. Postpartum women also report lower rates of intention to accept the COVID-19 vaccine than nonpregnant women (Sutton et al., 2021), citing similar reasons for COVID-19 vaccine refusal as those reported during pregnancy, such as concerns over safety and efficacy (Goncu-Ayhan et al., 2021; Oluklu et al., 2021). With the significant risk of severe COVID-19 in both pregnant and recently postpartum individuals, as well as the maternal–infant benefits of vaccination, it is crucial to understand psychological contributors to perinatal COVID vaccine uptake. COVID-19 Vaccine Hesitancy among Perinatal Populations Vaccine hesitancy is a leading contributor to low vaccination coverage across a range of diseases (Dubé et al., 2013), contributing to as many as 1.5 million deaths worldwide. Health care decision-making models, such as the theory of planned behavior (Ajzen, 1991), the health belief model (Rosenstock, 1974), and the behavioral model for vulnerable populations (Gelberg, Andersen, & Leake, 2000) include environmental, cultural, and systems-level factors that inform the engagement or rejection of various health behaviors, including vaccination. Theoretically guided research focused specifically on vaccination decisions during pregnancy and postpartum, using these and other vaccination-specific models of health behavior (e.g., the five Cs; Betsch et al., 2018), could help to clarify the reasons for vaccination decisions among perinatal women. Importantly, vaccine hesitancy research with other vaccines (before the COVID-19 pandemic) suggests that pregnant and postpartum individuals’ vaccine-specific and disease-specific beliefs, attitudes, and other psychological characteristics represent critical factors in predicting vaccine hesitancy (Kilich et al., 2020). However, despite the recognition of the important role of vaccine hesitancy in determining final behaviors regarding uptake and refusal, little is known about the psychological determinants of COVID-19 vaccine hesitancy among perinatal populations. Recent reviews examining determinants of recommended vaccinations during pregnancy (e.g., pertussis, influenza) have found that perceived maternal–infant risk, questions regarding vaccine efficacy, susceptibility to illness, and lack of knowledge are commonly reported concerns (Adeyanju et al., 2021; Kilich et al., 2020; Qiu, Bailey, & Thorne, 2021). The current research on perinatal COVID-19 vaccine hesitancy suggests that it is complex and that some vaccine-related attitudes and beliefs may even lead to seemingly inconsistent behavioral choices (Truong, Bakshi, Wasim, Ahmad, & Majid, 2021). For example, although most pregnant individuals report fears of COVID-19 infection and an overwhelming desire to protect one's unborn child, these sentiments can result in either acceptance or refusal of COVID-19 vaccination (Battarbee et al., 2021; Geoghegan et al., 2021). Of relevance to inquiry on this topic are reports of concerns regarding lack of confidence or mistrust in the development and dissemination of the COVID-19 vaccine that are not accounted for by previous vaccine attitudes or behaviors (Ceulemans et al., 2021; Goncu Ayhan et al., 2021; Palamenghi, Barello, Boccia, & Graffigna, 2020; Tram et al., 2021; Walker, Head, Owens, & Zimet, 2021). Medical mistrust among perinatal samples has been observed regarding medical interventions believed to be understudied and may contribute to health care decisions that run counter to provider-based recommendations (Denton, Creeley, Stavola, Hall, & Foltz, 2020). For example, there is a longstanding history of the exclusion of pregnant people from vaccine clinical trials, despite numerous calls from within academic medicine and obstetric providers and researchers for the inclusion of pregnant participants in early COVID-19 vaccine trials and accountability in the case of vaccine-related injuries (Bardají et al., 2021; Beigi et al., 2021; Halabi, Heinrich, & Omer, 2020). Thus, beliefs about the legitimacy and transparency of medical research may perpetuate maternal distrust of conventional medicine (Hornsey, Lobera, & Díaz-Catalán, 2020). Further, there is evidence that vaccine refusal and hesitancy during pregnancy may predict pediatric vaccine hesitancy (Cunningham et al., 2018; Fuchs, 2016). Therefore, it is essential to understand beliefs and attitudes about the COVID-19 vaccine during the perinatal period, because these may potentially extend to concerns regarding the vaccination of one's child and the subsequent associated health outcomes related to vaccine-preventable disease. Vaccine Hesitancy and Racial/Ethnic Health Disparities in Perinatal Populations Understanding mistrust and barriers to COVID-19 vaccination may be particularly salient for people of color, who have historically faced discriminatory medical treatment, subsequently influencing health care decision-making and creating greater risk for poor health outcomes (Gerend & Pai, 2008; Richard-Davis, 2021). In the United States, health disparities are perhaps most pronounced during the perinatal period, where Black and American Indian women are two to three more times likely to die from pregnancy-related complications as compared with non-Hispanic White women (Petersen et al., 2019). These disparities take on greater urgency in the context of the COVID-19 pandemic, during which higher proportions of Black and Hispanic pregnant individuals have tested positive for COVID-19 in comparison with those who are White ( Ellington et al., 2020; Jering et al., 2021). In one Southern U.S. state, for example, a recent report found that Black and Hispanic women accounted for 80% of COVID-19–related deaths among pregnant women, all of whom were unvaccinated (Kasehagen et al., 2021). Beliefs about vaccinations can vary significantly across racial groups (Wooten, Wortley, Singleton, & Euler, 2013), and it is possible that safety concerns about vaccines and/or medical distrust may disproportionately contribute to COVID-19 vaccine hesitancy among perinatal women of color in comparison with White women. Specific to the COVID-19 vaccines, one study found that Black Americans in general were more likely to believe that the vaccines are unsafe and endorse mistrust of the vaccine than other racial groups (Kricorian, Civen, & Equils, 2021). There is also evidence that greater medical mistrust is associated with greater COVID-19 vaccine hesitancy among Black immune-compromised individuals (Bogart et al., 2021). Additionally, racial inequities in health care could potentially influence health care decision-making among some perinatal groups. For example, women of color are more likely to report awareness of provider biases that contribute to increased maternal mortality or childbirth related trauma and, in turn, seek alternatives to medical interventions during pregnancy (Proujansky, 2021). To date, there have been no published studies examining the psychological determinants of COVID-19 vaccine hesitancy among perinatal people, nor any that examined specific factors that may drive vaccine hesitancy or behavior among perinatal individuals from racial/ethnic groups with higher rates of nonvaccination. Without a greater understanding of the drivers of vaccine hesitancy and refusal among perinatal populations, creating sensitive and effective approaches to addressing these issues will be challenging. Strategies for Developing Acceptable and Effective Vaccine-Related Interventions Although the first step in improving the rates of vaccine uptake is conducting research to identify factors influencing vaccine hesitancy among perinatal women—including factors that may be particularly salient for women of color—the ultimate challenge will be developing and implementing evidence-based interventions that lead to vaccine uptake. Interventions to promote vaccine acceptance across the perinatal period would be most effective when using empirically informed targets; that is, psychological factors that are specific to perinatal COVID-19 vaccine hesitancy. Before the COVID-19 pandemic, strategies developed to increase vaccination uptake among perinatal groups have included provider-based communications, education, bedside vaccine administration after childbirth, and evidence-based interviewing techniques; results have been mixed (Brewer et al., 2020; Cheng, Huang, Su, Peng, & Chang, 2015; Gagneur et al., 2018; Hutchinson & Smith, 2020; Mohammed, McMillan, Roberts, & Marshall, 2019; Wong, Lok, & Tarrant, 2016). Provider-delivered vaccine recommendations are consistently cited as significant drivers of vaccine behavior (Beel, Rench, Montesinos, Mayes, & Healy, 2013; Castillo, Patey, & MacDonald, 2021; Wiley, Cooper, Wood, & Leask, 2015). However, there is at least one report of failed intervention efforts to increase COVID-19 vaccination via provider counseling and onsite vaccine access (Hirshberg et al., 2021), suggesting other individual factors likely contribute to COVID-19 vaccine behaviors. Thus, there is a great need for understanding determinants of COVID-19 vaccine hesitancy in the perinatal population to develop efficacious interventions specific to COVID-19 vaccination. Further, understanding perinatal care providers’ perspectives regarding helpful versus unhelpful communication strategies—and effective models for integrating vaccine education and communication as well as COVID-19 vaccination into routine clinical care—will provide a critical angle as the field seeks to improve vaccination uptake. Although major models of behavior and existing studies on COVID-19 vaccine hesitancy take into account individual factors, few studies have used community-based participatory research approaches to investigate psychological determinants that may vary widely across underserved communities. Intervention acceptability and efficacy among marginalized and understudied populations can be improved by using community-engaged, patient-centered research that includes key stakeholders in health care (Collins et al., 2018; Gonzalez et al., 2021). Across health care consumers and providers, there is evidence of trends in COVID-19 vaccine hesitancy that are associated with sociodemographic characteristics (Momplaisir et al., 2021; Waring et al., 2022). Traditional psychology designs, while controlling for sample characteristics, may be enhanced by the input of community members, to fine tune the characteristics that are accounted for in the study design. Specifically, including patients, health care providers, and other stakeholders with representative views across a range of ethnic/racial groups in all phases of the research process can inform equitable and effective culturally adaptable interventions and health care policy. Further, using narrative and qualitative approaches to data collection can provide expanded and nuanced insight/understanding into complex and often understudied phenomena. Finally, the iterative co-creation of interventions through formative qualitative work and community-based participatory research may provide much-needed flexibility when conducting research within the rapidly changing landscape of COVID-19 prevention, mitigation, and treatment. As such, research that directly engages community members and stakeholders may improve the typically sluggish lines of communication that can exist between bench science and the community at large. More streamlined communication can ultimately result in a deeper understanding and prioritization of community needs, which can increase trust and improve care. Addressing COVID-19 Vaccine Hesitancy to Promote More Equitable Health Care Use To develop a robust and effective health care system that provides equitable care to all perinatal patients, it is essential to understand how trust and other psychological determinants of COVID-19 vaccine hesitancy may shape pregnant individuals' vaccination intentions and behavior—which could potentially impact future vaccination decisions with their children. Listening to and understanding patient concerns and provider insights—both in the clinical realm and through dedicated research—will help to uncover patient's experiences and attitudes that may shape care decisions in the era of COVID-19, including any specific concerns, fears, or misinformation that could serve as barriers to vaccination. Such research is essential to improve the care for all perinatal people, and we believe it may be particularly critical in strengthening systems of care for women of color who face greatest risk for poor outcomes due to COVID-19. Micheline R. Anderson, PhD, is a postdoctoral fellow in the Department of Psychiatry and Human Behavior, Alpert Medical School of Brown University and the Postpartum Day Hospital, Women & Infants Hospital. Her research interests include perinatal stress, coping, and related intervention development. Erica J. Hardy, MD, MMSc, is a Clinical Assistant Professor of Medicine and OBGYN, Alpert Medical School of Brown University and Director of Infectious Disease, Women & Infants Hospital. Her research interests include infectious disease and trauma-informed care in pregnancy. Cynthia L. Battle, PhD, is a clinical psychologist and Professor of Psychiatry & Human Behavior, Alpert Medical School of Brown University. Her research is focused on women's perinatal mental health, including development of novel, behavioral intervention approaches. Funding Statement: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. ==== Refs References Adeyanju G.C. Engel E. Koch L. Ranzinger T. Shahid I.B.M. Head M.G. …Betsch C. Determinants of influenza vaccine hesitancy among pregnant women in Europe: A systematic review European Journal of Medical Research 26 2021 116 34583779 Ajzen I. The theory of planned behavior Organizational Behavior and Human Decision Processes 50 2 1991 179 211 Allotey J. Stallings E. Bonet M. Yap M. Chatterjee S. Kew T. …Zhou D. Clinical manifestations, risk factors, and maternal and perinatal outcomes of coronavirus disease 2019 in pregnancy: Living systematic review and meta-analysis BMJ 370 2020 m33220 Angelidou A. Sullivan K. Melvin P.R. Shui J.E. Goldfarb I.T. Bartolome R. …Singh R. Association of maternal perinatal SARS-CoV-2 infection with neonatal outcomes during the COVID-19 pandemic in Massachusetts JAMA Network Open 4 4 2021 e217523 33890989 Bardají A. Sevene E. Cutland C. Menéndez C. Omer S.B. Aguado T. Muñoz F.M. The need for a global COVID-19 maternal immunisation research plan Lancet 397 2021 e17 e18 33508228 Battarbee A. Stockwell M. Varner M. Newes-Adey G. Daugherty M. Gyamfi-Bannerman C. …Subramaniam A. Attitudes toward COVID-19 illness and COVID-19 vaccination among pregnant women: A cross-sectional multicenter study during August-December 2020 American Journal of Perinatology 39 2021 75 83 34598291 Beel E.R. Rench M.A. Montesinos D.P. Mayes B. Healy C.M. Knowledge and attitudes of postpartum women toward immunization during pregnancy and the peripartum period Human Vaccines & Immunotherapeutics 9 9 2013 1926 1931 23782490 Beigi R.H. Krubiner C. Jamieson D.J. Lyerly A.D. Hughes B. Riley L. …Karron R. The need for inclusion of pregnant women in COVID-19 vaccine trials Vaccine 39 6 2021 868 33446385 Betsch C. Schmid P. Heinemeier D. Korn L. Holtmann C. …Bo hm R. Beyond confidence: Development of a measure assessing the 5C psychological antecedents of vaccination PLoS ONE 13 12 2018 e0208601 30532274 Blakeway H. Prasad S. Kalafat E. Heath P.T. Ladhani S.N. Le Doare K. …Khalil A. COVID-19 vaccination during pregnancy: Coverage and safety American Journal of Obstetrics and Gynecology 226 2021 236.e1 236.e14 34389291 Bogart L.M. Ojikutu B.O. Tyagi K. Klein D.J. Mutchler M.G. Dong L. …Kellman S. COVID-19 related medical mistrust, health impacts, and potential vaccine hesitancy among Black Americans living with HIV Journal of Acquired Immune Deficiency Syndromes 86 2 2021 200 33196555 Brewer S.E. Cataldi J.R. Fisher M. Glasgow R.E. Garrett K. O’Leary S.T. Motivational Interviewing for Maternal Immunisation (MI4MI) study: A protocol for an implementation study of a clinician vaccine communication intervention for prenatal care settings BMJ Open 10 11 2020 e040226 Castillo E. Patey A. MacDonald N. Vaccination in pregnancy: Challenges & evidence-based solutions Best Practice & Research Clinical Obstetrics & Gynaecology 76 2021 83 95 34090801 Centers for Disease Control and Prevention (CDC) COVID data tracker weekly review Available: www.cdc.gov/coronavirus/2019-ncov/covid-data/covidview/index.html 2021 Ceulemans M. Foulon V. Panchaud A. Winterfeld U. Pomar L. Lambelet V. …Richardson J.L. Vaccine willingness and impact of the COVID-19 pandemic on women’s perinatal experiences and practices—A multinational, cross-sectional study covering the first wave of the pandemic International Journal of Environmental Research and Public Health 18 7 2021 3367 33805097 Cheng P.-J. Huang S.-Y. Su S.-Y. Peng H.-H. Chang C.-L. Increasing postpartum rate of vaccination with tetanus, diphtheria, and acellular pertussis vaccine by incorporating pertussis cocooning information into prenatal education for group B streptococcus prevention Vaccine 33 51 2015 7225 7231 26549365 Chinn J. Sedighim S. Kirby K.A. Hohmann S. Hameed A.B. Jolley J. …Nguyen N.T. Characteristics and outcomes of women with COVID-19 giving birth at US academic centers during the COVID-19 pandemic JAMA Network Open 4 8 2021 e2120456 34379123 Collins S.E. Clifasefi S.L. Stanton J. Straits K.J. Gil-Kashiwabara E. Rodriguez Espinosa P. …Wallerstein N. Community-based participatory research (CBPR): Towards equitable involvement of community in psychology research American Psychologist 73 7 2018 884 29355352 Cunningham R.M. Minard C.G. Guffey D. Swaim L.S. Opel D.J. Boom J.A. Prevalence of vaccine hesitancy among expectant mothers in Houston, Texas Academic Pediatrics 18 2 2018 154 160 28826731 Denton L.K. Creeley C.E. Stavola B. Hall K. Foltz B.D. An analysis of online pregnancy message boards: Mother-to-mother advice on medication use Women and Birth 33 1 2020 e48 e58 30545755 Dubé E. Laberge C. Guay M. Bramadat P. Roy R. Bettinger J.A. Vaccine hesitancy: An overview Human Vaccines & Immunotherapeutics 9 8 2013 1763 1773 23584253 Ellington S. Strid P. Tong V.T. Woodworth K. Galang R.R. Zambrano L.D. …Gilboa S.M. Characteristics of women of reproductive age with laboratory-confirmed SARS-CoV-2 infection by pregnancy status—United States, January 22–June 7, 2020 Morbidity and Mortality Weekly Report 69 25 2020 769 32584795 Fuchs E.L. Self-reported prenatal influenza vaccination and early childhood vaccine series completion Preventive Medicine 88 2016 8 12 27002252 Gagneur A. Lemaître T. Gosselin V. Farrands A. Carrier N. Petit G. …de Wals P. A postpartum vaccination promotion intervention using motivational interviewing techniques improves short-term vaccine coverage: PromoVac study BMC Public Health 18 1 2018 811 29954370 Gelberg L. Andersen R.M. Leake B.D. The behavioral model for vulnerable populations: Application to medical care use and outcomes for homeless people Health Services Research 34 6 2000 1273 1302 10654830 Geoghegan S. Stephens L.C. Feemster K.A. Drew R.J. Eogan M. Butler K.M. “This choice does not just affect me.” Attitudes of pregnant women toward COVID-19 vaccines: A mixed-methods study Human Vaccines & Immunotherapeutics 17 2021 1 6 33529127 Gerend M.A. Pai M. Social determinants of Black-White disparities in breast cancer mortality: A review Cancer Epidemiology and Prevention Biomarkers 17 11 2008 2913 2923 Goncu Ayhan S. Oluklu D. Atalay A. Menekse Beser D. Tanacan A. Moraloglu Tekin O. …Sahin D. COVID-19 vaccine acceptance in pregnant women International Journal of Gynecology & Obstetrics 154 2021 291 296 33872386 Gonzalez C. Ramirez M. Diaz A. Duran M. Areàn P. Expanding virtual postpartum mental health care for latina women: a participatory research and policy agenda Womens Health Issues 31 2 2021 96 99 33250342 Gray K.J. Bordt E.A. Atyeo C. Deriso E. Akinwunmi B. Young N. …Edlow A.G. Coronavirus disease 2019 vaccine response in pregnant and lactating women: A cohort study American Journal of Obstetrics and Gynecology 225 3 2021 303.e1 303.e17 33775692 Halabi S. Heinrich A. Omer S.B. No-fault compensation for vaccine injury—The other side of equitable access to COVID-19 vaccines New England Journal of Medicine 383 23 2020 e125 33113309 Hirshberg J.S. Huysman B.C. Oakes M.C. Cater E.B. Odibo A.O. Raghuraman N. …Kelly J.C. Offering onsite COVID-19 vaccination to high-risk obstetrical patients: Initial findings American Journal of Obstetrics & Gynecology MFM 3 6 2021 100478 34481996 Hornsey M.J. Lobera J. Díaz-Catalán C. Vaccine hesitancy is strongly associated with distrust of conventional medicine, and only weakly associated with trust in alternative medicine Social Science & Medicine 255 2020 113019 32408085 Hutchinson A.F. Smith S.M. Effectiveness of strategies to increase uptake of pertussis vaccination by new parents and family caregivers: A systematic review Midwifery 87 2020 102734 32470666 Jering K.S. Claggett B.L. Cunningham J.W. Rosenthal N. Vardeny O. Greene M.F. …Solomon S.D. Clinical characteristics and outcomes of hospitalized women giving birth with and without COVID-19 JAMA Internal Medicine 181 5 2021 714 717 33449067 Kachikis A. Englund J.A. Singleton M. Covelli I. Drake A.L. Eckert L.O. Short-term reactions among pregnant and lactating individuals in the first wave of the COVID-19 vaccine rollout JAMA Network Open 4 8 2021 e2121310 34402893 Kasehagen L. Byers P. Taylor K. Kittle T. Roberts C. Collier C. …Dobbs T. COVID-19–associated deaths after SARS-CoV-2 infection during pregnancy—Mississippi, March 1, 2020–October 6, 2021 Morbidity and Mortality Weekly Report 70 47 2021 1646 34818319 Kilich E. Dada S. Francis M.R. Tazare J. Chico R.M. Paterson P. Larson H.J. Factors that influence vaccination decision-making among pregnant women: A systematic review and meta-analysis PLoS One 15 7 2020 e0234827 32645112 Kricorian K. Civen R. Equils O. COVID-19 vaccine hesitancy: Misinformation and perceptions of vaccine safety Human Vaccines & Immunotherapeutics 18 2021 1950504 34325612 Metz T.D. Clifton R.G. Hughes B.L. Sandoval G. Saade G.R. Grobman W.A. …Clark K. Disease severity and perinatal outcomes of pregnant patients with coronavirus disease 2019 (COVID-19) Obstetrics and Gynecology 137 4 2021 571 33560778 Mohammed H. McMillan M. Roberts C.T. Marshall H.S. A systematic review of interventions to improve uptake of pertussis vaccination in pregnancy PLoS One 14 3 2019 e0214538 30921421 Momplaisir F.M. Kuter B.J. Ghadimi F. Browne S. Nkwihoreze H. Feemster K.A. …Green-McKenzie J. Racial/ethnic differences in COVID-19 vaccine hesitancy among health care workers in 2 large academic hospitals JAMA Network Open 4 8 2021 e2121931 34459907 Murphy J. Vallières F. Bentall R.P. Shevlin M. McBride O. Hartman T.K. …Hyland P. Psychological characteristics associated with COVID-19 vaccine hesitancy and resistance in Ireland and the United Kingdom Nature Communications 12 1 2021 29 Oluklu D. Goncu Ayhan S. Menekse Beser D. Uyan Hendem D. Ozden Tokalioglu E. Turgut E. …Sahin D. Factors affecting the acceptability of COVID-19 vaccine in the postpartum period Human Vaccines & Immunotherapeutics 17 2021 4043 4047 34714190 Palamenghi L. Barello S. Boccia S. Graffigna G. Mistrust in biomedical research and vaccine hesitancy: The forefront challenge in the battle against COVID-19 in Italy European Journal of Epidemiology 35 8 2020 785 788 32808095 Petersen E.E. Davis N.L. Goodman D. Cox S. Syverson C. Seed K. Shapiro-Mendoza C. …Barfield W. Racial/Ethnic Disparities in Pregnancy-Related Deaths - United States, 2007-2016 MMWR. Morbidity and Mortality Weekly Report 68 35 2019 762 765 31487273 Proujansky A. Why Black women are rejecting hospitals in search of better births. The New York Times Available: www.nytimes.com/2021/03/11/nyregion/birth-centers-new-jersey.html 2021 Qiu X. Bailey H. Thorne C. Barriers and facilitators associated with vaccine acceptance and uptake among pregnant women in high income countries: A mini-review Frontiers in Immunology 12 2021 1246 Richard-Davis G. The pipeline problem: Barriers to access of Black patients and providers in reproductive medicine Fertility and Sterility 116 2 2021 292 295 34353571 Rosenstock I.M. The health belief model and preventive health behavior Health Education Monographs 2 4 1974 354 386 Shimabukuro T.T. Kim S.Y. Myers T.R. Moro P.L. Oduyebo T. Panagiotakopoulos L. …Meaney-Delman D.M. Preliminary findings of mRNA Covid-19 vaccine safety in pregnant persons New England Journal of Medicine 384 24 2021 2273 2282 33882218 Sutton D. D’Alton M. Zhang Y. Kahe K. Cepin A. Goffman D. …Coletta J. COVID-19 vaccine acceptance among pregnant, breastfeeding and non-pregnant reproductive aged women American Journal of Obstetrics & Gynecology MFM 3 2021 100403 34048965 Tram K.H. Saeed S. Bradley C. Fox B. Eshun-Wilson I. Mody A. Geng E. Deliberation, dissent, and distrust: Understanding distinct drivers of coronavirus disease 2019 vaccine hesitancy in the United States Clinical Infectious Diseases 16 2021 ciab633 Trostle M.E. Aguero-Rosenfeld M.E. Roman A.S. Jennifer L. High antibody levels in cord blood from pregnant women vaccinated against COVID-19 American Journal of Obstetrics & Gynecology MFM 3 2021 100481 34562636 Trostle M.E. Limaye M.A. Avtushka V. Lighter J.L. Penfield C.A. Roman A.S. COVID-19 vaccination in pregnancy: early experience from a single institution American Journal of Obstetrics & Gynecology MFM 3 6 2021 100464 34411758 Truong J. Bakshi S. Wasim A. Ahmad M. Majid U. What factors promote vaccine hesitancy or acceptance during pandemics? A systematic review and thematic analysis Health Promotion International 37 2021 daab105 Villar J. Ariff S. Gunier R.B. Thiruvengadam R. Rauch S. Kholin A. …Cardona-Perez J.A. Maternal and neonatal morbidity and mortality among pregnant women with and without COVID-19 infection: The INTERCOVID multinational cohort study JAMA Pediatrics 175 2021 817 826 33885740 Walker K.K. Head K.J. Owens H. Zimet G.D. A qualitative study exploring the relationship between mothers’ vaccine hesitancy and health beliefs with COVID-19 vaccination intention and prevention during the early pandemic months Human Vaccines & Immunotherapeutics 17 10 2021 3355 3364 34187310 Waring M.E. Pagoto S.L. Rudin L.R. Ho C. Horkachuck A. Kapoor I.A. …Foye Q. Factors associated with mothers’ hesitancy to receive a COVID-19 vaccine Journal of Behavioral Medicine 2022 1 6 34379236 Wei S.Q. Bilodeau-Bertrand M. Liu S. Auger N. The impact of COVID-19 on pregnancy outcomes: A systematic review and meta-analysis CMAJ 193 16 2021 E540 E548 33741725 Wiley K.E. Cooper S.C. Wood N. Leask J. Understanding pregnant women’s attitudes and behavior toward influenza and pertussis vaccination Qualitative Health Research 25 3 2015 360 370 25246330 Wooten K.G. Wortley P.M. Singleton J.A. Euler G.L. Perceptions matter: Beliefs about influenza vaccine and vaccination behavior among elderly White, Black and Hispanic Americans Vaccine 30 48 2012 6927 6934 22939908 Wong V.W.Y. Lok K.Y.W. Tarrant M. Interventions to increase the uptake of seasonal influenza vaccination among pregnant women: A systematic review Vaccine 34 1 2016 20 32 26602267 Zambrano L.D. Ellington S. Strid P. Galang R.R. Oduyebo T. Tong V.T. …Gilboa S.M. Update: characteristics of symptomatic women of reproductive age with laboratory-confirmed SARS-CoV-2 infection by pregnancy status—United States, January 22–October 3, 2020 Morbidity and Mortality Weekly Report 69 44 2020 1641 33151921
PMC009xxxxxx/PMC9005362.txt	==== Front J Am Med Dir Assoc J Am Med Dir Assoc Journal of the American Medical Directors Association 1525-8610 1538-9375 Published by Elsevier Inc S1525-8610(22)00296-1 10.1016/j.jamda.2022.04.003 Original Study Mortality of Care Home Residents and Community-Dwelling Controls During the COVID-19 Pandemic in 2020: Matched Cohort Study Gulliford Martin C. FRCP, FFPH ab∗ Prevost A. Toby PhD c Clegg Andrew MD de Rezel-Potts Emma PhD ab a School of Population and Life Course Sciences, King’s College London, Guy’s Campus, London, United Kingdom b NIHR Biomedical Research Centre at Guy’s and St Thomas’ Hospitals London, Great Maze Pond, London, United Kingdom c Nightingale-Saunders Clinical Trials and Epidemiology Unit, Cicely Saunders Institute, King’s College London, London, United Kingdom d Bradford Institute for Health Research, Bradford Teaching Hospitals NHS Foundation Trust, Bradford, United Kingdom e Academic Unit for Ageing and Stroke Research, University of Leeds, Leeds, United Kingdom ∗ Address correspondence to Martin C. Gulliford, FRCP, FFPH, School of Population and Life Course Sciences, King’s College London, Guy’s Campus, London SE1 1UL United Kingdom. 13 4 2022 6 2022 13 4 2022 23 6 923929.e2 © 2022 The Authors 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Objective This study aimed to estimate and compare mortality of care home residents, and matched community-dwelling controls, during the COVID-19 pandemic from primary care electronic health records in England. Design Matched cohort study. Setting and Participants Family practices in England in the Clinical Practice Research Datalink Aurum database. There were 83,627 care home residents in 2020, with 26,923 deaths; 80,730 (97%) were matched on age, sex, and family practice with 300,445 community-dwelling adults. Methods All-cause mortality was evaluated and adjusted rate ratios by negative binomial regression were adjusted for age, sex, number of long-term conditions, frailty category, region, calendar month or week, and clustering by family practice. Results Underlying mortality of care home residents was higher than community controls (adjusted rate ratio 5.59, 95% confidence interval 5.23‒5.99, P < .001). During April 2020, there was a net increase in mortality of care home residents over that of controls. The mortality rate of care home residents was 27.2 deaths per 1000 patients per week, compared with 2.31 per 1000 for controls. Excess deaths for care home residents, above that predicted from pre-pandemic years, peaked between April 13 and 19 (men, 27.7, 95% confidence interval 25.1‒30.3; women, 17.4, 15.9‒18.8 per 1000 per week). Compared with care home residents, long-term conditions and frailty were differentially associated with greater mortality in community-dwelling controls. Conclusions and Implications Individual-patient data from primary care electronic health records may be used to estimate mortality in care home residents. Mortality is substantially higher than for community-dwelling comparators and showed a disproportionate increase in the first wave of the COVID-19 pandemic. Care home residents require particular protection during periods of high infectious disease transmission. Keywords COVID-19 SARS-CoV-2 mortality care home nursing home primary care pandemic ==== Body pmcThe COVID-19 pandemic had major impacts during 2020.1 The first wave of infections peaked during April 2020 in the United Kingdom (UK), with more than 1000 deaths per day within 28 days of a positive COVID -19 test. In the second wave, with more widespread testing, the number of people in the UK with a positive COVID-19 test result peaked at 81,525 on December 29, 2020.1 Early studies identified deprivation,2 household overcrowding,3 older age, male sex, obesity, comorbidity, and ethnic minority status as being important risk factors for severe disease and mortality.4 Residents of care homes, which in the UK include residential homes providing support with personal care, and nursing homes providing support with personal care and assistance from qualified nurses, were severely affected by the pandemic. Contributing factors included the discharge of hospital patients to care homes with risks of disease transmission,5 limited availability of COVID-19 testing,6 limited supply of personal protective equipment,7 and delayed development of guidance to ensure protection of the care home popualtion.8 Data from the Office for National Statistics showed that weekly counts of deaths of care home residents in England and Wales increased from 2799 in the last week of February to 8476 and 9015 in the last 2 weeks of April 2020.9 Analysis of data reported to the Care Quality Commission in England suggested that excess deaths represented about 6.5% of care home beds.10 Care home residents typically have multiple risk markers of vulnerability for severe COVID-19, but susceptibility to infection may also have been increased because the care home environment had potential to facilitate transmission of COVID-19 and outbreaks were frequent. However, rigorous epidemiologic analysis has been limited. An editorial observed that “the COVID-19 pandemic has placed a spotlight on how little is known about this sector, and the lack of easily accessible, aggregated data on the UK care home population.”11 To address this gap, we aimed to explore whether primary care electronic health records could be used to evaluate care home mortality during the pandemic.12 We aimed to use primary care electronic health records to estimate all-cause mortality of care home residents in comparison with matched community-dwelling controls in England during 2020. Methods Data Source and Participant Selection The study drew on data from the Clinical Practice Research Datalink (CPRD) Aurum database, a database of longitudinal primary care electronic health records in England,13 including a total of 1473 general practices in England with approximately 14.8 million registered patients at January 1, 2020. The protocol for the study was approved by the CPRD Independent Scientific Advisory Committee protocol number 20_000214. This study used data from the March 2021 release of CPRD Aurum, including all 215,110 patients registered in CPRD Aurum general practices in England between January 1, 2015 and December 31, 2020 who were recorded as being resident in a care home. The most frequently recorded index care home codes were “lives in a nursing home [or] care home” (Supplementary Table 1). There were 28,531 (13%) patients with index codes of “patient died in a nursing home [or] care home.” For these, patients we assumed that they were resident in the care home for 90 days before death. The median length of stay is 2 years for care home residents, and 1 year for nursing home residents,14 but we assumed that patients with first codes for “died in nursing/care home” would have lower than average lengths of stay. In sensitivity analyses, we found that varying this assumed duration between 14 and 365 days had negligible influence on estimates. For each patient, the start date was the latest of the patient’s start of registration or the first care home code. The end of the patient’s record was the earliest of the end of patient registration, the death date recorded by CPRD and the last data collection date for the practice. There were 7584 care home residents and 16,861 controls whose records were censored by end of registration rather than by death or last data collection date. We included patients age 18‒104 years of age. For 83,627 care home residents contributing person-time during 2020, a matched comparison cohort of community-dwelling adults was sampled from the list of all patients registered in the CPRD Aurum March 2021 release after excluding care home residents. Control patients were matched for general practice, sex, and year of birth, and had a start date that was no later than 18 months after the start date for matched cases. Up to 4 community-dwelling control participants were randomly sampled with replacement15 for each care home resident. Care home residents were omitted from this analysis where there were no eligible matched controls. Main Measures The primary measure of interest was mortality from any cause based on the CPRD death date. Covariates were age, sex, region in England, multiple morbidity, and frailty category. Age in 2020 and was divided into the age-groups of 18‒64, 65‒74, 75‒84, 85‒94, and 95‒104 years. Multiple morbidity was represented by a count of 20 conditions, ever recorded in each patient’s record up to the end of 2020, from the list of atrial fibrillation, cancer, chronic kidney disease, chronic obstructive pulmonary disease, dementia, depression, diabetes mellitus, epilepsy, frailty fractures, heart failure, hemorrhagic stroke, hypertension, ischemic heart disease, ischemic stroke, other mental health diagnoses, peripheral arterial disease, palliative care, rheumatoid arthritis, or transient ischemic attack. Frailty was evaluated from coded records of deficits noted in CPRD Observation files according to the e-Frailty index, as described by Clegg et al.16 The e-Frailty index is informed by the cumulative deficit model of frailty and includes 36 deficits across physical, mental, cognitive, and social functioning. Coded records of frailty index scores and frailty index categories were also analyzed to inform frailty classification with the highest recorded value being employed. Statistical Analyses We initially analyzed eligible care home patient records between January 1, 2015 and December 31, 2020. We divided records into calendar months, calculating the number of deaths, and person time at risk for each month. We fitted a negative binomial regression model using data up to the end of 2019 as the training dataset, with counts of observed deaths as dependent variable and age-group, sex, region, multiple morbidity, frailty category, calendar month, and calendar year as predictors. Month was fitted as a categorical variable, while year was fitted as a continuous predictor. Multiple morbidity was fitted with categories from 1 to 9 or more morbidities, with a separate category for “none recorded.” Frailty category was fitted as a categorical variable. The categories of “nonfrail,” “mild frailty,” “moderate frailty,” and “severe frailty” were employed for analysis. Robust standard errors were employed to allow for general practice clustering. The general practice effect was allowed to differ between care home residents and controls, by representing the care home residents and community controls of each practice as separate clusters, because the former were clustered within care homes. We estimated predicted deaths by month for pre-pandemic and pandemic periods (2015‒ 2020), comparing predicted and observed deaths graphically. To evaluate mortality in 2020 in more detail, we analyzed data for care home residents and community controls, evaluating counts of deaths and persons at risk by calendar week. We fitted a negative binomial model, with robust standard errors, now including interaction terms that allowed the associations of long-term conditions and frailty with mortality to differ between care home residents and community controls. To summarize the results, we fitted models separately for the periods of January to March, April, and May to December. However, we also present a difference-in-difference analysis that estimated the main effect of group (care home residence), time (January to March, April, and May to December) and the group by time interaction. Analyses were performed using the “statsmodels”17 package in Python 3.8.3 (Python Software Foundation). The “matplotlib”18 package in Python and the “ggplot2”19 package in the R program (R Foundation for Statistical Computing) were used for data visualization. Results We analyzed data for 215,110 patients who were registered at general practices in England and were recorded as resident in a care home, who contributed follow-up between January 1, 2015 and December 31, 2020. There were 137,024 (64%) women; 97,192 (45%) were age 85‒94 years and 24,685 (11%) were age 95 years or older; 180,390 (84%) had 2 or more morbidities. Figure 1 shows the distribution of observed deaths (red line) by month from 2015 to 2020, compared with predicted values estimated from 2015 to 2019 data (blue line). It was clear that there was a substantial excess of observed over predicted deaths in early 2020, with a peak in April 2020.Fig. 1 Monthly counts of observed deaths of care home residents between 2015 and 2020 (red) with predicted deaths from Poisson model fitted to 2015 to 2019 data (blue). Analyses were then restricted to 83,627 care home residents who were registered during 2020, of whom 80,730 (97%) were matched with 300,445 community-dwelling controls. Characteristics of the sample are shown in Supplementary Table 2. Care home residents and community controls were similar with respect to matching variables of sex and age-group, but care home residents generally showed higher counts of long-term conditions and greater levels of frailty. There was a peak in observed deaths between April 6, 2020 and April 26, 2020 (Figure 2 , upper panel). Mortality rates were higher in men than women and increased in successive age-groups. The highest age-specific mortality rate was 63.6 (95% confidence interval 47.8‒79.4) per 1000 patients per week in men age 95‒104 years between April 13 and 19, 2020. Excess deaths, calculated as the difference between observed and predicted deaths, were summed across all age-groups (Figure 2, lower panel). Across all ages, excess deaths peaked in men between April 13 and 19, 2020 (men, 27.7, 95% confidence interval 25.1‒30.3; women, 17.4, 15.9‒18.8 per 1000 per week). During the first wave of the pandemic (weeks 12‒24), there were an estimated 2125 excess deaths in men (137 per 1000 patients) and 2727 (89 per 1000 patients) in women.Fig. 2 Total deaths per 1000 patients per week during 2020 by age-group and sex (upper panel). Excess deaths (observed minus predicted) across all ages per 1000 patients per week (lower panel). Figure 3 shows mortality rates per 1000 patients per week for each week of 2020 for care home residents (red) and community controls (blue). Data are presented separately by number of long-term conditions (upper panel) and frailty category (lower panel). There was a peak in observed deaths between April 6, 2020 and April 26, 2020 that was evident in both care home residents and community controls. Mortality of care home residents was always higher than for community controls. Mortality also increased with number of long-term conditions and frailty category. However, the effect of increasing long-term condition count or frailty category was greater for community controls than for care home residents.Fig. 3 Total deaths per 1000 patients per week during 2020 by number of long-term conditions (upper panel) and frailty category (lower panel). Red symbols, care homes residents; blue symbols, community dwelling controls. Points represent crude rates; lines represent predictions from adjusted regression model. Table 1 shows data aggregated for the periods January to March, April, and May to December 2020. Mortality of care home residents was higher in the April period than the other periods; this increase was evident at each level of frailty with absolute risks of mortality increasing with frailty level. Mortality of community controls was also higher in April compared with the other periods; the increase was proportionately smaller than for care home residents but, in absolute terms, the increase was greatest for patients with the most advanced level of frailty. Comparing care home residents and community controls, the adjusted relative mortality rate decreased with increasing level of frailty, reflecting the higher mortality of frail community controls. However, relative risks were higher in April period than in other periods of 2020.Table 1 Deaths and Person-Time for Care Home Residents and Matched Controls by Year Months Frailty Category Care Home Residents Community-Dwelling Controls Adjusted Rate Ratio∗ (95% CI) Deaths Time at Risk (Patient Wk) Rate per 1000 Patient Wk Deaths Time at Risk (Patient Wk) Rate per 1000 Patient Wk RR LL UL January to March 2020 All patients† 6134 595,571.9 10.30 5291 3,737,963 1.42 6.22 5.79 6.69 Nonfrail 323 29,897.0 10.80 257 776,819.7 0.33 36.04 28.10 46.22 Mild 845 97,690.0 8.65 823 1,039,730.0 0.79 11.59 10.46 12.84 Moderate 1474 159,954.7 9.22 1420 942,300.1 1.51 6.39 5.92 6.91 Severe 3459 307,435.0 11.25 2740 846,803.6 3.24 3.55 3.28 3.83 April 2020 All patients† 5169 189,975.7 27.2 2776 1,200,610 2.31 11.1 10.1 12.2 Nonfrail 195 9971.8 19.56 125 253,107.3 0.49 44.78 31.74 63.19 Mild 691 31,542.3 21.91 486 336,445.6 1.44 18.51 15.49 22.13 Moderate 1385 51,132.3 27.09 776 301,813.3 2.57 12.21 10.67 13.97 Severe 2871 97,116.7 29.56 1362 266,028.7 5.12 6.17 5.74 6.62 May to December 2020 All patients† 14,340 1,617,320 8.87 11,452 9,379,935 1.22 6.17 5.85 6.51 Nonfrail 675 107,097.3 6.30 528 2,025,972.0 0.26 29.01 24.60 34.22 Mild 1707 288,639.6 5.91 1849 2,664,259.0 0.69 9.96 9.04 10.97 Moderate 3401 439,903.3 7.73 3266 2,348,550.0 1.39 6.19 5.74 6.66 Severe 8496 779,945.0 10.89 5735 1,993,801.0 2.88 3.88 3.75 4.02 CI, confidence interval; LL, lower bounds of 95% confidence interval; RR, adjusted rate ratio; UL, upper bounds of 95% confidence interval. ∗ Care home residents compared to community controls, adjusted for age-group, sex, frailty, number of long-term conditions, region, month and log of patient months as offset. † Includes patients with “not classified” frailty category (273 deaths and 525,420.3 person weeks). Table 2 presents the unadjusted and adjusted estimates from a difference-in-difference analysis. After adjustment for covariates, the rate ratio for care home residence overall was 5.59 (5.23‒5.99). Mortality for controls showed a 66% (55%‒78%) relative increase during April 2020 compared with January to March 2020. After allowing for the underlying difference between care home residents and controls, and the increase shown by controls in April 2020, care home residents showed a further 76% (60%‒93%) relative increase in mortality during April 2020.Table 2 Results of Poisson Regression Models Showing Main Effects of Group and Time Period and Group by Time Interaction Unadjusted Rate Ratio Adjusted Rate Ratio 95% CI P value 95% CI P Value Net effect of care home residence during 2020∗,† January: March 2020 Ref. Ref. April 2020 1.68 (1.53‒1.85) <.001 1.76 (1.60‒1.93) <.001 May: December 2020 1.00 (0.93‒1.07) .966 1.05 (0.98‒1.12) .183 Controls during 2020‡ January: March 2020 Ref. Ref. April 2020 1.64 (1.53‒1.77) <.001 1.66 (1.55‒1.78) <.001 May: December 2020 0.86 (0.81‒0.91) <.001 0.90 (0.85‒0.95) <.001 Overall difference between care home residents and controls§ Control Ref. Ref. Care home residents 7.38 (6.90‒7.88) <.001 5.59 (5.23‒5.99) <.001 CI, confidence interval; RR, adjusted rate ratio. Estimates were adjusted for age-group, sex, region, frailty category, and number of long-term conditions. ∗ Group by time interaction. † Additional effect of care home residence during 2020, net of underlying difference between care home residence and controls and rate in controls in same period. ‡ “time” effect. § “group” effect. Supplementary Figure 1 presents a forest plot of the adjusted mortality rate ratios, comparing care home residents with community controls. At each level of morbidity or frailty, the relative risk of mortality for care home residents was higher during the COVID-related peak of mortality in April 2020 compared with the mostly prepandemic period of January to March or later-pandemic period of May to December. Discussion Main Findings This analysis shows that primary care electronic health records have potential to provide timely and relevant information concerning the care home population. There was evidence of a substantial underlying mortality difference between care home residents and community-dwelling controls that were matched for age, sex, and general practice. This difference persisted after further adjustment for frailty category, number of long-term conditions and region. We caution that, because of residual confounding from unmeasured and incompletely measured confounders, this analysis cannot determine to what extent the underlying mortality difference between care home residents and community controls is determined by the health status of residents, or the shared environment of the care home. Analyses quantified the first wave of COVID-19 mortality in April 2020 and showed that mortality peaked between April 6 and 26, 2020, being strongly associated with advanced age, male sex, multiple morbidity, and frailty category. Compared with community-dwelling control patients, mortality for care home residents was 4 to 5 times higher before the onset of the pandemic. Care home residents were disproportionately affected and during the month of April 2020 after allowing for differences in case-mix; mortality of care home residents was more than 10 times higher than for community-dwelling patients overall. Mortality remained high during the remainder of 2020 while the pandemic continued. The level of frailty and number of long-term conditions were found to be effect modifiers, being more strongly associated with mortality of community-dwelling patients than those living in care homes. Strengths and Limitations We drew on a well-described database,13 and the quality of data offered by electronic health records has been shown to be generally high.20 However, we acknowledge that there could be misclassification of care home status and it is possible that care home residence might be under-recorded. Community controls were matched on a small number of well recorded variables including age, sex, and general practice. Community controls were exactly matched with care home residents on year of birth, to allow for the important confounder of age, results were summarized over age groups. Controls might have been more closely matched for health status, but this might lead to problems of bias from over-matching. We compared unadjusted and covariate-adjusted estimates, as well as stratifying analyses by health status. We also acknowledge that limited testing for COVID-19, and recording of COVID-19 diagnoses, might have underestimated the burden of illness during the early stages of the pandemic. We addressed this by comparing the mortality of care home residents in 2020, with the mortality experienced in the preceding 5-year period (2015‒2019). We also evaluated mortality for each week from January 1, 2020 onward. For control participants, the Office for National Statistics Coronavirus (COVID-19) Infection Survey showed that at the height of the first wave of infection from April 27 to May 10, 2020, an average of 0.27% (95% confidence interval: 0.17%‒0.41%) of the general population had COVID-19.21 We did not have data concerning whether control participants were receiving social or nursing care support in their own homes, which might have been associated with frailty status. We included a count of important long-term conditions as well as analyzing frailty category. In the cumulative deficit model, frailty and multiple morbidity are closely related concepts,22 but more accurate phenotypic characterization of patients frailty status over time would have added to the study.23 Deprivation is associated with reduced healthy life expectancy, which could lead to care home admission. Patients were matched for general practice, so it was not possible to adjust for deprivation at the general practice-level. We did not employ individual postcode-level deprivation scores as these might have presented difficulties if the care home postcode did not reflect deprivation exposures over the life-course. COVID-19 mortality is associated with deprivation, as well as age, but the effect of deprivation diminishes with age.24 Ethnic minorities make up about 3% of the English population age 80 years and over25 and, while ethnic minorities may be under-represented in care homes, mortality of minorities from COVID-19 was generally higher than in the white population.2 Future studies should aim to include ethnicity and socioeconomic measures. Control sampling was with replacement and duplicated controls were included to reduce bias.15 Matching for family practice ensured that care home cases and community controls were resident in similar local areas and exposed to similar community prevalence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Comparison with Other Studies Previous studies of care home mortality during the COVID-19 pandemic have mainly drawn on data from care home records.10 , 26 Morciano et al10 analyzed data for numbers of deaths reported to the care quality commission and estimated that over the first 7 months of 2020, deaths accounted for 6.5% of care home beds. The estimates from our analyses are not directly comparable because we estimated the mortality rate per 1000 residents per week. Dutey-Magni et al26 analyzed data collected by care homes for incidence of COVID-19 and mortality. Their findings, like our study, suggested that deaths were frequent among residents who were probably infected with SARS-CoV-2 but were not tested. Burton et al27 found that outbreaks of COVID-19 were frequent within care homes and most deaths occur in the context of outbreaks.10 , 27 In the United States, mortality in care homes was consistently associated with facility-size, community-incidence of COVID-19, and poverty.28 We did not have data to identify individuals at the same care homes and the possible clustering of deaths at care homes could not be investigated in our data. Hollinghurst et al29 analyzed linked primary care and administrative records for the population of Wales and found that care homes showed increased mortality during the first wave of the pandemic. Other studies confirm that background mortality is very high in care home residents. Vossius et al30 found that annual mortality of nursing home residents was 31.8%. Shah et al31 analyzing the The Health Improvement Network primary care database for 2009 found that the age and sex standardized mortality ratio for nursing home residents was 419 and for residential home residents was 284, consistent with the elevated relative rates observed in the present analyses. Conclusions and Implications This study shows that individual-patient data from primary care electronic health records may be used to estimate mortality in care home residents in comparison with community-dwelling comparators. Mortality of care home residents is substantially higher than for community-dwelling comparators and showed a disproportionate increase in the first wave of the COVID-19 pandemic. Care home residents require particular protection during periods of high infectious disease transmission. Supplementary Data Supplementary Table 1 First Recorded Medical Codes for Care Home Residence for 215,110 Patients Registered from 2015 to 2020 (14 Most Frequent Codes Shown). Medical Terms Frequency Lives in a nursing home 78,897 Lives in care home 54,217 Patient died in nursing home 25,652 Care home visit 9383 Seen in nursing home 8198 Care home visit for initial patient assessment 6724 Nursing home visit note 5715 Care home enhanced services administration 5054 Care home visit for follow-up patient review 4097 Weekly care home ward round 3509 Patient died in care home 2879 Admission to nursing home 2717 Nursing home 2642 Discharge to nursing home 1312 All other codes 4114 Supplementary Table 2 Characteristics of Participants Variables Category Care Home Residents Community Controls Total 80,730 300,445 Gender Female 51,873 (64) 191,523 (64) Male 28,857 (36) 108,922 (36) Age-group (y) 18‒64 9250 (11) 36,801 (12) 64‒74 8234 (10) 32,699 (11) 74‒84 20,777 (26) 81,402 (27) 84‒94 34,277 (42) 126,977 (42) 94‒104 8190 (10) 22,563 (8) Number of LTCs 1 10,466 (13) 64,481 (21) 2 14,264 (18) 64,935 (22) 3 16,199 (20) 51,240 (17) 4 14,448 (18) 33,622 (11) 5 10,084 (12) 18,773 (6) 6 5986 (7) 8996 (3) 7 3052 (4) 3849 (1) 8 1288 (2) 1387 (0) 9+ 614 (1) 560 (0) None recorded 4349 (5) 52,602 (18) Frailty category Nonfrail 5284 (7) 61,829 (21) Mild 13,467 (17) 82,645 (28) Moderate 21,400 (27) 75,600 (25) Severe 40,367 (50) 69,337 (23) Not recorded 212 (0) 11,034 (4) Figures are frequencies (percent of column total). Supplementary Fig. 1 Forest plot showing adjusted mortality rate ratios, comparing care home residents and community controls, at each level number of long-term conditions and frailty category. January to March 2020 (green); April 2020 (red); May to December 2020 (green). Acknowledgments MCG wrote the study protocol and conducted data analyses; ERP provided advice on study design, data analysis and interpretation. ATP advised on statistical modeling and interpretation of results. AC advised on frailty classification and interpretation of results. All authors contributed to drafting the article and approved the final draft. MCG is guarantor. The protocol for the study was approved by the Clinical Practice Research Datalink (CPRD Independent Scientific Advisory Committee protocol number 20_000214. The study is based in part on data from the CPRD obtained under license from the UK Medicines and Healthcare products Regulatory Agency. However, the interpretation and conclusions contained in this report are those of the authors alone. Requests for data sharing should be directed to the corresponding author. The authors were supported by the National Institute for Health Research Biomedical Research Centre at Guy’s and St Thomas’ Hospitals and 10.13039/100009360 King’s College London . The views expressed are those of the authors and not necessarily those of the National Health Service, the National Institute for Health Research, or the Department of Health. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The authors had full access to all the data in the study and shared final responsibility for the decision to submit for publication. All authors declare no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous three years, and no other relationships or activities that could appear to have influenced the submitted work. ==== Refs References 1 UK Government Coronavirus (COVID-19) in the UK London: UK Government https://coronavirus.data.gov.uk/ 2021 Accessed May 11, 2022 2 Patel A.P. Paranjpe M.D. Kathiresan N.P. Race, socioeconomic deprivation, and hospitalization for COVID-19 in English participants of a national biobank Int J Equity Health 19 2020 114 32631328 3 Harris R. Exploring the neighbourhood-level correlates of COVID-19 deaths in London using a difference across spatial boundaries method Health Place 66 2020 102446 33045672 4 Williamson E.J. Walker A.J. Bhaskaran K. Factors associated with COVID-19-related death using OpenSAFELY Nature 584 2020 430 436 32640463 5 NHS England Next steps on NHS response to COVID-19 London: NHS England https://www.england.nhs.uk/coronavirus/wp-content/uploads/sites/52/2020/03/20200317-NHS-COVID-letter-FINAL.pdf 2020 Accessed May 11, 2022 6 Oliver D. Let’s be open and honest about covid-19 deaths in care homes BMJ 369 2020 m2334 32554433 7 Burki T. England and Wales see 20,000 excess deaths in care homes The Lancet 395 2020 1602 8 Department for Health and Social Care Coronavirus (COVID-19): adult social care action plan London: Department for Health and Social Care https://www.gov.uk/government/publications/coronavirus-covid-19-adult-social-care-action-plan 2020 Accessed May 11, 2022 9 Office for National Statistics Care home resident deaths registered in England and Wales, provisional London: Office for National Statics https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/carehomeresidentdeathsregisteredinenglandandwalesprovisional 2021 Accessed May 11, 2022 10 Morciano M. Stokes J. Kontopantelis E. Excess mortality for care home residents during the first 23 weeks of the COVID-19 pandemic in England: a national cohort study BMC Med 19 2021 71 33663498 11 Hanratty B. Burton J.K. Goodman C. COVID-19 and lack of linked datasets for care homes BMJ 369 2020 m2463 32581008 12 Jain A. van Hoek A.J. Walker J.L. Identifying social factors amongst older individuals in linked electronic health records: an assessment in a population based study PLoS One 12 2017 e0189038 29190680 13 Wolf A. Dedman D. Campbell J. Data resource profile: Clinical Practice Research Datalink (CPRD) Aurum Int J Epidemiol 48 2019 1740 30859197 14 Forder J. Fernandez J.L. Length of stay in care homes Report commissioned by Bupa Care Services, PSSRU Discussion Paper 2769. Canterbury: PSSRU https://eprints.lse.ac.uk/33895/1/dp2769.pdf 2011 Accessed May 11, 2022 15 Heide-Jørgensen U. Adelborg K. Kahlert J. Sampling strategies for selecting general population comparison cohorts Clin Epidemiol 10 2018 1325 1337 30310326 16 Clegg A. Bates C. Young J. Development and validation of an electronic frailty index using routine primary care electronic health record data Age Ageing 45 2016 353 360 26944937 17 Seabold S.P.J. Statsmodels: econometric and statistical modeling with python. Proceedings of the 9th Python in Science Conference https://www.statsmodels.org/stable/index.html 2010 Accessed May 11, 2022 18 Hunter J.D. Matplotlib: a 2D graphics environment Computing Sci Eng 9 2007 90‒95 19 Wickham H. ggplot2: Elegant graphics for data analysis 2009 Springer Heidelberg 20 Herrett E.L. Thomas S.L. Smeeth L. Validity of diagnoses in the general practice research database Br J Gen Pract 61 2011 438‒439 21 Office for National Statistics Coronavirus (COVID-19) Infection Survey pilot: England, 14 May 2020 London: Office for National Statistics https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/coronaviruscovid19infectionsurveypilot/england14may2020 2020 Accessed May 11, 2022 22 Clegg A. Young J. Iliffe S. Frailty in elderly people The Lancet 381 2013 752 762 23 Fried L.P. Tangen C.M. Walston J. Frailty in older adults: evidence for a phenotype J Gerontol Series A Biol Sci Med Sci 56A 2001 M146 M157 24 Tinson A. What geographic inequalities in COVID-19 mortality rates and health can tell us about levelling up London: Health Foundation https://www.health.org.uk/news-and-comment/charts-and-infographics/what-geographic-inequalities-in-covid-19-mortality-rates-can-tell-us-about-levelling-up#:∼:text=The%20'deprivation%20gap'%20in%20COVID,those%20aged%2065%20and%20older 2021 Accessed May 11, 2022 25 NOMIS Official Labour Market Statistics LC2109 - Ethnic group by age London: Office for National Statistics https://www.nomisweb.co.uk/census/2011/LC2109EWLS/view/2092957703?rows=c_age&cols=c_ethpuk11 2011 Accessed May 11, 2022 26 Dutey-Magni P.F. Williams H. Jhass A. COVID-19 infection and attributable mortality in UK care homes: Cohort study using active surveillance and electronic records (March-June 2020) medRxiv 2021 2020.07.14.20152629 27 Burton J.K. Bayne G. Evans C. Evolution and effects of COVID-19 outbreaks in care homes: a population analysis in 189 care homes in one geographical region of the UK The Lancet Healthy Longevity 1 2020 e21 e31 34173614 28 Abrams H.R. Loomer L. Gandhi A. Characteristics of U.S. Nursing Homes with COVID-19 Cases J Am Geriatr Soc 68 2020 1653 1656 32484912 29 Hollinghurst J. Lyons J. Fry R. The impact of COVID-19 on adjusted mortality risk in care homes for older adults in Wales, UK: a retrospective population-based cohort study for mortality in 2016–2020 Age Ageing 50 2020 25 31 30 Vossius C. Selbæk G. Šaltytė Benth J. Mortality in nursing home residents: A longitudinal study over three years PloS One 13 2018 e0203480 30226850 31 Shah S.M. Carey I.M. Harris T. Mortality in older care home residents in England and Wales Age Ageing 42 2013 209 215 23305759
PMC009xxxxxx/PMC9005363.txt	==== Front J Surg Res J Surg Res The Journal of Surgical Research 0022-4804 1095-8673 The Author(s). Published by Elsevier Inc. S0022-4804(22)00196-2 10.1016/j.jss.2022.04.014 Article Experiences of Acute Surgical Care During the Coronavirus Disease 2019 Pandemic Among Patients and Their Next of Kin Torbjörnsson Eva RN, PhD ab∗ Fagerdahl Ann-Mari RN, PhD ac Älgå Andreas MD, PhD ab a Department of Clinical Science and Education, Södersjukhuset, Karolinska Institutet, Stockholm, Sweden b Department of Surgery, Södersjukhuset, Stockholm, Sweden c Wound Centre, Södersjukhuset, Stockholm, Sweden ∗ Corresponding author. Department of Clinical Science and Education, Södersjukhuset, Karolinska Institutet, Södersjukhuset, Sjukhusbacken 10, 118 83 Stockholm, Sweden. Tel.: +46 733 544 335; fax: + 46 8 616 2460. 13 4 2022 9 2022 13 4 2022 277 163170 21 12 2021 10 2 2022 5 4 2022 © 2022 The Author(s) 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Introduction Since March 2020, the coronavirus disease 2019 pandemic has affected healthcare systems worldwide. It is largely unknown how acutely ill surgical patients and their next of kin have perceived the hospital care during the ongoing pandemic. Therefore, we aimed to explore their experiences. Material and methods We performed 12 interviews with patients who had undergone acute abdominal surgery in a public acute care hospital in Sweden during March to June 2020. In addition, we interviewed 10 of the patients’ next of kin. We analyzed the interviews using content analysis. Results Our analysis resulted in two themes: “Worries about seeking acute care” and “The surgical care worked adequately, even though the system was overloaded.” The participants experienced that the hospital maintained its functionality during the ongoing pandemic. Both the patients and their next of kin experienced insufficient information by the hospital, especially during the initial acute phase and at discharge, which led to a perceived loss of control. The implemented ban on visitors was found to have had both positive and negative effects for the patients, whereas the next of kin's experiences focused on the difficulties with not being able to visit. Conclusions Our findings indicate that the challenges of communication with patients and their next of kin are exacerbated during a crisis such as a pandemic. In addition, a ban on visitors might have both positive and negative aspects. Therefore, we propose individualized routines for visits to acute surgical patients when possible. Keywords COVID-19 Surgery Ban on visitors Patient experiences Next of kin ==== Body pmcIntroduction The coronavirus disease 2019 (COVID-19) pandemic has affected healthcare systems worldwide, including patients and healthcare providers in the acute surgical care.1 The effects on surgical care can be divided into direct, that is related to the COVID-19 infection, and indirect, not directly associated with the infection itself. Direct effects include prolonged time from hospital admission to surgery2 and a high mortality among patients with perioperative COVID-19 infection.3 , 4 Indirect effects of the pandemic include crowding-out effects and altered healthcare seeking behaviours which have led to a delay of care.5, 6, 7 Fear of contracting COVID-19 has been shown to alter patient behaviour early during the pandemic.8 Worldwide, countries have handled the COVID-19 pandemic with various levels of restrictions and lockdowns. The World Health Organization (WHO) has recommended a global ban on visitors within healthcare systems as a way to reduce the spread of the coronavirus.9 The few available studies have highlighted the potentially negative effects of visiting bans during the COVID-19 pandemic.10 A recent questionnaire study that compared patients who underwent general surgery before and after the pandemic found that those who had undergone surgery during the pandemic were more likely to be dissatisfied with their total experience of the hospital.11 Previous research on experiences among surgical patients during the pandemic has mainly focused on the access to elective general and orthopaedic surgery.8 , 12 , 13 Qualitative studies on the experiences among patients undergoing emergency surgery are scarce. In addition, there is a dearth of literature on the experiences among surgical patients’ next of kin. Therefore, we aimed to explore the experiences of acute surgical care during the COVID-19 pandemic among patients and their next of kin. Materials and Methods Study setting The study was conducted at a public acute care hospital in Stockholm, Sweden. The hospital serves a population of half a million inhabitants with both elective and acute abdominal surgery. During the study period, only acute and imperative (e.g., life or limb threatening) surgery was performed at the hospital due to the COVID-19 pandemic. In early 2020, the hospital was organized with dedicated wards for COVID-19 positive patients. A total ban on all visitors was imposed in accordance with the WHO recommendations. Throughout the course of the pandemic, the Swedish people received continuous updates on the situation at the hospital through television and newspaper reports. Participants Patients were selected from the hospital's operation list using a purposeful sampling14 and were invited to participate if they were ≥18 y old, had undergone acute abdominal surgery during March to June 2020, were able to verbally express their experience, and were Swedish-speaking. Maximum variation in age, sex, diagnosis, and type of surgery was taken into consideration.15 Fourteen patients were invited to participate and 12 consented to do so. Participants were asked if they had a next of kin who had been involved in their care who would consider participation. In total, 10 of the patients had a next of kin that was eligible for inclusion, and all agreed to participate. Demographic data and type of surgery are provided in Table 1 .Table 1 Demographic data. Patients (n = 12) Sex, male 9 (75%) Age, median (IQR) 56 (35-71) Number of days in hospital, median (IQR) 7 (4-9) Type of habitation Living together 4 (33%) Single household 7 (58%) Adult living at home 1 (8%) Type of surgical procedure Exploratory laparotomy 8 (66%) Appendectomy 2 (17%) Cholecystectomy 2 (17%) COVID-19 positive 1 (8%) Next of kin (n = 10) Sex, male 5 (50%) Age, median (IQR) 47 (31-55) Relationship Married/partner 3 (30%) Child 3 (30%) Other (friend, parent) 4 (40%) IQR, interquartile range. All data are presented as n (%) if not otherwise is written. Data collection Semistructured individual interviews were performed during September to November 2020 using an interview guide created by the research group (Appendix 1). Owing to the ongoing pandemic, the participants were given the opportunity to choose where the interview would take place. Three of the interviews were conducted face-to-face (with adequate distance and protective gear), and the remaining 19 were conducted by telephone. The interviews were conducted by E.T. (operating room–nurse) or A.Ä. (general surgeon), who had not been involved in the care of the patients. All interviews were recorded digitally and started with an open-ended question: ‘Please tell me about the experiences of having had acute surgery during the COVID-19 pandemic’ (patients) or ‘Please tell me about the experiences of being a next of kin to a patient who needed acute care during the COVID-19 pandemic’ (next of kin). Additional questions were asked for clarification, along with probing questions.14 Pilot interviews with one patient and one next of kin were performed to test the interview guide. The research group concluded that the interview guide did not need to be revised, and the pilot interviews were included in the analysis. The interview duration ranged from 7 to 44 min (patients) and from 8 to 20 min (next of kin). Analysis The interviews were transcribed verbatim and analyzed with qualitative content analysis as per Graneheim and Lundman.16 As a first step, each transcribed interview was read and re-read several times to gain a sense of the entirety. The second step was to find meaning units corresponding to the aim of the study while still preserving the core meaning and to label the content with a code. The approach was influenced by the description of coding data as per Gibbs.17 During the third step, the codes were compared and sorted as per differences and similarities. The last step was to discuss the findings within the research group, while focusing on the underlying latent meaning of the text. In this final step, two themes with four subthemes were identified (Table 2 ). The analysis was performed by E.T. and the proceedings were continuously discussed within the research group. Chosen quotes were translated from Swedish to English, and each quotation was identified with a participant number. Initially, 10 interviews were performed with patients and then analyzed. After performing and analyzing two additional interviews, no additional information was obtained and the research group judged that saturation of the material was reached.14 As the last step, the interviews with the next of kin were analyzed. The interviews with the patients and the next of kin were first analyzed separately, whereupon the results from both analyses were merged.Table 2 Illustration of the analysis process. Illustration of the analysis process Condensed unit Code Subtheme Theme “I suspect that the doctor and the nurse in that tent were not so eager to get close to the patients. It was not the situation in the tent that caused the delay but perhaps the distance we kept. They were about 4-5 meters away. I only answered questions that was probably the reason why they missed the inflammation [the appendicitis].” Risk for delayed care A feeling that there was a risk of an insecure emergency care Worries about seeking acute care “He told me about the protective gear. They were very careful and changed their equipment depending on the room they went into. Everything worked really, really well. So, I was not worried at all during the whole time.” Protective gear Having control over the situation The surgical care worked adequately, although the system was overloaded Rigor The analysis was discussed within the research group as a way of enhancing the credibility of the findings.18 To optimize the trustworthiness of our findings we chose participants with various experiences of surgical care to obtain maximum variation.14 To decrease the effect of the interviewer the two interviewers discussed the interview guide during data collection, aiming at dependability of the findings. To increase the transferability of the results to other groups the selection of participants and the study setting have been thoroughly described.16 Ethical considerations The study conforms to the principles outlined in the Declaration of Helsinki15 and was approved by the Swedish Ethical Review Authority (Dnr 2020-01,572). A written informed consent was obtained from all participants before the interviews were performed, and the participants were informed that they could withdraw from the study at any time. Results We identified two main themes describing the participants’ experiences: “Worries about seeking acute care” and “The surgical care worked adequately, even though the system was overloaded” (Table 3 ). The themes describe the experiences by both patients and their next of kin.Table 3 Overview of the results. Overview of the results Subthemes Themes Thoughts on the decision to seek hospital care Worries about seeking acute care A feeling that there was a risk of an insecure emergency care Interaction with the healthcare personnel The surgical care worked acceptable despite an overloaded system Having control over the situation Worries about seeking acute care This theme describes the participants’ experiences of seeking acute surgical care and of the care provided at the emergency department. Most patients had decided of their own accord to visit the emergency department, whereas some had been redirected from a walk-in clinic. Thoughts on the decision to seek hospital care The patients and their next of kin described how worries regarding their acute medical condition overshadowed their concerns with the ongoing pandemic.“I don't think we thought all that much about the pandemic, eh, none of us, neither my wife nor my two kids, eh, but rather worried about what I was in hospital for” (Patient 9). Some of the patients had worried about needing acute care during the pandemic and had refrained from seeking care due to a fear of contracting COVID-19. On the other hand, one patient expressed no worries at all, pointing to a greater risk of getting infected in the supermarket or at work than at the hospital. Some of the patients had been worried about their medical condition contributing to overloading the healthcare system and expressed that they felt reluctant to take up the healthcare providers’ time. A feeling that there was a risk of an insecure emergency care The patients expressed that the acute care management worked well, but some had experienced delays at the emergency department. The next of kin described difficulties getting admitted for acute care and therefore thought that the care was at a risk of being delayed. During the study period, the primary triage was performed outside the hospital, due to the restrictions. Some of the patients and their next of kin explained that the healthcare personnel had only asked the patient questions, refraining from any physical examination. After this primary triage, some patients were sent back home from the hospital without any further investigation. One of the next of kin expressed that the patient's diagnosis (appendicitis) was missed because of the physical distance between the healthcare personnel and the patient. Patients described feeling that the personnel were afraid of being infected themselves with the virus and therefore refrained from a physical examination. Others expressed that the reason why they were sent home might have been that the healthcare personnel were afraid of overloading the healthcare system with patients not in need of immediate care. One of the patients thought that the health providers’ only focus was on COVID-19 and expressed a feeling that they had forgotten about the ‘normal patients’.“Well, I do think that if there hadn't been a pandemic, they would have brought me in for examination… checked me and I believe they would have swiftly discovered that I had an appendicitis” (Patient 7). Others, both patients and their next of kin, expressed that the acute care functioned without any delay. Some were surprised by this experience, as they had not expected this high level of functioning.“I thought it would be pretty chaotic, but it was just me there, so I walked in through the doors to the emergency department, and I was there all by myself, no waiting line or anything. It was very quiet, hardly any people there at all, patients that is” (Patient 9). Owing to the ban on visitors, the next of kin were not able to accompany the patients into the emergency department. Although accepting that, the next of kin described how hard it was to leave their loved one by the door, or in the ambulance, and wait at home alone without knowing what was happening.“I wasn't allowed to enter any further than the entrance, and ‘NN’ was taken care of after that. So, I was like, ‘hey, wait, can I come along, oh, wait!’ and you don't know what is going to happen, and so many questions popped up inside my head” (Next of kin 6). The surgical care worked adequately, although the system was overloaded This theme covers the participants’ experiences of the care provided in the surgical ward. Overall, the patients were surprised by the maintained hospital functionality, despite the high number of COVID-19 patients admitted to the hospital. Interaction with the healthcare personnel The patients were overwhelmed by how well the surgical inpatient care functioned, despite the overloaded healthcare system. They described how the healthcare personnel took their time to inform and to calm them during their hospital stay, despite the obvious stress among the personnel.“I thought that everybody, well, acted very professionally and I saw that it was, that they had a lot to do, but they were polite and helped out” (Patient 7). Some of the patients explained that they did what they could to minimize the workload of the healthcare personnel. For example, they waited a bit longer before asking for analgesia, or tried to solve their problems themselves, to reduce the workload at the ward. In addition, one patient expressed that the hospital's ban on visitors granted the healthcare providers more time to focus on the patients, as they did not have to make time for the next of kin.“Then, the personnel could spend more time on the patients rather than the kind of caring for a lot of worried next of kin who just... Well yes, I understand that they are worried, I get that, but you often have to spend even more time with them than with the patients” (Patient 9). The next of kin were to a large extent informed by the patients and not by the healthcare personnel. As per the participants, this worked well most of the time. However, the next of kin of patients who was elderly, or had undergone more advanced surgery, expressed that they would have preferred receiving information directly from the personnel. These next of kin experienced that the patient had a hard time understanding all the information given and described how it was challenging for them to provide support without any direct contact with the treating physician or not being present at the time of hospital discharge. Some of the next of kin called and talked with the nurses, whereas others did not wish to disturb the personnel. The ones who had called the hospital described that they had received good and detailed information.“Information is, I think, really important, from the personnel to the next of kin. That you get some information on what has happened at the hospital, what has been done, what things to think about in the future, rehab and so on. I think they (the patients, authors’ comment) have trouble understanding and knowing exactly what to do, so there you can be supportive and help with rehab and so on” (Next of kin 4). The next of kin found the shortage of information especially challenging during the time until their loved ones had recovered enough postoperatively to be able to make a phone call.“I know that my family found it very troublesome that they didn't know what was happening. It wasn't until after my second surgery, maybe three days later, that they contacted (the hospital, authors’ comment) because I wasn't well enough to tell them myself” (Patient 7). The next of kin of patients who arrived at the hospital by ambulance found the ban particularly troublesome. As they had not been able to accompany the patient to the hospital, they often did not know how to locate the patient. These participants described how their only option was to wait until their loved one was healthy enough to contact them, which in some cases took several days.“Well, I was kind of awake there the whole night, because I couldn't get hold of anyone and she was undergoing surgery for many hours, as far as I understood. So, it wasn't until the next morning that I could get hold of someone who knew where she was. But she was in a really bad condition when she was admitted to the hospital, so I thought that the worst thing possible might have happened. But I also thought that as long as they don't call me, it might not have happened” (Next of kin 1). Most patients and their next of kin experienced receiving insufficient information at the time of hospital discharge. They lacked guidelines to follow, including information on what to be aware of following surgery, or the information was delayed. For example, one patient received a written document by a mail several weeks after discharge.“I didn't receive any, any help or information on, like, what to eat or what would come afterwards and so, because I was really ill for a long time afterwards” (Patient 9). Having control over the situation Both the patients and their next of kin experienced that the restrictions during the pandemic affected their ability to control the situation. The factor with the greatest impact on their experience was the ban on visitors. The patients felt that it was difficult to get the support they wanted when their next of kin were not allowed to visit them. The next of kin expressed that the ban made it difficult to obtain sufficient information, which caused an experience of not having control over the situation. In general, both the patients and their next of kin explained that the ban was a burden but they believed it was a necessary measure to lower the risk of spreading the coronavirus. Some of the patients explained that it was emotionally challenging not being able to see their next of kin. One patient, who was experiencing severe pain, described that the personnel did not have time for comforting. Consequently, the patient longed for the psychological support that the next of kin could have provided if there had not been a ban on visitors. Some patients were so negatively affected by the ban that the healthcare personnel made an exception and allowed for visits by the patients' next of kin. Other patients described the ban on visitors as something positive as it gave them more time to focus on themselves—they appreciated the peace and tranquility. Some patients expressed feelings of relief not having to expose their next of kin to their deranged condition. The ban was also perceived by some patients as a chance to maintain their integrity, describing how it would have been difficult to meet their next of kin in their bad shape. The patients expressed that they believed that the ban on visitors was worst for their next of kin, as the ban made it difficult for them to create an accurate image of the situation at the hospital.“It was much worse for my daughter, she thought it was terrible as she didn't know how I was doing or anything. But I did explain and she spoke to the doctors but, but of course she was worried” (Patient 2). The next of kin found it challenging not being able to visit the hospital. They expressed that the pandemic had made their situation more worrisome than it would otherwise have been, and they experienced a feeling of a lack of control due to the insufficient information. However, the next of kin all expressed an understanding for the ban on visitors. Interestingly, some described the ban as an excuse for not having to visit the hospital, experiencing the ban as a sort of relief. One next of kin pointed to negative experiences from a previous hospital visit and another next of kin had such a fear of contracting COVID-19 that a visit to the hospital would not have been an option even without the ban. The use of video calls was described as a good surrogate for physical visits by both patients and their next of kin. This technique gave them an opportunity to see each other, which made it possible for the next of kin to form their own picture of how their loved ones were doing. The patients experienced that the technique made it easier for their next of kin to support them and that this made them feel less lonely. Some patients were aware of the ban on visitors before seeking hospital care and had consequently brought devices for making video calls. Others who had not, suggested that such devices might be something that the hospital could offer for rent. The patients felt assured that the personnel followed the existing guidelines to prevent the spread of the coronavirus and did not worry about their own risk of contracting the virus. The next of kin described that, at first, they were worried about their loved ones developing COVID-19 at the hospital. However, receiving information from the patient regarding the hospital guidelines had a calming effect.“He told me that this particular thing with the protective gear, you had to change gear depending on what room you entered, etc. All of this seemed to work out really, really great, so I wasn't at any time worried the slightest that he would be infected” (Next of kin 6). Discussion Our findings indicate that both the interviewed patients and their next of kin experienced maintained in-hospital functionality for patients undergoing acute surgical care during the COVID-19 pandemic. During this time, the hospital had imposed a ban on visitors as a way to reduce the risk of spreading the coronavirus. The ban was perceived as both positive and negative by the patients, whereas their next of kin generally found it challenging not to be able to visit. Furthermore, we have identified areas in need of improvement, particularly regarding the communication between the healthcare personnel and the patients and their next of kin. Some of the participants had experienced delays, especially at the emergency department. The perceived reason included a lack of resources at the hospital and a feeling that the healthcare providers' main focus was on COVID-19. Recent studies from the COVID-19 pandemic have identified a delay in healthcare seeking behaviours and a decreased number of hospitalizations due to acute surgical conditions such as appendicitis and biliary tract pathology.6 , 7 , 19, 20, 21 Reasons for such altered healthcare seeking behaviour could include a reluctance to overload the healthcare system and a fear of exposure to COVID-19 at the hospital.5 , 22 Surprisingly, a fear of developing COVID-19 was not expressed by the patients in our study. Instead, our participants described how their worries regarding their own symptoms overshadowed the concerns of the ongoing pandemic. These results are in line with a study of patients undergoing elective surgery during the pandemic, in which the patents’ fear was primarily associated with anxiety in relation to their primary pathology, or the waiting time until surgery, rather than that of developing COVID-19.23 The participants in our study expressed divided views on the hospital's ban on visitors. Some of the patients felt that the ban was positive as it gave them more time to focus on themselves. In addition, they thought that it generated more time for the healthcare personnel to focus on the patients. Others felt that it was hard not being able to see their next of kin. Previous studies of patients undergoing elective surgery during the COVID-19 pandemic found that a ban on visitors affected the patients' postoperative experience negatively, especially by generating feelings of isolation and loneliness.11 , 24 Our results offer a more diverse picture. The ban on visitors was challenging for the next of kin. Not being able to create their own image of the patient's situation generated a feeling of having a lack of control. Many of the next of kin felt that they had not received enough information by the healthcare personnel, especially during the initial acute phase and at discharge. In a study of terminal care during the pandemic, the patients' next of kin reported poor communication with the healthcare teams, and the authors concluded that innovative strategies are needed to improve the communication quality during visitation restrictions.25 Furthermore, a previous study on patients undergoing elective surgery during the COVID-19 pandemic showed that the participants experienced that their postoperative preferences were not adequately addressed upon discharge.11 A factor contributing to this may be the ban on visitors. The fact that the next of kin could not attend at discharge from the hospital made it harder for them to be actively involved in the postoperative care. Loss of information has been shown to have a negative effect on spouses to patients with a chronic disease, as it makes it harder for them to maintain control over the situation.26 Therefore, there might be a need for healthcare professionals to evaluate routines and to develop strategies for providing sufficient support and clear information in times of visitation restrictions. Inpatient surgical care is an essential component of any functioning healthcare system. The need to provide care for patients with acute surgical conditions will remain, also during a pandemic and its associated challenges with reduced resources. Our study indicates that there is a need for improving the information given to the patients and their next of kin and to develop strategies to improve the communication quality during visitation restrictions, both in the acute phase and at hospital discharge. Furthermore, there might be a conflict between the patients' need of calm and privacy and the next of kin's need of being able to participate in the care. The potential advantages of limiting visits to patients in acute surgical care also in a nonpandemic situation could be investigated in future studies, with a focus on the need of individualized routines for visitors. Strengths and limitations To the best of our knowledge, this is the first COVID-19–related study exploring the experiences from acute surgical care among patients and their next of kin. The study has some limitations that need to be taken into consideration. The interviews were performed several months after hospital discharge, which may have led to the introduction of recall bias.27 However, owing to the nature of the events and the extraordinary conditions during the pandemic, we judge this risk to be small. This study covers the surgical care of a selected group of patients admitted to one acute care hospital. Therefore, the transferability of our findings can be considered limited. To increase the readers’ ability to interpret and transfer the result, we have aimed to give a clear description of the context and the participants together with appropriate quotations as per the recommendations of Graneheim and Lundman.16 Conclusion We identified a need to improve the transfer of information from the personnel to the patients and their next of kin, especially in the period from hospital admission to the point where the patients are fit to communicate themselves and at the time of discharge from the hospital. Interestingly, the participants expressed both positive and negative aspects of the ban on visitors. Therefore, we propose individualized routines for visits to acute surgical patients when possible. Author Contributions Conception and design: E.T., A.F., and A.Ä. Data collection: E.T. and A.Ä. Analysis and interpretation: E.T. Writing the article: E.T., A.F., and A.Ä. Critical revision of the article: E.T., A.F., and A.Ä. Final approval of the article: E.T., A.F., and A.Ä. Obtained funding: E.T. Overall responsibility: A.Ä. Disclosure The authors report no conflicts of interest. The authors alone are responsible for the content and the writing of this article. Funding This work was supported by Stiftelsen Uppsala Sjuksköterskehem and by a donation from Kjell Gunnar Sten & Anne-Lie Rydé. The funding sources were not involved in the study. Supplementary Data Appendix 1 Acknowledgments We thank Siri Hellberg for assistance with transcription. Supplementary data related to this article can be found at https://doi.org/10.1016/j.jss.2022.04.014. ==== Refs References 1 Fagerdahl A.M. Torbjörnsson E. Gustavsson M. Älgå A. Moral distress among operating room personnel during the COVID-19 pandemic: a qualitative study J Surg Res 273 2021 110 118 35033820 2 Fisher N.D. Bi A.S. Aggarwal V. Leucht P. Tejwani N.C. McLaurin T.M. A Level 1 Trauma Center's response to the COVID-19 pandemic in New York City: a qualitative and quantitative story Eur J Orthop Surg Traumatol 31 2021 1451 1456 33616766 3 Brown W.A. Moore E.M. Watters D.A. Mortality of patients with COVID-19 who undergo an elective or emergency surgical procedure: a systematic review and meta-analysis ANZ J Surg 91 2021 33 41 33369009 4 Collaborative C. Mortality and pulmonary complications in patients undergoing surgery with perioperative SARS-CoV-2 infection: an international cohort study Lancet 396 2020 27 38 32479829 5 Cano-Valderrama O. Morales X. Ferrigni C.J. Acute care surgery during the COVID-19 pandemic in Spain: changes in volume, causes and complications. A multicentre retrospective cohort study Int J Surg 80 2020 157 161 32679205 6 Scheijmans J.C.G. Borgstein A.B.J. Puylaert C.A.J. Impact of the COVID-19 pandemic on incidence and severity of acute appendicitis: a comparison between 2019 and 2020 BMC Emerg Med 21 2021 61 33980150 7 Lidin M. Lyngå P. Kinch-Westerdahl A. Nymark C. Patient delay prior to care-seeking in acute myocardial infarction during the outbreak of the coronavirus SARS-CoV2 pandemic Eur J Cardiovasc Nurs 20 2021 752 759 34718511 8 Johnson C.L. Schwartz H. Greenberg A. Patient perceptions on barriers and facilitators to accessing low-acuity surgery during COVID-19 pandemic J Surg Res 264 2021 30 36 33744775 9 WHO Maintaing Essential Health Services: Operational Guidance for the COVID-19 Context 2020 10 Available at: https://www.who.int/publications/i/item/WHO-2019-nCoV-essential_health_services-2020.2 10 Hugelius K. Harada N. Marutani M. Consequences of visiting restrictions during the COVID-19 pandemic: an integrative review Int J Nurs Stud 121 2021 104000 34242976 11 Zeh R.D. Santry H.P. Monsour C. Impact of visitor restriction rules on the postoperative experience of COVID-19 negative patients undergoing surgery Surgery 168 2020 770 776 32943203 12 Pietrzak J.R.T. Maharaj Z. Erasmus M. Sikhauli N. Cakic J.N. Mokete L. Pain and function deteriorate in patients awaiting total joint arthroplasty that has been postponed due to the COVID-19 pandemic World J Orthop 12 2021 152 168 33816142 13 Rivard S.J. Vitous C.A. Cocroft S. Colorectal surgery patient perspectives on healthcare during the CoVID-19 pandemic Am J Surg 222 2021 759 765 33812662 14 Patton M.Q. Qualitative Research & Evaluation Methods 2002 SAGE London 15 World Medical Association World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects JAMA 310 2013 2191 2194 24141714 16 Graneheim U.H. Lundman B. Qualitative content analysis in nursing research: concepts, procedures and measures to achieve trustworthiness Nurse Educ Today 24 2004 105 112 14769454 17 Gibbs G. Thematic coding and categorizing Analyzing Qualitative Data 2nd ed. 2018 SAGE Publications Ltd London Available at: https://methods.sagepub.com/book/analyzing-qualitative-data-2e 18 Graneheim U.H. Lindgren B.-M. Lundman B. Methodological challenges in qualitative content analysis: a discussion paper Nurse Educ Today 56 2017 29 34 28651100 19 Hessheimer A.J. Morales X. Ginestá C. Where have all the appendicitis gone? Patterns of urgent surgical admissions during the COVID19 pandemic Br J Surg 107 2020 e545 e546 32866298 20 Gao Z. Li M. Zhou H. Complicated appendicitis are common during the epidemic period of 2019 novel coronavirus (2019-nCoV) Asian J Surg 43 2020 1002 1005 32863144 21 Orthopoulos G. Santone E. Izzo F. Increasing incidence of complicated appendicitis during COVID-19 pandemic Am J Surg 221 2021 1056 1060 33012500 22 Lee G. Clough O.T. Walker J.A. Anakwe R.E. The perception of patient safety in an alternate site of care for elective surgery during the first wave of the novel coronavirus pandemic in the United Kingdom: a survey of 158 patients Patient Saf Surg 15 2021 11 33712059 23 Doglietto F. Vezzoli M. Biroli A. Anxiety in neurosurgical patients undergoing nonurgent surgery during the COVID-19 pandemic Neurosurg Focus 49 2020 E19 24 Shannon A.B. Roberson J.L. Clapp J.T. What is the patient experience of surgical care during the coronavirus disease 2019 (COVID-19) pandemic? A mixed-methods study at a single institution Surgery 170 2020 550 557 33715849 25 Feder S. Smith D. Griffin H. “Why Couldn't I Go in To See Him?” Bereaved families' perceptions of end-of-life communication during COVID-19 J Am Geriatr Soc 69 2021 587 592 33320956 26 Ericsson A. Carlson E. Ching S.S.-Y. Molassiotis A. Kumlien C. Partners' experiences of living with men who have screening-detected abdominal aortic aneurysms: a qualitative descriptive study J Clin Nurs 29 2020 3711 3720 32619284 27 Althubaiti A. Information bias in health research: definition, pitfalls, and adjustment methods J Multidiscip Healthc 9 2016 211 217 27217764
PMC009xxxxxx/PMC9005364.txt	==== Front J Geriatr Oncol J Geriatr Oncol Journal of Geriatric Oncology 1879-4068 1879-4076 Elsevier Ltd. S1879-4068(22)00080-7 10.1016/j.jgo.2022.04.006 Article Remaining Agile in the COVID-19 pandemic healthcare landscape – How we adopted a hybrid telemedicine Geriatric Oncology care model in an academic tertiary cancer center Chen Matthew a Mohd Said Noorhanah b Mohd Rais Nydia Camelia c Ho Francis d Ling Natalie a Chun Meiling e Ng Yean Shin f Eng Wan Nghee b Yao Yao g Korc-Grodzicki Beatriz h Pang Angela f⁎ a Division of Geriatric Medicine, Department of Medicine, National University Hospital, Singapore b Department of Oncology Nursing, National University Cancer Institute, Singapore c Division of Geriatric Medicine, Department of Medicine, Ng Teng Fong General Hospital, Singapore d Department of Radiation Oncology, National University Cancer Institute, Singapore e Department of Surgery, Ng Teng Fong General Hospital, Singapore f Department of Haematology Oncology, National University Cancer Institute, Singapore g Department of Pharmacy, National University Hospital, Singapore h Department of Geriatrics, Memorial Sloan Kettering Cancer Center, New York ⁎ Corresponding author at: Department of Haematology-Oncology, National University Cancer Institute, Singapore (NCIS), Tower Block Level 7, 1E Kent Ridge Road, 119228, Singapore. 13 4 2022 13 4 2022 19 1 2022 21 2 2022 7 4 2022 © 2022 Elsevier Ltd. All rights reserved. 2022 Elsevier Inc. Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Introduction The COVID-19 pandemic has impacted healthcare on an unprecedented scale, with healthcare resources being channeled into managing the devastating effects of the outbreak. Healthcare provision for vulnerable older adults has also been affected by lockdowns and suspension of selected medical services worldwide. In our tertiary cancer center, the National University Cancer Institute, Singapore (NCIS), our Geriatric Oncology (GO) service for older adults with cancer was halted for five months. In this paper, we describe the adoption of a hybrid telemedicine model by our GO service to continue care provision for older adults in the midst of the pandemic. Materials and Methods Comprehensive geriatric assessments (CGA) were done via telemedicine and virtual multidisciplinary discussions were held prior to the patients' clinic visits. A hybrid telemedicine consultation allowed geriatricians and oncologists, segregated in different sites during the pandemic, to provide a hybrid physical and video geriatric oncology consultation. Scheduled phone follow ups by GO nurses helped to monitor patients for treatment-related toxicities and geriatric syndromes. Results Two hundred fifty patients were enrolled in the program from July 2020 to August 2021. All were assessed with a CGA, with 240 receiving interventions in the one-stop clinic. The average amount of time spent per visit was shortened from four hours to two and a half hours with an average of three interventions on the same day, versus one previously. Of the patients who received interventions, 84.8% were satisfied with the hybrid telemedicine model and 80.8% of them had reported a maintained or improved quality of life after being enrolled in the program. Discussion Telemedicine has been widely adopted during the pandemic, but older adults with limited digital literacy may find it a challenge. Our hybrid telemedicine model has allowed us to continue to provide cancer care, identify issues brought about by social isolation, and render timely assistance. It has become imperative to adapt, prepare and plan for the challenges we may face amid the ongoing COVID-19 pandemic and similar future outbreaks. Only by doing so can we remain agile and resilient, to continue providing quality care to our older patients with cancer. Keywords Geriatric oncology COVID19 Telemedicine Pandemic Telehealth ==== Body pmc1 Introduction The World Health Organization (WHO) declared the coronavirus disease 2019 (COVID-19) outbreak a pandemic on 11 March 2020. To date, COVID-19 has affected more than 230 regions/countries, with more than 242 million confirmed cases, and close to five million deaths globally [1]. The impact of the disease is multifaceted and devastating on a global scale, with countries adopting varying strategies and approaches to management as the understanding of the disease evolves, and vaccinations and treatments become available. Older adults with COVID-19 are at particularly risk of adverse outcomes including severe illness, hospitalization, and death [2,3]. Those who are frail or have underlying cancer or multimorbidity are more likely to develop critical illness [4]. Public health measures such as safe distancing, hand hygiene, and mask wearing are key to safeguarding this vulnerable population. These measures are especially relevant for countries with rapidly aging populations like Singapore, a multi-ethnic independent city-state with the second densest population in the world. Leveraging prior experience following the Severe Acute Respiratory Syndrome (SARS) epidemic in 2003, Singapore entered the preparedness phase, which saw the development of communicable disease detection and surveillance programs, the planning of infrastructure including healthcare facilities that can cope with a major infectious disease outbreak, and the stockpiling of medications and personal protective equipment (PPE) [5]. This was followed by the readiness phase when the first guidance was distributed by the Ministry of Health (MOH) in January 2020, and various processes in the healthcare system were reviewed, with disease testing put in place when it became available. In early February 2020, Singapore entered the response phase with heightened surveillance, border control, containment protocols, and widespread contact tracing. A nationwide “Circuit Breaker” measure was implemented from 7 April 2020 to 1 June 2020 with shutdown of multiple services to further curb community transmission. Our healthcare system had to actively reduce physical ambulatory clinic visits and turn to alternatives such as telemedicine. This led to a direct impact on our tertiary cancer center, the National University Cancer Institute, Singapore (NCIS), which is a comprehensive academic cancer center that manages 7000 adult and pediatric outpatients per month [6] While telemedicine is an effective solution to continuing medical care during the pandemic, there are also challenges with implementation given lower levels of digital literacy in a proportion of older adults. The resilience of our health system, which is the ability to prepare for and respond to crises by reorganizing systems to manage new conditions while maintaining core functions, was put to the test [6]. In this paper, we will describe how our Geriatric Oncology (GO) service was disrupted, and how we adapted our care model using a hybrid telemedicine workflow in order to remain agile and resilient in the pandemic, and continue to provide quality cancer care to our older patients with cancer. With more than 80% of the population now vaccinated against COVID-19 and the announcement on 9 October 2021 of Singapore heading towards COVID-resiliency through the streamlining of protocols in the healthcare system [7], we will also share our plans moving forward in order to provide a sustainable and effective service for older adults with cancer in the post-vaccination era. 2 The Geriatric Oncology Longitudinal End to eNd (GOLDEN) Program The NCIS GO service is made up of a multidisciplinary team comprising geriatricians, oncologists, advanced nurse practitioner, nurses, care coordinators, pharmacists, medical social workers (MSW), physiotherapists (PT), occupational therapists (OT), and dietitians. The Geriatric Oncology Longitudinal End to End (GOLDEN) Program was started in two restructured hospitals, NCIS and Ng Teng Fong General Hospital (NTFGH) in 2019. It is a service providing end-to-end clinical care from cancer diagnosis to treatment by the NCIS GO team. The program comprises:1. A Comprehensive Geriatric Assessment (CGA) of an older patient's health status and fitness for cancer treatment, 2. A Management and Innovation for Longevity in Elderly Surgical Patients (MILES) Surgical pre-rehabilitation program for patients planned for cancer surgery, 3. Tailored interventions by a multidisciplinary team of geriatricians, oncologists, oncology nurses, care coordinators, medical social workers, dietitians, pharmacists, physiotherapists and occupational therapists, 4. Transfer of care to the NCIS GO team post-surgery with a reassessment CGA post-surgery prior to initiation of chemotherapy or systemic therapy. 5. Telemedicine follow-up by nurses on patients' health status and monitoring of treatment side effects during cancer treatment. 3 GOLDEN Workflow Prior to the Pandemic All patients with cancer aged 65 years and above who were seen in either NCIS or NTFGH would be screened on their first visit with a Geriatric 8 (G8) questionnaire [8,9] to identify patients who might benefit from a CGA with directed interventions and care. An electronic memo would be sent to the primary oncologist (surgical oncology, medical oncology and radiation oncology) through the hospital's electronic medical records if the G8 score ≤ 14. Additionally, the primary oncologists could also directly refer their patients to the GO team for a geriatric oncology consultation. Patients who were referred to the GOLDEN program would undergo a CGA during their initial visit to the NCIS and NTFGH GO Clinics.1. Patients and the accompanying next-of-kin or caregiver would be interviewed by the GO case manager or nurse on-site and undergo several screening assessments. 2. This was followed by a medication review and reconciliation by the GO pharmacist. 3. The patient would then proceed with a medical consultation with both an oncologist and a geriatrician. 4. A referral would be made to the allied health professionals by the doctors if necessary, who would review the patient thereafter. This entire process was time-consuming and tiring for the older person and their caregivers as each visit could last four hours or more. Consequently, a significant number of patients declined referrals to allied health, which were planned only after the doctors' consultation. When the COVID-19 pandemic first broke out in 2020, NCIS had to adopt a segregated team workflow and cease inter-hospital staff movement to minimize transmission risk. The GOLDEN program had to be suspended for five months between January 2020 to June 2020 as a result. As these restrictions gradually lifted, the GOLDEN program in NUH and NTFGH resumed in July 2020. Over the subsequent months, restrictions in inter-hospital movement of healthcare workers and segregation of healthcare teams within the hospital were intermittently imposed, depending on the epidemiological trend of COVID-19 infection nationwide. The GO multidisciplinary team had to be agile and continuously modified existing workflows to continue caring for the geriatric oncology population in two healthcare institutions amidst a changing landscape. 4 Materials and Methods 4.1 Adopting Telemedicine for the GOLDEN Care Model during the COVID-19 Pandemic 4.1.1 Pre-Clinic Assessment and Medication Reconciliation conducted via Telemedicine Telemedicine was adopted to minimize time spent by patients and their caregivers in the GO clinic (Fig. 1 ). In Singapore, telemedicine refers to the provision of healthcare services using information and communications technology across physically separate environments, and it includes the exchange of information for clinical purposes between healthcare providers and patients/caregivers over telephone, text messages, and video/audio platforms [10].Fig. 1 Change in clinic assessment workflow before (left) and after (right) telemedicine adoption. Fig. 1 A selected portion of the CGA was conducted remotely via a telephone or video call for selected patients by the GO nurse, coordinator and pharmacist up to two days prior to the GO clinic appointment. The GO coordinator contacted the patient prior to the teleconsult to arrange a suitable timing for the consult with the patients and their caregivers. Verbal consent was obtained and the patient/caregiver was interviewed on his/her medical comorbidities, adherence to medications, functional ability, nutritional and emotional status and social circumstances. This telemedicine pre-clinic evaluation took approximately 30 to 45 min to complete. The GO pharmacist would also obtain the patient's active medication list and vaccination status from the National Electronic Health Records (NEHR), and check if the patient was taking the medications as prescribed. With the information from the initial assessment, a patient's risk of chemotherapy related toxicities would be calculated using the Cancer and Aging Research (CARG) [11,12] or Cancer and Aging Research - Breast Cancer (CARG-BC) [13] chemotoxicity calculators. Objective assessments that could only be done in person (e.g. cognitive assessments, timed up and go (TUG), postural blood pressure measurement) were performed on the day of the GO clinic consultation itself to complete the remainder of the CGA. 4.1.2 Virtual Pre-clinic GO Multidisciplinary Team discussion A weekly multidisciplinary team meeting, using a secure online video communication platform, was held before every GO clinic session. The team discussed all patients who completed the pre-clinic telemedicine assessment and made preliminary plans for every patient before they were physically seen. After the initial assessment, the cases were again discussed in the virtual team meeting prior to their follow-up visit, which typically occurred within a month. 4.1.3 Hybrid Telemedicine Clinic Model Originally, the doctors had to shuttle between sites to run GO clinics in two geographically separate healthcare institutions. During the pandemic, restrictions were imposed on the movement of healthcare workers between hospitals in Singapore to minimize transmission risk, which posed a significant challenge to the program. A hybrid tele-consultation model was piloted and implemented to overcome this limitation. Instead of having both a geriatrician and an oncologist present physically on site, only one doctor was required to be physically present during the clinic consultation with the patient. The other doctor would join in the consultation remotely via a video communication platform set up concurrently in the physical clinic. The patient would hence be able to consult both the geriatrician and oncologist in the same clinic session and be involved in the joint decision-making process. In this way, this hybrid teleconsultation model allowed the GOLDEN program to continue supporting patients during the pandemic despite the movement restrictions imposed on healthcare workers between institutions. 4.1.4 Scheduled GO Nurse Phone follow-up on Treatment-Related Toxicities and Geriatric Syndromes During the initial four months from the commencement of cancer therapy, the GO nurses performed follow-up phone calls every two weeks to assess for treatment-related toxicities graded with the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0 [14]. The patients and caregivers were also assessed to ensure that they were coping well in the community. Each call lasted approximately twenty minutes. GO patients who developed grade 3 or higher toxicities and worsening frailty during the phone follow-ups would be identified by the GO nurses, who then informed their primary oncologists for an early review in the clinic. If patients had grade 1–2 toxicities, they were given advice on how to manage and red flags to watch out for. As there were no follow-up phone calls by the treating oncology team, it is crucial for our GO nurses to detect the grade 3 toxicities, as older adults have limited physiological reserves and may deteriorate rapidly if timely treatment is not instituted. Patients who were deemed to be at a higher risk of grade 3–5 toxicities continued to be followed up by the GO nurses beyond the first four months of treatment. Our nurses also evaluated patients' quality of life using the EORTC QLQ-C30 questionnaire [15] in the first and third month to ensure that they were coping appropriately and adapting well to any treatment-induced changes to their lifestyle. An overview of the GO hybrid telemedicine workflow is shown below in Fig. 2 .Fig. 2 Overview of Geriatric Oncology Hybrid Telemedicine Clinic Visit and Follow-Up. Fig. 2 This study was approved by the National University Hospital (NUH) Institutional Ethics Review Board. 5 Results The GOLDEN program resumed in NCIS in July 2020 and in NTFGH in August 2020. Over a period of one year since resuming the program during the pandemic, 690 patients were screened with G8, of which 510 (73.9%) patients were eligible and 250 (36.2%) patients were referred to the GOLDEN program (Fig. 3 ).Fig. 3 Screening and recruitment of patients into the GOLDEN program. Fig. 3 Table 1 shows the characteristics of the patients referred to the GOLDEN program. The mean age was 75.3 +/− 6.2 years. Most of the patients were Chinese (86.8%). Two-thirds of the patients had early stage cancer, with lower gastrointestinal cancers being the most commonly diagnosed tumor type. Almost three-quarters were pre-frail (72.8%), one-fifth were frail (19.2%) and the remainder were fit (8.0%). The median G8 score was 11 (range 2–16). About three-quarters of the patients (73.8%) received treatment, of which 134 (54.0%) were of curative intent and 49 (19.8%) were of palliative intent.Table 1 Patient characteristics. Table 1Characteristics Patients referred to GOLDEN program (n = 250) Age 65–69 years 34 (13.6%) 70–74 years 93 (37.2%) 74–79 years 59 (23.6%) 80–84 years 47 (18.8%) ≥ 85 years 17 (6.8%) Age, mean 75.3 (±6.2 years) Gender Female 117 (46.8%) Male 133 (53.2%) Ethnicity Chinese 217 (86.8%) Malay 23 (9.2%) Indian 6 (2.4%) Others 4 (1.6%) Tumor type Lower gastrointestinal 102 (40.8%) Hepatobiliary 42 (16.8%) Thoracic 31 (12.4%) Genitourinary 20 (8.0%) Breast 17 (6.8%) Head and neck 13 (5.2%) Upper gastrointestinal 10 (4.1%) Gynecological 6 (2.4%) Sarcoma 5 (2.0%) Others 5 (2.0%) Metastatic disease No 171 (68.4%) Yes 79 (31.6%) Level of education None 31 (12.4%) Primary 116 (46.4%) Secondary 78 (31.2%) Tertiary 25 (10.0%) Living arrangements Alone 21 (8.4%) With family 221 (88.4) With others 8 (3.2%) G8 score, mean (SD) 11 (2.64) Frailty state Fit 20 (8.0%) Pre-frail 182 (72.8%) Frail 48 (19.2%) Abbreviations: G8: Geriatric 8 questionnaire, SD: standard deviation. The majority of patients (96.0%) required upstream interventions by our multidisciplinary allied health team of pharmacists, medical social workers, physiotherapist, and occupational therapist as recommended in the one-stop GOLDEN clinic (Table 2 ).Table 2 Patients with interventions by allied health professionals in the GOLDEN program. Table 2Interventions Allied health professionals Patients (N = 250) Medication reconciliation and review Pharmacist 237 (95.6%) Psychosocial and financial support Medical social worker 162 (65.3%) Nutrition advice Dietician 126 (50.8%) Falls prevention and activity adaptation Occupational therapist 72 (29.0%) Exercise recommendation Physiotherapist 70 (28.2%) Swallowing assessment and management Speech therapist 4 (1.6%) The average amount of time spent in the hospital per visit was shortened from four hours to two and one-half hours, with an average of three interventions on the same day versus one intervention previously in the one-stop clinic. A proportion of the older adults were unable to participate in telehealth due to barriers such as impaired hearing, cognitive impairment, and limited digital literacy, but this was overcome by taking a corroborative history from their caregivers. The majority of the patients who received interventions in the GOLDEN program (84.8%) were satisfied with the hybrid telemedicine model. The 227 patients who received interventions were followed up for quality of life (QOL) assessment using the EORTC QLQ-C30 questionnaire, of which 70 patients (33.9%) reported an overall improvement in their global health status, while 132 patients (58.1%) maintained their global health status after being enrolled in the GOLDEN program. 6 Discussion The population of older adults with cancer is heterogeneous and requires a careful and tailored approach to care that considers frailty, a syndrome which in itself has been associated with an elevated risk of adverse outcomes [16,17]. The COVID-19 pandemic has disproportionately impacted older adults globally due to ageism, social isolation, loneliness, and higher threats of illness including mental health problems, leading to multiple unmet needs [18,19]. With safe distancing measures and restrictions on ambulatory healthcare and rehabilitation services, the pandemic also created barriers to geriatric oncology care by potentially leading to cancelled, missed, or delayed appointments, making the performance of comprehensive geriatric assessments challenging and causing treatment delays [18]. Our GOLDEN multidisciplinary team had to rapidly adopt a hybrid telemedicine model of care to overcome these barriers in order to continue providing our service to older adults with cancer during the pandemic. We have learnt that well-established geriatric oncology centers like the University of Rochester SO-CARE clinic had also adapted a telehealth geriatric assessment model in response to the COVID-19 pandemic. [20] Despite screening all older adults in our cancer center for frailty with G8, only about half of the eligible patients were referred to the GOLDEN program. This could be attributed to the slow recognition of the need of a GO service by the primary physicians and patient factors, as many were reluctant to return for an additional visit to the hospital during the pandemic. With a selected portion of the CGA done prior to the physical clinic visit, and with both geriatricians and oncologists co-consulting in the hybrid telemedicine model, the amount of time spent by the patients and their caregivers in the clinic was shortened. This minimized the potential exposure and infection risk to patients and caregivers. The time saved also allowed for additional tailored interventions by other allied health professionals to be carried out in the same visit to the one-stop clinic. This saved the patients and their caregivers direct and indirect costs of multiple visits for assessments by allied health professionals, leading to improved value driven outcomes. Anxiety, while a normal response to a perceived threat, has been shown to be prevalent in older adults with cancer and is known to rise in response to additional threats such as surgery and disease progression [21,22]. In the context of COVID-19, with potential delays in diagnosis and initiation of treatment, disruptions in ongoing treatment and supportive care, this issue has become all the more relevant [23]. By engaging with patients and their caregivers earlier in the treatment journey starting with a telephone or video call, the GO team members were able to establish rapport earlier and elicit and address their concerns sooner. We have observed more than half of our patients reporting an improvement in their quality of life, though we are unable to make positive associations due to limitations of our study. While we have found some success with our hybrid telemedicine model for the GOLDEN program in terms of reduction of time spent in clinic and cost, and a high level of patient satisfaction, we also faced some limitations. Some older adults found it difficult to comprehend some of the interview questions that were posed over telemedicine owing to barriers such as impaired hearing and vision, cognitive impairment, and limited digital literacy. We overcame these limitations as best as we could by engaging family members and caregivers who could assist our older patients during telemedicine sessions.Our study was also not designed to investigate how these barriers affected participation in telemedicine in our population, and this should be looked at in future studies. Of note, digital literacy remains a significant challenge amongst older adults globally, with the proportion of older adults using digital technology being less than younger adults, though it is rising exponentially [24]. According to a 2019 survey by the Infocomm Media Development and Authority (IMDA), which is a statutory board under the Singaporean government, 58% of Singaporean residents aged 60 and above were internet users compared to 89% for all residents [25]. These figures are comparable to the findings of the Organization for Economic Cooperation and Development (OECD), with 63% of internet users being aged 55–74 years compared to 97% in 16–24-year-olds [26]. There has been a nationwide push to strengthen digital literacy across all age groups through the Digital Media and Information Literacy Framework as part of Singapore's Digital Readiness vision to be a Smart Nation, which has accelerated due to the pandemic [27]. The IMDA launched a “Seniors Go Digital” initiative at the height of the pandemic in May 2020 to equip older adults with skills such as using digital tools for video calls, with digital ambassadors being made readily available for them to do so [28], which will hopefully help them to overcome the barrier of digital illiteracy in the future. As part of the “Seniors Go Digital” initiative, a “Mobile Access for Seniors” scheme has also been put in place to provide subsidized smartphone and mobile plans to older adults who wish to go digital but are unable to afford them [29]. As we have begun to see the benefits of our hybrid telemedicine model during this pandemic, we perceive these initiatives to be an important step in the development of a pandemic-resilient healthcare system. Our hybrid telemedicine model shows promise in terms of improving patient satisfaction and efficiency of the clinic consult and will continue to evolve in the post-vaccination era as the population becomes more familiar with telemedicine. Factors that have been previously described, such as the provision of an overarching architecture and infrastructure, strong program management, and a thorough needs analysis, will be key to success leading to a sustainable model of care [30]. 7 Conclusion It has become imperative to adapt, prepare, and plan for the challenges that we may face amid the ongoing COVID-19 pandemic and similar future outbreaks. The GOLDEN program has been able to continue caring for our patients with the implementation of the hybrid teleconsultation model during this period. Our model of care will continue to evolve and be sustainable in the post-vaccination era as the use of telemedicine gains traction in our population. Ethnics Approval This study was approved by the National University Hospital (NUH) Institutional Ethnics Review Board. Consent for Publication Yes. Author Contribution Form as attached. Availability of Data and Materials Yes. Authors' Contribution Conception and design: Angela Pang and Matthew Chen. Data collection: Ng Yean Shin, Chun Meiling, Yao Yao and Eng Wan Nghee. Analysis and interpretation of data: Angela Pang, Matthew Chen, Nydia Camelia and Yao Yao. Manuscript writing: Angela Pang, Matthew Chen, Nydia Camelia, Natalie Ling, Francis Ho, Noorhanah Binte Mohammad, Yao Yao and Beatriz Korc-Grodzicki. Declaration of Competing Interest The authors indicated no potential conflicts of interest. Acknowledgements The GOLDEN program is funded by the Jurong Health Fund Grant. ==== Refs References 1 World Health Organisation WHO Coronavirus (COVID-19) Dashboard URL: https://COVID-19.who.int/ 2021 Accessed 25 October 2021 2 Shahid Z. Kalayanamitra R. McClafferty B. COVID-19 and older adults: what we know J Am Geriatr Soc 68 5 2020 926 929 32255507 3 Singhal S. Kumar P. Singh S. Saha S. Dey A.B. Clinical features and outcomes of COVID-19 in older adults: a systematic review and meta-analysis BMC Geriatr 21 1 2021 321 34011269 4 Cavalcanti I.D.L. Soares J.C.S. Impact of COVID-19 on cancer patients: a review Asia-Pac J Clin Oncol 17 2021 186 192 32970923 5 Fisher D. Mak K. Exiting the pandemic: Singapore style BMC Med 19 2021 238 34530835 6 Nuzzo J.B. Meyer D. Snyder M. What makes health systems resilient against infectious disease outbreaks and natural hazards? Results from a scoping review BMC Public Health 19 2019 1310 31623594 7 Government of Singapore Updates on the COVID-19 situation in Singapore URL: https://www.gov.sg/features/covid-19 2021 Accessed 25 October 2021 8 Bellera C. Rainfray M. Mathoulin-Pélissier S. Screening older cancer patients: first evaluation of the G-8 geriatric screening tool Ann Oncol 23 8 2012 Aug 2166 2172 22250183 9 Soubeyran P. Bellera C. Goyard J. Screening for vulnerability in older cancer patients: the ONCODAGE Prospective Multicenter Cohort Study PLoS One 9 12 2014 e115060 Published 2014 Dec 11 10 Ministry of Health Singapore National Telemedicine Guidelines 2015 11 Hurria A. Togawa K. Mohile S. Predicting chemotherapy toxicity in older adults with cancer: a prospective 500 patient multi-center study J Clin Oncol 29 25 2011 3457 3465 21810685 12 Hurria A. Mohile S.G. Gajra A. Validation of a prediction tool for chemotherapy toxicity in older adults with cancer J Clin Oncol 34 20 2016 2366 2371 27185838 13 Magnuson A. Sedrak M. Gross C. Development and validation of a risk tool for predicting severe toxicity in older adults receiving chemotherapy for early-stage breast cancer J Clin Oncol 39 6 2021 608 618 33444080 14 National Cancer Institute National Institutes of Health US Department of Health and Human Services Common terminology criteria for adverse events (CTCAE), Version 4.0. NIH publication 09–7473. Published May 29, 2009 Revised June 14 2010 15 Aaronson N.K. Ahmedzai S. Bergman B. Bullinger M. Cull A. Duez N.J. The European Organisation for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology J Natl Cancer Inst 85 1993 365 376 8433390 16 Clegg A. Young J. Iliffe S. Frailty in elderly people Lancet 381 9868 2013 752 762 23395245 17 Handforth C. Clegg A. Young C. The prevalence and outcomes of frailty in older cancer patients: a systematic review Ann Oncol 26 6 2015 1091 1101 25403592 18 Van Jaarsveld G. The effects of COVID-19 among the elderly population: a case for closing the digital divide Front Psych 11 2020 1211 19 Lebrasseur A. Fortin-Bédard N. Lettre J. Impact of the COVID-19 pandemic on older adults: rapid review JMIR Aging 4 2 2021 e26474 20 DiGiovanni G. Mousaw K. Lloyd T. Development of a telehealth geriatric assessment model in response to the COVID-19 pandemic J Geriatr Oncol 11 5 2020 761 763 32327321 21 Trevino K.M. Saracino R.M. Roth A.J. Symptomatology, assessment, and treatment of anxiety in older adults with cancer J Geriatr Oncol 12 2 2021 316 319 32565145 22 Stark D.P. House A. Anxiety in cancer patients Br J Cancer 83 10 2000 1261 1267 11044347 23 Büntzel J. Klein M. Keinki C. Oncology services in corona times: a flash interview among German cancer patients and their physicians J Cancer Res Clin Oncol 1–3 2020 24 Oh S.S. Kim K.A. Kim M. Measurement of Digital Literacy Among Older Adults: Systematic Review [published correction appears in J Med Internet Res. 2021 Mar 3;23(3):e28211] [published correction appears in J Med Internet Res. 2021 Jun 15;23(6):e30828] J Med Internet Res 23 2 2021 e26145 Published 2021 Feb 3 25 Infocomm Media Development and Authority. Annual survey on infocomm usage in households and by individuals for 2019. URL: https://www.imda.gov.sg/-/media/Imda/Files/Infocomm-Media-Landscape/Research-and-Statistics/Survey-Report/2019-HH-Public-Report_09032020.pdf. Assessed 1 December 2021. 26 OECD Secretariat OECD Digital Economy Outlook 2017. Organisation for economic cooperation and development (OECD) URL: https://www.oecd.org/digital/oecd-digital-economy-outlook-2017-9789264276284-en.htm 2017 Oct 11 27 Ministry of Communications and Information Digital media and information literacy framework URL: https://www.mci.gov.sg/literacy 2021 Assessed 1 December 2021 28 Infocomm Media Development and Authority Seniors Go Digital URL: https://www.imda.gov.sg/digitalforlife/About-Us 2021 29 Infocomm Media Development and Authority Mobile access for seniors URL: https://www.imda.gov.sg/ma 2021 30 Moehr J.R. Schaafsma J. Anglin C. Success factors for telehealth--a case study Int J Med Inform 75 10−11 2006 Oct-Nov 755 763 16388982
PMC009xxxxxx/PMC9005365.txt	==== Front J Infect Chemother J Infect Chemother Journal of Infection and Chemotherapy 1341-321X 1437-7780 Japanese Society of Chemotherapy and The Japanese Association for Infectious Diseases. Published by Elsevier Ltd. S1341-321X(22)00119-2 10.1016/j.jiac.2022.04.010 Original Article Epidemiology of SARS-CoV-2 infection in nursing facilities and the impact of their clusters in a Japanese core city Shimizu Koki abc Maeda Haruka de Sando Eiichiro f Fujita Ayumi g Tashiro Masato g Tanaka Takeshi g Izumikawa Koichi g Motomura Katsuaki a Morimoto Konosuke d∗ a Nagasaki City Public Health Center, Nagasaki, Japan b School of Tropical Medicine and Global Health, Nagasaki University, Nagasaki, Japan c Department of Clinical Medicine, Institute of Tropical Medicine, Nagasaki University, Nagasaki, Japan d Department of Respiratory Infections, Institute of Tropical Medicine, Nagasaki University, Nagasaki, Japan e Department of Clinical Tropical Medicine, Nagasaki University Graduate School of Biomedical Sciences, Nagasaki, Japan f Department of General Internal Medicine and Clinical Infectious Diseases, Fukushima Medical University, Fukushima, Japan g Infection Control and Education Center, Nagasaki University Hospital, Nagasaki, Japan ∗ Corresponding author. Department of Respiratory Infections, Institute of Tropical Medicine, Nagasaki University, 1-12-4, Sakamoto, Nagasaki, 852-8523, Japan. 13 4 2022 7 2022 13 4 2022 28 7 955961 28 12 2021 4 4 2022 8 4 2022 © 2022 Japanese Society of Chemotherapy and The Japanese Association for Infectious Diseases. Published by Elsevier Ltd. All rights reserved. 2022 Japanese Society of Chemotherapy and The Japanese Association for Infectious Diseases Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Introduction Nursing facilities are vulnerable to coronavirus disease 2019 (COVID-19) due to the congregate nature of their housing, the older age of the residents, and the variety of their geriatric chronic conditions. Little is known about the impact of nursing facility COVID-19 on the local health system. Methods We collected data of COVID-19 cases in Nagasaki city from April 15, 2020 to June 30, 2021. We performed universal screening of the healthcare workers (HCWs) and the users of nursing facilities, once the first case of COVID-19 was detected within that facility. The community-dwelling people received testing if they had symptoms or if they were suspected of having close contact with the positive cases. The epidemiological survey for each COVID-19 case was performed by the public health officers of the local public health center. Results Out of 111,773 community-dwelling older adults (age ≥ 65 years) and 20,668 nursing facility users in Nagasaki city, we identified 358 and 71 COVID-19 cases, and 33 and 12 COVID-19 deaths, respectively, during the study period. The incidence rate ratios (IRRs) for COVID-19 and its deaths among the nursing facility users were 1.07 (95% confidence interval (CI), 0.82–1.39) and 1.97 (95%CI, 0.92–3.91) compared with the community-dwelling older adults. Four clusters, which had more than 10 COVID-19 cases, accounted for 60% (65/109) of the overall cases by the HCWs and the users. Conclusions The prevention of COVID-19 clusters is important to reduce the number of COVID-19 cases and deaths among the nursing facility population. Keywords Coronavirus SARS-CoV-2 COVID-19 Long-term care Nursing homes Abbreviations ADL activities of daily living CI confidence interval COVID-19 coronavirus disease 2019 HCW healthcare worker IQR interquartile range IRR incidence rate ratio LAMP loop-mediated isothermal amplification LTCF long-term care facility MHLW Ministry of Health, Labour, and Welfare PCR polymerase chain reaction SARS-CoV-2 severe acute respiratory syndrome coronavirus 2 ==== Body pmc1 Introduction Nursing facilities suffer from a disproportionate impact of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) mediated disease, COVID-19 [1]. The congregate settings of the nursing facilities, the older ages of the residents, and their widespread comorbidities constitute high risk environments for COVID-19 prevalence and severity [2]. In the U.S, over 20% of COVID-19 deaths were attributed to nursing home residents [3]. Similarly, nursing home residents were reported to account for at least 14% of COVID-19 deaths in Japan during the earliest phase of the pandemic [4]. Large-scale outbreaks of COVID-19 in nursing homes have also been reported globally [[5], [6], [7]]. However, to date, little is known about the impact of COVID-19 clusters in nursing facilities on the local health system. Multiple strategies have been proposed to prevent the outbreak of COVID-19 in nursing facilities, including source control through the use of masks, enhancement of physical distancing measures, visitor restrictions, vaccination, symptom surveillance, and routine and outbreak testing [8]. However, infection control in a nursing facility remains challenging as described by the following examples: 1) residents with dementia or depression may have difficulty complying with the obligation of wearing masks and other countermeasures, 2) visitor restrictions and the resulting social isolation may aggravate both their physical and mental conditions, 3) symptom surveillance is inadequate because older residents may be asymptomatic or sometimes manifest atypical symptoms [9,10]. Vaccines for COVID-19 have so far greatly reduced the risk of infection, but the long-term effectiveness of the vaccines and their protective effects against emerging SARS-CoV-2 variants are still of concern [11,12]. Here, we aimed to describe the epidemiology of COVID-19 in nursing facilities and the impact of their clusters on a Japanese local health system through the epidemiological data of 43 nursing facilities in Nagasaki city between July 2020 and June 2021. 2 Methods 2.1 Setting Nagasaki city is the capital of Nagasaki prefecture, located on the northwest coast of Kyushu island. Its population is approximately 400,000 (as of January 2021, 31% of the 1.3 million people living in Nagasaki prefecture), and has the 15th largest population among the 62 core cities with populations over 200,000 [[13], [14], [15]]. The population of older adults who are ages above 65 is roughly 130,000 (33% of the population of Nagasaki city) [16]. Of those, 20,668 (16%) rely on care services, which are provided under long-term care insurance [17]. On April 15, 2020, Nagasaki city had its first case of COVID-19, three months after the first entry of COVID-19 into Japan [18]. Nagasaki city subsequently underwent four waves of a COVID-19 pandemic (Fig. 1 ), and the alpha variant (B.1.1.7) was introduced to Nagasaki city during the fourth wave. The delta variant (B.1.617.2) has not been detected in Nagasaki as of June 30, 2021.Fig. 1 COVID-19 cases in Nagasaki city from April 15, 2020 to June 30, 2021. The number of new COVID-19 cases in Nagasaki city is drawn in blue bars, and the number of new COVID-19 cases related to nursing facilities is drawn in orange bars. The green line indicates the bed occupancy of medical facilities in Nagasaki city in percentage. Information of bed occupancy was available only from December 15, 2020 to June 30, 2021. Nagasaki city underwent four waves of a COVID-19 pandemic since April, 2020. The peak of the pandemic in the nursing facility population coincides with the peak of the pandemic in the general population. Abbreviations: COVID-19, coronavirus disease 2019; HCWs, healthcare workers. Fig. 1 Since the beginning of the pandemic, the Ministry of Health, Labour, and Welfare (MHLW) in Japan has made various attempts to protect the nursing facilities. By January 31, 2020, MHLW informed the nursing facilities about infection control measures for residents and health care workers (HCWs) [19]. On February 24, 2020, MHLW informed the residential care facilities to consider visitor restrictions and lockdown of the facilities [20]. On May 17, 2021, the government recommended periodic testing of the HCWs to prevent them from introducing COVID-19 into the facilities [21]. In Japan, COVID-19 vaccines were provided in the order of priority listing. In Nagasaki city, COVID-19 vaccination was initiated for medical staffs on March 8, 2021, and for older adults on April 12, 2021. As of July 4, 2021 in Nagasaki city, 70% of the older adults have received the first dose of COVID-19 vaccine, and 27% have received the second dose. 2.2 Testing strategies In Nagasaki city, community-dwelling people received testing for COVID-19 either voluntarily through purchasing a commercial kit or at medical facilities, or at a testing facility of a local health center if they were recognized through epidemiological survey as a close contact of a COVID-19 case. All COVID-19 cases were confirmed with either polymerase chain reaction (PCR) or loop-mediated isothermal amplification (LAMP) assays [22,23]. PCR universal screening tests were performed at nursing facilities for the HCWs and users if a COVID-19 case was diagnosed within the facility. The targets for screening were determined by the capacity of the testing facility and the expected spread of COVID-19 within the facility based on the epidemiological survey. Screening tests were done within one to two days after the confirmation of the first case. 2.3 Definitions The term “nursing facility” includes the long-term care facilities (LTCFs), short-stay services (short-term admission in residential care facilities), home care services, and day care services. The LTCFs include welfare facilities for older adults (nursing homes), healthcare facilities for older adults, sanatorium-type medical care facilities for older adults, and adult care homes. The community-based services include short-stay services, home care services, and day care services. We included monasteries as nursing facilities because they accommodate older adults similar to the LTCFs. The term “users” of the nursing facilities indicates those who are ages 40 or above and receive care services under the long-term care insurance. We did not take into account those who receive care services without the qualification of the long-term care insurance because their activities of daily living (ADLs) are sufficient to not qualify for the insurance and are rare. People who are ages 40 to 64 can qualify for long-term care insurance if they are diagnosed with specific diseases such as early dementia, neurodegenerative disease, or stroke [17]. People who are ages 65 or above can qualify for long-term care insurance depending on the limitations in their ADLs. They can receive different types of care services at various nursing facilities under the long-term care insurance. The term “older adults” refers to those who are ages 65 or above. The term “community-dwelling older adults” indicates people who are ages 65 or above and do not rely on care services. The term “cluster” refers to two or more COVID-19 cases that are epidemiologically linked. For instance, if a HCW or a user visits several facilities and introduces COVID-19 to the facilities, we count the sum of all COVID-19 cases in the facilities as one cluster. 2.4 Data collection and sources We included data from the first case of COVID-19 in a nursing facility in Nagasaki city on July 15, 2020 onward to June 30, 2021. We collected information of age, sex, comorbidities, number of close contacts, date of symptom onset, date of diagnosis, cycle threshold value in real-time RT-PCR, status of hospital admission, duration of hospital stay, and the outcome of death for each individual from the epidemiological survey, which was performed by the Nagasaki city public health center under the Infectious Disease Control Law. We obtained the information on the number of positive cases in Nagasaki city and Nagasaki prefecture, their population by age groups, and the number of nursing facilities in Nagasaki city from the website of Nagasaki city and Nagasaki prefecture, which are publicly available. The information on the bed occupancy of medical facilities in Nagasaki city, the past outbreak of infections in the nursing facilities, the number of people who receive care services under the long-term care insurance, the vaccine status of the older adults in Nagasaki city, and the details of the COVID-19 outbreak investigation at nursing facilities were obtained from the records of the Nagasaki city public health center. 2.5 Statistical analysis We estimated incidence rate ratios (IRRs) of COVID-19 cases and deaths in the nursing facility users compared with those in the community-dwelling older adults who were ages 65 or above. We used Fisher's exact test to compare the proportions of COVID-19 occurrence in each facility type (LTCFs, day care services, home care services, and short-stay services). We used Kruskal-Wallis H test to compare the median numbers of COVID-19 cases in each facility type (α = 0.05). All statistical analyses were conducted using STATA ver. 16.1 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.). 2.6 Ethics statement The data collection and analysis were conducted under the Infectious Diseases Control Law in Japan. The ethical review board of the Institute of Tropical Medicine, Nagasaki University waived the need for informed consent based on the condition that the anonymities of each individual and the facility were assured. Publishing this study was approved by the ethical committee of the Institute of Tropical Medicine, Nagasaki University (No. 210603261). 3 Results 3.1 Overview of COVID-19 cases in Nagasaki city From April 15, 2020 to June 30, 2021, Nagasaki city underwent four waves of a COVID-19 pandemic (Fig. 1). During that period, Nagasaki city had 1527 COVID-19 cases (48% of the 3197 cases in Nagasaki prefecture), and 160 cases of those originated in nursing facilities [24,25]. Fig. 1 demonstrates that the peaks of the pandemic within the general population and in the nursing facilities occurred almost simultaneously. The incidence rates of COVID-19 and death in Nagasaki city were 311/100,000 person-year and 9.4/100,000 person-year (Table 1 ). The number of COVID-19 cases was 429 for older adults (28% of the overall cases, 429/1527), 358 for the community-dwelling older adults (23% of the overall cases, 358/1527), and 71 for the nursing facility users (4.6% of the overall cases, 71/1527). The IRRs of COVID-19 and its deaths among the nursing facility users was 1.07 (95% confidence interval (CI), 0.82–1.39) and 1.97 (95%CI, 0.92–3.91) compared with the community-dwelling older adults.Table 1 IRRs of COVID-19 and its death in Nagasaki city by different age groups and status of the nursing facility use. Table 1 Population (no.) COVID-19 cases (no.) Incidence rate (/100,000 person-year) IRR (vs community-dwelling) (95%CI) Total population 406,313 1527 311.0 1.17 (1.05–1.32) ≤ 64 years 267,204 1096 339.5 1.28 (1.14–1.45) ≥ 65 years 132,441 429 268.1 1.01 (0.88–1.17) Nursing facility users 20,668 71 284.3 1.07 (0.82–1.39) Community-dwelling older adults 111,773 358 265.1 1 Population (no.) COVID-19 deaths (no.) Incidence rate (/100,000 person-year) IRR (vs community-dwelling) (95%CI) Total population 406,313 46 9.37 0.38 (0.24–0.62) ≤ 64 years 267,204 1 0.31 0.01 (0.0003–0.08) ≥ 65 years 132,441 45 28.1 1.15 (0.72–1.86) Nursing facility users 20,668 12 48.1 1.97 (0.92–3.91) Community-dwelling older adults 111,773 33 24.4 1 The population data is based on Japan national census data as of October 1, 2020. Age data of 6668 people in Nagasaki city and 2 COVID-19 cases are missing. The follow-up period from April 15, 2020 to June 30, 2021 (14.5 months) was used to calculate the incidence rate. Abbreviations: IRR, incidence rate ratio; COVID-19, coronavirus disease 2019; CI, confidence interval. 3.2 Overview of COVID-19 cases in the nursing facilities in Nagasaki city Since the first case of COVID-19 in a nursing facility on July 15, 2020, 160 COVID-19 cases were identified that were either of the users (n = 71), the HCWs (n = 38), or their family members (n = 51). They accounted for 10% of the overall cases in Nagasaki city (160/1527), 16% of all hospitalized cases (101/629), and 28% of COVID-19 deaths (13/46). The average days of hospital stay was 16.6 days for the nursing facility users, while they were 14.5 days for the overall older adults in Nagasaki city. No positive case had a previous administration of a COVID-19 vaccine. The proportions of hospital admission and the case fatality of the users were 100% and 17% (Table 2 ). There was no occasion in which the users could not be admitted to a hospital due to a high bed occupancy.Table 2 Demographics and clinical outcomes of COVID-19 cases among the users of nursing facilities from July 15, 2020 to June 30, 2021. Table 2 Users (n = 71) Age - median (IQR), years 86 (81–90) Female - no. (%) 52 (73.2) Comorbidities - no. (%) Dementia 23 (32.4) Hypertension 36 (50.7) Diabetes mellitus 16 (22.5) Cardiac disease 22 (31.0) Cerebrovascular disease 25 (35.2) Malignancy 12 (16.9) Renal disease 2 (2.8) Pulmonary disease 5 (7.0) Hospitalized - no. (%) 71 (100.0) Admission days - no. (%) ≤ 10 days 13 (18.3) 11– - 30 days 46 (64.8) ≥ 31 days 10 (14.1) Unknown 2 (2.8) Death - no. (%) 12 (16.9) Data presented in median (interquartile range) and no. (%). Abbreviations: COVID-19, coronavirus disease 2019; IQR, interquartile range. Table 3 shows the transmission characteristics of SARS-CoV-2 among the users and the HCWs in the nursing facilities. The number of identified close contacts from one COVID-19 case was one (range 0–32) when the COVID-19 case was a user, and three (range 0–32) when the COVID-19 case was a HCW. However, 82% (58/71) of users and 82% (31/38) of HCWs were not previously recognized as the close contact of another COVID-19 case before testing positive for SARS-CoV-2. The users were asymptomatic in 23 of 71 cases (32.4%) when diagnosed with COVID-19.Table 3 Transmission characteristics of SARS-CoV-2 among the users and the HCWs of nursing facilities from July 15, 2020 to June 30, 2021. Table 3 Users (n = 71) HCWs (n = 38) The number of identified close contacts from one COVID-19 case - median (range) 1 (0–32) 3 (0–32) The number of COVID-19 cases among close contacts- median (range) 0 (0–11) 0 (0–2) Asymptomatic when diagnosed with COVID-19 - no. (%) 23 (32.4) 5 (13.2) Asymptomatic throughout the observation period - no. (%) a 4 (7.1) 0 (0.0) Data presented in median (range) and no. (%). For “asymptomatic throughout the observation period”, the number of cases and percentages were calculated irrespective of missing data. Abbreviations: SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; HCWs, healthcare workers; COVID-19, coronavirus disease 2019. a The data of 15 users and 3 HCWs were missing for “asymptomatic throughout the observation period.” There were 43 nursing facilities, which had more than one COVID-19 case during the study period. Of the 43 facilities, 17 were LTCFs (40%), 14 were day care services (33%), 6 were home care services (14%), 3 were short-stay services (7%), and 3 were monasteries (7%). Out of 277 LTCFs, 257 day care services, 173 home care services, and 94 short-stay services in Nagasaki city, COVID-19 occurred in 6% of LTCFs (17/277), 5% of day care services (14/257), 3% of home care services (6/173) and 3% of short-stay services (3/94) [26,27]. There was no statistical difference among the four proportions (p-value = 0.56). The number of COVID-19 cases among the HCWs and the users in each facility is shown in Fig. 2 . Out of 43 facilities, 23 facilities (53%) had no secondary transmission of SARS-CoV-2 within their facilities (Supplementary Table S1). The median numbers of COVID-19 cases per facility were 2 (interquartile range (IQR) 1–3) for LTCFs, 1 (IQR 1–3) for day care services, 1.5 (IQR 1–2) for home care services, 4 (IQR 3–14) for short-stay services, and 1 (IQR 1–4) for monasteries. There was no statistical difference among the median numbers of COVID-19 cases in each facility type (p-value = 0.22).Fig. 2 Number of COVID-19 cases per facility according to facility type among 43 nursing facilities. The graph shows the number of COVID-19 cases per facility according to their facility types. The COVID-19 cases include the users and the HCWs. Twenty-three facilities (53%, 23/43) had only one COVID-19 case within their facilities. Two LTCFs and one short-stay service had 14 cases in each of their facilities, that marked the highest number of COVID-19 cases per facility. Abbreviations: COVID-19, coronavirus disease 2019; HCWs, healthcare workers; LTCFs, long-term care facilities. Fig. 2 COVID-19 cases of the HCWs and the users are grouped into clusters based on the epidemiological links among the cases (Table 4 ). The index case was a user for 5 of 12 clusters (42%), a HCW for 5 of 12 clusters (42%), and a family member for 2 of 12 clusters (17%). The cross-facility movement of HCW or user was observed in 6 of 12 clusters (50%) (Supplementary Fig. S1). A single large cluster involved 28 HCWs and users (28/109, 26%) deriving from 7 different facilities (2 short-stay services, 2 day care services 2 LTCFs, and 1 home care service) (Supplementary Fig. S2). Four clusters (A-D), which had more than ten positive cases, composed 60% (65/109) of all COVID-19 cases among the HCWs and the users.Table 4 The number of COVID-19 cases and deaths among users and HCWs, index case, presence of cross-facility movement, number of facilities and types of facilities in each cluster. Table 4Cluster Number of COVID-19 cases Users HCWs Deaths Index casea Cross-facility movement (HCW/User) Number of facilities Types of facilitiesb (Number of cases within the facility) A 28 19 9 7 User Yes (HCW/User) 7 LTCF(14), LTCF(1), SS(14), SS(4), DC(11), DC(1), HC(2) B 14 12 2 2 HCW Yes (HCW) 2 LTCF(14), LTCF(1) C 12 7 5 0 HCW Yes (User) 2 DC(12), DC(1) D 11 8 3 1 HCW Yes (HCW) 6 LTCF(8), LTCF(2), MO(4), MO(1), MO(1), HC(3) E 9 7 2 0 User Yes (User) 5 DC(7), DC(1), DC(1), DC(1), HC(2) F 6 2 4 0 HCW No 1 LTCF(6) G 4 4 0 1 Family Yes (User) 2 DC(3), LTCF(2) H 3 2 1 1 User No 1 SS(3) I 3 1 2 0 Family No 1 LTCF(3) J 3 1 2 0 HCW No 1 LTCF(3) K 2 1 1 0 User No 2 DC(1), HC(1) L 2 2 0 0 User No 1 LTCF(2) Non 1 1 0 0 User No 1 LTCF(1) Non 1 1 0 0 User No 1 DC(1) Non 1 0 1 0 HCW No 1 HC(1) Non 1 0 1 0 HCW No 1 HC(1) Non 1 0 1 0 HCW No 1 LTCF(1) Non 1 1 0 0 Family No 1 DC(1) Non 1 0 1 0 HCW No 1 LTCF(1) Non 1 1 0 0 User No 1 DC(1) Non 1 0 1 0 Family No 1 LTCF(1) Non 1 0 1 0 HCW No 1 LTCF(1) Non 1 0 1 0 Family No 1 LTCF(1) Non 1 1 0 0 User No 1 DC(1) Total 109 71 38 12 43 Abbreviations: COVID-19, coronavirus disease 2019; HCWs, healthcare workers; LTCF, long-term care facility; SS, short-stay service; DC, day care service; HC, home care service; MO, monastery. a Index case was defined as the first person who was tested positive for SARS-CoV-2 within the facility. If there was more than one person who was tested positive on the same date, the person who first showed symptoms was identified as the index case. b For facilities that had cross-facility movement, the person (HCW or user) who belonged to two or more facilities was counted separately in each facility. 4 Discussion We described the epidemiology of COVID-19 in 43 nursing facilities in Nagasaki city. COVID-19 cases in the nursing facilities composed 10% of all positive cases, and 28% of COVID-19 deaths in Nagasaki city, which were comparable with global epidemiological data [1]. The incidence rates of COVID-19 were similar between the nursing facility users and the community-dwelling older adults (Table 1). However, the incidence rate of COVID-19 deaths in the nursing facility users was two times higher than deaths in the community-dwelling older adults. The use of the nursing facility may not increase the chance of contracting COVID-19, but the prognosis of the disease may be more severe among nursing facility users than community-dwelling older adults [28]. The fluctuations of COVID-19 cases in the nursing facilities coincided with the community prevalence of COVID-19 (Fig. 1); and the peaks of the COVID-19 cases in the nursing facilities and overall cases were almost simultaneous. The high community prevalence of COVID-19 influenced the number of cases in the nursing facilities by increasing the chance that the HCWs or the users would be infected with COVID-19 in the community and then introduce the infection into the facilities [29]. The COVID-19 cases among the nursing facility users also had long average duration of the hospital stay (16.6 days) and high proportions of hospitalization (100%), which resulted in high bed occupancy and straining of the local health system (Fig. 1). The results from Table 3 indicate the difficulty of applying manual contact tracing to identify close contact (high-risk contact) in the nursing facilities and the limited capability of symptom-based screening. More than 80% of COVID-19 cases by users or HCWs were not recognized as the close contacts of another COVID-19 case before their diagnosis of COVID-19. The manual contact tracing based on the memory of users and HCWs may be insufficient to promptly and correctly identify close contacts in the nursing facilities [30]. The congregate nature of the nursing facilities makes it difficult to recall individual encounters, and some users may have memory impairment to recall their close contacts. Table 3 shows that 32.4% of users were asymptomatic when diagnosed with COVID-19. Asymptomatic transmission is a known contributor of COVID-19 transmission especially in the nursing facilities [31]. Considering the limitations of manual contact tracing and symptom-based screening, mass screening of users and HCWs is needed to detect asymptomatic cases and prevent further spread of COVID-19 [32]. We observed that various types of facilities were affected by COVID-19 (Fig. 2). We expected community-based services (short-stay, home care, and day care services) to have higher proportions of COVID-19 than LTCFs under the lockdown (specifically, restrictions of visitors and activities outside of the facility); and LTCFs could have links to the community only through HCWs or newly admitted cases. However, the proportions in all facility types were similar. We should be mindful of numerous interactions with the community that persist even after the implementation of lockdown [33,34]. The LTCFs and short-stay services, which were residential facilities, had higher median numbers of cases per facility than day care services and home care services, which are outpatient-based. Although there was no statistical significance, the median numbers of cases per facility were higher in the residential facilities because the containment of infection may be easier to implement in outpatient-based services than the residential facilities (Fig. 2). In outpatient-based services, it is possible to isolate the close contacts or the positive cases in each of their homes; and their services can be temporarily closed down or be substituted in the case of a COVID-19 event. However, in the residential facilities, the close contacts and the positive cases remain within the facility, and they sometimes have difficulty complying with the infection control measures. Four large clusters of COVID-19 had a large impact on the local health system (Table 4, Supplementary Fig. S2). Clusters (A-D), which had more than 10 positive cases, involved 17 different facilities, 19 HCWs and 46 users, and accounted for 60% (65/109) of the overall cases of the HCWs and the users. Due to the limited sample size, this study could not identify the factors that triggered the secondary transmissions of SARS-CoV-2 within and out of the nursing facilities. However, we observed a higher percentage of symptomatic index cases (57.9% vs. 34.8%) in the facilities that had secondary transmissions (Supplementary Table S1). Symptomatic index cases could cause higher secondary attack rates than asymptomatic index cases [35,36]. Cross-facility movement provoked a spread of SARS-CoV-2 beyond one facility, and was observed in 50% of the COVID-19 clusters (Table 4). Further research is required to examine the benefits of avoiding cross-facility movement, which should outweigh the disadvantages of limiting the access of users to various care services [37]. There are three main limitations to our study. Firstly, our sample size was limited for adequate comparison of the characteristics of each facility type and individuals. Secondly, our calculation of the incidence rate of COVID-19 did not take into account the possible asymptomatic cases among the community-dwelling people who were not tested. We could have underestimated the incidence rate of COVID-19 in the community. Third is the difference in the virus lineage of the fourth pandemic wave, which was mainly due to the alpha variant known for its high lethality and infectivity compared to the pre-existing lineages [38]. This difference could have influenced the transmission pattern between each individual and transmission into the nursing facility. 5 Conclusion We described the magnitude of the COVID-19 pandemic in Nagasaki city nursing facilities and the impact of their clusters. The nursing facility users had more severe outcomes of COVID-19 than the community-dwelling older adults. Prevention of COVID-19 clusters is crucial to reduce the number of COVID-19 cases and deaths within nursing facilities and to sustain the local health system. Author contributions Koki Shimizu (KS) and Konosuke Morimoto (KoM) proposed the study idea. Haruka Maeda (HM), Eiichiro Sando (ES), KoM, and Katsuaki Motomura supervised the research. KS performed the analysis. KS, HM, ES, and KoM interpreted the findings. KS drafted the first report. All authors contributed to the writing of the final report. All authors have approved the final version to be submitted. Role of the funder/sponsor We did not receive any fund for the research. Declaration of competing interest The authors declare no conflicts of interest. This article has not been published previously nor is not being considered for publication in other journals. Appendix A Supplementary data The following is the Supplementary data to this article:Multimedia component 1 Multimedia component 1 Acknowledgments We would like to thank the help and guidance received from the Nagasaki City Public Health Center (Dr. Katamine, Mr. Mizuashi, Dr. Kinoshita, Mr. Yamaguchi, Ms. Ikeyama, Ms. Nakamoto), Welfare department of Nagasaki City Hall (Mr. Yamaguchi, Ms. Takebu), Nagasaki City Laboratory of Public Health and Environment (Dr. Shimasaki), Nagasaki Prefectural Government (Dr. Hasegawa, Dr. Ando), Infection Control and Education Center of Nagasaki University Hospital, Department of Clinical Medicine, Institute of Tropical Medicine of Nagasaki University (Dr. Ariyoshi), and the Department of Laboratory Medicine of Nagasaki University Hospital (Dr. Yanagihara, Dr. Ota). Appendix A Supplementary data to this article can be found online at https://doi.org/10.1016/j.jiac.2022.04.010. ==== Refs References 1 Salcher-Konrad A. Jhass A. Naci H. Tan M. El-Tawil Y. Comas-Herrera A. COVID-19 related mortality and spread of disease in long-term care: first findings from a living systematic review of emerging evidence medRxiv [Preprint]. [posted 2020 Apr 1; cited 2021 Jul 21]. Available from: https://www.medrxiv.org/content/10.1101/2020.06.09.20125237v3 2022 10.1101/2020.06.09.20125237 2 Williamson E.J. Walker A.J. Bhaskaran K. Bacon S. Bates C. Morton C.E. Factors associated with COVID-19-related death using OpenSAFELY Nature 584 7821 2020 Aug 430 436 10.1038/s41586-020-2521-4 32640463 3 Data.CMS.gov [Internet] COVID-19 nursing home data [cited 2021 July 21]. Available from: https://data.cms.gov/stories/s/COVID-19-Nursing-Home-Data/bkwz-xpvg/ 4 The kyodo news service [Internet] COVID-19 deaths in the nursing facilities account for 14% of overall COVID-19 deaths [cited 2021 Jul 21]. 47 NEWS. Available from: https://47news.jp/4808143.html 5 McMichael T.M. Currie D.W. Clark S. Pogosjans S. Kay M. Schwartz N.G. Epidemiology of COVID-19 in a long-term care facility in King Country, Washington N Engl J Med 382 21 2020 May 21 2005 2011 10.1056/NEJMoa2005412 32220208 6 Graham N.S.N. Junghans C. Downes R. Sendall C. Lai H. McKirdy A. SARS-CoV-2 infection, clinical features and outcome of COVID-19 in the United Kingdom nursing homes J Infect 81 3 2020 Sep 411 419 10.1016/j.jinf.2020.05.073 32504743 7 ECDC Public Health Emergency TeamDanis K. Fonteneau L. Georges S. Daniau C. Bernard-Stoecklin S. High impact of COVID-19 in long-term care facilities, suggestion for monitoring in the EU/EEA, May 2020 Euro Surveill 25 22 2020 Jun 2000956 10.2807/1560-7917.ES.2020.25.22.2000956 32524949 8 Centers for disease control and prevention [Internet] Interim infection prevention and control recommendations to prevent SARS-CoV-2 spread in nursing homes [cited 2021 Aug 11]. Available from: https://www.cdc.gov/coronavirus/2019-ncov/hcp/long-term-care.html 9 Wang H. Li T. Barbarino P. Gauthier S. Brodaty H. Molinuevo J.L. Dementia care during COVID-19 Lancet 395 10231 2020 Apr 11 1190 1191 10.1016/S0140-6736(20)30755-8 32240625 10 Kimball A. Hatfield K.M. Arons M. James A. Taylor J. Spicer K. Asymptomatic and presymptomatic SARS-CoV-2 infections in residents of a long-term care skilled nursing facility – king Country, Washington, March 2020 MMWR Morb Mortal Wkly Rep 69 13 2020 Apr 3 377 381 10.15585/mmwr.mm6913e1 32240128 11 Keehner J. Horton L.E. Binkin N.J. Laurent L.C. SEARCH AlliancePride D. Resurgence of SARS-CoV-2 infection in a highly vaccinated health system workforce N Engl J Med 385 14 2021 Sep 30 1330 1332 10.1056/NEJMc2112981 34469645 12 Brown C.M. Vostok J. Johnson H. Burns M. Gharpure R. Sami S. Outbreak of SARS-CoV-2 infections, including COVID-19 vaccine breakthrough infections, associated with large public gatherings – barnstable County, Massachusetts, July 2021 MMWR Morb Mortal Wkly Rep 70 31 2021 Aug 6 1059 1062 10.15585/mmwr.mm7031e2 34351882 13 Web Japan [Internet] Japan fact sheet, local self-government [cited 2021 Aug 19]. Available from: https://web-japan.org/factsheet/en/pdf/e10_local.pdf 14 Ministry of Internal affairs and communications [Internet] Core cities in Japan (as of Apr 1, 2021) [cited 2021 Aug 19]. Available from: https://www.soumu.go.jp/cyukaku/ 15 Nagasaki prefectural government [Internet] Statistics of Nagasaki (April, 2021) [cited 2021 Oct 17]. Available from: https://www.pref.nagasaki.jp/bunrui/kenseijoho/toukeijoho/idojinko/493981.html 16 Nagasaki city Hall [Internet] Population by sex and age group based on the national census (as of Oct 1 2020) [cited 2021 Jul 21]. Available from: https://www.city.nagasaki.lg.jp/syokai/750000/752000/p023439.html 17 Ministry of health, Labour and welfare [Internet] Long-term care insurance in Japan [cited 2021 July 28]. Available from: https://www.mhlw.go.jp/english/topics/elderly/care/ 18 Nagasaki city Hall [Internet] Press conference by the mayor (Apr 15, 2020) [cited 2021 July 21]. Available from: https://www.city.nagasaki.lg.jp/syokai/710000/713001/p034490.html 19 Ministry of health, Labour and welfare [Internet] Nursing facilities’ response against COVID-19 [cited 2021 Jul 21]. Available from: https://www.mhlw.go.jp/content/11920000/000601572.pdf 20 Ministry of health, Labour and welfare [Internet] Infection control precautions in the nursing facilities [cited 2021 Jul 21]. Available from: https://www.mhlw.go.jp/content/000603941.pdf 21 Ministry of health, Labour and welfare [Internet] Periodic testing of staffs at the nursing facilities [cited 2021 Jul 21]. Available from: https://www.mhlw.go.jp/content/000781628.pdf 22 Notomi T. Okayama H. Masubuchi H. Yonekawa T. Watanabe K. Amino N. Loop-mediated isothermal amplification of DNA Nucleic Acids Res 28 12 2000 Jun 15 e63 10.1093/nar/28.12.e63 10871386 23 Yoshikawa R. Abe H. Igasaki Y. Negishi S. Goto H. Yasuda J. Development and evaluation of a rapid and simple diagnostic assay for COVID-19 based on loop-mediated isothermal amplification PLoS Neglected Trop Dis 14 11 2020 Nov 4 e0008855 10.1371/journal.pntd.0008855 24 Nagasaki city Hall [Internet] Confirmation of COVID-19 case (1527th case of Nagasaki city) [cited 2021 Jul 21]. Available from: https://www.city.nagasaki.lg.jp/fukushi/450000/454000/p034301_d/fil/20210630.pdf 25 Nagasaki prefectural government [Internet] Updated data on COVID-19 infections in Nagasaki [cited 2021 Jul 21]. Available from: https://dev-covid19-nagasaki.netlify.app/cards/number-of-confirmed-cases/ 26 Nagasaki city Hall [Internet] List of nursing facilities [cited 2021 Jul 21]. Available from: https://www.city.nagasaki.lg.jp/fukushi/420000/422000/p007400.html 27 Nagasaki city Hall [Internet] Information on nursing homes [cited 2021 Jul 21]. Available from: https://www.city.nagasaki.lg.jp/jigyo/380000/388000/p002145.html 28 Fisman D.N. Bogoch I. Lapointe-Shaw L. McCready J. Tuite A.R. Risk factors associated with mortality among residents with coronavirus disease 2019 (COVID-19) in long-term care facilities in Ontario, Canada JAMA Netw Open 3 7 2020 Jul 1 e2015957 10.1001/jamanetworkopen.2020.15957 29 White E.M. Kosar C.M. Feifer R.A. Blackman C. Gravenstein S. Ouslander J. Variation in SARS-Cov-2 prevalence in U.S. skilled nursing facilities J Am Geriatr Soc 68 10 2020 Oct 2167 2173 10.1111/jgs.16752 32674223 30 Wilmink G. Summer I. Marsyla D. Sukhu S. Grote J. Zobel G. Real-time digital contact tracing: development of a system to control COVID-19 outbreaks in nursing homes and long-term care facilities JMIR Public Health Surveill 6 3 2020 Aug 25 e20828 10.2196/20828 31 Feaster M. Goh Y.Y. High proportions of asymptomatic SARS-CoV-2 infections in 9 long-term care facilities, Pasadena, California, USA, April 2020 Emerg Infect Dis 26 10 2020 Oct 2416 2419 10.3201/eid2610.202694 32614768 32 Louie J.K. Scott H.M. DuBois A. Sturtz N. Lu W. Stoltey J. Lessons from mass-testing for coronavirus disease 2019 in long-term care facilities for the elderly in San Francisco Clin Infect Dis 72 11 2021 Jun 1 2018 2020 10.1093/cid/ciaa1020 32687150 33 Gandhi M. Yokoe D.S. Havlir D.V. Asymptomatic transmission, the Achilles' heel of current strategies to control COVID-19 N Engl J Med 382 22 2020 May 28 2158 2160 10.1056/NEJMe2009758 32329972 34 Vandael E. Latour K. Islamaj E. Panis L.I. Callies M. Haarhuis F. COVID-19 cases, hospitalizations and deaths in Belgian nursing homes: results of a surveillance conducted between April and December 2020 Arch Publ Health 80 1 2022 Jan 29 45 10.1186/s13690-022-00794-6 35 Madewell Z.J. Yang Y. Longini I.M. Jr. Halloran M.E. Dean N.E. Household transmission of SARS-CoV-2: a systematic review and meta-analysis JAMA Netw Open 3 12 2020 Dec 1 e2031756 10.1001/jamanetworkopen.2020.31756 36 Ko Y.K. Furuse Y. Ninomiya K. Otani K. Akaba H. Miyahara R. Secondary transmission of SARS-CoV-2 during the first two waves in Japan: demographic characteristics and overdispersion Int J Infect Dis 116 2022 Jan 20 365 373 10.1016/j.ijid.2022.01.036 35066162 37 Chen M.K. Chevalier J.A. Long E.F. Nursing home staff networks and COVID-19 Proc Natl Acad Sci USA 118 1 2021 Jan 7 e2015455118 10.1073/pnas.2015455118 38 Davies N.G. Jarvis C.I. CMMID COVID-19 Working GroupEdmunds W.J. Jewell N.P. Diaz-Ordaz K. Increased mortality in community-tested cases of SARS-Cov-2 lineage B.1.1.7 Nature 593 7858 2021 May 270 274 10.1038/s41586-021-03426-1 33723411
PMC009xxxxxx/PMC9005366.txt	==== Front J Biosaf Biosecur J Biosaf Biosecur Journal of Biosafety and Biosecurity 2588-9338 Published by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. S2588-9338(22)00004-8 10.1016/j.jobb.2022.03.001 Research Article Strengthening biological security after COVID-19: Using cartoons for engaging life science stakeholders with the Biological and Toxin Weapons Convention (BTWC) Novossiolova Tatyana a Whitby Simon b Dando Malcolm bd Shang Lijun cd⁎ a The Law Program of the Center for the Study of Democracy, Bulgaria b Division of Peace Studies University of Bradford, Bradford, United Kingdom c School of Human Sciences, London Metropolitan University, London, United Kingdom d Biological Security Research Centre, London Metropolitan University, London, United Kingdom ⁎ Corresponding author at: School of Human Sciences, London Metropolitan University, London, United Kingdom. 13 4 2022 6 2022 13 4 2022 4 1 6874 1 3 2022 24 3 2022 31 3 2022 © 2022 Published by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. The devastating effects of the COVID-19 pandemic have acutely shown the need for maintaining robust international and national systems for biological security and ensuring that life sciences are used only for peaceful purposes. Life science stakeholders can play an important role in safeguarding scientific and technological advances in biology and related fields against accidental or deliberate misuse, not least because they are on the frontlines of driving innovation. In this paper, we argue that enhancing awareness and understanding of the risk of deliberate disease is essential for effective biological security. We first discuss the issue of ‘dual use’ in science and technology as it relates to disarmament and non-proliferation of weapons of mass destruction. Second, we review how scientist engagement with dual-use risks has been addressed in the context of the Biological and Toxin Weapons Convention (BTWC). Third, we report on the development of an innovative awareness-raising tool, a cartoon series, that can be used for engaging life science stakeholders with BTWC issues. Finally, we outline a set of practical considerations for promoting sustainable life science engagement with the BTWC. Keywords Cartoon Life sciences Biological security Biological and Toxin Weapons Convention Awareness Engagement ==== Body pmc1 Introduction The rapid progress made in the life sciences over the past few decades offers tremendous prospects for health, socio-economic development, and prosperity. At the same time, this progress could profoundly redefine the international security landscape by enabling novel misuse capabilities and multiplying the range of actors with access to such capabilities.20, 49 When addressing the Security Council regarding the implications of COVID-19 on the maintenance of international peace and security, the United Nations (UN) Secretary-General noted that ‘the pandemic also highlights the risks of bioterrorist attacks, and has already shown some of the ways in which preparedness might fall short if a disease were to be deliberately manipulated to be more virulent, or intentionally released in multiple places at once’.45 The Secretary-General has further pointed out that improving the response to future disease threats requires serious attention to preventing the deliberate use of diseases as weapons.45 Life scientists can play an important role in addressing the dual-use potential of modern life science and related technologies, not least because they are on the frontlines of driving innovation that seeks to benefit humanity. This paper argues that enhancing awareness and understanding of the risk of deliberate disease is essential for effective biological security. It first discusses the issue of ‘dual use’ in science and technology as it relates to disarmament and non-proliferation of weapons of mass destruction (WMD). Second, the paper reviews how scientist engagement with dual-use risks has been addressed in the context of the Biological and Toxin Weapons Convention (BTWC). Third, it reports on the development of an innovative awareness-raising tool, a cartoon series, that can be used to engage life science stakeholders with BTWC issues. The conclusion outlines a set of practical considerations for promoting sustainable life science engagement with the BTWC. 2 Biological security and the question of dual use in the life sciences Dual-use life science research is benignly intended life science research that has the potential to provide knowledge, information, products, or technologies that could be directly misapplied to create a serious threat with potential consequences for public health and safety, agricultural species, and other plants, animals, and the environment.55 This includes development of biological weapons and the risk of bioterrorism. International efforts to prohibit biological weapons reflect a shared commitment among states to ensure that advances in biology and related fields are used only for peaceful ends and for the benefit of humanity. The 1975 Biological and Toxin Weapons Convention (BTWC), the first international agreement to outlaw an entire class of WMD, is indicative in this regard. The prohibition of biological and toxin weapons is a responsibility incumbent upon all those engaged in the life sciences, whether in government, industry, or academia. States Parties to the BTWC have recognised the importance of engaging life science stakeholders as a way of promoting the in-depth implementation of the Convention. For example, when considering national implementation of the Convention, the Second Review Conference noted the importance of the following:- ‘Legislation regarding the physical protection of laboratories and facilities to prevent unauthorised access to and removal of pathogenic or toxic material. - ‘Inclusion in textbooks and in medical, scientific and military educational programmes of information dealing with the prohibition of bacteriological (biological) and toxin weapons and the provisions of the Geneva Protocol.’40 Similar language has been agreed in subsequent Review Conferences of the BTWC. The Sixth Review Conference of the Convention held in 2006 ‘encourage[d] States Parties to take necessary measures to promote awareness amongst relevant professionals of the need to report activities […] that could constitute a violation of the Convention or related national criminal law. In this context, the Conference recognise[d] the importance of codes of conduct and self-regulatory mechanisms in raising awareness, and call[ed] upon States Parties to support and encourage their development, promulgation and adoption’.42 The Seventh and Eighth Review Conferences of the BTWC, held in 2011 and 2016, respectively, reiterated the value of scientist engagement with the Convention: ‘The Conference notes the value of national implementation measures, as appropriate, in accordance with the constitutional process of each State Party, to:(a) encourage the consideration of development of appropriate arrangements to promote awareness among relevant professionals in the private and public sectors and throughout relevant scientific and administrative activities; (b) promote amongst those working in the biological sciences awareness of the obligations of States Parties under the Convention, as well as relevant national legislation and guidelines; (c) promote the development of training and education programmes for those granted access to biological agents and toxins relevant to the Convention and for those with the knowledge or capacity to modify such agents and toxins; (d) encourage the promotion of a culture of responsibility amongst relevant national professionals and the voluntary development, adoption and promulgation of codes of conduct”.41, 7 Efforts to enhance engagement among science stakeholders with security issues are observed in other areas of non-proliferation and disarmament. For example, the scope of the Nuclear Security Programme of the International Atomic Energy Agency (IAEA) is to ‘contribute to global efforts to achieve effective nuclear security, by establishing comprehensive nuclear security guidance and, upon request, promoting its use through peer reviews and advisory services and capacity building, including education and training’.14 Since 2010, the International Nuclear Security Education Network (INSEN) has been functioning as a partnership through which the IAEA, educational and research institutions, as well as other stakeholders cooperate to promote sustainable nuclear security education.12 To this end, INSEN utilises a comprehensive approach comprising activities on the development, implementation, and evaluation of nuclear security education programmes.12 The IAEA also cooperates with States on the establishment of national Nuclear Security Support Centres (NSSCs) to strengthen the sustainability of nuclear security.13 NSSCs are intended to serve as national coordination hubs for promoting nuclear security culture through a multiple stakeholder engagement underpinned by human resource development, technical support services, and scientific support services.13, 44 Likewise, States Parties to the Chemical Weapons Convention (CWC) have acknowledged the value of broad stakeholder engagement in the area of chemical security stressing their:- ‘Determination to maintain the Convention’s role as a bulwark against chemical weapons; to that end to promote, inter alia, outreach, capacity building, education, and public diplomacy. - Desire to improve interaction with chemical industry, the scientific community, academia, and civil society organisations engaged in issues relevant to the Convention, and cooperate as appropriate with other relevant international and regional organisations, in promoting the goals of the Convention’ [original emphasis].34 The Scientific Advisory Board (SAB) of the Organisation for the Prohibition of Chemical Weapons (OPCW) reviews and assesses developments in scientific and technological fields that are relevant to the Convention and provides advice on technical matters related to its implementation.30 The work of the SAB is critical for ensuring that the CWC keeps pace with rapidly advancing and converging science and technology. The Hague Ethical Guidelines—a set of guiding ethical principles for responsible conduct in chemistry—were developed by an independent group of international experts and established under the OPCW in 2015.31 The Hague Ethical Guidelines apply to all stakeholders in chemistry and related fields and aim to support the development of a professional science culture that helps prevent the re-emergence of chemical weapons. Established in 2015, the Advisory Board on Education and Outreach (ABEO) of the OPCW is a multidisciplinary body comprising 15 independent experts.33, 32 Its primary function is to provide advice on the development of education and outreach strategies, key messages, and partnerships that support the implementation of the Convention.1 3 Scientist engagement with the Biological and Toxin weapons Convention (BTWC) During the 2005 BTWC Meeting of States Parties, Russia tabled a Working Paper entitled ‘Basic Principles (Core Elements) of the Codes of Conduct of Scientists Majoring in Biosciences’ which defined professional duties and responsibilities for biologists with regard to the BTWC, as follows: ‘Scientists should:i. Be well informed of, and apply in their practice, international and national regulatory legal instruments on the prohibition of biological and toxin weapons ii. Be involved in raising biologists’ awareness of international and national obligations related to the prohibition of biological weapons, including criminal liability for their violation iii. Assist in improving and strengthening international legally binding arrangements banning biological weapons and their proliferation iv. Participate, within their competence, in the development of national regulatory legal acts aimed at using scientific and practical results of biological research solely for peaceful purposes v. Contribute to the reduction of new risks and threats which may affect the enforcement of the BTWC vi. Avoid referring to the results of the work, which may be used in violation of the BTWC provisions, in their scientific papers and statements to the mass media vii. Take measures to ensure that transfers of biological agents, toxins, equipment and technologies to any natural or legal person are performed in compliance with the BTWC requirements and national legislation enforcing such measures’.39 Codes of conduct are sets of principles that denote acceptable modes of behaviour within a given social group. Codes of conduct also determine the ways in which the members of a particular group relate to their broader social environment. In this sense, codes of conduct shape professional responsibility and assign corresponding duties.27 Codes of conduct and codes of ethics for life scientists are common but few explicitly address the issue of dual-use research.3, 48, 9 A well-known example of a professional ethical code is the Hippocratic Oath which specifies the main principles of the medical profession that physicians should abide by in their everyday practice. The World Medical Association (WMA)—a professional standard-setting body in the area of medicine—has adopted a set of key documents that specifically examine the question of medical involvement in the development of chemical and biological weapons.6 The WMA Declaration of Washington on Biological Weapons, which was adopted in 2002 and last reaffirmed in 2012, draws attention to the special responsibilities of all those concerned with health care and biomedical research as regards ‘the growing threat that biological weapons might be used to cause devastating epidemics’ (Box 1).56 Box 1 World Medical Association Declaration of Washington on Biological Weapons* ✓ […] The release of organisms causing smallpox, plague, anthrax or other diseases could prove catastrophic in terms of the resulting illnesses and deaths compounded by the panic such outbreaks would generate. At the same time, there is a growing potential for production of new microbial agents, as expertise in biotechnology grows and methods for genetic manipulation of organisms become simpler. These developments are of special concern to medical and public health professionals because it is they who best know the potential human suffering caused by epidemic disease and it is they who will bear primary responsibility for dealing with the victims of biological weapons. Thus, the World Medical Association believes that medical associations and all who are concerned with health care bear a special responsibility to lead in educating the public and policy makers about the implications of biological weapons and to mobilise universal support for condemning research, development, or use of such weapons as morally and ethically unacceptable. ✓ […] Nonproliferation and arms control measures can diminish but cannot completely eliminate the threat of biological weapons. Thus, there is a need for the creation of and adherence to a globally accepted ethos that rejects the development and use of biological weapons. ✓ All who participate in biomedical research have a moral and ethical obligation to consider the implications of possible malicious use of their findings. Through deliberate or inadvertent means, genetic modification of microorganisms could create organisms that are more virulent, are antibiotic-resistant, or have greater stability in the environment. ✓ Research specifically for the purposes of creating biological weapons is to be condemned. As scientists and humanitarians, physicians have a societal responsibility to decry scientific research for the development and use of biological weapons and to express abhorrence for the use of biotechnology and information technologies for potentially harmful purposes. ✓ Physicians and medical organisations have important societal roles in demanding a global prohibition on biological weapons and stigmatising their use, guarding against unethical and illicit research, and mitigating civilian harm from use of biological weapons. Source: World Medical Association56; emphases added. In 2005, the Inter-Academy Partnership, an organisation that brings together more than 140 national, regional, and global science academies, published the Statement on Biosecurity, which underscores ‘scientists’ special responsibility regarding problems of ‘dual use’ and the misuse of science and technology’.11 The Statement contains core guiding principles that are intended to inform the development of codes of conduct that address the dual-use potential of life sciences (Box 2). This statement was issued ahead of the 2005 BTWC Meeting of States Parties and endorsed by over 60 national science academies.11 Box 2 IAP Statement on Biosecurity – Guiding Principles 1. Awareness. Scientists should always bear in mind the potential consequences – possibly harmful – of their research and refuse to undertake research that has only harmful consequences for humankind. 2. Safety and Security. Scientists have a responsibility to use good, safe and secure laboratory procedures, whether codified by law or common practice. 3. Education and Information. Scientists should be aware of, disseminate information about and teach national and international laws and regulations, as well as policies and principles aimed at preventing the misuse of biological research. 4. Accountability. Scientists who become aware of activities that violate the Biological and Toxin Weapons Convention or international customary law should raise their concerns with appropriate people, authorities and agencies. 5. Oversight. Scientists with responsibility for oversight of research or for evaluation of projects or publications should promote adherence to these principles by those under their control, supervision or evaluation and act as role models in this regard. Source: Inter-Academy Partnership.11 Low levels of awareness regarding the BTWC and dual-use issues among life scientists has been recognised as a challenge to the development and adoption of relevant codes of conduct,37, 36, 43, 50 Two Working Papers submitted to the 2008 BTWC Meeting of Experts by China and Japan, respectively, and a Working Paper submitted to the 2008 Meeting of States Parties by Pakistan, highlighted the need to consider codes of conduct and biological security education as mutually reinforcing sets of measures for promoting responsible life science practice.5, 16, 35 In 2018, China and Pakistan tabled a joint proposal for the development of a model code of conduct for biological scientists under the BTWC. This proposal built upon an earlier Working Paper submitted by China in 2015, ahead of the Eighth Review Conference of the BTWC. The proposed model code comprised 10 elements, including:8. ‘(Education and training) Scientific community and professional associations should play an active role in education and training, increase public awareness of the Convention, and establish a safety education and training system for all parties involved in biotechnology research. Biological scientists should be encouraged to engage in dialogue and cooperation with social scientists, philosophers, and anthropologists, so as to have a better understanding of the possible ethical and social implications of relevant biological research and its outcome. 9. (Awareness and engagement) Biological scientists should be fully aware of the potential threats of dual-use research to human society, the ecological environment, and economic security. It is advocated to promote the peaceful application of biological research achievements, to prevent the abuse and misuse of biological products, scientific knowledge, technology and equipment, and to consciously resist any unethical scientific conducts that are harmful to human society’.4 At the same BTWC Meeting of Experts, the French delegation delivered a technical presentation on the development of a national code of conduct (charter) to promote good practice and govern dual-use research in biology and biotechnology.8 The national code is being developed by the National Consultative Council for Biosecurity and features education on dual-use issues for life scientists as a key element. In 2019, the United States tabled a Working Paper, Approaches to Risk and Benefit Assessment for Advances in Life Scientists, which recognised the value of biological security awareness-raising in preventing the misuse of life sciences:19. ‘In the coming years, it is certain that there will be remarkable biotechnology research advances with dual-use potential. Science-based assessment and evaluation tools can help to assess potential risks and benefits and to direct oversight attention and resources towards the most likely or concerning threats. When combined with additional tools like biorisk management and social awareness, these risk assessment and evaluation tools can help reduce the risk of misuse of biology’47; emphasis added). Another Working Paper tabled in 2019 by the United Kingdom points out the need for promoting awareness about biological security issues and interactions among all relevant life science sectors and communities to ensure effective biological risk assessment and management.46 The Tianjin Biosecurity Guidelines for Codes of Conduct for Scientists developed in 2021 set out 10 elements for strengthening responsible conduct in life sciences and safeguarding research and technologies against hostile misuse.17 One of the elements of the Tianjin Biosecurity Guidelines specifically focuses on the role of education and training in the process of ensuring that the life sciences are used only for peaceful purposes: ‘Scientists, along with their professional associations in industry and academia, should work to maintain a well-educated, fully trained scientific community that is well versed in relevant laws, regulations, international obligations and norms. Education and training of staff at all levels should consider the input of experts from multiple fields, including social and human sciences, to provide a more robust understanding of the implications of biological research. Scientists should receive ethical training on a regular basis.’17 4 Using cartoons for engaging life science stakeholders with the BTWC It is important that biosecurity principles, such as those contained in the Tianjin Biosecurity Guidelines, are internalised in life science professional practice. This process requires a concerted effort from multiple stakeholders, such as researchers, funders of life science research, publishers, policy-makers, and end-users.52, 26, 25 This section reports on the development of an innovative awareness-raising tool for engaging life science stakeholders with biological security issues. The tool comprises a cartoon series available in 13 languages. Earlier research in the area of biological security education and outreach has revealed some of the challenges to fostering awareness about dual-use issues, as well as the utility of active learning methods for promoting reflection on such issues in a more engaging and easily assimilated form.36, 21, 22, 23, 24 The choice of using cartoons has been informed by several considerations. The use of illustrative material—graphs, histograms, pie charts, and even graphical abstracts—is widespread in scientific settings. The cartoon format also has applications for presenting complex processes and concepts in a simplified way.29 Additionally, cartoons are used for educational purposes, including for promoting awareness about biological security issues.2, 38, 15 The cartoon series entitled Strengthening the Web of Prevention against Chemical and Biological Weapons features five two-page cartoons, each focusing on a specific concept of relevance to biological security.18 Cartoon 1 examines the issue of dual-use research and preventing threats owing to biological weapons. Cartoon 2 examines the issue of codes of conduct. Cartoon 3 examines the issue of education and awareness-raising. Cartoon 4 examines the process of fostering a biological security culture. And Cartoon 5 examines the concept of ‘One Health’ security. The cartoon series is envisaged as an integrated five-part story whereby some of the characters appear in several cartoons. Each cartoon can be used separately, as well. The cartoons are set in a way that makes it possible for life scientists to relate to the scenario presented. For example, the first page of Cartoon 1 focuses on a dialogue between life scientists during a conference. The second page of this cartoon looks into the role that life scientists could play in raising awareness about biological security issues (Fig. 1 ). Cartoon 1 further aims to highlight existing resources that could be used for biological security education and training, such as the Inter-Academy Partnership guide Doing Global Science, the training guide Preventing Biological Threats: What You Can Do, and the active learning manual, Biological Security Education Handbook: The Power of Team-Based Learning.10, 51, 28 Fig. 1 Cartoon 1 – Preventing Threats from Biological Weapons. Source: LMU.18 To facilitate the dissemination of the cartoons, video has been developed that focuses on the different situations that are depicted in the cartoon.19 Additionally, the cartoon series has been translated into 12 languages, including the 6 official UN languages (Arabic, Armenian, Chinese, French, German, Greek, Italian, Japanese, Russian, Spanish, Ukrainian, and Urdu). The translations were carried out by biological security education practitioners from around the world. The cartoons have been presented at several international events and have been used as part of biological security awareness-raising efforts.25 5 Conclusion Enhancing life scientist engagement with the BTWC, including through codes of conduct and education, requires both nationally and internationally coordinated effort. At the time of this writing, the World Health Organisation is coordinating the development of a framework document—Global Guidance Framework for the Responsible Use of Life Sciences. Mitigating Biorisks Governing Dual-Use Research—with the aim to consolidate and integrate existing good practices and lessons learned for the responsible conduct of science.53, 54 The life science community can play a crucial role in promoting a shared recognition of the need to internalise relevant modes of behaviour and reasoning in the everyday practice of life scientists. There are at least five areas of action to which the life science community can actively contribute so as to enhance understanding of the risk of deliberate disease and strengthen dual-use risk management; these include:• Design and implementation of biological security education and training programmes. • Standardisation of biological security competence, practice, and relevant infrastructure. • Promulgation of biological security practices, including through the development of codes of conduct. • Development of methodologies, tools, and instruments for science and technology assessment, including cost and benefit analysis. • Promoting biosecurity capacity building, including through scientific cooperation, dialogue, and exchange of lessons learned, good practices, and experiences. These five areas of action are intertwined and cross-cutting and should be implemented in an integrated manner. Considerations related to the risk of deliberate disease should be built-in at every stage of the life science research process to ensure comprehensive risk assessment and management. At the international level, the interaction between the BTWC and other disarmament agreements and ethics-related initiatives that underscore the important role of human resource development for maintaining scientific integrity should be deepened. These could include, for example, the International Nuclear Security Education Network, the International Nuclear Security Training and Support Centres (NSSCs Network), and the Advisory Board on Education and Outreach of the OPCW. In addition, the World Commission on the Ethics of Scientific Knowledge and Technology (COMEST) could be consulted and the experience and expertise of the OPCW in industry engagement with safety and security issues could be leveraged to identify viable mechanisms for promoting engagement with the BTWC within the life sciences. CRediT authorship contribution statement Tatyana Novossiolova: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. Simon Whitby: Conceptualization, Investigation, Writing – review & editing. Malcolm Dando: Conceptualization, Investigation, Writing – review & editing. Lijun Shang: Conceptualization, Investigation, Writing – review & editing, Funding acquisition. Conflict of interest The authors declare that there is no conflict of interests. Acknowledgement We would like to thank the reviewers for their comments and suggestions. The graphic design of the Cartoon Series and the publication of this manuscript have been funded by a grant provided by the UK Research and Innovation Strategic Priorities Fund and HEIF Rescaling Fund through London Metropolitan University, UK. ==== Refs References 1 Advisory Board on Education and Outreach, 2018. Report on the Role of Education and Outreach in Preventing the Re-Emergence of Chemical Weapons. ABEO-5/1, OPCW, The Hague, 12 February, available at https://www.opcw.org/sites/default/files/documents/ABEO/abeo-5-01_e_pdf (accessed 19 March 2022). 2 Bickford JH, 2010. Uncomplicated Technologies and Erstwhile Aids: How PowerPoint, the Internet, and Political Cartoons Can Elicit Engagement and Challenge Thinking in New Ways. The History Teacher;44(1):51–66, available at https://www.jstor.org/stable/25799396 (accessed 22 March 2022). 3 BWC Implementation Support Unit, 2008. Developments in Codes of Conduct since 2005, BWC/MSP/2008/MX/INF.2, 26 June, available at https://meetings.unoda.org/section/bwc-mx-2008-documents/ (accessed 19 March 2022). 4 China and Pakistan, 2018. Proposal for the Development of a Model Code of Conduct for Biological Scientists under the Biological Weapons Convention, BWC/MSP/2018/MX.2/WP.9, BTWC Meeting of Experts on Review of developments in the field of science and technology related to the Convention, 9 August, available at https://undocs.org/en/BWC/MSP/2018/MX.2/WP.9 (accessed 19 March 2022). 5 China, 2008. Oversight of Science, Education and Awareness Raising, Codes of Conduct, BWC/MSP/2008/MX/WP.18, BTWC Meeting of Experts, 14 August, https://meetings.unoda.org/section/bwc-mx-2008-documents/ (accessed 19 March 2022). 6 Crowley M, Nathanson V, 2018. The Role of the Non-Governmental Medical Community in Combatting the Development, Proliferation and Use of Chemical Weapons’ in Crowley, M. et al. (Eds.) Preventing Chemical Weapons: Arms Control and Disarmament as the Sciences Converge, pp. 560–579. 7 Eighth Review Conference of the States Parties to the Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction, 2017. Final Document, BWC/CONF.VIII/4, 7-25 November 2016, Geneva, available at https://meetings.unoda.org/section/bwc-revcon-2016-documents/ (19 March 2022). 8 France, 2018. Technical Briefing: Ethics and Scientific Integrity in France, BTWC Meeting of Experts on Review of Developments in the Field of Science and Technology Related to the Convention, 9-10 August, Geneva, available at https://meetings.unoda.org/section/bwc-mx-2018-mx2-presentations/ (accessed 19 March 2022). 9 Green S, Taub S, Morin K, Higginson D., 2006. ‘Guidelines to Prevent Malevolent Use of Biomedical Research’, Cambridge Quarterly of Healthcare Ethics, 15, 4, pp. 432–447, available at 10.1017/S0963180106210569 (accessed 19 March 2022). 10 Inter-Academy Partnership, 2016. Doing Global Science: A Guide to Responsible Conduct in the Global Research Enterprise, Princeton University Press, https://www.interacademies.org/publication/doing-global-science-guide-responsible-conduct-global-research-enterprise (accessed 19 March 2022). 11 Inter-Academy Partnership, 2005. IAP Statement on Biosecurity, 7 November, available at https://www.interacademies.org/publication/iap-statement-biosecurity (accessed 19 March 2022). 12 International Atomic Energy Agency, 2022. International Nuclear Security Education Network (INSEN), available at https://www.iaea.org/services/networks/insen (accessed 19 March 2022). 13 International Atomic Energy Agency, 2022, International Network for Nuclear Security Training and Support Centres, available at https://www.iaea.org/services/networks/nssc (accessed 19 March 2022).. 14 International Atomic Energy Agency, Board of Governors (IAEA-BG), 2017. Nuclear Security Plan 2018-2021, GC(61)/24, 14 September, available at https://www.iaea.org/gc-archives/gc (accessed 19 March 2022). 15 Japan, 2018. Technical Briefing: Cutting Edge Life Science and Dual-Use Issues – How Should We Have a Dialogue with Society?, BTWC Meeting of Experts on Review of Developments in the Field of Science and Technology Related to the Convention, 9-10 August, Geneva, available at https://meetings.unoda.org/section/bwc-mx-2018-mx2-presentations/ (accessed 19 March 2022). 16 Japan, 2008. Oversight, Education, Awareness Raising, and Codes of Conduct for Preventing the Misuse of Bio-Science and Bio-Technology, BWC/MSP/2008/MX/WP.21, BTWC Meeting of Experts, 14 August, available at https://meetings.unoda.org/section/bwc-mx-2008-documents/ (accessed 19 March 2022). 17 Johns Hopkins Center for Health Security and Tianjin University, 2021. Tianjin Biosecurity Guidelines for Codes of Conduct for Scientists, available at https://www.interacademies.org/publication/tianjin-biosecurity-guidelines-codes-conduct-scientists (accessed 19 March 2022). 18 London Metropolitan University (LMU), 2021. Heightened Risk of Disease as a Means of Terrorism, say international security experts, Press release, 30 June, available at https://www.londonmet.ac.uk/news/spotlight/heightened-risk-of-disease-as-a-means-of-terrorism-say-international-security-experts/ (accessed 19 March 2022). 19 London Metropolitan University (LMU), 2021. Strengthening the Web of Prevention against Chemical and Biological Weapons, [video resource], https://www.londonmet.ac.uk/research/centres-groups-and-units/biological-security-research-centre/ (accessed 19 March 2022). 20 Meselson M, 2000. Averting the hostile exploitation of biotechnology. Chem Biol Weapons Convent Bull, 48, 16-19. 21 National Academies of Sciences, Engineering and Medicine, 2018. Governance of Dual Use Research in the Life Sciences: Advancing Global Consensus on Research Oversight: Proceedings of a Workshop, National Academies Press, Washington, D.C., available at https://www.nap.edu/catalog/25154/governance-of-dual-use-research-in-the-life-sciences-advancing (accessed 19 March 2022). 22 National Academies of Sciences, Engineering, and Medicine, 2018. How People Learn II: Learners, Contexts, and Cultures, National Academies Press, Washington, D.C., available at https://www.nap.edu/catalog/24783/how-people-learn-ii-learners-contexts-and-cultures (accessed 19 March 2022). 23 National Research Council, 2011. Challenges and Opportunities for Education about Dual-Use Issues in the Life Sciences, National Academies Press, Washington DC, available at https://www.nap.edu/catalog/12958/challenges-and-opportunities-for-education-about-dual-use-issues-in-the-life-sciences (accessed 19 March 2022). 24 National Research Council, 2000. How People Learn: Brain, Mind, Experience, and School, National Academies Press, Washington D.C., available at https://www.nap.edu/catalog/9853/how-people-learn-brain-mind-experience-and-school-expanded-edition (accessed 19 March 2022). 25 Novossiolova T, Shang L, Dando M, 2021. Biological Security Education, Awareness, and Outreach as Essential Elements of Strengthening the Review of Science and Technology under the BTWC, CBW Magazine, July-December, available at https://www.idsa.in/system/files/page/2015/cbw_w_KALEIDOSCOPE_2021.pdf (accessed 19 March 2022). 26 Novossiolova T. et al., 2021. Addressing Emerging Synthetic Biology Threats: The Role of Education and Outreach in Fostering Effective Bottom-Up Grassroots Governance in Trump, B. et al (eds.), Emerging Threats of synthetic biology and biotechnology: addressing security and resilience issues, Springer, available at https://link.springer.com/book/10.1007/978-94-024-2086-9 (accessed 22 March 2022). 27 Novossiolova T, Martellini M, 2019. ‘Promoting Responsible Science and CBRN Security Through Codes of Conduct and Education’, Biosafety and Health, 1, 2, pp. 59–64, available at 10.1016/j.bsheal.2019.08.001 (accessed 19 March 2022). 28 Novossiolova T, 2016. Biological Security Education Handbook: The Power of Team-Based Learning, University of Bradford, available at https://bradscholars.brad.ac.uk/handle/10454/7822 (accessed 19 March 2022). 29 Office of Science and Technology, 2022. Getting your research into the UK Parliament. Available at http://bit.ly/researchin parliament (accessed 22 March 2022). 30 Organisation for the Prohibition of Chemical Weapons, 2022. Scientific Advisory Board, available at https://www.opcw.org/about-us/subsidiary-bodies/scientific-advisory-board (accessed 19 March 2022). 31 Organisation for the Prohibition of Chemical Weapons, 2022. The Hague Ethical Guidelines, available at https://www.opcw.org/hague-ethical-guidelines (accessed 19 March 2022). 32 Organisation for the Prohibition of Chemical Weapons, 2022, Advisory Board on Education and Outreach (ABEO), available at https://www.opcw.org/about-us/subsidiary-bodies/advisory-board-education-and-outreach (accessed 19 March 2022). 33 Organisation for the Prohibition of Chemical Weapons – Conference of the States Parties (OPCW-CSP), 2015. Report of the Twentieth Session of the Conference of the States Parties, C-20/5, 4 December, available at https://www.opcw.org/sites/default/files/documents/CSP/C-20/en/c2005_e_pdf (accessed 19 March 2022). 34 Organisation for the Prohibition of Chemical Weapons – Conference of the States Parties (OPCW-CSP), 2013. Report of the Third Special Session of the Conference of the States Parties to Review the Operation of the Chemical Weapons Convention, RC-3/3, 19 April, available at https://www.opcw.org/sites/default/files/documents/CSP/RC-3/en/rc303__e_pdf (accessed 19 March 2022). 35 Pakistan, 2008. Perspective on Oversight, Codes of Conduct, Education and Awareness Raising, BWC/MSP/2008/WP.5, Meeting of the States Parties to the Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction, 5 December, available at https://meetings.unoda.org/section/bwc-msp-2008-documents/ (accessed 19 March 2022). 36 Rappert, B. (eds.), 2010. Education and Ethics in the Life Sciences: Strengthening the Prohibition of Biological Weapons, Canberra: ANU Press, available at https://press-files.anu.edu.au/downloads/press/p51221/pdf/book.pdf (accessed 19 March 2022). 37 Rappert B, Chevrier M, Dando MR, 2006. In-Depth Implementation of the BTWC: Education and Outreach, BTWC Review Conference Paper No 18, available at https://bradscholars.brad.ac.uk/handle/10454/856 (accessed 19 March 2022). 38 Royal Society of Chemistry, 2022. Science Concept Cartoons [online resource], available at https://edu.rsc.org/resources/science-concept-cartoons/4012180.article (accessed 19 March 2022). 39 Russian Federation, 2005. Basic Principles (Core Elements) of the Codes of Conduct of Scientists Majoring in Biosciences, BWC/MSP/2005/WP.2, Meeting of the States Parties to the Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction, 5 December, available at https://documents-dds-ny.un.org/doc/UNDOC/GEN/G05/644/61/PDF/G0564461.pdf?OpenElement (accessed 19 March 2022). 40 Second Review Conference of the Parties to the Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction, 1986. Final Document, BWC/CONF.II/13, 30 September, available at https://meetings.unoda.org/section/bwc-revcon-1986-documents/ (accessed 19 March 2022). 41 Seventh Review Conference of the States Parties to the Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction, 2012. Final Document, BWC/CONF.VII/7, 13 January, available at https://meetings.unoda.org/section/bwc-revcon-2011-documents/ (accessed 19 March 2022). 42 Sixth Review Conference of the States Parties to the Convention on the Prohibition of the Development, Production and Stockpiling of Bacteriological (Biological) and Toxin Weapons and on Their Destruction, 2006. Final Document, BWC/CONF.VI/6, 20 November – 8 December, available at https://undocs.org/en/BWC/CONF.VI/6 (accessed 19 March 2022). 43 Stearns T, 2017. Moving Beyond Dual Use Research of Concern Regulation to an Integrated Responsible Research Environment, paper commissioned for the National Academies Committee on Dual Use Research of Concern, Options for Future Management Workshop, available at https://www.nap.edu/resource/24761/Stearns_Paper_021717.pdf (accessed 19 March 2022). 44 Technical Meeting: Nuclear Security Plan 2014-2017 – Implementation of the International Network for Nuclear Security Training and Support Centres (NSSC), 2014. Chairman’s Report, IAEA Headquarters, Vienna, Austria 19-21 February, available at https://www.iaea.org/sites/default/files/18/01/nssc-network-annual-meeting-report-2014.pdf (accessed 19 March 2022). 45 United Nations Secretary-General, 2020. Remarks to Security Council Open Video-Teleconference on the Maintenance of International Peace and Security: Implications of COVID-19, 2 July, New York, available at https://www.un.org/sg/en/content/sg/speeches/2020-07-02/remarks-security-council-maintenance-of-international-peace-and-security-implications-of-covid-19 (accessed 19 March 2022). 46 United Kingdom, 2019. Biological Risk Assessment and Management: Some Further Considerations, BWC/MSP/2019/MX.2/WP.6, BTWC Meeting of Experts on the Review of Developments in the Field of Science and Technology Related to the Convention, (26 July) 31 July and 2 August, Geneva, available at https://undocs.org/en/bwc/msp/2019/mx.2/wp.6 (accessed 19 March 2022). 47 United States of America, 2019. Approaches to Risk and Benefit Assessment for Advances in the Life Sciences, BWC/MSP/2019/MX.2/WP.3, BTWC Meeting of Experts on the Review of Developments in the Field of Science and Technology Related to the Convention, (11 July) 31 July and 2 August, Geneva, available at https://undocs.org/en/bwc/msp/2019/mx.2/wp.3 (accessed 19 March 2022). 48 Virtual Biosecurity Center, 2019. Biosecurity Codes, available at https://www.virtualbiosecuritycenter.org/codes-of-ethics/list/ (accessed 19 March 2022). 49 Wheelis, M. Rozsa, L. and Dando, M.R. (Eds.), 2006. Deadly Cultures: Biological Weapons Since 1945. Harvard University Press, Cambridge, Mass. 50 Whitby S, Tang C, Shang L, Dando MR, 2020. After COVID-19: Time to agree a biosecurity code of conduct under the biological and toxin weapons convention. J Chem Biol Weapons, Jan-Jun, available at https://idsa.in/cbwmagazine/summer (accessed 19 March 2022). 51 Whitby S, et al., 2015. Preventing Biological Threats: What You Can Do, University of Bradford, available at https://bradscholars.brad.ac.uk/handle/10454/7821 (accessed 19 March 2022). 52 Whitby S, Dando M, 2010. Biosecurity Awareness-Raising and Education for Life Scientists: What Should Be Done?, in Rappert, B. (eds.) Education and Ethics in the Life Sciences: Strengthening the Prohibition of Biological Weapons, Canberra: ANU Press, available at https://press-files.anu.edu.au/downloads/press/p51221/pdf/book.pdf (accessed 22 March 2022). 53 World Health Organisation, 2022. Ensuring Responsible Use of Life Science Research, available at https://www.who.int/activities/ensuring-responsible-use-of-life-sciences-research (accessed 19 March 2022). 54 World Health Organisation, 2022. WHO Guidance Framework for the Responsible Use of Life Sciences, 23 February, available at https://www.who.int/news-room/articles-detail/call-for-comments–-who-global-guidance-framework-for-the-responsible-use-of-the-life-sciences (accessed 19 March 2022). 55 World Health Organisation, 2020. Laboratory Biosafety Manual, 4th ed., available at https://www.who.int/publications/i/item/9789240011311 (accessed 19 March 2022). 56 World Medical Association, 2002. WMA Declaration of Washington on Biological Weapons, 15 February, available at https://www.wma.net/policies-post/wma-declaration-of-washington-on-biological-weapons/ (accessed 19 March 2022).
PMC009xxxxxx/PMC9005367.txt	==== Front Am J Emerg Med Am J Emerg Med The American Journal of Emergency Medicine 0735-6757 1532-8171 Elsevier Inc. S0735-6757(22)00238-8 10.1016/j.ajem.2022.04.006 Article The impact of COVID-19 on incidence and outcomes from out-of-hospital cardiac arrest (OHCA) in Texas Chavez Summer DO, MPH, MPM ab⁎ Huebinger Ryan MD ab Chan Hei Kit MS ab Gill Joseph MD ab White Lynn MS c Mendez Donna MD ab Jarvis Jeffrey L. MD bd Vithalani Veer D. MD e Tannenbaum Lloyd MD f Al-Araji Rabab MPH g Bobrow Bentley MD ab a Texas Emergency Medicine Research Center, McGovern Medical School, Houston, TX, United States of America b McGovern Medical School at The University of Texas Health Science Center at Houston (UTHealth), Department of Emergency Medicine, Houston, TX, United States of America c Global Medical Response, Greenwood Village, CO, United States of America d Williamson County EMS, Georgetown, TX, United States of America e JPS/Medstar, Fort Worth, TX, United States of America f Brooke Army Medical Ctr/Uniform Services Univ of the Health Sciences, San Antonio, TX, United States of America g Emory University Rollins School of Public Health, Atlanta, GA, United States of America ⁎ Corresponding author at: Department of Emergency Medicine, McGovern Medical School of UTHealth at Houston, 6431 Fannin Street, JJL 475, Houston, TX 77030, UT Office, United States of America. 13 4 2022 7 2022 13 4 2022 57 15 31 1 2022 27 3 2022 3 4 2022 © 2022 Elsevier Inc. All rights reserved. 2022 Elsevier Inc. Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Introduction Emerging research demonstrates lower rates of bystander cardiopulmonary resuscitation (BCPR), public AED (PAD), worse outcomes, and higher incidence of OHCA during the COVID-19 pandemic. We aim to characterize the incidence of OHCA during the early pandemic period and the subsequent long-term period while describing changes in OHCA outcomes and survival. Methods We analyzed adult OHCAs in Texas from the Cardiac Arrest Registry to Enhance Survival (CARES) during March 11–December 31 of 2019 and 2020. We stratified cases into pre-COVID-19 and COVID-19 periods. Our prehospital outcomes were bystander cardiopulmonary resuscitation (BCPR), public AED use (PAD), sustained ROSC, and prehospital termination of resuscitation (TOR). Our hospital survival outcomes were survival to hospital admission, survival to hospital discharge, good neurological outcomes (CPC Score of 1 or 2) and Utstein bystander survival. We created a mixed effects logistic regression model analyzing the association between the pandemic on outcomes, using EMS agency as the random intercept. Results There were 3619 OHCAs (45.0% of overall study population) in 2019 compared to 4418 (55.0% of overall study population) in 2020. Rates of BCPR (46.2% in 2019 to 42.2% in 2020, P < 0.01) and PAD (13.0% to 7.3%, p < 0.01) decreased. Patient survival to hospital admission decreased from 27.2% in 2019 to 21.0% in 2020 (p < 0.01) and survival to hospital discharge decreased from 10.0% in 2019 to 7.4% in 2020 (p < 0.01). OHCA patients were less likely to receive PAD (aOR = 0.5, 95% CI [0.4, 0.8]) and the odds of field termination increased (aOR = 1.5, 95% CI [1.4, 1.7]). Conclusions Our study adds state-wide evidence to the national phenomenon of long-term increased OHCA incidence during COVID-19, worsening rates of BCPR, PAD use and survival outcomes. Keywords Cardiac arrest Out-of-hospital cardiac arrest Prehospital care COVID-19 ==== Body pmc1 Introduction Out-of-hospital cardiac arrest (OHCA) is a leading cause of death in the United States, accounting for more than 300,000 occurrences per year [1,2]. Decades of research has shown few interventions that consistently improve outcomes–early initiation of bystander cardiopulmonary resuscitation (BCPR) and automated external defibrillator use (AED)-- yet survival rates have remained relatively constant (<10%) [[3], [4], [5], [6]]. Prior research estimates that close to half of OHCA patients receive BCPR while public AED rates hover around 9–12% [[7], [8], [9], [10]]. As such, these are high-yield targets for public health efforts to increase interventions that are key to improving OHCA survival rates. Emerging studies are beginning to show the tremendous impact of the SARS-CoV-2 pandemic (COVID-19) on OHCA outcomes. These previous studies demonstrate lower rates of BCPR (8.94% decrease from 2020 to 2019) and public AED (PAD) usage, corresponding to lower rates of sustained return of spontaneous circulation (ROSC) (23.0% in 2020 versus 29.8% in 2019; adjusted rate ratio, 0.82 [95% CI, 0.78–0.87]; P < 0.001), lower survival to hospital discharge (6.6% in 2020 versus 9.8% in 2020; adjusted RR, 0.83 [95% CI, 0.69–1.00]; P = 0.048) and in general, higher incidence of OHCA during the COVID-19 pandemic compared to prior years [[11], [12], [13]]. However, most of these studies were conducted during the early months of COVID-19 and it is unclear whether these trends continue to persist on a statewide level. In this study, we aim to characterize the incidence of OHCA during the pandemic period compared to the prior year and describe changes, if any, in OHCA outcomes and survival. This study can help fill the gap regarding long-term impacts of the pandemic beyond the initial first wave on cardiac arrest care and patient outcomes. 2 Methods 2.1 Data source We analyzed adult OHCAs in Texas from the Cardiac Arrest Registry to Enhance Survival (CARES) during March 11–December 31 of 2019 and 2020. The CARES database is a collaboration between Emory University and the Centers for Disease Control and Prevention, which aims to improve cardiac arrest outcomes by using OHCA surveillance data. Data is voluntarily reported by EMS agencies and hospitals from across the country. CARES has standardized quality measurements for benchmarking and quality improvement efforts on the local, state and national level [14,15]. The national CARES database captures 51% about the U.S. population, approximately 167 million people [14,15]. The study was approved by the University of Texas Health Science Center at Houston IRB and the CARES Data Sharing Committee (HSC-MS-19-0601). 2.2 Selection of subjects We stratified cases into different groups of interest: pre-COVID-19 (March 11, 2019 to December 31, 2019) and COVID-19 period (March 11, 2020 to December 31, 2020), representing the period after the emergence and persistence of COVID-19 and a matched period during the prior years. We used March 11, 2020 as the cutoff date as this is when COVID-19 was declared a pandemic by the World Health Organization [16]. We excluded cases witnessed by 9–1-1 responders, arrests occurring at healthcare facilities and pediatric cases (<18 years old). We excluded 2 EMS agencies who did not participate in CARES for the entire study period, representing 0.003% of the original data set. 2.3 Study variables and outcomes Cardiac arrest characteristics were defined as age, gender, race/ethnicity (White, Black, Hispanic, Other), witnessed arrest, initial rhythm type (shockable, non-shockable) and location type (home/residence, public). Our prehospital outcomes of interest were BCPR, public AED use, sustained ROSC, and prehospital termination of resuscitation (TOR). Our hospital survival outcomes of interest were survival to hospital admission, survival to hospital discharge, good neurological outcomes (CPC Score of 1 or 2) and Utstein bystander survival rate (patients surviving to hospital discharge with witnessed bystander arrest, initial shockable rhythm receiving BCPR and/or AED) [17]. 2.4 Statistical analysis We analyzed the patient arrest characteristics and prehospital care data using descriptive statistics stratified by pre-pandemic and pandemic periods. We calculated the median and interquartile ranges for patient ages. We used Pearson's χ 2 tests to determine association of variables with the two study years. Similarly, we used Fisher's exact tests for variables with expected frequencies less than 5 to evaluate their association with the two study years. We built mixed multivariable logistic regression models to estimate the adjusted odds ratio (aOR) of outcomes of interest comparing the pre-pandemic and pandemic periods, and stratifying by EMS agency as the random intercept. The models were controlled for the following confounders: age, gender, race/ethnicity, witnessed arrest, initial rhythm type and location type. All statistical analyses were conducted with STATA 16.1. 3 Results As shown in Figs. 1 , 8037 patients were included in the analysis. (Fig. 1). There were 3619 OHCAs (45.0%) in 2019 compared to 4418 (55.0%) in 2020 for a total of 8037 cases (Table 2). Across the two study periods, median age, race/ethnicity and witnessed arrest were consistent. From 2019 to 2020, the proportion of arrests occurring at home or at a residence increased from 80.9% to 86.7% (P < 0.01) (Table 1 ). Additionally, there were slightly more cardiac arrests occurring in the COVID-19 period thought to be related to respiratory causes (8.2% in 2019 vs 8.4% in 2020) while OHCAs secondary to a cardiac etiology decreased (85.1% in 2019 vs 84.6% in 2020) (P = 0.04) (Table 1).Fig. 1 Patient Selection Flowchart. Fig. 1 Table 1 Patient cardiac arrest characteristics, stratified by year. Table 1 2019 2020 Total Arrests 3619 (45.0%) 4418 (55.0%) Median Age (Years) (IQR)1 63 (51–74) 63 (51–74) P = 0.7 Male Gender2 2307 (63.8%) 2781 (63.0%) P = 0.5 Race & Ethnicity2 White 1591 (44.0%) 1857 (42.0%) P = 0.2 Black/African-American 901 (24.9%) 1128 (25.5%) Hispanic/Latino 935 (25.8%) 1217 (27.6%) Other 192 (5.3%) 216 (4.9%) Location Type2 Home/Residence 2926 (80.9%) 3831 (86.7%) P < 0.01 Public 693 (19.2%) 587 (13.3%) Arrest Witnessed Status2 Unwitnessed 1938 (53.5%) 2338 (52.9%) P = 0.6 Witnessed by Bystander 1682 (46.5%) 2080 (47.1%) Presumed Cardiac Arrest Etiology2 Drowning/Submersion 18 (0.5%) 26 (0.6%) P = 0.04 Drug Overdose 168 (4.6%) 169 (3.8%) Electrocution 3 (0.1%) 8 (0.2%) Exsanguination/Hemorrhage 15 (0.4%) 33 (0.7%) Other 41 (1.1%) 76 (1.7%) Presumed Cardiac Etiology 3079 (85.1%) 3736 (84.6%) Respiratory/Asphyxia 295 (8.2%) 370 (8.4%) 1 P-value determined using Wilcoxon rank sum test. 2 P-value determined using Pearson's χ2 test. Comparing prehospital characteristics from 2019 to 2020, the proportion of prehospital CPR (EMS 20.1% in 2019, 22.5% in 2020; first responder 33.7% in 2019, 35.3% in 2020) and AED (EMS: 88.8% in 2019, 94.1% in 2020) initiated by EMS or first responders increased (p < 0.01). There were decreases in bystander CPR (46.2% in 2019 vs 42.2% in 2020) and bystander AED application (13.0% in 2019 vs 7.3% in 2020). Patients were more likely to present in asystole than any other rhythm (51.3% in 2019 vs 58.2% in 2020) (P < 0.01), and the proportion of non-shockable rhythms increased (79.9% in 2019 vs 84.7% in 2020) (P < 0.01). In terms of outcome, patients were less likely to have sustained ROSC during the pandemic period compared to 2019 (28.8% in 2019 vs 21.1% in 2020) (P < 0.01). They were more likely to have an advanced airway placed in 2020 (90.7% in 2019 vs 92.6% in 2020) (P < 0.01). There was also an increase in field terminations in 2020 (37.3% in 2019 vs 46.7% in 2020) (P < 0.01). Comparing patient survival outcomes, we found both patient survival to hospital admission (27.2% in 2019 vs 21.0% in 2020) (p < 0.01) and patient survival to hospital discharge (10.0% in 2019 to 7.4% in 2020) (P < 0.01) decreased in 2020. The proportion of patients with good neurologic outcomes remained consistent (70.0% in 2019 vs 67.6% in 2020) (P = 0.5). While the Utstein bystander survival rate decreased in 2020 compared to 2019, this did not reach statistical significance (38.5% in 2019 vs 31.4% in 2020) (P = 0.06) (Table 2 ).Table 2 Prehospital Care & Hospital outcomes, stratified by year. Table 2 2019 2020 Party Initiating CPR1 Bystander 1670 (46.2%) 1863 (42.2%) P < 0.01 EMS 728 (20.1%) 995 (22.5%) First Responder 1221 (33.7%) 1559 (35.3%) Who First Applied AED1 Bystander 57 (5.0%) 17 (1.4%) P < 0.01 Family Member 3 (0.3%) 1 (0.1%) Healthcare Provider (Non-911 Responder) 32 (2.8%) 25 (2.0%) Law Enforcement First Responder 37 (3.2%) 31 (2.5%) Non-Law Enforcement First Responder 1020 (88.8%) 1182 (94.1%) Public AED1 90/693 (13.0%) 43/587 (7.3%) P < 0.01 First Monitored Rhythm1 Asystole 1856 (51.3%) 2571 (58.2%) P < 0.01 Idioventricular/PEA 871 (24.1%) 1057 (23.9%) Unknown Shockable Rhythm 110 (3.0%) 64 (1.5%) Unknown Unshockable rhythm 166 (4.6%) 115 (2.6%) Ventricular Fibrillation 579 (16.0%) 572 (13.0%) Ventricular Tachycardia 37 (1.0%) 39 (0.9%) Initial Rhythm Type1 Non-Shockable 2893 (79.9%) 3743 (84.7%) P < 0.01 Shockable 726 (20.1%) 675 (15.3%) Sustained ROSC1 2582 (28.8%) 3503 (21.1%) P < 0.01 Advanced Airway Successfully Placed2 No 316 (9.3%) 302 (7.4%) P < 0.01 Used existing tracheostomy 0 (0.0%) 1 (0.0%) Yes 3094 (90.7%) 3776 (92.6%) Pre-Hospital Outcome1 Dead in Field 1349 (37.3%) 2063 (46.7%) P < 0.01 Ongoing Resuscitation in ED 1496 (41.3%) 1513 (34.3%) Pronounced Dead in ED 774 (21.4%) 842 (19.1%) Survival to Hospital Admission1 981 (27.2%) 923 (21.0%) P < 0.01 Survival to Hospital Discharge1 360 (10.0%) 326 (7.4%) P < 0.01 Good CPC Score1 250/357 (70.0%) 220/326 (67.6%) P = 0.5 Utstein Bystander Survival Rate1 124/322 (38.5%) 87/279 (31.4%) P = 0.06 1 P-value determined using Pearson's χ2 test. 2 P-value determined using Fisher's exact test. Utilizing mixed multivariable logistic regression to compare care between 2019 and 2020, we found patients were slightly less likely to have bystander CPR in 2020 (aOR = 0.9, 95% CI [0.8, 0.95]). OHCA patients were about half as likely to benefit from PAD in 2020 (aOR = 0.5, 95% CI [0.4, 0.8]). Patients were also less likely to have sustained ROSC during the COVID-19 period compared to 2019 (aOR = 0.7, 95% CI [0.6, 0.8]). In terms of survival, the odds of field termination were 1.5 times greater during COVID-19 (aOR = 1.5, 95% CI [1.4, 1.7]). Patients were also less likely to survive to hospital admission (aOR = 0.7, 95% CI [0.7, 0.8]) and hospital discharge (aOR = 0.8, 95% CI [0.7, 0.96]) during COVID-19. While odds of patients having good CPC scores remained consistent (aOR = 0.99, 95% CI [0.7, 1.5]), the odds Ustein bystander survival slightly decreased during the COVID-19 period (aOR = 0.8, 95% CI [0.5, 1.0]) (Table 3 ).Table 3 Generalized outcomes, stratified by year. Table 3 aOR [95% CI] Bystander CPR aOR = 0.9 [0.8, 0.95] Public AED aOR = 0.5 [0.4, 0.8] Sustained ROSC aOR = 0.7 [0.6, 0.7] Field Termination aOR = 1.5 [1.4, 1.7] Survival to Hospital Admission aOR = 0.7 [0.7, 0.8] Survival to Hospital Discharge aOR = 0.8 [0.7, 0.96] Good CPC Score aOR = 0.99 [0.7, 1.4] Utstein Bystander Survival Rate aOR = 0.7 [0.5, 1.0] 4 Discussion Emerging data suggests worsening OHCA outcomes on a national level during the COVD-19 pandemic, but it is not clear if this trend persists on the state level [13]. We hypothesized public interventions, especially those shown to improve OHCA outcomes such as BCPR and PAD would decrease while survival outcomes would worsen. To date, no other state-wide studies have been comparable in terms of geographic size, patient population or temporal length. By utilizing such a comprehensive reach, this study adds to the evidence that OHCA care and outcomes has continued to worsen in the COVID-19 pandemic period. Better characterization of which aspects of pre-hospital cardiac care have deteriorated due to the pandemic can help drive quality improvement efforts and restore the chain of the survival. Between the two study periods, the patient demographics (age, gender, race & ethnicity) were similar. However, the total number of cardiac arrests increased from 3627 in 2019 to 4443 in 2020, an absolute 22.5% increase. Because the same EMS agencies participated in both study periods, this suggests an absolute increase in cardiac arrests as it cannot be attributed to agency participation. While this general increase is consistent with other studies, estimates of differences in OHCA incidence range from −3% to 191.4% during the early months of the pandemic [18]. In a systematic review by Teoh, et al., the pooled annual OHCA incidence increased by 39.5% (P < 0.001), which is more in-line with our calculations [19]. In this study, we used a series of mixed effect logistic regression models, with EMS agency as the random intercept, adjusting for covariates, to calculate adjusted odds ratios and thus the impact of the pandemic on OHCA outcomes of interest. We found that life-saving bystander interventions, BCPR and PAD, worsened relative to the pre-pandemic period and arrests were more likely to occur at home during the COVID-19 period. This is comparable to other estimates which also found decreases in AED usage (OR = 1.78, 95% CI 1.06–2.98) in the pre-COVID-19 period [12,18]. However, study of BCPR has yielded mixed results, with some research suggesting BCPR did not differ significantly during the pandemic period to prior, while others found a decrease (COVID-19: 33.0% versus non-COVID-19: 41.3%, P < 0.001) [12,13,18]. A study from Lim, et al. found that OHCAs during the COVID-19 pandemic were more likely to occur at home (aOR = 1.48, 95% CI 1.24–1.75) and less likely to receive BCPR (aOR = 0.70, 95% CI 0.61–0.81) although 65% of witnessed arrests were by a family member [18]. We would expect for a pandemic spread by a virus with close physical contact, that the frequency of interventions requiring laypersons to be in close proximity to patients (i.e. performing CPR or applying an AED) would decrease due to hesitancy of virus transmission [11,12,20]. Because there are less arrests occurring in public settings, the likelihood of a trained layperson able to perform CPR and/or apply an AED would also decrease [11,21,22]. In addition to lower rates of BCPR and PAD, we found lower rates of sustained ROSC, increased odds of termination and worsening survival outcomes [13,[22], [23], [24]]. Our estimates of sustained ROSC are consistent with what has been reported in the literature thus far (aOR = 0.61–0.67) [13,19,21,22]. Our study found the odds of field termination to be slightly lower than that reported in Detroit (OR = 2.36, 95% CI 1.36–4.07) or other studies (OR = 2.46, 95% CI 1.62–3.74) [19,22]. Across the US, Chan et al. found the rate of field termination to be higher in 2020, (53.9% in 2020 versus 39.9% in 2019) increasing not only in areas with high and very high COVID-19 mortality, but even those with the lowest COVID-19 mortality rates [13]. This increase in field termination is similar to our results. We found survival to hospital admission and hospital discharge to decrease, while good neurologic outcomes remained unchanged. Our study found rates of survival to hospital admission (OR = 0.56–0.65) and survival to hospital discharge (OR = 0.46–0.68) to be slightly higher than what is reported in the literature [13,19,24,25]. While the rates of good neurologic scores did not change significantly in our study, conflicting research exists, demonstrating worse or unchanged CPC scores compared to the pre-COVID-19 period [12,26]. In Texas, public interventions shown to improve OHCA survival were less frequent during the pandemic and survival outcomes worsened. Odds of field termination increased. Although more arrests occurred at home and non-shockable rhythm was more common, good neurologic outcomes remained unchanged. Future research efforts should identify contributing factors to decreased BCPR and PAD in order to develop targeted education to restrengthen the cardiac chain of survival. Clarifying reasoning behind the increase in field terminations may also shed light on decision making in resource-limited situation or with limited information. This study helps to provide additional explanations as to why OHCA worsened during the pandemic. 5 Limitations Our study had several limitations. Our study was limited to those participants within the Texas CARES catchment, which covers approximately 40.1% of the population. Those participants served by EMS agencies not participating in CARES may have different characteristics leading to a change in study results. The data set does not assess for any variations in quality or processes that may have led to differences in outcomes. For instance, any instance of BCPR is included, but measurements related to compression quality or type of CPR (i.e. compressions only) are not reported. We did not account for individual COVID seropositivity or medical comorbidities, which could affect survival outcomes. Because we conducted our analysis on the aggregate state level, we were unable to account for the microlevel impacts of human geography, which could mask trends on a smaller level. 6 Conclusion Our study adds state-wide evidence to the national phenomenon of increased OHCA incidence during COVID-19, along with worsening rates of BCPR, PAD use and outcomes. The significance of this study lies in its novelty of geographic size, temporal scope and comprehensive nature of EMS agencies by being part of the CARES database. Our results corroborate prior findings that bystander interventions, field terminations and survival outcomes worsened during the pandemic. Funding details This study was funded by a grant by the Zoll Foundation. Prior presentations Presented at the Resuscitation Science Symposium Annual Meeting, November 12–14, 2021. Credit authorship contribution statement Summer Chavez: Writing – review & editing, Writing – original draft, Conceptualization, Formal analysis, Funding acquisition, Methodology. Ryan Huebinger: Writing – review & editing, Formal analysis, Conceptualization. Hei Kit Chan: Formal analysis. Joseph Gill: Data curation. Lynn White: Data curation. Donna Mendez: Writing – review & editing. Jeffrey L. Jarvis: Data curation. Veer D. Vithalani: Data curation. Lloyd Tannenbaum: Data curation. Rabab Al-Araji: Data curation. Bentley Bobrow: Conceptualization, data curation, Study conception and design, data acquisition. Declaration of Competing Interest None. ==== Refs References 1 Mozaffarian D. Executive summary: heart disease and stroke statistics—2015 update Circulation 131 4 2015 434 441 10.1161/CIR.0000000000000157 2 Virani S.S. Heart disease and stroke statistics—2021 update Circulation 143 8 2021 e254 e743 10.1161/CIR.0000000000000950 33501848 3 Leong B.S.H. Bystander CPR and survival Singapore Med J 52 8 2011 573 575 21879214 4 Xu F. Zhang Y. Chen Y. Cardiopulmonary resuscitation training in China: current situation and future development JAMA Cardiol 2 5 2017 469 470 10.1001/jamacardio.2017.0035 28297007 5 Yan S. The global survival rate among adult out-of-hospital cardiac arrest patients who received cardiopulmonary resuscitation: a systematic review and meta-analysis Crit Care 24 1 2020 61 10.1186/s13054-020-2773-2 32087741 6 Out-of-hospital cardiac arrest: a unique medical emergency The Lancet 391 10124 2018 911 10.1016/S0140-6736(18)30552-X https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(18)30552-X/fulltext (accessed Nov. 12, 2021) 7 Bystander CPR rates rising, but survival chances worse for women www.heart.org https://www.heart.org/en/news/2018/09/24/bystander-cpr-rates-rising-but-survival-chances-worse-for-women (accessed Nov. 12, 2021) 8 Malta Hansen C. Association of Bystander and First-Responder Intervention with Survival after out-of-Hospital Cardiac Arrest in North Carolina, 2010-2013 JAMA 314 3 2015 255 264 10.1001/jama.2015.7938 26197186 9 Huebinger R. Community variations in out-of-hospital cardiac arrest care and outcomes in Texas Prehospital emergency care : official journal of the National Association of EMS Physicians and the National Association of State EMS Directors 2021 10.1080/10903127.2021.1907007 10 Cardiac Arrest Registry to Enhance Survival, “2020 CARES Annual Report,” CARES. https://mycares.net/sitepages/uploads/2021/2020_flipbook/index.html?page=6 (accessed 04, 2022). 11 Shekhar A.C. Campbell T. Blumen I. Decreased pre-EMS CPR during the first six months of the COVID-19 pandemic Resuscitation 162 2021 312 313 10.1016/j.resuscitation.2021.03.031 33838167 12 Nishiyama C. Influence of COVID-19 pandemic on bystander interventions, emergency medical service activities, and patient outcomes in out-of-hospital cardiac arrest in Osaka City, Japan Resusc Plus 5 2021 100088 10.1016/j.resplu.2021.100088 13 Chan P.S. Girotra S. Tang Y. Al-Araji R. Nallamothu B.K. McNally B. Outcomes for Out-of-Hospital Cardiac Arrest in the United States During the Coronavirus Disease 2019 Pandemic JAMA Cardiol 2020 10.1001/jamacardio.2020.6210 14 TX-CARES.com CARES \| TX-CARES https://tx-cares.com/cares/ 2019 15 About CARES « MyCares https://mycares.net/sitepages/aboutcares.jsp (accessed Nov. 08, 2020) 16 Cucinotta D. Vanelli M. WHO declares COVID-19 a pandemic Acta Biomed 91 1 2020 157 160 10.23750/abm.v91i1.9397 32191675 17 May S. Improvement in non-traumatic, out-of-hospital cardiac arrest survival in Detroit from 2014 to 2016 J Am Heart Assoc 7 16 2018 e009831 10.1161/JAHA.118.009831 18 Lim Z.J. Ponnapa Reddy M. Afroz A. Billah B. Shekar K. Subramaniam A. Incidence and outcome of out-of-hospital cardiac arrests in the COVID-19 era: a systematic review and meta-analysis Resuscitation 157 2020 248 258 10.1016/j.resuscitation.2020.10.025 33137418 19 Teoh S.E. Impact of the COVID-19 pandemic on the epidemiology of out-of-hospital cardiac arrest: a systematic review and meta-analysis Ann Intensive Care 11 1 2021 169 10.1186/s13613-021-00957-8 34874498 20 Bray J. Cartledge S. Scapigliati A. Bystander CPR in the COVID-19 pandemic Resusc Plus 4 2020 100041 10.1016/j.resplu.2020.100041 21 Lim S.L. Impact of COVID-19 on out-of-hospital cardiac arrest in Singapore Int J Environ Res Public Health 18 7 2021 3646 10.3390/ijerph18073646 33807454 22 Mathew S. Effects of the COVID-19 pandemic on out-of-hospital cardiac arrest care in Detroit Am J Emerg Med 46 2021 90 96 10.1016/j.ajem.2021.03.025 33740572 23 Lai P.H. Characteristics associated with out-of-hospital cardiac arrests and resuscitations during the novel coronavirus disease 2019 pandemic in new York City JAMA Cardiol 5 10 2020 1154 10.1001/jamacardio.2020.2488 32558876 24 Ball J. Nehme Z. Bernard S. Stub D. Stephenson M. Smith K. Collateral damage: hidden impact of the COVID-19 pandemic on the out-of-hospital cardiac arrest system-of-care Resuscitation 156 2020 157 163 10.1016/j.resuscitation.2020.09.017 32961304 25 Jaguszewski M.J. Szarpak L. Filipiak K.J. Impact of COVID-19 pandemic on out-of-hospital cardiac arrest survival rate Resuscitation 159 2021 40 41 10.1016/j.resuscitation.2020.12.013 33383099 26 Bielski K. The influence of COVID-19 on out-hospital cardiac arrest survival outcomes: an updated systematic review and Meta-analysis J Clin Med 10 23 2021 5573 10.3390/jcm10235573 34884289
PMC009xxxxxx/PMC9005368.txt	==== Front Int J Drug Policy Int J Drug Policy The International Journal on Drug Policy 0955-3959 1873-4758 Elsevier B.V. S0955-3959(22)00100-1 10.1016/j.drugpo.2022.103680 103680 Research Paper Evaluating interventions to facilitate opioid agonist treatment access among people who inject drugs in Toronto, Ontario during COVID-19 pandemic restrictions Bouck Zachary ab Scheim Ayden I. ac Gomes Tara defg Ling Vicki e Caudarella Alexander h Werb Dan agi⁎ a Centre on Drug Policy Evaluation, Unity Health Toronto, Toronto, ON, Canada b Epidemiology Division, Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada c Epidemiology and Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, PA, United States d Ontario Drug Policy Research Network, Unity Health Toronto, Toronto, ON, Canada e ICES, Toronto, ON, Canada f Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, ON, Canada g Institute for Health Policy, Management and Evaluation, University of Toronto, Toronto, ON, Canada h Department of Community and Family Medicine, University of Toronto, Toronto, ON, Canada i Division of Infectious Diseases and Global Public Health, UC San Diego, La Jolla, CA, United States ⁎ Corresponding author at: Centre on Drug Policy Evaluation, Unity Health Toronto, Li Ka Shing Knowledge Institute, Room 343, 209 Victoria Street, Toronto, ON, M5B 1T8, Canada. 13 4 2022 6 2022 13 4 2022 104 103680103680 © 2022 Elsevier B.V. All rights reserved. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background In March 2020, following a provincial COVID-19 emergency declaration, modifications to opioid agonist treatment (OAT) were introduced in Ontario, Canada to promote treatment access amid the pandemic and ongoing opioid overdose crisis. Modifications included federal exemptions to facilitate OAT prescription re-fills, extensions, and deliveries and interim treatment guidance emphasizing take-home (non-observed) doses and reduced urine drug screening for OAT patients. Methods We conducted an interrupted time series study using health administrative data from September 17th, 2019–September 21st, 2020, on 359 people who inject drugs with suspected opioid use disorder in Toronto, Ontario. We used segmented regression analyses to evaluate the joint effects of the provincial COVID-19 emergency declaration, federal OAT exemptions, and interim treatment guidance—all implemented between March 17th–23rd, 2020—on the weekly proportion of participants enrolled in OAT (i.e., ≥1 day(s) covered with methadone or buprenorphine/naloxone), with an opioid-related overdose (based on emergency department visits and hospitalizations), and who died (all-cause), and the weekly proportion of OAT-enrolled participants receiving take-home doses (i.e., ≥1 day(s) covered) and undergoing urine drug screening. Results Post-implementation, the interventions were associated with immediate absolute changes in OAT enrollment (+1.95%; 95% CI=0.04%–3.85%), receipt of take-home doses (+18.3%; 95% CI=13.2%–23.4%), and urine drug screening (-22.4%; 95% CI=[-26.9%]–[-17.9%]) and a gradual absolute increase of 0.56% in urine drug screening week-to-week (95% CI=0.27%–0.86%) beyond the pre-implementation trend. At 26 weeks post-implementation, OAT enrollment and urine drug screening approached pre-implementation levels whereas the increase in take-home doses was largely sustained (+15.0%; 95% CI=4.33%–25.6%). No post-implementation increases in opioid-related overdoses were observed. Death was not modelled (low event frequency). Conclusion Changes to OAT provision following provincial COVID-19 restrictions were associated with an immediate and sustained increase in take-home dose coverage among OAT-enrolled participants, without corresponding increases in opioid-related overdoses among all participants. Keywords Methadone Buprenorphine/naloxone Take-home doses Urine drug screening Medication for opioid use disorder Overdose ==== Body pmcIntroduction To curb the transmission of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and incidence of coronavirus disease 2019 (COVID-19), the Government of Ontario, Canada declared a provincial state of emergency on March 17th, 2020 and implemented associated measures to close non-essential services, mandate mask-wearing in indoor spaces, and facilitate physical distancing (Lawson et al., 2021; Ontario Agency for Health Protection and Promotion (Public Health Ontario), 2020). The COVID-19 pandemic emerged amidst an ongoing opioid overdose crisis in Canada, which has resulted in 20,470 hospitalizations for opioid poisoning and 15,820 opioid-related deaths from January 2016 through March 2020, with approximately 41% and 35% of these hospitalizations and deaths occurring in the province of Ontario (Public Health Agency of Canada, 2021a,b). Over the six months preceding the COVID-19 emergency declaration in Ontario, the provincial rate of opioid-related deaths increased steadily month-to-month from 5.2 per 100,000 population in September 2019 to 11.2 per 100,000 population in February 2020 (Public Health Ontario, 2021). Opioid agonist treatment (OAT) with methadone or buprenorphine/naloxone is the recommended first-line therapy for moderate to severe opioid use disorder across Canada due to its effectiveness in reducing the risk of overdose-related mortality and morbidity in treated patients, and thus represents a key intervention in the opioid overdose crisis response (Bruneau et al., 2018; Centre for Addiction and Mental Health, 2021; Eibl et al., 2017; Government of Ontario, 2018). In Ontario, as in most of Canada, OAT is primarily dispensed daily for observed ingestion in community pharmacies (Eibl et al., 2017; Government of Ontario, 2018). Before the COVID-19 pandemic, only stabilized OAT patients who regularly attended daily observed doses over a sufficient period (typically ≥2 months) and routinely cleared urine drug screening (to rule out other opioid use) could receive days’ to weeks’ worth of take-home (non-observed) methadone or buprenorphine/naloxone doses per dispensation (Eibl et al., 2017; Government of Ontario, 2018). However, during the pandemic, compliance with provincial COVID-19 emergency orders and public health guidelines by OAT dispensaries could result in reduced operating hours, reduced capacity, or temporary closures (e.g., due to outbreaks), thereby hindering treatment access at these sites (Canadian Centre on Substance Use and Addiction, 2020; Friesen et al., 2021; Nguyen & Buxton, 2021). Even without such service interruptions, Ontarians with opioid use disorder may be deterred from initiating or maintaining OAT during the pandemic to mitigate their risk of SARS-CoV-2 infection or transmission, as treatment requires regular in-person clinical encounters (Eibl et al., 2017; Government of Ontario, 2018). In response to anticipated pandemic-related barriers to treatment access and continuity of care, which might further exacerbate the opioid overdose crisis, several modifications to OAT practices were introduced shortly after the provincial COVID-19 emergency declaration (Centre for Addiction and Mental Health et al., 2020). First, Health Canada implemented temporary federal exemptions to the Controlled Drugs and Substances Act on March 19th, 2020 (presently effective through September 30th, 2026) (Health Canada, 2020). These exemptions allowed prescribers across Canada to refill or extend OAT prescriptions by phone, pharmacists to extend and transfer OAT prescriptions, and pharmacy employees to deliver OAT prescriptions to patients self-isolating at home or other locations (Health Canada, 2020). Second, on March 22nd, 2020, interim guidance on providing OAT during the COVID-19 pandemic was released for Ontario prescribers and pharmacists (Centre for Addiction and Mental Health et al., 2020). To maintain treatment access while minimizing OAT patients’ risk of SARS-CoV-2 infection, the COVID-19 OAT guidance recommended scheduling virtual versus in-person visits (where possible), reducing the frequency of urine drug screening, and facilitating take-home doses, primarily by re-evaluating patients deemed ineligible under pre-pandemic guidelines using a relaxed set of eligibility criteria (Centre for Addiction and Mental Health et al., 2020; College of Physicians & Surgeons of Ontario, 2011). Specifically, the interim guidance recommended that OAT patients who continue to use substances (including opioids) can receive take-home doses during the pandemic unless they meet ‘high-risk’ criteria: (1) intoxicated or sedated at clinical assessment; (2) unstable psychiatric comorbidity (acutely suicidal or psychotic); (3) recent overdose; or (4) current high-risk use of illicit substances (e.g., injecting high-dose intravenous illicit opioids) (Centre for Addiction and Mental Health et al., 2020). Both the exemptions and interim guidance were intended to mitigate pandemic-related barriers to OAT access in Ontario without contradicting provincial COVID-19 emergency orders and physical distancing recommendations. However, whether or to what degree these interventions facilitated OAT access among structurally vulnerable people with opioid use disorder during the pandemic is largely unknown. Due to the brief time elapsed between the implementation of the provincial COVID-19 emergency declaration (March 17th, 2020), federal OAT exemptions (March 19th, 2020), and interim COVID-19 OAT guidance (March 22nd, 2020), we cannot estimate and compare the independent effects of each intervention on OAT access (Penfold & Zhang, 2013). Therefore, our primary objective was to evaluate the joint effects of these co-occurring interventions on OAT enrollment (i.e., the proportion actively receiving treatment) within a cohort of structurally vulnerable people who inject drugs (PWID) with suspected opioid use disorder in Toronto, Ontario. We additionally assessed whether the interventions affected receipt of take-home doses and the frequency of urine drug screening among OAT-enrolled participants—targets of the interim treatment guidance—which might influence overall enrollment. Lastly, we investigated concurrent pre- and post-implementation trends in opioid-related overdoses and all-cause mortality within the study population. Methods Design and setting We conducted an interrupted time series study between September 17th, 2019, and September 21st, 2020, to assess the effects of the provincial COVID-19 emergency declaration (and associated public health measures), federal OAT exemptions, and provincial COVID-19 OAT guidance on OAT enrollment and treatment-related outcomes within a prospective community-based cohort of PWID living in Toronto, Ontario with suspected opioid use disorder. The study was divided into two 26-week-long periods—pre-implementation (September 17th, 2019–March 16th, 2020) and post-implementation (March 24th, 2020–September 21st, 2020)—around the non-calendar week in which the interventions were consecutively implemented (March 17th–23rd, 2020) (Centre for Addiction and Mental Health et al., 2020; Health Canada, 2020; Ontario Agency for Health Protection and Promotion (Public Health Ontario), 2020). Study conduct and reporting were guided by published recommendations for interrupted time series designs (Ramsay et al., 2003; Turner et al., 2020b). For context, eFigure 1 displays the daily incidence of COVID-19 cases reported in Toronto during the study period. Within our pre-implementation period, a total of 117 COVID-19 cases were reported, with the first cases reported in Toronto on January 23rd, 2020 (Public Health Ontario, 2022). From January 23rd through March 16th, 2020, a median of 0 COVID-19 cases were reported daily in the city (interquartile range = 0 to 1 cases/day) (Public Health Ontario, 2022). In contrast, our post-implementation period encompasses the majority of “Wave 1” of the COVID-19 pandemic in Toronto (exponential increase in daily case counts beginning late March 2020) as well as the start of “Wave 2” in September 2020 (Ontario Agency for Health Protection and Promotion (Public Health Ontario), 2020, 2021; Public Health Ontario, 2022). Within the post-implementation period, a total of 16,223 COVID-19 cases were reported (median [interquartile range]=71 [32 to 136] cases/day) (Public Health Ontario, 2022). Data sources Data were drawn from the ongoing Ontario integrated Supervised Injection Services study in Toronto (OiSIS-Toronto), which aims to evaluate how supervised consumption services influence health care service use and clinical outcomes among local PWID (Scheim et al., 2021a,b). At recruitment, all OiSIS-Toronto participants are ≥18 years old, live in Toronto, report injection drug use in the past six months, provide written informed consent, and complete a baseline questionnaire that collects data on their sociodemographic information, drug use behaviours, and history of treatment for substance use disorders (Scheim, Sniderman, et al., 2021). Recruitment, which began November 5th, 2018, is achieved through self-referral, snowball sampling, and community or street outreach (Scheim et al., 2021b). We identified OiSIS-Toronto as a suitable source cohort for this study for several reasons. First, the high baseline prevalence of self-reported overdose and frequent (i.e., daily or near daily) opioid injection drug use among OiSIS-Toronto participants suggests that many cohort members may be eligible for, and benefit from, OAT (Scheim et al., 2021a). Second, most OiSIS-Toronto participants are experiencing structural vulnerabilities that pose serious challenges to OAT initiation and retention. For example, over 90% of participants reported recent homelessness or unstable housing at baseline (Scheim et al., 2021b), a structural vulnerability that has been previously associated with difficulty accessing OAT (Prangnell et al., 2016). Third, OiSIS-Toronto participants are asked at baseline for additional consent to having their questionnaire data transferred and linked at ICES—a non-profit research institute authorized under Ontario's health information privacy law to collect and analyze health care and demographic data for health system evaluation and improvement (Bouck et al., 2022; Scheim et al., 2021b). The linkage process has been summarized previously in greater detail (Bouck et al., 2022). Briefly, 74% (521/701) of OiSIS-Toronto participants recruited by March 19th, 2020, consented to and were successfully linked at ICES. For these linked participants, we can access their routinely collected health care administrative data (e.g., prescription medication claims and hospitalization records) and demographic data at ICES, which enables repeated assessment of OAT enrollment and treatment-related outcomes at more regular intervals (e.g., weekly) and potentially with greater accuracy versus cohort questionnaires (completed semi-annually after baseline, all data participant-reported) (Bouck et al., 2022; Scheim et al., 2021b). We used OiSIS-Toronto participant data from the following health care administrative and demographic databases, which were linked using encoded identifiers and analyzed at ICES: the Registered Persons Database, which includes sociodemographic information and vital statistics on anyone ever issued an Ontario Health Insurance Plan (OHIP) number (OHIP is the province's publicly-funded health insurance plan); the OHIP database, which captures billing claims submitted to OHIP by physicians in Ontario; the Ontario Drug Benefit database, which captures dispensation claims for prescription medications covered, fully or partially, under Ontario's public drug insurance plan; the National Ambulatory Care Reporting System, which captures diagnoses and procedures during emergency department visits in Ontario; the Discharge Abstract Database, which captures diagnoses and procedures during inpatient hospitalizations in Ontario; the Ontario Mental Health Reporting System, which captures inpatient mental health services received by Ontarians; and the Narcotics Monitoring System, which captures all dispensations for controlled substances (including methadone and buprenorphine/naloxone) from community pharmacies across Ontario, irrespective of payer. This study was approved by Research Ethics Boards at Unity Health Toronto and Toronto Public Health. Participants We constructed a cohort comprising all OiSIS-Toronto participants who: (1) consented to and had their baseline questionnaire data linked at ICES (required for outcome measurement); (2) completed their baseline questionnaire by September 16th, 2019 (i.e., the day before our study period began); (3) self-reported non-medical opioid use in the past six months on their baseline questionnaire (taken to suggest a potential opioid use disorder); and (4) were alive as of September 16th, 2019 (according to the Registered Persons Database). Participants were followed until death or September 22nd, 2020, whichever came first. Outcomes Over the 53-week study period (September 17th, 2019–September 21st, 2020), we collected health administrative data to measure the following outcomes on a weekly basis among all remaining participants (i.e., all participants alive at the end of the preceding week): (1) OAT enrollment, defined as having ≥1 day(s) that week covered with methadone or buprenorphine/naloxone based on prescription dispensation records (eligible formulations listed in eTable 1) from that week and the prior 30 days (Fig. 1 ); (2) opioid-related overdose, defined as ≥1 emergency department visit (any diagnosis type) or inpatient hospitalization (pre-admission diagnosis) for opioid poisoning (ICD-10-CA codes T40.0 to T40.4 or T40.6) in that week (Gomes et al., 2018; Gomes et al., 2021a); and (3) death (all-cause). Among OAT-enrolled participants in a given week, we additionally measured: (1) receipt of take-home doses, defined as having ≥1 day(s) in that week covered with a take-home dose of methadone or buprenorphine/naloxone based on prescription dispensation records from that week and the prior 30 days and (2) urine drug screening, defined as ≥1 OHIP billing claim with fee code G040 to G043 that week (Morin et al., 2020; Moss et al., 2018).Fig. 1 Ascertainment of participant enrollment in opioid agonist treatment per week. Notes: OAT = opioid agonist treatment. A participant was deemed enrolled on OAT in week t if they had ≥1 eligible dispensation(s) in that week (i.e., between [d, d+6]) (e.g., panel A) or if they had a dispensation in the 30-day window preceding week t (i.e., between [d-30, d-1]) where the quantity (i.e., days supplied) dispensed on the most recent dispensation date provided coverage minimally through the first day of week t (d) (e.g., panel B). Participants were considered to not be enrolled in OAT in week t if they had no eligible dispensations during week t and (i) had no dispensation in the prior 30 days (e.g., panel C) or (ii) had a dispensation in the prior 30 days but the quantity dispensed on the most recent dispensation date did not provide coverage through the first day of week t (d) (e.g., panel D). Fig 1 Characteristics To describe the study cohort over time, we measured several characteristics using administrative data on the day before the study period commenced and the day before the provincial COVID-19 emergency declaration: age (in years); sex (male or female); Ontario Drug Benefit plan coverage, defined as age≥65 years (i.e., eligible due to age) or ≥1 prescription dispensation claim in the past 180 days in the Ontario Drug Benefit database; acute psychiatric comorbidity, defined as ≥1 emergency department visit in the past 30 days for schizophrenia (including delusional disorders) or deliberate self-harm (Gomes et al., 2021a); alcohol use disorder, defined as ≥1 emergency department visit, hospitalization, or OHIP billing claim in the past 180 days indicating alcohol use disorder (Gomes et al., 2021a); and recent opioid-related overdose, defined as ≥1 emergency department visit or hospitalization for opioid poisoning in the past 7 days (eAppendix 1). Many of these characteristics were selected as they approximate ‘high-risk’ criteria that could disqualify patients from receiving take-home doses, even under the relaxed COVID-19 OAT guidance (Centre for Addiction and Mental Health et al., 2020). Statistical analysis For each outcome, we pooled weekly data among participants and analyzed the resulting weekly proportions (expressed as percentages) using a segmented linear regression model with first-order autoregressive errors and terms for time (t [in weeks]; treated as a continuous variable), implementation (I=1 if post-implementation, I=0 if pre-implementation), and time since implementation (t-26 if post-implementation and 0 if pre-implementation; in weeks, treated as a continuous variable) (Wagner et al., 2002). Model parameters were estimated using restricted maximum likelihood estimation (Turner et al., 2020a). Week 27 (March 17th–23rd, 2020) was excluded from analysis as all interventions (provincial COVID-19 emergency declaration, federal OAT exemptions, and interim COVID-19 OAT guidance) were implemented that week on different dates. Estimated coefficients for the implementation and time since implementation terms were respectively interpreted as the collective immediate effect (level change) and gradual effect (slope change) of the interventions on the modelled outcome, expressed as absolute differences with 95% confidence intervals (CI) (Wagner et al., 2002). We additionally estimated the overall effect (i.e., the combined immediate and gradual intervention effects) on each outcome at 26 weeks post-implementation, expressed as an absolute difference with 95% CI, by comparing the predicted outcome response for the last week of observation (week 53: September 15th–21st, 2020) with the extrapolated response for that same week assuming no interventions occurred (Wagner et al., 2002). We performed all analyses using SAS V9.4 software (SAS Institute Inc.; Cary, NC). Results Of the 701 OiSIS-Toronto participants recruited by March 19th, 2020, 521 (74.32%) had their baseline questionnaire and administrative data linked at ICES. Of these participants, 359 (68.91%) met our remaining eligibility criteria and were included in the study cohort (Fig. 2 ). On the day before the study period commenced, the mean (standard deviation [SD]) age of these 359 participants was 41.27 (SD, 10.62) years, 66.57% were male, 77.44% had Ontario Drug Benefit coverage, 14.48% had a recent alcohol use disorder diagnosis, and 1.95% had an emergency department visit or hospitalization for an opioid-related overdose in the past week (Table 1 ). Compared to the day before the study commenced (September 16th, 2019), the distribution of these characteristics was largely similar among participants on the day before the provincial COVID-19 emergency declaration (March 16th, 2020), except a larger proportion (6.53% vs 1.95%) had an opioid-related overdose in the past week. Over the 26 weeks between these two dates, 7 participants died. Over the remainder of the study period (week 27–53), an additional ≤5 participants died (exact number of post-implementation deaths suppressed to prevent re-identification of individual participants per ICES policies). Due to the low number of events, death was not subsequently modelled.Fig. 2 Flow of participants into the study. Notes: OiSIS-Toronto = Ontario integrated Supervised Injection Services Toronto. Fig 2 Table 1 Characteristics of participants on day before study period commenced and day before provincial COVID-19 state of emergency declaration – the Ontario integrated Supervised Injection Services study in Toronto. Table 1Characteristic Day before study period commenced (September 16th, 2019) Day before provincial COVID-19 emergency declaration (March 16th, 2020) No. of remaining participants 359 352 Age (y), mean (SD) 41.27 (10.62) 41.83 (10.66) Male, n (%) 239 (66.57) 234 (66.48) ODB coveragea, n (%) 278 (77.44) 273 (77.56) Acute psychiatric comorbidityb, n (%) ≤5 (NR) ≤5 (NR) Alcohol use disorderc, n (%) 52 (14.48) 46 (13.07) Recent opioid-related overdosed, n (%) 7 (1.95) 23 (6.53) Notes: SD = standard deviation; ODB = Ontario Drug Benefit; NR = not reported. Cell counts between 1-5 were suppressed (reported as ‘≤5’) and corresponding proportions were not reported in accordance with ICES policies to prevent back calculation of these values and possible identification of individual participants. a Defined as age≥65 or ≥1 dispensation(s) in the ODB database in the past 180 days. b Defined as ≥1 emergency department visit for schizophrenia (including delusional disorders) or deliberate self-harm in the past 30 days. c Defined as ≥1 emergency department visit, hospitalization, or physician claim with an alcohol use disorder diagnostic code in the past 180 days. d Defined as ≥1 emergency department visit or hospitalization for opioid poisoning in the past 7 days. OAT enrollment and opioid-related overdoses Our measures of weekly OAT enrollment and opioid-related overdoses share a common denominator, i.e., the number of participants remaining (still alive) in the cohort at the end of the preceding week. The average weekly denominator for these outcomes was 353 participants (SD, 2.99). Fig. 3 plots the observed and predicted weekly proportions of participants enrolled in OAT.Fig. 3 Weekly proportion of participants enrolled in OAT between September 17th, 2019 and September 21st, 2020 – Ontario integrated Supervised Injection Services study in Toronto. Notes: Observed proportions represented by blue ‘x's, the solid blue lines are the fitted regression pre- and post-implementation trendlines, and the hatched blue line represents the projected trend had the interventions not been implemented (i.e., counterfactual). The fitted trendlines and counterfactual were obtained from a segmented linear autoregressive error regression model (equation provided in figure). The vertical hatched red line indicates the week in which the interventions were implemented (week 27: March 17th–23rd, 2020), which was excluded from all analyses. Fig 3 Overall, 36.49% (131/359) of participants were enrolled in OAT the first week. Based on the fitted model, there was a non-statistically significant decrease of 0.08% per week in OAT enrollment (95% CI -0.23% to 0.07%) during the pre-implementation period. Comparing the first full week post-implementation (week 28: March 24th–30th, 2020) and the last week pre-implementation (week 26: March 10th–16th, 2020), the interventions were associated with an immediate increase of 1.95% in OAT enrollment (95% CI 0.04% to 3.85%). Furthermore, post-implementation, OAT enrollment gradually declined an additional 0.17% week-to-week over and above the pre-implementation trend (95% CI -0.42% to 0.08%). At 26 weeks post-implementation, the interventions were not associated with a statistically significant difference in OAT enrollment compared to if they had not been implemented (-2.57% difference; 95% CI -9.23% to 4.09%). The weekly number of participants that experienced an opioid-related overdose (based on emergency department visit and hospitalization records) was low throughout the study period. We observed that ≤5 participants had an opioid-related overdose for 50 of 53 weeks measured (including week 1: ≤5/359 or ≤1.39% with an opioid-related overdose that week), with a maximum weekly value of 11 participants. Correspondingly, we did not plot the observed data for this outcome to prevent possible re-identification of individual participants based on small cell sizes (i.e., numerator values between 1 to 5) in accordance with ICES policies; however, we fit a segmented linear regression model to the observed data. Based on the fitted model (Yt=0.33+0.03t−0.32I−0.05[t−26]I+εt), where εt=0.29εt−1+wtandwt∼N(0,0.44), the proportion of participants with an opioid-related overdose remained relatively constant during the pre-implementation period (0.03% increase per week; 95% CI -0.01% to 0.08%). Post-implementation, the proportion experiencing an opioid-related overdose immediately decreased by 0.32% (95% CI -1.24% to 0.60%) and gradually decreased by an additional 0.05% per week beyond the pre-implementation trend (95% CI -0.11% to 0.02%); however, neither change was statistically significant. At 26 weeks post-implementation, the interventions were not associated with a statistically significant change in the proportion of participants experiencing an opioid-related overdose versus had the interventions not been implemented (-1.53% difference; 95% CI -3.42% to 0.36%). Take-home doses and urine drug screening Fig. 4 plots the observed and predicted weekly proportions of OAT-enrolled participants who received take-home doses and underwent urine drug screening. The average weekly denominator was 116 OAT-enrolled participants (SD, 7.89) for both outcomes. On average, 90.07% of OAT-enrolled participants each week were last dispensed methadone (range = 87.10% to 92.86%).Fig. 4 Weekly proportion of participants enrolled in OAT that received take-home OAT doses (panel a) and underwent urine drug screening (panel b) between September 17th, 2019 and September 21st, 2020 – Ontario integrated Supervised Injection Services study in Toronto. Notes: Observed proportions represented by blue ‘x's, the solid blue lines are the fitted regression pre- and post-implementation trendlines, and the hatched blue line represents the projected trend had the interventions not been implemented (i.e., counterfactual). The fitted trendlines and counterfactual were obtained from segmented linear autoregressive error regression analyses (model equation provided in each figure). The vertical hatched red line indicates the week in which the interventions were implemented (week 27: March 17th–23rd, 2020), which was excluded from all analyses. Fig 4 Overall, 18.32% (24/131) of OAT-enrolled participants received take home doses in week 1. Based on the fitted model, this proportion remained relatively unchanged during the pre-implementation period (0.04% increase per week; 95% CI -0.21% to 0.29%) (Fig. 4 a). Post-implementation, the interventions were associated with a statistically significant immediate increase of 18.31% in the proportion of OAT-enrolled participants receiving take-home doses (95% CI 13.21% to 23.40%); however, no gradual effect was observed post-implementation (0.13% decrease per week beyond the pre-implementation trend; 95% CI -0.49% to 0.24%). At 26 weeks post-implementation, the interventions were associated with a statistically significant increase of 14.98 additional OAT-enrolled participants receiving take-home doses per 100 (95% CI 4.33% to 25.62%) versus if the interventions were never implemented. We observed that 68.70% (90/131) of OAT-enrolled participants underwent urine drug screening in the first week. Based on the fitted model, this proportion was relatively stable during the pre-implementation period (0.15% decrease per week; 95% CI -0.36% to 0.06%) (Fig. 4 b). Post-implementation, the interventions were associated with a statistically significant immediate decrease of 22.38% in the proportion of OAT-enrolled participants undergoing urine drug screening (95% CI -26.89% to -17.88%) and a statistically significant gradual increase of 0.56% in urine drug screening per week beyond the pre-implementation trend (95% CI 0.27% to 0.86%). At 26-weeks post-implementation, the interventions were associated with 7.72 fewer OAT-enrolled patients per 100 undergoing urine drug screening (95% CI -16.53% to 1.10%) compared to if the interventions were never implemented; however, this difference was not statistically significant. Discussion We evaluated the effects of a provincial COVID-19 emergency declaration, federal exemptions, and interim treatment guidance on OAT enrollment and related outcomes—measured weekly from September 17th, 2019 through September 21st, 2020—among 359 Toronto-based PWID with suspected opioid use disorder. Post-implementation, the interventions were collectively associated with a slight immediate increase in OAT enrollment among all participants (1.95%) and substantial immediate changes in receipt of take-home doses (any quantity; 18.31% increase) and urine drug screening (22.38% decrease) among OAT-enrolled participants. By the final week of observation, OAT enrollment and urine drug screening reverted towards expected levels had the interventions never occurred whereas the increase in receipt of take-home doses was largely sustained (14.98% increase). The interventions were not associated with any changes in opioid-related overdoses among all participants. Due to the low number of deaths, we could not evaluate the joint impact of the interventions on all-cause mortality (outcome not modelled). These findings suggest that rapid modifications to OAT delivery at the beginning of the COVID-19 pandemic in Ontario may have helped mitigate anticipated pandemic-related barriers to treatment access within the study cohort. Although OAT enrollment did not meaningfully increase among participants, the absence of post-implementation decreases in enrollment is arguably a success of the COVID-19-related OAT modifications (federal exemptions and interim treatment guidance), as pandemic restrictions (including provincial emergency orders) were expected to worsen OAT access and thereby enrollment in this population (Centre for Addiction and Mental Health et al., 2020). The relatively static level of OAT enrollment post-implementation is likely owed to the immediate and sustained increase in the proportion of OAT-enrolled participants receiving take-home doses and the immediate, albeit temporary, decrease in the proportion undergoing weekly urine drug screening. Specifically, the increased likelihood of receiving take-home doses following the OAT modifications and corresponding decreased likelihood of routine urine drug screening (at least initially) may have facilitated treatment retention despite pandemic restrictions by reducing OAT patients’ in-person clinical encounters and affording greater flexibility in their dosing schedules (Corace et al., 2022; Haasen & Brink, 2006; Sarasvita et al., 2012; Schaub et al., 2010). These post-implementation changes are likely attributable to the provincial interim treatment guidance, which explicitly recommended that prescribers and pharmacists reduce the frequency of urine drug screening and observed doses for OAT patients during the COVID-19 pandemic (Centre for Addiction and Mental Health et al., 2020). These inferences are supported by analogous findings from the broader OAT patient population in Ontario. A study by Kitchen et al. observed that the number of Ontarians actively being treated with methadone or buprenorphine/naloxone was unchanged following the provincial COVID-19 emergency declaration and interim OAT guidance (Kitchen et al., 2022). Though enrollment remained stable, as in our study, the interim treatment guidance was associated with immediate increases in the weekly proportions of OAT patients receiving extended supplies of take-home methadone or buprenorphine/naloxone doses (i.e., ≥7 days’ worth per dispensation) (Kitchen et al., 2022). Another population-based analysis of OAT-enrolled Ontarians found that individuals who received increased take-home doses in the first 30 days following the interim treatment guidance (e.g., transitioned from daily observed dosing to any take-home doses) were significantly less likely to pause or discontinue treatment in the next six months versus those without increased take-home doses (Gomes et al., 2022). Taken together with our findings, these results suggest that, as intended, the provincial interim treatment guidance led to increased provision of take-home doses to OAT patients both in the general population and our study cohort of structurally vulnerable PWID, which facilitated treatment retention in the early stages of the COVID-19 pandemic (Gomes et al., 2022; Kitchen et al., 2022). Increases in take-home doses for OAT patients have also been observed in other jurisdictions that released similar guidance de-emphasizing observed doses during the COVID-19 pandemic. In the United States, a federal exemption and national guidelines were implemented in March 2020 to allow prolonged take-home doses of ≤28 days for stable patients or ≤14 days for less stable patients on methadone for opioid use disorder (Substance Abuse and Mental Health Services Administration (SAMHSA), 2020). A pre-post analysis of 194 methadone patients in Spokane, Washington found that patients received, on average, an additional 41.4 days’ worth of take-home doses over the three months following the guidelines versus the preceding three months (Amram et al., 2021). Similar to Ontario, the Australian government released national guidelines emphasizing virtual visits, less frequent urine drug screening, and reducing observed dosing for OAT patients during the pandemic (Lintzeris et al., 2021). Based on three public treatment services in Sydney, Australia, Lintzeris and colleagues found that the proportion of OAT patients receiving any take-home methadone or sublingual buprenorphine doses increased considerably in the four months after guideline-based service changes (67% or 210/314) versus the preceding four-month period (23% or 86/378) (Lintzeris et al., 2021). Lastly, in Ukraine, the national Ministry of Health released interim guidance in March 2020 relaxing an existing requirement of six months of sobriety for OAT patients to receive take-home doses (Meteliuk et al., 2021). As in our study, OAT enrollment increased negligibly post-guidance but the proportion of OAT patients in Ukraine receiving take-home doses increased substantially in the first 60-days post-guidance (82.2% or 10,766/13,097) versus the last 60-days pre-guidance (57.5% or 7,381/12,837) (Meteliuk et al., 2021). While the impact of interim treatment guidance promoting take-home doses on opioid-related overdoses was not evaluated in the preceding international studies, annualized mortality among OAT patients in Ukraine was comparably low between the post- and pre-guidance periods (Meteliuk et al., 2021). We found no evidence of immediate or gradual increases in opioid-related overdoses within the overall cohort following the provincial COVID-19 emergency declaration and corresponding OAT modifications. This important result is somewhat surprising given the elevated overdose risk in the source cohort (i.e., 38.6% of OiSIS-Toronto participants reported a recent non-fatal overdose at baseline) (Scheim et al., 2021a) and our restriction to participants with a suspected moderate or severe opioid use disorder. In contrast with our findings, prior analyses have demonstrated significant increases in the rate of fatal opioid-related overdoses in Toronto and Ontario during the first months of the pandemic overlapping with our post-implementation period (Gomes et al., 2021b; Toronto Public Health, 2021). These conflicting trends may be because most OiSIS-Toronto participants—all of whom were PWID—did not qualify for take-home OAT doses under pre-pandemic guidelines (College of Physicians & Surgeons of Ontario, 2011). In other words, compared to the average OAT patient in Ontario, cohort members may have been more likely to initiate take-home doses following the relaxed, interim treatment guidance, as evidenced by the drastic post-guidance increase in take-home dose coverage among OAT-enrolled participants. Relatedly, Gomes and colleagues found that Ontarians who transitioned from daily dispensed methadone to any quantity of take-home doses in the 30 days after the interim guidance were 27% less likely to experience an opioid-related overdose over the next six months of the pandemic versus methadone patients who did not initiate take-home doses (Gomes et al., 2022). Given this protective association and the prominence of methadone (versus buprenorphine/naloxone) dispensing in our study, the absence of post-implementation increases in opioid-related overdoses for the overall cohort could be due to increases in take-home dose provision among OAT-enrolled participants (with corresponding increases in treatment retention and decreases in overdose risk) (Gomes et al., 2022), which offset increases in overdose risk within the non-OAT-enrolled subset over time. Limitations Several limitations of our study merit discussion. First, our opioid-related overdose measure relied on data from emergency departments and inpatient hospital stays, and therefore does not capture overdoses attended to in the community or confirmed opioid-related deaths where the individual was not transported to hospital (Gomes et al., 2018). Therefore, this outcome underestimates the true incidence of these events. Second, our findings may not be generalizable to the broader population of PWID in Toronto, as the source cohort (OiSIS-Toronto) is a convenience sample primarily composed of supervised consumption service clients (Scheim et al., 2021b). Third, the health administrative databases used in this study lack information on participant characteristics that might influence their access to OAT and eligibility for take-home doses even under relaxed pandemic criteria (e.g., homeless and unable to safely store take-home doses). Fourth, in using self-reported non-medical opioid use at OiSIS-Toronto baseline to identify participants with a suspected opioid use disorder, some individuals may have ceased non-medical opioid use before the study period began; including these non-OAT-eligible participants would underestimate overall treatment enrollment. Fifth, due to the global nature of the COVID-19 pandemic and similar public health responses undertaken elsewhere, we could not identify a concurrent, external control group for analysis, which could have strengthened (or challenged) our attribution of post-implementation outcome changes to measured interventions (Jandoc et al., 2015; Lopez Bernal et al., 2016). Conclusions Although provincial COVID-19 emergency measures were expected to worsen treatment access in Toronto, Ontario, it appears that rapid changes to OAT provision (via federal exemptions and interim treatment guidance) resulted in an immediate and lasting increase in take-home dose coverage among OAT-enrolled participants in our study, without corresponding increases in opioid-related overdoses among all participants. Therefore, it may be worthwhile to consider long-term adoption of these OAT modifications beyond the COVID-19 pandemic in populations that are comparable to our study cohort (i.e., structurally vulnerable people who inject drugs). Data availability The dataset used in this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g., healthcare organizations and government) prohibit ICES from making the dataset publicly available, access may be granted to those who meet pre-specified criteria for confidential access, available at www.ices.on.ca/DAS (email: das@ices.on.ca). The full dataset creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification. Declarations of Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Appendix Supplementary materials Image, application 1 Acknowledgments The authors would like to express their gratitude to all participants and staff involved in the Ontario integrated Supervised Injection Services Toronto (OiSIS-Toronto) cohort study. We acknowledge the land on which we conducted this research is the traditional territory of many nations including the Mississaugas of the Credit, the Anishnabeg, the Chippewa, the Haudenosaunee, and the Wendat Peoples, and home to many diverse First Nations, Inuit, and Métis Peoples. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC). This study also received funding from a COVID-19 Rapid Research–Social Policy and Public Health Responses Operating Grant (application year: 2020; application #: 447989) awarded by the Canadian Institutes of Health Research. Parts of this material are based on data and information compiled and provided by Ontario Ministry of Health (MOH) and the Canadian Institute for Health Information. The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred. We thank IQVIA Solutions Canada Inc. for use of their Drug Information File. Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.drugpo.2022.103680. ==== Refs References Amram O. Amiri S. Thorn E.L. Lutz R. Joudrey P.J. Changes in methadone take-home dosing before and after COVID-19 Journal of Substance Abuse Treatment 108552 2021 10.1016/j.jsat.2021.108552 Bouck Z. Tricco A.C. Rosella L.C. Ling V. Gomes T. Tadrous M. Fox M.P. Scheim A.I. Werb D. Validation of self-reported opioid agonist treatment among people who inject drugs using prescription dispensation records Epidemiology 33 2 2022 10.1097/EDE.0000000000001443 Bruneau J. Ahamad K. Goyer M.-È. Poulin G. Selby P. Fischer B. Wild T.C. Wood E. Management of opioid use disorders: A national clinical practice guideline Canadian Medical Association Journal 190 9 2018 E247 E257 10.1503/cmaj.170958 29507156 Canadian Centre on Substance Use and Addiction. (2020). Impacts of the COVID-19 pandemic on substance use treatment capacity in Canada. https://www.ccsa.ca/sites/default/files/2020-12/CCSA-COVID-19-Impacts-Pandemic-Substance-Use-Treatment-Capacity-Canada-2020-en.pdf. Centre for Addiction and Mental Health. (2021). Opioid agonist therapy: A synthesis of canadian guidelines for treating opioid use disorder. 52. Centre for Addiction and Mental Health, Ontario Medication Association, & META:PHI. (2020). COVID-19 opioid agonist treatment guidance. https://www.metaphi.ca/assets/documents/provider%20tools/COVID19_OpioidAgonistTreatmentGuidance.pdf. College of Physicians & Surgeons of Ontario. (2011). Methadone maintenance treatment program standards and clinical guidelines. Corace K. Suschinsky K. Wyman J. Leece P. Cragg S. Konefal S. …Hutton B. Evaluating how has care been affected by the Ontario COVID-19 Opioid Agonist Treatment Guidance: Patients’ and prescribers’ experiences with changes in unsupervised dosing International Journal of Drug Policy 102 2022 103573 10.1016/j.drugpo.2021.103573 Eibl J.K. Morin K. Leinonen E. Marsh D.C. The state of opioid agonist therapy in Canada 20 years after federal oversight Canadian Journal of Psychiatry. Revue Canadienne de Psychiatrie 62 7 2017 444 450 10.1177/0706743717711167 28525291 Friesen E.L. Kurdyak P.A. Gomes T. Kolla G. Leece P. Zhu L. Toombs E. O'Neill B. Stall N.M. Jüni P. Mushquash C.J. Mah L. The impact of the COVID-19 pandemic on opioid-related harm in Ontario 2021 Ontario COVID-19 Science Advisory Table 10.47326/ocsat.2021.02.42.1.0 Gomes T. Campbell T.J. Kitchen S.A. Garg R. Bozinoff N. Men S. Tadrous M. Munro C. Antoniou T. Werb D. Wyman J. Association between increased dispensing of opioid agonist therapy take-home doses and opioid overdose and treatment interruption and discontinuation Journal of the American Medical Association 327 9 2022 846 855 10.1001/jama.2022.1271 35230394 Gomes T. Campbell T. Tadrous M. Mamdani M.M. Paterson J.M. Juurlink D.N. Initial opioid prescription patterns and the risk of ongoing use and adverse outcomes Pharmacoepidemiology and Drug Safety 30 3 2021 379 389 10.1002/pds.5180 33300138 Gomes T. Khuu W. Martins D. Tadrous M. Mamdani M.M. Paterson J.M. Juurlink D.N. Contributions of prescribed and non-prescribed opioids to opioid related deaths: Population based cohort study in Ontario, Canada British Medical Journal 2018 k3207 10.1136/bmj.k3207 30158106 Gomes, T., Kitchen, S. A., & Murray, R. (2021b). Measuring the burden of opioid-related mortality in Ontario, Canada, during the COVID-19 pandemic. 4(5), e2112865. 10.1001/jamanetworkopen.2021.12865. Government of Ontario. (2018). Opioid use disorder: Care for people 16 years of age and older. 56. Haasen C. Brink W. Innovations in agonist maintenance treatment of opioid-dependent patients Current Opinion in Psychiatry 19 2006 10.1097/01.yco.0000245759.13997.9d Health Canada. (2020). CDSA exemption and interpretive guide for controlled substances. https://abpharmacy.ca/sites/default/files/CDSA_Exemption_and_interpretive_guide_for_controlled_substances.pdf. Jandoc R. Burden A.M. Mamdani M. Lévesque L.E. Cadarette S.M. Interrupted time series analysis in drug utilization research is increasing: Systematic review and recommendations Journal of Clinical Epidemiology 68 8 2015 950 956 10.1016/j.jclinepi.2014.12.018 25890805 Kitchen S.A. Campbell T.J. Men S. Bozinoff N. Tadrous M. Antoniou T. …Gomes T. Impact of the COVID-19 pandemic on the provision of take-home doses of opioid agonist therapy in Ontario, Canada: A population-based time-series analysis International Journal of Drug Policy 103 2022 103644 10.1016/j.drugpo.2022.103644 Lawson, T., Nathans, L., Goldenberg, A., Fimiani, M., & Boire-Schwab, D. (2021). COVID-19: Emergency measures tracker. COVID-19: emergency measures tracker. https://www.mccarthy.ca/en/insights/articles/covid-19-emergency-measures-tracker. Lintzeris N. Deacon R.M. Hayes V. Cowan T. Mills L. Parvaresh L. Harvey Dodds L. Jansen L. Dojcinovic R. Leung M.C. Demirkol A. Finch T. Mammen K. Opioid agonist treatment and patient outcomes during the COVID-19 pandemic in south east Sydney, Australia Drug and Alcohol Review 2021 10.1111/dar.13382 n/a(n/a) Lopez Bernal J. Cummins S. Gasparrini A Interrupted time series regression for the evaluation of public health interventions: A tutorial International Journal of Epidemiology 2016 dyw098 10.1093/ije/dyw098 Meteliuk A. Galvez de Leon S.J. Madden L.M. Pykalo I. Fomenko T. Filippovych M. Farnum S.O. Dvoryak S. Islam Z.M. Altice F.L. Rapid transitional response to the COVID-19 pandemic by opioid agonist treatment programs in Ukraine Journal of Substance Abuse Treatment 121 2021 108164 10.1016/j.jsat.2020.108164 Morin K.A. Prevost C.R. Eibl J.K. Franklyn M.T. Moise A.R. Marsh D.C. A retrospective cohort study evaluating correlates of deep tissue infections among patients enrolled in opioid agonist treatment using administrative data in Ontario, Canada PLOS One 15 4 2020 e0232191 10.1371/journal.pone.0232191 Moss E. McEachern J. Adye-White L. Priest K.C. Gorfinkel L. Wood E. Cullen W. Klimas J. Large variation in provincial guidelines for urine drug screening during opioid agonist treatment in Canada The Canadian Journal of Addiction 9 2 2018 6 9 10.1097/CXA.0000000000000015 30410962 Nguyen T. Buxton J.A. Pathways between COVID-19 public health responses and increasing overdose risks: A rapid review and conceptual framework International Journal of Drug Policy 93 2021 103236 10.1016/j.drugpo.2021.103236 Ontario Agency for Health Protection and Promotion (Public Health Ontario). (2020). COVID-19 in Ontario: A summary of wave 1 transmission patterns and case identification. (p. 13). https://www.publichealthontario.ca/-/media/documents/ncov/epi/2020/08/covid-19-wave-1-transmission-patterns-epi-summary.pdf?la=en. Ontario Agency for Health Protection and Promotion (Public Health Ontario). (2021). Trends of COVID-19 incidence in Ontario. https://www.publichealthontario.ca/-/media/documents/ncov/epi/covid-19-epi-trends-incidence-ontario.pdf?la=en. Penfold R.B. Zhang F. Use of interrupted time series analysis in evaluating health care quality improvements Academic Pediatrics 13 6 2013 S38 S44 10.1016/j.acap.2013.08.002 24268083 Prangnell A. Daly-Grafstein B. Dong H. Nolan S. Milloy M.-J. Wood E. Kerr T. Hayashi K. Factors associated with inability to access addiction treatment among people who inject drugs in Vancouver, Canada Substance Abuse Treatment, Prevention, and Policy 11 1 2016 9 10.1186/s13011-016-0053-6 Public Health Agency of Canada. (2021a). Apparent opioid and stimulant toxicity deaths—Surveillance of opioid- and stimulant-related harms in Canada. https://health-infobase.canada.ca/src/doc/SRHD/UpdateDeathsSep2021.pdf. Public Health Agency of Canada. (2021b). Opioid and stimulant poisoning hospitalizations—Surveillance of opioid- and stimulant-related harms in Canada. https://health-infobase.canada.ca/src/doc/SRHD/UpdateHospitalizationsSep2021.pdf. Public Health Ontario. (2021). Interactive opioid tool. Interactive opioid tool – Opioid-related morbidity and mortality in Ontario. https://www.publichealthontario.ca/en/data-and-analysis/substance-use/interactive-opioid-tool. Public Health Ontario. (2022). Ontario COVID-19 data tool. Ontario COVID-19 Data Tool. https://www.publichealthontario.ca/en/data-and-analysis/infectious-disease/covid-19-data-surveillance/covid-19-data-tool. Ramsay C.R. Matowe L. Grilli R. Grimshaw J.M. Thomas R.E. Interrupted time series designs in health technology assessment: Lessons from two systematic reviews of behavior change strategies International Journal of Technology Assessment in Health Care 19 4 2003 613 623 10.1017/S0266462303000576 15095767 Sarasvita R. Tonkin A. Utomo B. Ali R. Predictive factors for treatment retention in methadone programs in Indonesia Journal of Substance Abuse Treatment 42 2012 10.1016/j.jsat.2011.07.009 Schaub M. Chtenguelov V. Subata E. Weiler G. Uchtenhagen A. Feasibility of buprenorphine and methadone maintenance programmes among users of home made opioids in Ukraine International Journal of Drug Policy 21 2010 10.1016/j.drugpo.2009.10.005 Scheim A.I. Bouck Z. Tookey P. Hopkins S. Sniderman R. McLean E. Garber G. Baral S. Rourke S.B. Werb D. Supervised consumption service use and recent non-fatal overdose among people who inject drugs in Toronto, Canada International Journal of Drug Policy 87 2021 102993 10.1016/j.drugpo.2020.102993 Scheim A.I. Sniderman R. Wang R. Bouck Z. McLean E. Mason K. Bardwell G. Mitra S. Greenwald Z.R. Thavorn K. Garber G. Baral S.D. Rourke S.B. Werb D. The Ontario integrated supervised injection services cohort study of people who inject drugs in Toronto, Canada (OiSIS-Toronto): Cohort profile Journal of Urban Health 2021 10.1007/s11524-021-00547-w Substance Abuse and Mental Health Services Administration (SAMHSA). (2020). Opioid Treatment Program (OTP) guidance. https://www.samhsa.gov/sites/default/files/otp-guidance-20200316.pdf. Toronto Public Health. (2021). Toronto overdose information system—deaths. https://public.tableau.com/app/profile/tphseu/viz/TOISDashboard_Final/Deaths. Turner S.L. Forbes A.B. Karahalios A. Taljaard M. McKenzie J.E. Evaluation of statistical methods used in the analysis of interrupted time series studies: A simulation study [Preprint] 2020 Public and Global Health 10.1101/2020.10.12.20211706 Turner S.L. Karahalios A. Forbes A.B. Taljaard M. Grimshaw J.M. Cheng A.C. Bero L. McKenzie J.E. Design characteristics and statistical methods used in interrupted time series studies evaluating public health interventions: A review Journal of Clinical Epidemiology 122 2020 1 11 10.1016/j.jclinepi.2020.02.006 32109503 Wagner A.K. Soumerai S.B. Zhang F. Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research Journal of Clinical Pharmacy and Therapeutics 27 4 2002 299 309 10.1046/j.1365-2710.2002.00430.x 12174032
PMC009xxxxxx/PMC9005369.txt	==== Front Biomed Signal Process Control Biomed Signal Process Control Biomedical Signal Processing and Control 1746-8094 1746-8094 Elsevier Ltd. S1746-8094(22)00237-3 10.1016/j.bspc.2022.103715 103715 Article IoMT-fog-cloud based architecture for Covid-19 detection Khelili M.A. a⁎ Slatnia S. a Kazar O. ab Harous S. c a Department of Computer Science, Smart Computer Science Laboratory, (University of Mohamed Khider, Biskra, Algeria), Biskra, Algeria b Department of Information Systems and Security, College of Information Technology, (United Arab Emirate University, United Arab Emirate), United Arab Emirates c Department of Computer Science, College of Computing and Informatics, (University of Sharjah, Al Ain, United Arab Emirates), Al Ain, United Arab Emirates ⁎ Corresponding author. 13 4 2022 7 2022 13 4 2022 76 103715103715 10 1 2022 23 2 2022 9 4 2022 © 2022 Elsevier Ltd. All rights reserved. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Limitations of available literature Nowadays, coronavirus disease 2019 (COVID-19) is the world-wide pandemic due to its mutation over time. Several works done for covid-19 detection using different techniques however, the use of small datasets and the lack of validation tests still limit their works. Also, they depend only on the increasing the accuracy and the precision of the model without giving attention to their complexity which is one of the main conditions in the healthcare application. Moreover, the majority of healthcare applications with cloud computing use centralization transmission process of various and vast volumes of information what make the privacy and security of personal patient’s data easy for hacking. Furthermore, the traditional architecture of the cloud showed many weaknesses such as the latency and the low persistent performance. Method proposed by the author with technical information In our system, we used Discrete Wavelet transform (DWT) and Principal Component Analysis (PCA) and different energy tracking methods such as Teager Kaiser Energy Operator (TKEO), Shannon Wavelet Entropy Energy (SWEE), Log Energy Entropy (LEE) for preprocessing the dataset. For the first step, DWT used to decompose the image into coefficients where each coefficient is vector of features. Then, we apply PCA for reduction the dimension by choosing the most essential features in features map. Moreover, we used TKEO, SHEE, LEE to track the energy in the features in order to select the best and the most optimal features to reduce the complexity of the model. Also, we used CNN model that contains convolution and pooling layers due to its efficacity in image processing. Furthermore, we depend on deep neurons using small kernel windows which provide better features learning and minimize the model's complexity. The used DWT-PCA technique with TKEO filtering technique showed great results in terms of noise measure where the Peak Signal-to-Noise Ratio (PSNR) was 3.14 dB and the Signal-to-Noise Ratio (SNR) of original and preprocessed image was 1.48, 1.47 respectively which guaranteed the performance of the filtering techniques. The experimental results of the CNN model ensure the high performance of the proposed system in classifying the covid-19, pneumonia and normal cases with 97% of accuracy, 100% of precession, 97% of recall, 99% of F1-score, and 98% of AUC. Advantages and application of proposed method The use of DWT-PCA and TKEO optimize the selection of the optimal features and reduce the complexity of the model. The proposed system achieves good results in identifying covid-19, pneumonia and normal cases. The implementation of fog computing as an intermediate layer to solve the latency problem and computational cost which improve the Quality of Service (QoS) of the cloud. Fog computing ensure the privacy and security of the patients’ data. With further refinement and validation, the IFC-Covid system will be real-time and effective application for covid-19 detection, which is user friendly and costless. Keywords Covid-19 Internet of Medical Things (IoMT) Cloud computing Fog computing Deep learning Quality of Service (QoS) ==== Body pmc1 Introduction The World Health Organization (WHO) considered Coronavirus Disease of 2019 as a highly contagious viral illness caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and Middle Eastern Respiratory Syndrome (MERS). After the first appearance of COVID-19, on December 31, 2019, in Wuhan, Hubei Province, China, over than 4.8 million confirmed cases were reported because of their rapid prevalence in the whole world [1]. Meanwhile, the virus has the ability of genetic evolution with the development of mutations over time, which create new variants with different characteristics. According to WHO, four variants of concern (VOC) were discovered [2]. The first variant of the virus was in December 2020 in the United Kingdom (UK) called Alpha (B.1.1.7). During the same period, two other variants called Gamma (P.1 lineage) and Delta (B.1.617.2) were detected in Brazil and India respectively. While South Africa has discovered Beta (B.1.351) variant. Experts from various organizations proposed various types of vaccines such as: Whole Virus, Protein Subunit, Viral Vector, and Nucleic Acid (RNA and DNA). The main objective of these categories is to get immunity to the virus, and stop transmission by smuggling the antigen into the body, or by using the body's cells to make the viral antigen [3]. WHO approved several vaccines for emergency or full use, such as Oxford–AstraZeneca [4], Pfizer-BioNTech [5], Sinopharm-BBIBP [6], Moderna [7], Sinovac [8], Janssen [9], and Sputnik V [10]. Although vaccines can help prevent most people from getting sick with the COVID-19 virus, they cannot protect everyone. In most cases, people who take all of the recommended doses still have the possibility to get infected [11]. As result, Nowadays, the detection of COVID-19 in its first stages is still a necessity of time within the catastrophic impact of this pandemic. The World Health Organization declared that Real-Time Reverse Transcriptase Chain Reaction or RT-PCR is the gold standard method for detecting COVID-19 because of their possibility in detecting the ribonucleic acid (RNA) of the virus [12]. In the literature, many researchers used Xray [13], [14] and CT scan [15], [16] images for Covid-19 detection and diagnosing with the help of AI technics. These techniques have resulted into very promising outcomes. 2 Background Table 1 and Table 2 summarize several techniques that use ML and DL approaches to detect Covid-19.Table 1 State-of-art-work that used ML and DL for Covid-19. Ref. Dataset Method used Evaluation Metrics Research challenges [17] Dataset of X-ray images and CT scan images provided by Dr. Joseph Cohen from the GitHub repository + COVID-19 Chest X-ray Dataset Initiative + Chest X-ray image dataset provided by Kaggle VGG16, InceptionV3, ResNet50, Xception, DenseNet121 Decision Tree, Random Forest, AdaBoost, Bagging, SVM Accuracy, Sensitivity, Specificity, F1-Score, AUC Depend only on the classification of COVID-19 and normal X-ray images. They did not include other chest disease images such as bacterial pneumonia, viral pneumonia. [18] Chest X-ray images of COVID-19 obtained from the Kaggle database PA (prophet algorithm), ARIMA, and LSTM + VGG16 Precision, Recall, and F-measure, Accuracy, Root Mean Square Error (RMSE) Depend only on the classification of COVID-19 and normal X-ray images. They did not include other chest disease images such as bacterial pneumonia, viral pneumonia. [19] X-ray images of the chest from Cohen and Kaggle repositories ResNet101 and ResNet-50 Precision, Recall, Accuracy, The limited number of COVID-19 images [20] ChestX-ray14 dataset extracted from National Institutes of Health Clinical Centre, USA + COVID-19 X-ray data set curated by a group of researchers from the University of Montreal ResNet-101 Sensitivity, Specificity, Accuracy and AUC Lack of images used in the testing phase. [21] Chest x-ray scans Kaggle repository Inception V3, Xception, and ResNeXt Accuracy. Precision, Recall, F1-Score Small dataset. [22] Public Dataset of X-ray images from healthy, pneumonia and covid-19 patients VGG16 AUC, Accuracy Small dataset [23] Publicly available COVID-19 chest X-ray image repository pneumonia, and normal\healthy individuals ResNet, VGG, Inception and Efficient Net Accuracy Small number of images in validation [24] COVIDx CT-2A and COVIDx CT-2B datasets COVID-Net-CT ResNet-v2 Sensitivity, Specificity, Positive Predictive Value, and Negative Predictive Value, Accuracy [25] A total of 1100 chest X-ray images were randomly selected from three different open sources: the GitHub repository shared by Joseph Cohen, Kaggle, Bachir, and Mooney. The chest X-ray images in the datasets The datasets contain chest X-ray images of confirmed COVID-19 cases, other pneumonia, and no-findings (normal) SVM Sensitivity, Specificity, AUC, Positive Predictive Value, Negative Predictive Value, and Accuracy. Lack of COVID-19 X-ray images [26] Chest CT (CCT) images from local hospitals CCSHNet Micro-averaged (MA) F1 score, Sensitivity, Precision, Small dataset [27] chest CT images FGCNet Fowlkes–Mallows index (FMI), Matthews correlation coefficient (MCC), F1, Sensitivity, Specificity, Precision and Accuracy. Lack of clinical testing Table 2 Comparative analysis with state-of-art-work that used cloud computing and DL for Covid-19. Work Application IoT Fog Cloud DL [47] An automatic COVID-19 detection system Yes No Yes ResNet50 [48] Corona virus detection Yes Yes Yes SVM, DNN, K-NN, LSTM, Naïve Bayes, One Rule, NN, Decision Table [49] COVID-19 risky behavior monitoring system Yes No Yes CNN [50] Yes Yes Yes No [51] Predict the potential threat of COVID-19 No No Yes ML Our approach (IFC-Covid) COVID-19 and Pneumonia detection Yes Yes Yes CNN On the other side, advancements in healthcare technology provide numerous benefits in terms of improving and safeguarding people's lives all over the world. Furthermore, modern communication technology has allowed medical services to be accessed from a distance. The goal of real-time monitoring and automated prediction and detection systems is to save lives and minimize medical costs. For example, combining cloud technology with the Internet of Things (IoT) or the Internet of Medical Things (IoMT) allows for real-time detection and diagnosis of a sickness, which leads to early intervention that saves lives and lowers healthcare costs [28], [29]. Cloud Computing is one of the major computing paradigms in the information technology domain [30]. The role of cloud computing is to serve the user’s requests for using the computing resources anytime and anywhere [31]. In general, the deployment in the cloud could be in Private Cloud, Public Cloud, Community Cloud, or Hybrid Cloud [32]. The Cloud provides three services: Software as a Service (SaaS), which provides software for users, Platform as a Service (PaaS), for software and application building, and Infrastructure as a Service (IaaS), for storage and computing process and network uses [30]. Fog Computing is an intermediate layer between the IoT devices and the cloud which aims to ameliorate the Quality of service (QoS) of the Cloud by providing several distributed nodes. Fog nodes are used to reduce traffic and latency issues, between users and the cloud, and energy consumption. Also, they provide computing and storage capacity and ensure secure communication between IoT devices and cloud [33]. 2.1 Cloud computing in healthcare application Cloud computing for medical purpose is based on a mobile device, cloud servers, and a network which provides real-time access to the resources in anytime and anywhere [34]. However, this traditional architecture of the cloud showed many weaknesses such as the time required for emergency cases [35]. Also, the increased amount of power consumption and the cost of data transmission to the cloud are some other short coming [36]. Moreover, the problems of latency and the low persistent performance are the limitations of the traditional structure of the cloud [37]. Furthermore, the high cost of the mobile environment needed for the medical scenarios of the patient [38]. 2.2 Fog computing in healthcare application Fog computing is a distributed structure of cloud computing that aims to make the data processing closer to the network edge which provides more suitable options to overcome the limitations of cloud computing [34] . First, it reduces the cost of memory usage, computational costs, and sensors power consumption. Fog computing offers a lower latency by increasing the number of fog nodes or using various edge mining techniques to reduce the data transmission time [39]. In addition, the edge computing applications include a high degree of protection and authentication which ensure that patient information remains private [40], [41]. Furthermore, instead of using detection sensors or GPS location systems, edge applications use specific localization techniques that detect the patient's position in a more accurate and efficient way [42]. The energy consumption is reduced due to the shared structure of edge computing, encryption methods and categorization approaches for health care applications [43], edge mining [44], and proper resource management [45], [46]. Healthcare applications, that are easy and simple to use, should be able to provide a variety of services to patients, without requiring any technical expertise or medical training. [43]. The main purpose of our contribution is to reduce the spread of this pandemic by early detection of the infected people and take the necessary actions as soon as possible. It has been proven that medical treatments are more effective if they are administrated as soon as the virus is detected. In order to implement such a system, we took advantage of the advanced technologies in the healthcare sector such as IoMT, Cloud computing, Fog computing. The composition of such components provides a real-time and secure system that could be easy to use for every user. 3 Material and methods Nowadays, the early detection of the Covid-19 and taking the necessary remedies could decrease the spread of this pandemic. For this reason, designing solutions using the power of artificial intelligence such as deep learning, the advantages of IoMT, fog computing and cloud computing could be very promising approach. 3.1 Deep learning approach The implementation of real-time systems is still one of the biggest challenges that face developers. On the other hand, the urgent situation worldwide, which is to find a way to detect covid-19 in a fast and efficient manner. These are the most reasons that motivate us to introduce IoMT-based fog-cloud for the detection of Covid-19. Our proposed IoMT–fog–cloud-based solution for Covid-19 detection (IFC-Covid) consists of three layers: user layer, fog layer and cloud layer. Fig. 1 presents the general architecture of our proposed model.Fig. 1 The proposed IoMT based Fog-Cloud architecture for Covid-19. In general, Cloud and the Fog-based solution requires IoT devices, sensors, user devices, and different nodes that ensure the cloud and fog services. In addition, different communication protocols, such as Bluetooth, IEEE 802.15.4, 3G, 4G, 5G, and IEEE 802.11, are required.I. User Layer Several models of mobile or portable radiography are used in this layer as IoT devices. These devices are specially designed to adapt to the needs of intensive care units and emergency departments. They are perfectly maneuverable in any hospital environment because of their robustness and reliability as well as their levels of safety and comfort. In addition, they allow obtaining high quality images, even under difficult conditions. 3.1.1 End-user's device The user has the opportunity to use a variety of devices such as smartphones, personal computers (PC), tablets. These devices are used to receive data from IoT devices and send it to the fog layer. Also, they visualize the results and notifications obtained from the fog or cloud, for the end-users. An important feature of designing a handy medical system is ensuring that users' gadgets have an adequate degree of visibility. 3.1.2 End-User In general, the end-users in such medical application consist of the doctors, patients, nurses, and/or other users that are related to the patient. The role of the end-user is to send patient data to the fog (in our case X-ray images). After some processing steps in the fog and cloud, he/she receives the results of the data analytics that have been sent as input Fig. 7 .II. Fog Layer Fig. 7 Scenario of the proposed system. The secret of adding a fog computing layer to a cloud-based architecture is to bridge the gap of the cloud layer and ensure the real-time data analysis and the classification process. Moreover, it ensures several proprieties such as the privacy of the patient's data and its security. In addition, some preprocessing steps are done in the fog layer that helps and facilitate the analysis and classification process of the cloud layer, which reduces its latency and ensures its quality of service.III. Cloud Layer The cloud layer is the main layer in a smart healthcare system. It provides several services with high computing capacity that are used in the analytic process. In addition, it provides huge space for data storage called data center. 3.2 Detailed architecture In this part, we describe our system's design and show how the components of each tier communicate securely Fig. 2 . We begin with a description of the system architecture, followed by a description of the security mechanism that is utilized to safeguard the privacy of patient data at various stages of processing. Next, we show how we train our CNN model on the passed dataset hosted in a data center. Due to the similarity of Pneumonia X-ray and Covid-19 X-ray, we train our model to discriminate between the two for better classification results.I. End layer Fig. 2 The Composition of each layer of the IoMT based Fog-Cloud system for Covid-19. Regarding Fog-Cloud-Covid19 system, this layer is composed from radiology devices and smart devices. The radiology devices can record the x-ray image of the patient. These radiology devices are equipped with sensors that allow them to communicate with other devices via Bluetooth or WiFi, 3G, 4G, 5G. After recording the x-ray images, the doctor sends it to smart devices such as phones, tablets, or laptop computers. Then, the patient or the doctor that receives the x-ray image on his/her smart device has to transmit it and other personal data to the fog layer. The main advantage of using an intermediate layer such as fog layer is to reduce time consumption, and cost which improves the quality of service of our system.II. Fog layer In the majority of healthcare applications that use cloud computing, the privacy and security of personal data of the patient can be easily hacked because of the centralization transmission process of various and vast volumes of information. We introduced a fog layer in our system to ensure the privacy and security of the data using several services such as: Identification: The first process of the fog layer is identifying the user that are sending the data or allowing the creation of profiles for new users. Authentication: In this step, the fog service has to check the authenticity of the user using authentication protocols such as Authenticated key Agreement (AKA) [52], Certificate Revocation List (CRL) and Online Certificate Status Protocol (OCSP), [41]. Security: The fog computing provides an encryption service to encrypt the user data and ensure its security. Several encryption techniques are used in the literature such Elliptic-Curve Cryptography (ECC) [53], Privacy-Preserving Fog-Assisted Information Sharing Scheme (PFHD) [54], Bilinear pairing IBE [55], Modified Elliptic Curve Cryptography (MECC) [56], Fully Homomorphic Encryption Scheme (FHE) [57], and Enhanced Value Substitution (EVS) [58]. Filtering and normalization: After receiving the user data, some preprocessing steps are performed such as filtering and normalizing the data to make it ready for the processing process in the cloud layer. The purpose of this step is to reduce the processing steps of the cloud which minimizes its computing time. Data storage: In the fog layer and after preprocessing steps, part of the user data has to be stored in fog repository for security and authentication procedures to ease the access and lockout process, authenticity of data, and lower the access control delivery cost.III. Cloud layer The cloud layer is the main layer that consists of the analysis of the data and processing techniques then saving the data in the cloud storage and visualizing the resulting data. Preprocessing In this step and after receiving data segmentation from the fog layer, some process has to be done such normalization and preparing this data to be passed to the processing step. Cloud Processing In the processing layer several tasks have to be done such as preparing the Convolution Neural Network model to be trained on a dataset that contains three categories of samples: Covid-19 cases, Pneumonia cases, and Normal cases. CNN Model Over the last few decades, the community of specialists in the field of medical research has developed several health care systems which make it possible to specify the doctor's decision on the case of his patient. Also, due to human nature and the possibility of making mistakes in diagnosing cases which are not related to the degree of knowledge of doctors, but also to how they deal with patient problems and other aspects [59]. However, the revolution of Artificial Intelligence (AI) in various domains help to answer several questions in a semi-automatic or automatic way. Deep Learning (DL) is AI technique that is used in the detection and prediction issues. Similar to the other domains, the healthcare area used many DL applications that achieved great results in many medical cases because of their capability to learn from the context using supervised learning, semi-supervised learning, and unsupervised learning. CNN is DL approach that specialized in image recognition, image classification, image prediction, detecting the abnormality of the signal records, etc. [60], [61]. The auto extraction of the features from images and the deep analytics of CNN, are the main reasons that make us use the CNN for the training process in our system. Several researchers have used CNN for covid-19 detection [18], [62]. In our system, we used CNN model that contains convolution and pooling layers due to its efficacity in image processing. Furthermore, we depend on deep neurons using small kernel windows which provide better features learning and minimize the model's complexity. The Covid-19 and pneumonia detection systems use x-ray pictures to create a deep model that extracts the most characteristics from the image. In our model, there are a total of six convolution layers and three completely linked layers (Table 3 ). After the features map is built, it must be sent to the flatten layer to prepare it for the completely linked layers. The last completely linked layer is dedicated to the final classification results.Table 3 Summary of CNN model. Layer Feature Map Size Kernel Size Stride Activation Input Image 1 180×180×3 – – – 1 Convolution 2D 32 180×180×3 2×2 2×2 Tanh Max Pooling 2D 32 90×90×32 2×2 2×2 3 Convolution 2D 32 45×45×32 2×2 2×2 Tanh Max Pooling 2D 32 22×22×32 2×2 2×2 5 Convolution 2D 64 11×11×64 2×2 2×2 Softsign Max Pooling 2D 5×5×64 2×2 2×2 Flatten 3200 – – – – 8 FC 132 422532 – – Relu 9 FC 60 7980 – – Relu Output FC 3 183 – – Softmax 3.3 Dataset In our experimentation, we used a free big dataset that contains 6432 X-ray images extracted from Kaggle repository [63]. The dataset is divided in two folders for the training (5144 X-ray pictures) and test (1288 X-ray pictures). Each one of these folders contains three subfolders named Covid-19, Normal, Pneumonia. After the extraction of the dataset, we apply some pre-processing methods then generating new images for data augmentation. 3.4 Dataset preparation Medical Image processing is an essential and critical step in medical field which aims to visualize the abnormalities and the special issues contained in the image. Image segmentation is one of the image processing steps which divides the original image into regions depending on specific characteristics of the image such as brightness, and grey level. Wavelet transform is one of the powerful tools that used for different medical issues such as signal decomposition [64], [65] and image decomposition [66], [67], [68]. In our case we use Discrete wavelet transform (DWT) with Biorthogonal 1.3 for the segmentation level in order to reduce the unnecessary data and optimize the analysis effort. Moreover, the combination of wavelet transforms and reduction techniques such as Principal Component Analysis helped us in selecting the optimal features, reducing the complexity and ameliorating the accuracy of the neural network [69]. Principal Component Analysis (PCA) is linear algebra technique that generally used for feature extraction and dimensionality reduction. The purpose of PCA is to reduce the number of features whereas keeping most of the original ones in order to reduce the complexities of the model. The main steps of the PCA are: the standardization of the dataset into d-dimension, then building the covariance matrix, computing the eigenvectors and eigenvalues, selecting the k eigenvalues, in the end, generate the new k-dimensional features of the original dataset. [70]. In our case, we used PCA with only 80 of principle component following these steps:- Dividing the image into three components (Red, Green, Bleu) channels. - Applying PCA to each channel. - Applying inverse transform to transformed array. - Inversing the transformation in order to rebuild the original image with only 80 of principal components. Fig. 5 shows the resulting image using PCA with different values. After using DWT on the images Fig. 4 , we applied the entropy Fig. 6 for each coefficient in order to extract the optimal features using different kind of entropy such as Teager Kaiser Energy Operator, Log Energy Entropy, Shannon wavelet entropy energy.Fig. 3 Dataset samples. Fig. 4 The application of DWT on image with bior 1.3. Fig. 5 The application of PCA with different values. Fig. 6 The application of entropy on the image. Teager Kaiser Energy Operator (TKEO) is non-linear energy tracking operator which used for signal and image. The TKEO based on Amplitude Modulated- Frequency Modulated (AM-FM) for analyzing the data [71]. TKEO represented by Eq. (1).(1) ϕ2Ix,y=\|∇I(x,y)\|2-Ix,y.∇2[Ix,y] (2) ∇Ix,y=(∂I(x,y)∂x,∂I(x,y)∂y) (3) Ix,y=∑k=1Kakx,ycos(ξk(x,y)) Where akx,y represent the amplitude modulating of the image contrast in k narrowband component and ∇ξk(x,y) represent frequency modulation of image structure properties in the instantaneous phase component ξk(x,y). Shannon Wavelet Entropy Energy (SWEE) is combination between wavelet, shannon entropy and energy. This combination used for effective analysis which extract optimal features using time frequency [72]. The Shannon wavelet entropy energy could be presented by Eq (4):(4) ηd=E(d)ShannonEntropy(d) (5) Ed=∑k=1c\|Mk(d)\|2 Where E is energy of data d in each wavelet coefficient.c (6) ShannonEntropyd=-∑k=1cPklogPk Where Pk is the energy probability of each wavelet coefficient and.∑k=1cPk=1 (7) Pk=\|Mk(d)\|2E(d) Log Energy Entropy (LEE) is feature extraction method based on the energy and it is similar to Shannon entropy. The main role of it is to calculate the uncertainty of features in data according to Eq (8) [73].(8) LogEnergyEntropyd=∑k=1clog(Pk2) Where Pk is the energy probability of each wavelet coefficient and.∑k=1cPk=1 Classification Another processing step used in Computer-Aided Diagnosis (CAD) which considered as a key step in medical image processing. In the classification phase, the received images are passed to the CNN model to be classified. The output of this phase is the class of each image. In our case, we have three classes: Covid-19, Normal, and Pneumonia (Fig. 3). Various classification metrics were used for the evaluation of our model: True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN). These measures are used to calculate Recall, True Positive Rate (TPR), False Positive Rate (FPR), Precision, Specificity, Sensitivity, F1-score, and Accuracy.Recall=TPR=TPTP+FN Precision=TPTP+FP Specificity=TNFP+TN Sensitivity=TPTP+FN Accuracy=TP+TNTP+TN+FP+FN F1-score=2TP2TP+FP+FN FPR=FPFP+TN Cloud Storage The cloud layer has a powerful capacity of storage which stores the big data received from the fog layer and also creates the user profile and stores the results of the classified data. Visualize results This phase specialized in displaying analytical dashboards and the results of the classification. 4 Results and discussion In our implementation, we used Python with Keras and Tensorflow libraries which provide the needed tools for deep learning implementation. These libraries are built on top of the PyCharm environment. Lenovo PC with Windows 10 pro 64 -bit, CPU of Intel Core i7, 3.60 GHz, 16 GB of RAM, Intel HD Graphics 4600 GPU and 1 T of storage are used. Our model achieves efficient results in term of classification metrics: 97% of accuracy, 100% of precision, 97% of recall and 99% for F1-score. Table 4 summarizes the classification results of the model.Table 4 Classification results. Precision Recall F1-score Accuracy Covid-19 100% 97% 99% 97% Normal 94% 90% 92% 97% Pneumonia 96% 98% 97% 97% The application of DWT-PCA on the dataset during the preprocessing step has great impact in optimizing the model and reducing its complexity by reducing the trainable parameters (see Table 5 ). As shown in Table 5, 28.24% of parameters are reduced, which improves the training time and reduces the complexity of the full model.Table 5 The number of trainable parameters of the original dataset and the preprocessed dataset. Trainable parameters Original dataset 658,119 Pre-processed dataset 472,263 The measurement of the signal power and the noise power ratio (SNR) of the original image and the preprocessed image are shown in Table 6 . Moreover, we used peak signal-to-noise ratio (PSNR) to check the effect of the preprocessing step on the original dataset by using an original image and its preprocessed one. The result of PSNR is 3.14 dB which indicates that the maximum possible power and the corruption noise power do not have an effect on the resulting image and improves the effectiveness the use of DWT-PCA in the preprocessing step.Table 6 The SNR and PSNR values of the original image and the preprocessed image. SNR (dB) PSNR (dB) Original image 1.48 3.14 Pre-processed image 1.47 In addition, we instigated the energy mean for the identification process by estimating the pixel energy of the image using different kind of methods such TKEO, SHEE, LEE. As shown in Table 7 , TKEO has the minimum energy with only 3.26 bits/pixel which improves its performance in tracking the energy of non-linear features. The use of TKEO, in these critical cases such x-ray images, has great impact on the correct identification of the x-ray images. This improves the performance of the system in distinguishing between covid-19, normal, and pneumonia cases.Table 7 The result of application TKEO, SWEE and LEE on the image. TKEO (bits/pixel) SHEE (bits/pixel) LEE (bits/pixel) Image 3.26 6.75 4.68 We studied the performance of our system in the training process by comparing accuracy and loss metrics of the original dataset and the preprocessed dataset. However, the training process with preprocessed dataset showed better results in term of accuracy and loss where, the increasing green curve improves the high accuracy. Meanwhile, the loss function decreases until it reaches low values (see Fig. 8 ).Fig. 8 The training accuracy and loss of the CNN model on the original dataset and preprocessed dataset. 4.1 Confusion matrix A confusion matrix is machine learning metrics used in measuring the effectiveness and the performance of the system. Usually, it is used for binary or multiclass classification issues. The main role of the confusion matrix is to help understand how the classification model is confused when making predictions. Not only it captures how many errors were made, but also what kind of errors were made. The rows of the matrix represent occurrences of the actual class and the columns represent instances of the projected category [74]. In other words, it keeps track of each class's correct and wrong predictions. To study the performance of our system we calculate True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN). These metrics are used to visualize the results of predicting instances of each class. Fig. 9 shows that the model correctly predicted almost all the covid-19 samples used for the validation dataset, except one sample, predicted it as normal case and six samples as pneumonia. In addition, the system correctly predicted normal samples and missed a few samples (1 as covid-19 and 39 as pneumonia). Fig. 10 illustrates that 98% of pneumonia samples are correctly predicted by the model.Fig. 9 Confusion matrix of the predicted classes. Fig. 10 The percentage of predicted samples by classes. 4.2 Area Under the ROC Curve The Area Under Curve (AUC) is a description of the Receiver Operating Characteristic Curve (ROC curve), which assesses a classifier's ability to differentiate across classes. Meanwhile, the ROC curve is a graph that uses the True Positive Rate (TPR) and False Positive Rate (FPR) to indicate how well a classification model performs across all categorization levels (FPR) [75]. The ROC curve of our model is plotted to verify the classification performance of our model (Fig. 11 ). As we see in Fig. 10, the huge space between the green line and the blue line defines the AUC of the model and it reaches 0.98. In which the perfect diagnostic test of the classes where in the value 1.0. Also, the achieving AUC = 0.98 is reflection of the accuracy of the model in predicting classes what ensures the outperformance and the capacity of the model in classifying the classes.Fig. 11 The ROC curve of the model. 5 Conclusion Covid-19 is a serious topic that needs to be treated due to its impact on various domains such as industry, education, economic, etc. Because of the absence of a cure to this pandemic, the only way to deal with it is to reduce its spread by precaution methods and the early identification of the infected patients. With the revolution of artificial intelligence and its impact on the various domains. The health care sector, similar to other domains, adopted the AI approaches which resulted in great innovations in the medical field such as the Internet of medical things (IoMT). The IoMT helped to solve many medical issues that need a real-time response. In this paper, we propose IoMT-fog-cloud-based architecture for covid-19 detection. Our work aims to reduce the spread of the pandemic by early detection of the infected people. In addition, the proposed system facilitates the detection of covid-19 anywhere and anytime. The system is easy to use and user friendly. Furthermore, we focus on the Quality of Service (QoS) of the cloud by introducing an intermediate layer between the user and the cloud to reduce the latency and ensure real-time response. The privacy and security of the patient's data were other objectives of our study. For better classification and identification results we combined Discrete Wavelet Transform and Principal Component Analysis (DWT-PCA) to extract the optimal features and reduce the dimension. Also, we used different kind of energy tracking in the images to ameliorate the identification in the x-ray images. Our proposed filtering technique reduces the complexity of the model by reducing trainable parameters and minimizing the time. The evaluation of our CNN model in the classification showed an efficient performance which can classify the covid-19, pneumonia and normal cases with 97% and the precision rate reaches 100%. From the experimental results, our proposed system could be aid discission making in the identification of covid-19. As future work, we aim to use the ElectroCardioGraph (ECG) for the covid-19 detection since the latest variants of covid-19 have effect on cardiovascular system. In addition, CT-images and X-ray images are not suitable source for the identification of this new variants. Also, we aim to investigate the use of Empirical Wavelet Transform (EWT) and PCA for filtering process of the data. CRediT authorship contribution statement M.A. Khelili: Conceptualization, Methodology, Software, Writing – original draft, Visualization, Investigation, Formal analysis. S. Slatnia: Writing – review & editing, Supervision, Validation. O. Kazar: Writing – review & editing, Supervision, Validation. S. Harous: Writing – review & editing, Validation. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ==== Refs References 1 WHO, “World Health Organization,” [Online]. Available: https://www.who.int/fr. [Accessed 15 06 2021]. 2 C. Marco, R. Michael, A. Abdul, D. Scott and D. N. Raffaela, “Features, Evaluation, and Treatment of Coronavirus (COVID-19),” 2 09 2021. [Online]. Available: https://www.statpearls.com/ArticleLibrary/viewarticle/52171. [Accessed 17 10 2021]. 3 Gavi, “There are four types of COVID-19 vaccines: here’s how they work,” 2021. [Online]. Available: https://www.gavi.org/vaccineswork/there-are-four-types-covid-19-vaccines-heres-how-they-work. 4 M. Terry, “UPDATED Comparing COVID-19 Vaccines: Timelines, Types and Prices.(Biospace),” [Online]. Available: https://www.biospace.com/article/comparing-covid-19-vaccines-pfizer-biontech-moderna-astrazeneca-oxford-j-and-j-russia-s-sputnik-v/. [Accessed 06 08 2021]. 5 V. P. David Bême, “Vaccin Pfizer contre le covid-19: efficacité, allergies, effets secondaires. Doctissimo,” [Online]. Available: https://www.doctissimo.fr/sante/epidemie/coronavirus-chinois/pfizer-biontech-vaccin-coronavirus-covid. [Accessed 15 08 2021]. 6 Staff, “Sinopharm COVID-19 Vaccine (BBIBP-CorV). Precision Vaccinations,” [Online]. Available: https://www.precisionvaccinations.com/vaccines/sinopharm-covid-19-vaccine-bbibp-corv. [Accessed 30 08 2021]. 7 Moderna-NIAID, “SPIKEVAX - COVID-19 Vaccine Moderna. Mes Vaccins,” [Online]. Available: https://www.mesvaccins.net/web/vaccines/656-spikevax-covid-19-vaccine-moderna. [Accessed 30 08 2021]. 8 L. Sinovac Research and Development Co., “Coronavac.(Mes vaccins.net),” [Online]. Available: https://www.mesvaccins.net/web/vaccines/651-coronavac. [Accessed 05 09 2021]. 9 N. C. f. I. a. R. D. (NCIRD)., “Johnson & Johnson’s Janssen COVID-19 Vaccine Overview and Safety,” Centers for disease control and prevention, [Online]. Available: https://www.cdc.gov/coronavirus/2019-ncov/vaccines/different-vaccines/janssen.html. [Accessed 20 09 2021]. 10 A. Keown, “Russia Claims its Sputnik V Vaccine is 92% Effective Following Interim Analysis,” Biospace, [Online]. Available: https://www.biospace.com/article/russia-claims-its-sputnik-v-vaccine-is-92-percent-effective-following-interim-analysis/. [Accessed 03 10 2021]. 11 WHO, “Vaccine efficacy, effectiveness and protection,” 2021. [Online]. Available: https://www.who.int/news-room/feature-stories/detail/vaccine-efficacy-effectiveness-and-protection. 12 Imran A. Posokhova I. Qureshi H.N. Masood U. Riaz M.S. Ali K. Nabeel M. AI4COVID-19: AI enabled preliminary diagnosis for COVID-19 from cough samples via an app Informatics in Medicine Unlocked 20 2020 100378 13 Wang L. Lin Z.Q. Wong A. Covid-net: A tailored deep convolutional neural network design for detection of covid-19 cases from chest x-ray images Scientific Reports 10 1 2020 1 12 31913322 14 Albahli S. Efficient GAN-based Chest Radiographs (CXR) augmentation to diagnose coronavirus disease pneumonia International journal of medical sciences 10 17 2020 1439 15 Xiaowei X. Xiangao J. Chunlian M. Peng D. Xukun L. L. Shuangzhi and al., “A deep learning system to screen novel coronavirus disease 2019 pneumonia.” Engineering 10 6 2020 1122 1129 16 Li L. Qin L. Xu Z. Yin Y. Wang X. Kong B. Xia J. Artificial intelligence distinguishes COVID-19 from community acquired pneumonia on chest CT Radiology 2020 17 D. Wang, J. Mo, G. Zhou, L. Xu and Y. Liu, “An efficient mixture of deep and machine learning models for COVID-19 diagnosis in chest X-ray images,” PloS one, vol. 15, no. 11, 2020. 18 Alazab M. al. COVID-19 prediction and detection using deep learning International Journal of Computer Information Systems and Industrial Management Applications 12 2020 168 181 19 Jain G. Mittal D. Thakur D. Mittal M.K. A deep learning approach to detect Covid-19 coronavirus with X-ray images Biocybernetics and biomedical engineering 4 40 2020 1391 1405 20 M. Z. Che Azemin, R. Hassan, M. I. Mohd Tamrin and M. A. & Md Ali, “COVID-19 deep learning prediction model using publicly available radiologist-adjudicated chest X-ray images as training data: preliminary findings.,” International Journal of Biomedical Imaging., 2020. 21 Jain R. Gupta M. Taneja S. Hemanth D.J. Deep learning based detection and analysis of COVID-19 on chest X-ray images Applied Intelligence 51 3 2021 1690 1700 34764553 22 Civit-Masot J. Luna-Perejón F. Domínguez Morales M. Civit A. Deep learning system for COVID-19 diagnosis aid using X-ray pulmonary images Applied Sciences. 10 13 2020 4640 23 Dey S. Bacellar G.C. Chandrappa M.B. Kulkarni R. COVID-19 Chest X-Ray Image Classification Using Deep Learning medRxiv. 2021 24 Zhao W. Jiang W. Qiu X. Deep learning for COVID-19 detection based on CT images Scientific Reports 11 1 2021 1 12 33414495 25 Erdaw Y. Tachbele E. Machine Learning Model Applied on Chest X-Ray Images Enables Automatic Detection of COVID-19 Cases with High Accuracy International Journal of General Medicine 14 2021 4923 4931 34483682 26 Wang S.-H. Nayak D.R. Guttery D.S. Zhang X. Zhang Y.-D. COVID-19 classification by CCSHNet with deep fusion using transfer learning and discriminant correlation analysis Information Fusion 68 2021 131 148 33519321 27 Wang S.-H. Govindaraj V.V. Ǵorriz J.M. Zhang X. Zhang Y.-D. Covid-19 classification by FGCNet with deep feature fusion from graph convolutional network and convolutional neural network Information Fusion 67 2021 208 229 33052196 28 Alabdulatif A. Khalil I. Forkan A.R.M. Atiquzzaman M. Real-time secure health surveillance for smarter health communities IEEE Communications Magazine 1 57 2019 122 129 29 Azana H.M.A. Wan H.H. Shilan S. Zainab S.A. Mojtaba A. Liza A.L. IoMT amid COVID-19 pandemic: Application, architecture, technology, and security Journal of Network and Computer Applications. 174 2020 102886 30 Alam T. Cloud Computing and its role in the Information Technology IAIC Transactions on Sustainable Digital Innovation (ITSDI) 1 2021 108 115 31 Hussain M. Beg M. Alam M. Fog Computing for Big Data Analytics in IoT Aided Smart Grid Networks Wireless Pers Commun 114 2020 3395 3418 32 Namasudra S. Data access control in the cloud computing environment for bioinformatics International Journal of Applied Research in Bioinformatics (IJARB) 11 1 2021 40 50 33 Samann, E. F. Fady, R. Z. Subhi and A. Shavan, “IoT provisioning QoS based on cloud and fog computing,” Journal of Applied Science and Technology Trends, vol. 2, no. 1, pp. 29-40, 2021. 34 Hartmann M. Hashmi U.S. Imran A. Edge computing in smart health care systems: Review, challenges, and research directions Transactions on Emerging Telecommunications Technologies e3710 2019 25 35 Thiyagaraja S. Dantu R. Shrestha P. Thompson M. Smith C. Optimized and secured transmission and retrieval of vital signs from remote devices Proceedings of the Second IEEE/ACM International Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE) 2017 36 Kumar K. Lu Y. Cloud computing for mobile users: can offloading computation save energy? Computer 43 4 2010 51 56 37 Jackson K. Ramakrishnan L. Muriki K. al. Performance analysis of high performance computing applications on the Amazon web services cloud 2010 IEEE Second International Conference on Cloud Computing Technology and Science 2010 38 K. Bhargava and S. Ivanov, “A fog computing approach for localization in WSN.,” 2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC). Montreal, Canada., 2017. 39 Althebyan Q. Yaseen Q. Y. Jararweh and al., “Cloud support for large scale e-healthcare systems,” Annals of telecommunications 9 71 2016 503 515 40 Yu W. Liang F. X. He and al., “A survey on the edge computing for the Internet of Things,” IEEE Access 6 2017 6900 6919 41 Alrawais A. Alhothaily A. Hu C. Cheng X. Fog computing for the Internet of Things: security and privacy issues IEEE Internet Comput 2 21 2017 34 42 42 Maddumabandara A. Leung H. Liu M. Experimental evaluation of indoor localization using wireless sensor networks IEEE Sensors J 15 9 2015 5228 5237 43 Y. Cao, P. Hou, D. Brown, J. Wang and S. Chen, “Distributed analytics and edge intelligence: Pervasive health monitoring at the era of fogcomputing.,” Proceedings of the 2015 Workshop on Mobile Big Data (Mobidata); Hangzhou, China., 2015. 44 Gaura E. Brusey J. Allen M. Wilkins R. Goldsmith D. Rednic R. Edge mining the Internet of Things IEEE Sensors J 13 10 2013 3816 3825 45 H. Wang, J. Gong, Y. Zhuang, H. Shen and J. Lach, “HealthEdge: Task scheduling for edge computing with health emergency and human behavior consideration in smart homes.,” Paper presented at: 2017 IEEE International Conference on Big Data (Big Data). Boston, MA., 2017. 46 M. Al-khafajiy, L. Webster, T. Baker and A. Waraich, “Towards fog driven IoT healthcare: Challenges and framework of fog computing in healthcare,” Proceedings of the 2nd International Conference on Future Networks and Distributed Systems (ICFNDS). Amman, Jordan., 2018. 47 Nasser N. Emad-ul-Haq Q. Imran M. al., A smart healthcare framework for detection and monitoring of COVID-19 using IoT and cloud computing Neural Computing and Applications 2021 1 15 48 Aljumah A. Assessment of Machine Learning Techniques in IoT-Based Architecture for the Monitoring and Prediction of COVID-19 Electronics 10 15 2021 1834 49 Liang H.S. al. “An Interoperable Architecture for the Internet of COVID-19 Things (IoCT) Using Open Geospatial Standards—Case Study Workplace Reopening,“ Sensors 21 1 2021 50 50 Kallel A. Molka R. Mahdi K. IoT-fog-cloud based architecture for smart systems: Prototypes of autism and COVID-19 monitoring systems Software: Practice and Experience 51 1 2021 91 116 51 Shreshth T. Shikhar T. Rakesh T. Sukhpal S.G. Predicting the growth and trend of COVID-19 pandemic using machine learning and cloud computing Internet of Things 11, vol. 11 2020 100222 52 Jia X. He D. Kumar N. Choo K. Authenticated key agreement scheme for fog-driven IoT healthcare system Wireless Networks 25 2018 4737 4750 53 Giri D. Obaidat M. Maitra T. SecHealth: An efficient fog based sender initiated secure data transmission of healthcare sensors for e-medical system 2017 IEEE Global Communications Conference (GLOBECOM) 2017 54 W. Tang, K. Zhang, J. Ren and Y. S. X. Zhang, “Lightweight and privacy-preserving fog-assisted information sharing scheme for health big data,” 2017 IEEE Global Communications Conference (GLOBECOM). Singapore, 2017. 55 H. Al Hamid, S. Rahman, M. Hossain, A. Almogren and A. Alamri, “A security model for preserving the privacy of medical big data in a healthcare cloud using a fog computing facility with pairing-based cryptography.,” IEEE Access. 5, pp. 22313-22328. 56 Ghosh S. Jamthe A. Chakraborty S. Agrawal D. Secured wireless medical data transmission using modified elliptic curve cryptography The 3rd ACM MobiHoc Workshop on Pervasive Wireless Healthcare (MobileHealth) 2013 19 24 57 Sun X. Zhang P. Sookhak M. Yu J. Xie W. Utilizing fully homomorphic encryption to implement secure medical computation in smart cities Pers Ubiquit Comput. 21 5 2017 831 839 58 Elmisery A. Rho S. Botvich D. A fog based middleware for automated compliance with OECD privacy principles in Internet of Healthcare Things IEEE Access 4 2016 8418 8441 59 Sweta B. Praveen K.R.M. Quoc-Viet P. Thippa R.G. Siva R.K.S. Chiranji L.C. Mamoun A. Md J.P. Deep learning and medical image processing for coronavirus (COVID-19) pandemic: A survey Sustainable cities and society 65 2021 102589 60 Litjens G. Kooi T. Bejnordi B.E. Setio A.A.A. Ciompi F. Ghafoorian M. al. A survey on deep learning in medical image analysis,“ Medical Image Analysis 42 2017 60 88 28778026 61 Ker J. Wang L. Rao J. Lim T. Deep learning applications in medical image analysis IEEE Access 6 2017 9375 9389 62 Jain R. al. Deep learning based detection and analysis of COVID-19 on chest X-ray images Applied Intelligence 51 3 2021 1690 1700 34764553 63 Patel P. “Chest X-ray (Covid-19 & Pneumonia) Kaggle,“ [Online]. Available 13 [Accessed 2021, 02 https://www.kaggle.com/prashant268/chest-xray-covid19-pneumonia 64 Sadiq M.T. Yu X. Yuan Z. Fan Z. Rehman A.U. Li G. Xiao G. Motor Imagery EEG Signals Classification Based on Mode Amplitude and Frequency Components Using Empirical Wavelet Transform Ieee Access 7 2019 127678 127692 65 Sadiq M.T. Yu X. Yuan Z. Aziz M.Z. Motor imagery BCI classification based on novel two-dimensional modelling in empirical wavelet transform Electron. Lett 56 25 2020 1367 1369 66 B. Rinisha, W. Sulochana and W. Arun Kumar, “A Wavelet Transform and Neural Network Based Segmentation & Classification System For Bone Fracture Detection,” Optik, vol. 236, 2021. 67 X. Feng, Z. Wuxia, S. Xiuqin and Z. Xu, “Optical Remote Sensing Image Denoising and Super-Resolution Reconstructing Using Optimized Generative Network in Wavelet Transform Domain,” Remote Sensing, vol. 13, no. 9, 2021. 68 Una T. Dejan Z. Image Denoising by Discrete Wavelet Transform with Edge Preservation 13th International Conference on Electronics, Computers and Artificial Intelligence (ECAI) 2021 1 4 69 Tariq S.M. Yu X. Yuan Z. Exploiting dimensionality reduction and neural network techniques for the development of expert brain–computer interfaces Expert Systems with Applications 164 2021 114031 70 L. Li, “Principal Component Analysis for Dimensionality Reduction,” 2019. [Online]. Available: https://towardsdatascience.com/principal-component-analysis-for-dimensionality-reduction-115a3d157bad. [Accessed 13 02 2022]. 71 G. Gianto, “Multi-dimensional Teager-Kaiser signal processing for improved characterization using white light interferometry,” Signal and Image Processing, 2018. 72 Bafroui H.H. Ohadi A. Application ofwavelet energy and Shannon entropy for feature extraction in gearbox fault detection under varying speed conditions Neurocomputing 133 2014 437 445 73 Taiyong L. Zhou M. ECG Classification Using Wavelet Packet Entropy and Random Forests Entropy 18 08 2016 285 74 Bhandari A. “Everything you Should Know about Confusion Matrix for Machine Learning,” Analytics vidhya [Online]. Available 03 [Accessed 2021, 10 https://www.analyticsvidhya.com/blog/2020/04/confusion-matrix-machine-learning/ 75 A. Bhandari, “AUC-ROC Curve in Machine Learning Clearly Explained,” 16 06 2020. [Online]. Available: https://www.analyticsvidhya.com/blog/2020/06/auc-roc-curve-machine-learning/. [Accessed 21 10 2021].
PMC009xxxxxx/PMC9005381.txt	==== Front J Clin Neurosci J Clin Neurosci Journal of Clinical Neuroscience 0967-5868 1532-2653 Elsevier Ltd. S0967-5868(22)00168-0 10.1016/j.jocn.2022.04.010 Correspondence Contributing factors towards progression of migraines during the Covid-19 pandemic Shafeeq Ahmed H. 1 Bangalore Medical College and Research Institute, K.R. Road, Bangalore 560002, Karnataka, India 1 ORCID: 0000-0003-1671-8474. 13 4 2022 12 2022 13 4 2022 106 242242 31 3 2022 10 4 2022 © 2022 Elsevier Ltd. All rights reserved. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcTo the Editor, The research paper by Yuksel et al. [1] delved into the much important aspect of discussing topics related to the worsening of migraines due to Personal Protective Equipment (PPE), and the study findings give a brief but in-depth review of their effect on patients. The findings indicate that scalp contact masks, double masks and daily mask duration were higher in the worsening group. Adding to this, the frequency of personal disinfectant use was also higher in this group. The authors further demonstrated a relationship between migraine worsening and mask type, number of masks, and intensive disinfectant use. Though the study and its methodology have a novel approach, there are other reasons for worsening migraines that the authors could’ve considered while doing this study. During the pandemic, more than one third of the people had increased work-related stress, including stress due to unemployment, workplace stress, and other factors [2]. Furthermore, since the study had 184 (59.3%) unemployed participants, the stress due to the Covid-19 pandemic could have greatly affected the participants and may not have been purely out of PPE usage. Considering this, the study investigators could have taken this into account during the result discussion as a drawback. Another factor that should’ve been considered was the number of research participants that were diagnosed with a co-morbid condition. 78 (44%) of the total 177 candidates with worsening migraines were diagnosed with co-morbidities and 45 (25.4%) of them had a Covid-19 diagnosis. Several studies show that the progression of chronic migraines may be worsened due to co-morbidities [3]. To get a fair analysis, the investigators could’ve checked whether patients with co-morbidities more frequently used PPEs due to fear of severe Covid-19 complications. A supplementing argument to previous suggestions is that the authors indicate that the worsening group were more commonly diagnosed with allodynia. It is entirely possible that frequent use of masks caused pain in these subjects and hence was a contributing factor. It cannot be purely indicated that PPEs were the only contributing factor for worsening migraines in these candidates and that there are several underlying factors. It is my hope that the authors in future studies do a more in-depth analysis of the said study so that an appropriate and verifiable analysis can be obtained. The authors could’ve analysed other factors that contributed to the worsening of migraines including age, employment, and others. Moreover, an equitable population distribution among the three categories of migraine participants could have been more eye-opening. Since only 37 (11.9%) of the candidates were in the improving group compared to 177 (57%) of participants in the worsening group, this study surely has several limitations. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgement/Funding None. ==== Refs References 1 Yuksel H. Kenar S.G. Gursoy G.T. Bektas H. The impacts of masks and disinfectants on migraine patients in the COVID-19 pandemic J Clin Neurosci 97 2022 87 92 35066364 2 Al Dhaheri AS, Bataineh MF, Mohamad MN, Ajab A, Al Marzouqi A, Jarrar AH, et al. Impact of COVID-19 on mental health and quality of life: Is there any effect? A cross-sectional study of the MENA region. PLoS One [Internet]. 2021;16(3):e0249107. Available from: http://dx.doi.org/10.1371/journal.pone.0249107. 3 Buse D.C. Greisman J.D. Baigi K. Lipton R.B. Migraine progression: a systematic review Headache J Head Face Pain 59 3 2019 306 338
PMC009xxxxxx/PMC9005382.txt	==== Front J Am Med Dir Assoc J Am Med Dir Assoc Journal of the American Medical Directors Association 1525-8610 1538-9375 AMDA - The Society for Post-Acute and Long-Term Care Medicine. S1525-8610(22)00284-5 10.1016/j.jamda.2022.04.002 Research Letter Changing Dynamics of COVID-19 Deaths During the SARS-CoV2 B.1.617.2 (Delta Variant) Outbreak in England and Wales: Reduced COVID-19 Deaths Among the Care Home Residents Emani Venkata R. MD Central Valley Cardiovascular Associates, Inc, Stockton, CA, USA Reddy Raghunath MD Stockton Primary Care, Stockton, CA, USA Emani Shaila R. BS Central Valley Cardiovascular Associates, Inc, Stockton, CA, USA Goswami Kartik K. BS San Joaquin Critical Care Medical Group, Stockton, CA, USA Maddula Kailash R. BS Central Valley Cardiovascular Associates, Inc, Stockton, CA, USA Reddy Nikhila K. Nakka Abirath S. BS Reddy Nidhi K. BA Stockton Primary Care, Stockton, CA, USA Nandanoor Dheeraj MD Synergy Med, Stockton, CA, USA Goswami Sanjeev MD San Joaquin Critical Care Medical Group, Stockton, CA, USA 13 4 2022 6 2022 13 4 2022 23 6 950953 © 2022 AMDA - The Society for Post-Acute and Long-Term Care Medicine. 2022 AMDA – The Society for Post-Acute and Long-Term Care Medicine Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcTo the Editor: Care homes and long-term care facilities (LTCFs) worldwide plunged into crisis during the initial stages of COVID-19 pandemic caused by SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2).1 , 2 Numerous preventive measures were taken to reduce the COVID-19 infections among the care home residents and to improve the outcomes.3, 4, 5 The Office of National Statistics reported a sharp increase in the COVID-19 deaths among care home residents in England and Wales during the initial stages of the pandemic.6 The COVID-19 surge since early June 2021 was predominantly due to the Delta variant (SARS-CoV2 B.1.617.2), but the outcomes of COVID-19 deaths among care home residents have not been described. Methodology In this observational study, we analyzed the nationwide data of care home deaths in England and Wales between March 7, 2020, and November 26, 2021, during the COVID-19 pandemic using data from the UK Office of National Statistics.7 , 8 We performed an analysis of the COVID-19 deaths occurring at the places of death, including care homes, in terms of total weekly COVID-19 deaths during 3 comparative periods of March 7–August 28, 2020; August 29, 2020–May 28, 2021; and May 29–November 26, 2021. We also performed further analysis of the proportion of deaths occurring at care homes that were due to COVID-19 during each of those 3 periods. Statistical Analysis The relative risk (RR), 95% CI, and P values were calculated to compare the outcomes during the 3 study periods. Statistical comparisons of these outcomes were made between each preceding study period. Results Table 1 shows the distribution of the total COVID-19 deaths (n = 153,179) in England and Wales. There were 220,092 care home deaths from all causes and 33,329 deaths due to COVID-19 in care homes from March 7, 2020, to November 26, 2021.Table 1 Total SARS-CoV2 Deaths in England and Wales, Place of Death, and the Percentage of Deaths due to COVID-19 (at Each Place of Death) During March 7, 2020–November 26, 2021 Study Period March 7–August 28, 2020, n (%) August 29, 2020–May 28, 2021, n (%) RR (95% CI); P Value∗ May 29–November 26, 2021, n (%) RR (95% CI); P Value∗ Total deaths from all causes (n = 1,025,282) 299,844 (100) 456,847 (100) 268,591 (100) Total deaths due to COVID-19 (n = 153,179) 51,740 (17.3) 86,627 (19.0) 1.099 (1.088-1.110); P < .001 14,812 (5.5) 0.291 (0.286-0.296); P < .001 Place where the COVID-19 deaths occurred Hospital (acute or community, not psychiatric) 32,731 (63.3) 62,653 (72.3) 1.143 (1.135-1.152); P < .001 11,968 (80.8) 1.117 (1.107-1.127); P < .001 Care home 15,414 (29.8) 16,603 (19.2) 0.643 (0.631-0.656); P < .001 1312 (8.9) 0.462 (0.438-0.488); P < .001 Home 2432 (4.7) 5451 (6.3) 1.339 (1.278-1.403); P < .001 1297 (8.8) 1.392 (1.313-1.475); P < .001 Hospice 730 (1.4) 1315 (1.5) 1.076 (0.984-1.177); P = .11 129 (0.9) 0.574 (0.479-0.687); P < .001 Other communal establishment 228 (0.4) 292 (0.3) 0.765 (0.644, 0.909); P = .002 24 (0.2) 0.481 (0.317-0.729); P = .001 Elsewhere 205 (0.4) 313 (0.4) 0.912 (0.765-1.087); P = .30 82 (0.6) 1.532 (1.202-1.953); P = .001 Deaths due to COVID-19/deaths from all causes, n/n (%) Care home (n = 33,329) 15,414/76,906 (20.0) 16,603/89,954 (18.5) 0.921 (0.903-0.939); P < .001 1312/53,232 (2.5) 0.134 (0.126-0.141); P < .001 Hospital (n = 107,352) 32,731/120,273 (27.2) 62,653/206,411 (30.4) 1.115 (1.103-1.128); P < .001 11,968/116,191 (10.3) 0.339 (0.333-0.346); P < .001 Home (n = 9180) 2432/82,713 (2.9) 5451/129,926 (4.2) 1.427 (1.361-1.496); P < .001 1297/79,066 (1.6) 0.391 (0.368-0.415); P < .001 Hospice (n = 2174) 730/12,261 (6.0) 1315/18,317 (7.2) 1.206 (1.105-1.316); P < .001 129/12,126 (1.1) 0.148 (0.124-0.177); P < .001 Elsewhere (n = 600) 205/6441 (3.2) 313/10,612 (2.9) 0.927 (0.779-1.102); P = .39 82/7035 (1.2) 0.395 (0.311-0.503); P < .001 Other communal establishments (n = 544) 228/1250 (18.2) 292/1627 (17.9) 0.984 (0.841-1.151); P = .84 24/941 (2.6) 0.142 (0.095-0.214); P < .001 ∗ Statistical comparisons were performed with the prior comparison period. During March 7–August 28, 2020 (first wave), a total of 15,414 COVID-19 deaths (29.8%) occurred in care homes. There was a significant decrease in the percentage of COVID-19 deaths occurring in care homes relative to the total COVID-19 deaths [16,603 (19.2%); RR 0.64, 95% CI 0.63-0.65; P < .001] during August 29, 2020–May 28, 2021. Furthermore, COVID-19 deaths occurring in care homes decreased significantly [1312 (8.9%); RR 0.46, 95% CI 0.43-0.48; P < .001] during the Delta variant surge (May 29–November 26, 2021) compared with prior periods. Table 1 also shows that during March 7–August 28, 2020, COVID-19 was responsible for 20.0% (15,514/76,906) of all care home deaths, which significantly declined to 18.5% (RR 0.92, 95% CI 0.90-0.93; P < .001) during August 29, 2020–May 28, 2021, and then further to 2.5% (RR 0.13, 95% CI 0.12-0.14; P < .001) during May 29–November 26, 2021. During the first surge (Figure 1 ), up to 43.6% of the weekly COVID-19 deaths occurred at care homes, with associated decreases in deaths occurring in hospitals. The significant decline in the proportion of weekly COVID-19 deaths in care homes during the Delta variant surge compared with prior surges is associated with the significantly increased proportion of COVID-19 deaths occurring in hospitals (Table 1, Figure 1).Fig. 1 Proportion of weekly COVID-19 deaths based on the place of death (primary axis) and total weekly deaths (secondary axis). Since the Delta variant surge in June 2021, the proportion of COVID-19 deaths occurring in hospitals increased significantly. There is also a significant decline in the proportion of weekly COVID-19 deaths in care homes during the Delta variant surge compared with prior surges. During the first surge, up to 43.6% of weekly COVID-19 deaths occurred at care homes, with associated decreases in deaths occurring in hospitals. Discussion Our study indicates that during the first wave (March 7–August 28, 2020), COVID-19 had a devastating effect on care homes, with 29.8% of all COVID-19 deaths occurring in care homes. Overall, 20% of deaths in care homes were due to COVID-19. During the second wave, there was a slight, but significant, decrease in the COVID-19 deaths occurring in care homes. The deaths due to COVID-19 occurring in care homes showed a significantly sharp decline (from 19.2% to 8.9%) during the Delta variant surge, as well as since the second surge (18.5% vs 2.5%). The findings of our study for the first wave is similar to prior reports of the UK Office of National Statistics, which highlighted the higher number of deaths in care homes during the second wave than the first wave.6 The reduced death rates in the care homes during the Delta variant surge are most likely due to the infection control and protective measures implemented in the care homes by applying lessons learned from the previous surges.3 , 4 The greater adoption of COVID-19 vaccination among care home residents is also a major contributory factor for reduced deaths during the Delta variant surge.9 The immunity from previous SARS-CoV2 infections may also be a potential contributory factor for reduced deaths during the Delta variant surge.10 The increased hospital deaths we noted during the Delta variant surge are probably due to a relative increase in COVID-19 mortality among persons living in the community compared with persons living in care homes. Limitations of our study are that it is an observational study of publicly reported data and that the generalizability of the findings is limited to the England and Wales population. The other limitation of our study is that we are unable to determine the proportion of deaths that occurred in care homes after transfer from a hospital, as those data are not publicly available. In summary, we observed that a significantly lower number of COVID-19 deaths occurred in care homes and that during the Delta variant surge in England and Wales, the number of care home deaths caused by COVID-19 was significantly lower than the prior surges. Furthermore, care home residents are at the highest risk for mortality because of advanced age and comorbidities; therefore, continuous monitoring and research on COVID-19 preventive interventions is an absolute necessity to further improve the outcomes. The authors declare no conflicts of interest. ==== Refs References 1 Ouslander J.G. Grabowski D.C. COVID-19 in nursing homes: calming the perfect storm J Am Geriatr Soc 68 2020 2153 2162 32735036 2 Grabowski D.C. Mor V. Nursing home care in crisis in the wake of COVID-19 JAMA 324 2020 23 32442303 3 Vijh R. Prairie J. Otterstatter M.C. Evaluation of a multisectoral intervention to mitigate the risk of severe acute respiratory coronavirus virus 2 (SARS-CoV-2) transmission in long-term care facilities Infect Control Hosp Epidemiol 42 2021 1181 1188 33397533 4 Frazer K. Mitchell L. Stokes D. Lacey E. Crowley E. Kelleher C.C. A rapid systematic review of measures to protect older people in long-term care facilities from COVID-19 BMJ Open 11 2021 e047012 5 Dykgraaf S.H. Matenge S. Desborough J. Protecting nursing homes and long-term care facilities from COVID-19: a rapid review of international evidence J Am Med Dir Assoc 22 2021 1969 1988 34428466 6 UK Office for National Statistics Deaths involving COVID-19 in the care sector, England and Wales: deaths registered between week ending 20 March 2020 and week ending 2 April 2021 2021. Accessed May 10, 2021. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/articles/deathsinvolvingcovid19inthecaresectorenglandandwales/deathsregisteredbetweenweekending20march2020andweekending2april2021 May 11, 2021 7 UK Office for National Statistics Deaths registered in England and Wales 2021 . Accessed September 23, 2021 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/deathsregisteredinenglandandwalesseriesdrreferencetables 8 UK office of National Statistics Deaths registered weekly in England and Wales, provisional 2021 . Accessed October 6, 2021 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/weeklyprovisionalfiguresondeathsregisteredinenglandandwales/2021 9 Shrotri M. Krutikov M. Palmer T. Vaccine effectiveness of the first dose of ChAdOx1 nCoV-19 and BNT162b2 against SARS-CoV-2 infection in residents of long-term care facilities in England (VIVALDI): a prospective cohort study Lancet Infect Dis 21 2021 1529 1538 34174193 10 Jeffery-Smith A. Rowland T.A.J. Patel M. Reinfection with new variants of SARS-CoV-2 after natural infection: a prospective observational cohort in 13 care homes in England Lancet Healthy Longev 2 2021 e811 e819 34873592
PMC009xxxxxx/PMC9005383.txt	==== Front Med (N Y) Med (N Y) Med (New York, N.y.) 2666-6359 2666-6340 Elsevier Inc. S2666-6340(22)00167-2 10.1016/j.medj.2022.04.001 Clinical Advances Gastrointestinal symptoms and fecal shedding of SARS-CoV-2 RNA suggest prolonged gastrointestinal infection Natarajan Aravind 129 Zlitni Soumaya 129 Brooks Erin F. 29 Vance Summer E. 29 Dahlen Alex 3 Hedlin Haley 3 Park Ryan M. 12 Han Alvin 4 Schmidtke Danica T. 4 Verma Renu 5 Jacobson Karen B. 5 Parsonnet Julie 67 Bonilla Hector F. 6 Singh Upinder 5 Pinsky Benjamin A. 58 Andrews Jason R. 5 Jagannathan Prasanna 46 Bhatt Ami S. 1210∗ 1 Department of Genetics, Stanford University, 269 Campus Dr, CCSR 1155b, Stanford, CA, USA 2 Department of Medicine (Hematology, Blood and Marrow Transplantation), Stanford University, Stanford, CA, USA 3 Quantitative Science Unit, Stanford University, Stanford, CA, USA 4 Department of Microbiology and Immunology, Stanford University, Stanford, CA, USA 5 Department of Medicine (Infectious Diseases and Geographic Medicine), Stanford University, Stanford, CA, USA 6 Department of Medicine (Infectious Diseases), Stanford University, Stanford, CA, USA 7 Department of Medicine (Epidemiology and Population Health), Stanford University, Stanford, CA, USA 8 Department of Pathology, Stanford University, Stanford, CA, USA ∗ Corresponding author 9 These authors contributed equally 10 Lead contact 13 4 2022 10 6 2022 13 4 2022 3 6 371387.e9 27 1 2022 1 3 2022 5 4 2022 © 2022 Elsevier Inc. 2022 Elsevier Inc. Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background COVID-19 manifests with respiratory, systemic, and gastrointestinal (GI) symptoms.1, SARS-CoV-2 RNA is detected in respiratory and fecal samples, and recent reports demonstrate viral replication in both the lung and intestinal tissue.2, 3, 4 Although much is known about early fecal RNA shedding, little is known about long-term shedding, especially in those with mild COVID-19. Furthermore, most reports of fecal RNA shedding do not correlate these findings with GI symptoms.5 Methods We analyzed the dynamics of fecal RNA shedding up to 10 months after COVID-19 diagnosis in 113 individuals with mild to moderate disease. We also correlated shedding with disease symptoms. Findings Fecal SARS-CoV-2 RNA is detected in 49.2% [95% confidence interval, 38.2%–60.3%] of participants within the first week after diagnosis. Whereas there was no ongoing oropharyngeal SARS-CoV-2 RNA shedding in subjects at 4 months, 12.7% [8.5%–18.4%] of participants continued to shed SARS-CoV-2 RNA in the feces at 4 months after diagnosis and 3.8% [2.0%–7.3%] shed at 7 months. Finally, we found that GI symptoms (abdominal pain, nausea, vomiting) are associated with fecal shedding of SARS-CoV-2 RNA. Conclusions The extended presence of viral RNA in feces, but not in respiratory samples, along with the association of fecal viral RNA shedding with GI symptoms suggest that SARS-CoV-2 infects the GI tract and that this infection can be prolonged in a subset of individuals with COVID-19. Funding This research was supported by a Stanford ChemH-IMA grant; fellowships from the AACR and NSF; and NIH R01-AI148623, R01-AI143757, and UL1TR003142. Graphical abstract Context and significance Gastrointestinal symptoms and SARS-CoV-2 RNA shedding in feces point to the gastrointestinal tract as a possible site of infection in COVID-19. Researchers from Stanford University measured the dynamics of fecal viral RNA in patients with mild to moderate COVID-19 followed for 10 months post-diagnosis. The authors found that fecal viral RNA shedding was correlated with gastrointestinal symptoms in patients who had cleared their respiratory infection. They also observed that fecal shedding can continue to 7 months post-diagnosis. In conjunction with recent related findings, this work presents compelling evidence of SARS-CoV-2 infection in the gastrointestinal tract and suggests a possible role for long-term infection of the gastrointestinal tract in syndromes such as “long COVID.” Natarajan et al. perform a longitudinal study of fecal SARS-CoV-2 RNA shedding in patients with mild to moderate COVID-19, revealing that patients can shed RNA for up to 7 months after infection, that shedding is associated with gastrointestinal symptoms, and that the gastrointestinal tract may be infected even after the respiratory infection has cleared. Keywords COVID-19 SARS-CoV-2 fecal RNA gastrointestinal infection viral shedding Published: April 13, 2022 ==== Body pmcIntroduction COVID-19 is a disease with protean manifestations, ranging from respiratory to gastrointestinal to systemic. Although the primary site of SARS-CoV-2 infection is the respiratory tract, the presence of symptoms affecting other organ systems (e.g., abdominal pain, nausea, arthralgia), coupled with in vitro evidence of SARS-CoV-2 infectivity in a variety of other tissues, suggests that SARS-CoV-2 infection can extend beyond the respiratory system. Meta-analyses of studies that focus on hospitalized individuals with COVID-19 estimate the pooled incidence of gastrointestinal (GI) symptoms such as nausea, vomiting, and diarrhea to be between 11% and 18%.1 , 2 , 3 , 6, 7, 8 Additionally, within this moderate- to severe-disease group, SARS-CoV-2 RNA has been detected in 40%–85% of fecal samples (reviewed in Brooks EF and Bhatt AS9), indicating that SARS-CoV-2 viral RNA is found in feces nearly as frequently as in respiratory secretions.10 Patients with moderate to severe COVID-19 have been well studied; by contrast, much less is known about the clearance of SARS-CoV-2 RNA in the feces of patients with mild to moderate disease despite the fact that they make up ∼81% of those who contract COVID-19.11 , 12 Furthermore, most studies are cross-sectional, and the few reported longitudinal studies have focused on the early time period after diagnosis. Thus, a comprehensive understanding of the dynamics of fecal clearance of SARS-CoV-2 RNA in individuals with mild to moderate COVID-19 is both of crucial importance and lacking. Interestingly, in the few studies that have investigated longitudinal fecal samples, prolonged fecal shedding of SARS-CoV-2 RNA can occur even after respiratory shedding ceases. Indeed, in one notable pediatric case, fecal viral RNA shedding extended beyond 70 days after disease onset.8 If SARS-CoV-2 RNA shedding in the feces is indicative of a GI infection, this suggests that SARS-CoV-2 infection of the GI tract can continue after clearance from the respiratory tract. While the presence of SARS-CoV-2 RNA in feces is well established, whether live, infectious SARS-CoV2 is commonly shed in stool remains an outstanding question (reviewed in Guo M et al.13). Five studies have reported isolating infectious SARS-CoV-2 from stool samples collected from participants with severe COVID-19,14, 15, 16, 17, 18 whereas others have reported being unable to isolate infectious virions from stool.19 , 20 Therefore, it remains unclear whether the presence of infectious virions of SARS-CoV-2 in the stool is a rare or common phenomenon. However, there is mounting evidence of possible SARS-CoV-2 infection of the GI tract. Specifically, the presence of SARS-CoV-2 RNA,4 , 21, 22, 23 protein antigen,21 , 24 and virions4 , 23 , 25 in GI biopsies all point to a potential infection of the GI tract. Additionally, the presence of a gut immune response26 and inflammation markers such as fecal calprotectin27 , 28 in individuals with COVID-19 provides supporting evidence of a GI infection. Finally, in vitro experiments reveal that SARS-CoV-2 is able to successfully infect enteroid models of the gut29, 30, 31 and intestinal cell lines.32 This phenomenon of possible GI tract involvement is not surprising, as bovine coronavirus (BCoV) and human enteric coronavirus (HECoV-4408), both of the same genus as SARS-CoV-2 (Betacoronaviruses), can infect respiratory and GI tissues.33 Taken together, these data indicate that the GI tract may be an important site of SARS-CoV-2 infection.33 SARS-CoV-2 presence in the GI tract has additional relevance to patient health. The GI tract is a highly immunoactive tissue, and SARS-CoV-2 antigens in this body site may hone a humoral immune response against variants of the SARS-CoV-2 virus.21 Furthermore, prolonged presence of SARS-CoV-2 in the GI tissue may also have an impact on the hitherto mysterious phenomenon of post-acute sequelae of SARS-CoV-2 infection (PASC) or “Long COVID,” where individuals suffer from an unusual constellation of symptoms even after recovery from the respiratory SARS-CoV-2 infection.34 Taken together, it is critical that we understand whether or not the GI tract is infected and the dynamics of the infection in this tissue, from the standpoint of both the acute infection and the long-term sequelae of COVID-19. Here, we sought to better define the features of SARS-CoV-2 presence in the GI tract and its relevance for short- and long-term human health. We leveraged longitudinal fecal and respiratory samples from individuals enrolled in a randomized controlled study of Peg-interferon lambda-1a (IFN-λ) versus a placebo control for the treatment of mild to moderate COVID-19 (n = 120).35 While the intervention did not shorten the duration of oropharyngeal shedding of SARS-CoV-2 RNA (primary outcome) or disease symptoms (secondary outcome), the study provided a rich, prospectively collected dataset from which to evaluate fecal shedding dynamics and its relation to GI symptoms. Using fecal samples collected at regular intervals from the time of COVID-19 diagnosis to 10 months after diagnosis, we compared fecal viral RNA shedding with the presence of GI and other symptoms and found that it is positively correlated with GI symptoms. This constitutes the largest longitudinal analysis of paired fecal viral RNA shedding and disease symptomatology data in individuals with mild to moderate COVID-19, and it reveals important information about the pathophysiology of the disease. Results Description of study participants and sample collection The Peg-IFN-λ clinical trial (NCT04331899) enrolled 120 participants with mild to moderate COVID-19 between 25 April and 17 July 2020.35 Of these, 113 participants collected at least one stool sample at one of the six predefined stool collection time points. These collection time points centered around days 3 (range = 0–7 days), 14 (8–21), 28 (22–35), 120 (75–165), 210 (166–255), and 300 (>255 days) post-enrollment (Figure 1A). Of these 113 participants, 86 provided samples for at least three time points (summarized in Data S1; the overall Data S1 file includes additional data and analysis that informs methods and conclusions in the study and is related to Figures 1, 2 , 3 , and STAR Methods).Figure 1 Summary of study protocol and cohort demographics (A) Sample and data collection timeline represented in days. Day 0 marks the day of enrollment in the trial, within 72 h of a COVID-19 diagnosis. Each sample collection event is marked by a colored dot, where orange represents a blood draw and blue an oropharyngeal (OP) swab. Additionally, clinical appointments and symptom surveys are marked by yellow and green dots, respectively. Some of these events are marked by day ranges to represent collection time frames. The symptom survey at day 0 retrospectively collected symptomatology for 3 weeks prior to enrollment using a single questionnaire. Symptom surveys at time points centered around days 120, 210, and 300 retrospectively collected symptomatology for 1 week prior to the appointment using a single questionnaire at each timepoint. Collection of stool samples and their respective day ranges are marked below the timeline. Subjects were asked to provide samples in the OMNIgene GUT collection tube (OG) and the Zymo DNA/RNA shield fecal collection tube (ZY) at six time points. (B) Cohort characteristics. 120 participants were enrolled in the clinical trial. Participants had a COVID-19 infection of mild to moderate severity and were between the ages of 18 and 71. The age and sex distributions of the paticipants are reprented here. The x axis separates the groups by self-reported sex, and the y axis lists age in years. Each bar represents a range of 5 years. Figure 2 Fecal and oropharyngeal viral gRNA measurements over time (A) Summary of viral RNA positivity rates as determined by fecal and OP samples acquired from participants enrolled in the study for a period of around 28 days. The x axis lists time point categories since enrollment as days 3 (range 0–7), 14 (8–21), and 28 (22–35). The y axis lists the percentage of fecal samples (brown bar) and OP samples (gray bar) that tested positive at each of the time points. Fecal positivity rates are evaluated using the logistic GEE model described in the statistical methods section (see STAR Methods), which averages over all of the sample collection methods, gene types, and technical replicates. OP positivity rates are evaluated for the swab taken within 3 days of the fecal sample. Each bar also marks the 95% confidence interval (CI). Number of participants and percent positivity are listed as numbers at the top of the plot in black and red fonts, respectively, and summarized in Data S1. (B) Same as (A), except restricted to the subset of those who participated in the extended study, and following them through all six time points. As before, the x axis lists time point categories since enrollment: days 3 (range 0–7), 14 (8–21), 28 (22–35), 120 (75–165), 210 (166–255), and 300 (>255 days), and the y axis lists the percentage of participants with positive fecal samples (brown bar) and OP samples (gray bar) at each of the time points, with 95% CI. Number of participants and percent positivity are listed in black and red fonts and summarized in Data S1. (C) SARS-CoV-2 viral RNA concentration in stool samples collected in the ZY kit from participants (n = 104) with mild to moderate COVID-19 infection over a time period of 300 days from enrollment in the study. Note that the ZY kits had higher overall positivity rates than the OG kits, so positivity rates in this panel tend to be slightly larger than the numbers in the previous two panels, which average over kits and genes. Fecal viral RNA concentration was determined using RT-qPCR with primers/probes targeting the E, N1, N2, and RdRP genes in the SARS-CoV-2 genome, as indicated in the tab at the top of each panel. The x axis lists time point categories since enrollment. The y axis lists the percentage of participants with a given viral RNA concentration, as indicated by the color scheme in the stacked bar plot; dark blue refers to those with no detectable viral RNA, orange to viral RNA concentrations between 0 and one log10 copy/μL, yellow between one and two log10 copies/μL, green between two and three log10 copies/μL, and light blue over three log10 copies/μL. Number of participants per time point is listed above each bar in the stacked bar plot. (D) Fecal viral RNA concentration in stool samples collected in the ZY kit from participants (n = 104) with mild to moderate COVID-19 infection and assayed using RT-qPCR detecting the N1 gene (viral RNA concentration in log10 copies per μL) versus time (continuous variable; x axis). Time point categories are indicated by color scheme: yellow for days 3 (range 0–7), lavender for day 14 (8–21), red for day 28 (22–35), gray for day 120 (75–165), light blue for day 210 (166–255), and dark blue for day 300 (>255 days). A smoothed line generated using LOESS regression (span parameter = 0.75) and 95% CI is marked in the scatterplot. Note that all viral RNA concentration measurements are expressed on a logarithmic scale by applying the transformation log10 (viral RNA concentration+1). Figure 3 The effect of IFN-λ on fecal viral RNA shedding (A) Percentage of participants with detectable fecal SARS-CoV-2 RNA across each of the study arms, as evaluated using the logistic GEE model described in the statistical methods section (STAR Methods). The x axis marks the time point in the study: days 3 (range 0–7), 14 (8–21), and 28 (22–35). The y axis indicates the percentage of participants with detectable fecal SARS-CoV-2 RNA. The blue bar corresponds to participants in the placebo control arm, and the orange bar corresponds to participants in the IFN-λ intervention arm. Each bar also marks the 95% CI. Number of participants and percentage of participants that provided a positive stool sample are listed above each stacked bar in black and red fonts, respectively, and summarized in Data S1. (B) Odds ratio comparing detectable fecal SARS-CoV-2 RNA shedding in the IFN-λ intervention arm with the placebo arm at each time point in the first month of the study. The x axis marks the odds ratio adjusted for age, sex, collection kit type (OG or ZY), and target gene (E, N1, N2, or RdRP) (aOR). The y axis marks the time point in the study: days 3 (range 0–7), 14 (8–21), and 28 (22–35). The point marks the aOR, flanked by lines denoting the 95% CIs. The red dashed vertical line at aOR = 1.0 indicates no association. We originally started collecting stool samples in the OMNIGene GUT collection tube (OG), which is extensively used in gut microbiome studies.36 Parallel work from our group10 and one other group37 optimized and benchmarked stool collection and processing methods for the detection of fecal SARS-CoV-2 RNA; our group found that the Zymo DNA/RNA shield fecal collection tube (ZY) performs better than OG in viral RNA preservation. Therefore, starting on 14 May 2020, study participants were asked to provide samples in both the OG and ZY kits. Overall, a total of 326 samples were collected in the OG kit, and 347 in the ZY kit (sample collection compliance is summarized in Data S1 and the STAR Methods). In addition to these stool samples, oropharyngeal (OP) swabs were obtained daily during the initial part of the study, and at each study visit on days 120, 210, and 300; blood samples were drawn at days 0, 5, 14, 28, 120, 210, and 300 (Figure 1A). Clinical specimens were paired with self-reported symptom data collected through questionnaires administered on the day of enrollment and then daily from days 1 through 28, and on days 120, 210, and 300. Additionally, symptoms experienced in the 3 weeks preceding study enrollment were surveyed on the day of enrollment. Finally, long-term follow-up questionnaires on days 120, 210, and 300 collected symptoms occurring in the 7 days leading up to the appointment. Among the participants who returned at least one stool sample, the median age was 36 years (IQR = 29–51 years), 46 (41%) were female, and 72 (65%) were Hispanic (Figure 1B and Table 1 ). We describe the overall cohort, as well as two subsets: those reporting gastrointestinal (GI) symptoms (n = 54, 49%) at the first time point and those reporting no GI symptoms (i.e., exclusively respiratory symptoms or no symptoms at all) at that time point. Participants with GI symptoms at baseline are more likely to also experience a constellation of other symptoms, including myalgias (participants with GI symptoms, 78%; without GI symptoms, 30%; standard difference, −1.09), chills (59%, 21%, −0.84), decreased smell (63%, 30%, −0.7), headache (70%, 42%, −0.59), and joint pain (46%, 19%, −0.6). A comparison of those with and without GI symptoms, in terms of age, sex, ethnicity, and clinical measures at enrollment, including temperature, blood oxygen saturation, white blood cell count, blood alanine aminotransferase (ALT) concentration, and SARS-CoV-2 IgG seropositivity reveal no large differences and are presented in Table 1.Table 1 Cohort demographics and associated metadata Overall GI symptoms at enrollment Standardized difference Yes No n 111 54 57 Age, median (IQR) 36 (29–51) 36 (29–49) 37 (30–53) 0.05 Female, n (%) 46 (41%) 26 (48%) 20 (35%) −0.27 BMI (kg/m2), median (IQR) 27.7 (24.8–31.8) 28.2 (25.0–32.1) 27.4 (24.7–30.5) −0.25 Race/Ethnicity, n (%) Hispanic 72 (65%) 38 (70%) 34 (60%) −0.22 White 28 (25%) 12 (22%) 16 (28%) 0.13 Asian 4 (4%) 3 (6%) 1 (2%) −0.2 Unknown 6 (5%) 1 (2%) 5 (9%) 0.31 Symptomatology Asymptomatic at enrollment, n (%) 8 (7%) 0 (0%) 8 (14%) 0.56 Duration of symptoms in days prior to randomization, median (IQR) 5 (4–7) 6 (5–8) 5 (3–7) −0.61 GI symptoms at enrollment Any GI symptom 54 (49%) 54 (100%) 0 (0%) Abdominal pain 13.0 (12%) 13.0 (24%) 0 (0%) −0.8 Diarrhea 29.0 (26%) 29.0 (54%) 0 (0%) −1.53 Nausea 31.0 (28%) 31.0 (57%) 0 (0%) −0.8 Vomiting 5.0 (5%) 5.0 (9%) 0 (0%) −0.45 Other symptoms at enrollment Body aches (myalgias) 59.0 (53%) 42.0 (78%) 17.0 (30%) −1.09 Chest pain/pressure 21.0 (19%) 15.0 (28%) 6.0 (11%) −0.45 Chills 44.0 (40%) 32.0 (59%) 12.0 (21%) −0.84 Cough 62.0 (56%) 38.0 (70%) 24.0 (42%) −0.59 Decreased smell 51.0 (46%) 34.0 (63%) 17.0 (30%) −0.7 Fatigue 68.0 (61%) 43.0 (80%) 25.0 (44%) −0.78 Fever (>99.5°F) 10 (9%) 4 (7%) 6 (11%) 0.11 Headache 62.0 (56%) 38.0 (70%) 24.0 (42%) −0.59 Joint pain 36.0 (32%) 25.0 (46%) 11.0 (19%) −0.6 Shortness of breath 28.0 (25%) 17.0 (32%) 11.0 (19%) −0.28 Sore throat 43.0 (39%) 27.0 (50%) 16.0 (28%) −0.46 Rash 6.0 (5%) 4.0 (7%) 2.0 (4%) −0.17 Runny nose 24.0 (22%) 16.0 (30%) 8.0 (14%) −0.38 Laboratory values at enrollment, median (IQR) Absolute lymphocyte count (cells/μL) 1.5 (1.2–2.2) 1.4 (1.1–1.9) 1.6 (1.2–2.3) 0.33 Alanine aminotransferase (IU/L) 30.0 (22.0–48.5) 31.5 (22.0–47.8) 28.0 (22.0–50.0) 0.07 Aspartate aminotransferase (IU/L) 30.0 (25.0–39.0) 32.5 (26.0–41.0) 29.0 (24.0–34.0) −0.03 Seropositivity at enrollment, n (%) 46 (41%) 22 (41%) 24 (42%) 0.03 White blood cell count (cells/μL) 5.5 (4.2–7.1) 5.4 (3.8–7.1) 5.8 (4.7–7.1) 0.18 Longitudinal dynamics of SARS-CoV-2 RNA in stool A total of 673 stool samples collected from 113 participants over a period of 10 months were processed as per a recently optimized and benchmarked protocol10 outlined in the STAR Methods and summarized in Figure S1. Briefly, RNA was extracted from each of these stool samples and assayed for four target genes in the SARS-CoV-2 genomic RNA (gRNA) encoding the envelope protein (E), nucleocapsid protein (N1 and N2), and RNA-dependent RNA polymerase (RdRP) in technical duplicate, using RT-qPCR. We also assayed 278 of the 673 RNA samples, derived predominantly from samples collected in the first month of the study for the N1 and E gene, using multiplexed droplet digital PCR (ddPCR) assays because ddPCR is more robust to the presence of PCR inhibitors than RT-qPCR.38 We found the measurement of the N1 and E genes using ddPCR to be concordant with one another (Figure S2) and thus assayed the remainder of the samples (n = 395) only for the N1 gene. In total, 5,384 RT-qPCR assays and 951 ddPCR assays measuring the concentration of fecal SARS-CoV-2 gRNA were carried out. This dataset was then analyzed as summarized in the STAR Methods. SARS-CoV-2 viral RNA concentrations estimated by RT-qPCR and ddPCR targeting the N1 gene were found to be concordant (Figure S3; ZY, Pearson’s correlation, R = 0.98, p < 0.0001; OG, Pearson’s correlation, R = 0.9, p < 0.0001). Given the relative concordance between the RT-qPCR and ddPCR results, and the fact that that we had a richer dataset across four target genes in duplicate reactions using RT-qPCR, we decided to carry out the rest of our analyses using the RT-qPCR results alone; where relevant, associated analyses using ddPCR-derived viral RNA concentrations are included in Data S1 and are referenced below. We applied a logistic regression model that averaged RT-qPCR-derived viral RNA concentrations over all four target genes and both sample collection kits with fixed effects to correct for systematic differences. The model uses a generalized estimating equations (GEE) approach and is described in the STAR Methods; it was used in all our primary analyses except where noted. In study participants with uncomplicated COVID-19, the GEE model that considers RT-qPCR-derived viral RNA concentrations across all four target genes in the gRNA shows that 49% (95% confidence interval = 38%–60%) of participants (n = 102) were positive for fecal SARS-CoV-2 RNA at the first time point around day 3 (Figure 2A). The proportion of participants with fecal shedding of SARS-CoV-2 RNA gradually declined to 40% (95% confidence interval = 28%–53%, n = 86) on day 14 and 11.0% (6%–20%, 83) on day 28. To determine whether fecal SARS-CoV-2 RNA shedding continues after oropharyngeal shedding ceases, we compared the presence of SARS-CoV-2 RNA in fecal samples to that in OP samples from the same participant.35 At 4 months (120 days) post-enrollment, all participants (n = 57) who provided paired fecal and OP samples tested negative for SARS-CoV-2 RNA in their OP samples but 12.7% (95% confidence interval = 8.5%–18.4%) of their fecal samples were positive for SARS-CoV-2 RNA (Figure 2B). OP samples were not tested beyond the 4-month time point. However, at 7 months (210 days) post-enrollment, 3.8% (2.0%–7.3%) of the participants’ fecal samples were positive for SARS-CoV-2 RNA. Among the 23 fecal samples collected at 10 months (300 days), none were positive for SARS-CoV-2 RNA. It should be noted that the presence of viral RNA in the feces at the later time points could be the consequence of prolonged infection and viral RNA shedding or the consequence of a re-infection. We then calculated the absolute concentrations of fecal SARS-CoV-2 RNA using RT-qPCR of samples collected in the ZY kit (Figure 2C; corresponding data from samples collected in the OG kit are presented in Figure S4). In samples collected around day 3, between 54% and 77% of the participants shed viral RNA in their stool, depending on the gene targeted in the assay. At the first time point, looking at viral RNA concentrations derived from measuring the N1 gene, the gene that yielded the most number of SARS-CoV-2-positive fecal samples at this time point, we find that positive stool samples had between 0.32 and 3.97 log10 copies of viral RNA per microliter of eluate. We found that these viral RNA concentration data were concordant when measured using an orthogonal assay using ddPCR (Figure S5). Finally, to understand the temporal dynamics of shedding, we treated time since enrollment in the study as a continuous variable (Figure 2D), and we observed a decline in fecal viral gRNA concentration over the first month post-enrollment, with a few individuals demonstrating extended shedding versus evidence of a possible re-infection at the 4- and 7-month time points. Although gRNA is regularly used as an indicator of SARS-CoV-2 infection, this biomolecule does not mark an active infection, because non-infective viral particles can also harbor gRNA. Subgenomic RNA (sgRNA) is a possible indicator of an actively replicating virus, although there is ongoing debate about its specificity. Hence, we quantified sgRNA as previously described;39 23.8% (95% confidence interval = 15.2%–35.3%) of participants had detectable sgRNA (0.8–5.69 log10 copies of viral sgRNA per μL of eluate) in the first time point after diagnosis (Figure S6). This is in comparison with the 49.2% (38.2%–60.3%) of participants who had detectable gRNA in the first time point after diagnosis. Although there were samples that tested positive for gRNA that did not test positive for sgRNA, there were no samples where sgRNA was detected but gRNA was not. Finally, at the fourth time point, SARS-CoV-2 sgRNA had almost totally cleared, with 0.7% (0.2–3.0%) of samples remaining positive for sgRNA. Impact of interferon lambda on fecal shedding of SARS-CoV-2 RNA As samples from this study were collected from individuals on a randomized controlled trial of Peg-IFN-λ , we carried out an exploratory analysis to determine whether this intervention affected fecal SARS-CoV-2 RNA clearance in the first month after treatment. We found that there was no significant difference in the percentage of participants who shed SARS-CoV-2 RNA in their feces between the two arms of the study at the first three time points (Figure 3A). We went on to calculate the odds ratio adjusted for age, sex, collection kit type, and target gene (adjusted odds ratio, aOR) that a person who received the IFN-λ intervention would also be shedding viral RNA in stool at the first three time points (Figure 3B). At the first time point, around 3 days after enrollment in the study, we find that receiving the IFN-λ intervention was associated with lower odds of shedding viral RNA in stool (aOR = 0.32, 95% confidence interval = 0.12–0.89). While the association between exposure to IFN-λ and lower odds of fecal viral RNA shedding was intriguing and suggested that exposure to the intervention on day 1 might decrease short-term fecal viral RNA shedding, this association failed to replicate upon execution of several sensitivity analyses (Figure S7 and Data S1). In summary, in the current study, we did not observe a robust effect of a single 180-μg subcutaneous dose of IFN-λ on fecal SARS-CoV-2 RNA shedding. Subjects with detectable fecal SARS-CoV-2 RNA also manifest GI symptoms In limited recent studies, the presence of fecal SARS-CoV-2 RNA has been linked to the presence of GI symptoms. However, these studies are mostly cross-sectional in nature, collect symptomatology data retrospectively and do not use a uniform, benchmarked methodology for quantification of SARS-CoV-2 RNA in stool. To address the question of whether fecal viral RNA shedding is associated with GI symptoms, we collected comprehensive longitudinal symptomatology data, including information on GI symptoms from study participants in this interventional trial and compared these with absolute viral RNA concentrations measured in their feces (Figure 4A). Across the first month of the study, we found that participants who shed viral RNA in their stool were more likely to report nausea (aOR = 1.61, 95% confidence interval = 1.09–2.39), vomiting (3.20, 1.11–9.21), and abdominal pain (2.05, 1.09–3.86); no association was observed between viral RNA shedding and diarrhea (1.10, 0.63–1.91) or when considering any GI symptom (1.38, 0.94–2.04). Respiratory and systemic symptoms including runny nose (1.67, 1.05–2.66), headaches (1.56, 1.04–2.35), and body aches (2.21, 1.45–3.38) were also associated with the presence of fecal SARS-CoV-2 RNA. These results taken together, fecal SARS-CoV-2 RNA shedding is positively associated with most GI symptoms and with specific systemic and respiratory symptoms.Figure 4 Association between fecal viral RNA shedding and symptoms We present these results in the overall population, as well as stratified by the presence and absence of ongoing viral RNA shedding from the oropharynx (OP). (A) Summary of the association between viral RNA shedding and report of a given symptom in all participants. Shedding and symptom data from up to day 28 were included in this analysis. Adjusted odds ratios (aOR) for this association were evaluated using the logistic GEE model described in the statistical methods section (STAR Methods), which averages over collection kits (OG and ZY), target genes (E, N1, N2, and RdRP) and technical replicates and is adjusted for age, sex, collection kit, and target gene. The x axis indicates the aOR for the presence of a given symptom. The y axis lists symptoms divided into those associated with the GI tract and those not associated with the GI tract. The odds ratio for each symptom is indicated by the circle, and associated bars represent the 95% CI. The red dashed vertical line at aOR = 1.0 indicates no association. The percentage of surveys reporting each symptom is provided to the left of these bars. aOR and the 95% CIs are listed to the right of the bars. Analyses where sample size was insufficient are listed as “Too few reports.” (B and C) Identical data to (A) where (B) lists participants with negative paired OP swabs for SARS-CoV-2 RNA, and (C) lists participants with positive paired OP swabs for SARS-CoV-2 RNA. To determine whether the observed association between symptoms and fecal shedding was independent of respiratory shedding, we next divided the data into two subsets based on whether or not the participant was shedding virus in the oropharynx at the time the fecal sample was taken; specifically, we looked at participants whose OP swabs were collected within 3 days of the stool sample and (1) did not have any detectable SARS-CoV-2 RNA (n = 69; Figure 4B) or (2) had detectable SARS-CoV-2 RNA (n = 54; Figure 4C). Participants who were shedding viral RNA from the oropharynx had higher rates of almost all COVID-19-related symptoms, and we found no significant association between fecal shedding and symptoms for this subgroup. By contrast, participants who were not shedding viral RNA from the oropharynx had far lower rates of COVID-19-related symptoms in general. However, we found many significant associations between fecal shedding and symptoms in this subgroup. This is consistent with an interpretation that patients with an active infection of the respiratory system can experience an array of COVID-19-related symptoms independent of whether or not they are shedding viral RNA in their feces, but that patients whose respiratory infection has cleared could still be experiencing an active infection of the GI tract, which itself is associated with many different COVID-19-related symptoms. Taken together, these data suggest that fecal shedding of SARS-CoV-2 RNA is a possible indicator of an ongoing GI infection, and that this infection is accompanied by GI and other systemic symptoms. Discussion Severe SARS-CoV-2 infections can lead to a life-threatening hypoxemic respiratory failure. Therefore, much of the initial investigation of COVID-19 focused on the respiratory infection and related manifestations of the disease. This may be why, two years into the pandemic, we still do not definitively know whether SARS-CoV-2 infects the GI tract of humans. However, we know that SARS-CoV-2 can infect intestinal cells in vitro, both in cell lines32 and in human tissue-derived intestinal organoids.29, 30, 31 Additionally, the largest autopsy series of patients with COVID-19 to date recently demonstrated consistent evidence of infection of the small intestine by SARS-CoV-2; they also recovered live virus from these intestinal biopsies.4 This evidence suggests that SARS-CoV-2 can infect the GI tract, and perhaps when it does, it induces the GI symptoms observed in individuals with COVID-19. This postulated GI-tropism of SARS-CoV-2 is in keeping with the fact that other Betacoronaviruses that infect mammals can cause GI diseases. For example, BCoV causes severe GI diseases such as calf diarrhea and winter dysentery in cows., 40 What we have lacked in trying to understand whether the GI tract is commonly infected in COVID-19 is longitudinal samples that demonstrate prolonged shedding of fecal viral RNA after respiratory shedding has stopped. We have also lacked data that would enable us to clearly investigate whether or not there is a link between fecal viral RNA shedding and GI symptoms, both during and after respiratory infection by SARS-CoV-2. To address this gap, we leveraged one of the largest collections of longitudinal fecal samples from patients with mild to moderate COVID-19 to investigate fecal viral RNA shedding and its relationship to both OP viral RNA shedding and COVID-19 symptoms. Among the 113 participants who provided stool samples in this study, 49.2% (95% confidence interval = 38.2–60.3%) shed viral RNA in their feces within 6 days after their COVID-19 diagnosis. The fact that only a subset of individuals with COVID-19 exhibited fecal viral RNA shedding may be a consequence of a broad, nearly 1-week, window for the first sample collection from the time of diagnosis; alternatively, this may also be the result of physiological and genetic differences between individuals. Over the course of the first month in this study, the number of participants shedding fecal viral RNA decreased to 11% (6%−20%), and the viral RNA concentration among those still shedding decreased from up to ∼3 log10 copies per microliter to <1 log10 copies per microliter. At the first time point, we found that a larger proportion of participants shed viral RNA in their OP swab compared with their feces; however, this trend reversed in the rest of the time points. This suggests that clearance of SARS-CoV-2 is more rapid in the respiratory tissue than it is in the GI tissue and that the GI tract may be a site of longer-term infection. When considered in the context of previously documented evidence of a likely GI infection by SARS-CoV-2, our detection of SARS-CoV-2 sgRNA in fecal samples supports the model of an active infection in the GI tract. The presence of sgRNA, as opposed to gRNA, has been proposed as a marker of active infection and viral replication; however, subsequent work has now established that sgRNA outlives actively replicating virus in cell culture experiments and therefore may be an unreliable indicator of an ongoing, active infection.39 , 41 Therefore, although we detected sgRNA in stool up to 28 days after infection, whether or not this, on its own, is sufficient evidence of an ongoing infection remains unclear. Beyond informing our understanding of SARS-CoV-2 pathobiology, the information we present on the frequency, amount, and duration of viral RNA shed in stool is valuable for inferring population-level prevalence of COVID-19 from wastewater studies. This may, in turn, help inform public health measures. For example, long-term fecal viral RNA shedders may contribute to prolonged elevated levels of SARS-CoV-2 RNA in wastewater. If transmission occurs largely or entirely through respiratory secretions, the continued presence of fecal viral RNA in wastewater from a prolonged GI infection may be mistakenly interpreted as evidence of the prevalence of infectious individuals in a community. Since wastewater viral RNA levels are being considered for use in guiding community level policies (e.g., shutdowns and re-openings),42, 43, 44, 45, 46 it is critical that we understand how respiratory viral shedding and transmissibility of SARS-CoV-2 RNA are temporally related to fecal viral RNA shedding. Based on the available evidence, it is highly plausible that the presence of GI symptoms in patients with COVID-19 is due to infection of the GI tissues. With a comprehensive collection of clinical symptom data and fecal viral RNA concentrations, we find that over the course of the first month after enrollment, those who shed viral RNA in stool are more likely to also have GI symptoms including nausea, vomiting, and abdominal pain among other symptoms like runny nose, body aches, and headaches. It is notable that those who shed viral RNA in stool were not more likely to have diarrhea—this finding is contradicted by two prior studies (n = 59, 44), which found that patients with diarrhea were more likely to shed viral RNA in stool and, that too, at higher concentrations.2 Our finding of no association between diarrhea and fecal viral RNA shedding might be due to the relatively small number of participants who reported diarrhea in our study. When focusing on participants who had extended shedding of viral RNA in their stool even after their OP shedding had ceased, we found that fecal shedding of viral RNA is associated with a range of systemic and GI symptoms. On the other hand, for the duration that participants provided an OP swab positive for viral RNA, i.e., had an active respiratory infection, we did not find any association between fecal viral RNA shedding and symptomatology. We postulate that this is because participants who have an ongoing respiratory infection manifest classic COVID-19-related symptoms whether or not they have an infection in their GI tract. These observations support the hypothesis that there is likely a prolonged SARS-CoV-2 infection of the GI tract even after the upper respiratory infection is cleared. Since the GI tract is a highly immunoactive tissue,47 prolonged infections of the GI tissue may have consequences for patient health and may also be associated with the hitherto mysterious phenomenon of PASC or Long COVID. In fact, many studies following patients who have recovered from COVID-19 identify the prolonged presence of GI sequelae.48, 49, 50, 51, 52, 53, 54 In conclusion, we sought to address a key gap in our knowledge about the pathophysiology of a possible GI infection by SARS-CoV-2 by sampling stool over an extended period of time (10 months) and gathering paired symptomatology data. We have demonstrated the longest recorded shedding of fecal SARS-CoV-2 RNA in any COVID-19 patient: ∼210 days post-infection in two participants. Furthermore, we have found that extended shedding of SARS-CoV-2 RNA in participants who no longer have detectable viral RNA in OP swabs is closely associated with a host of systemic and GI symptoms, providing further evidence of a SARS-CoV-2 infection of the gut. Data presented here, when placed in the context of preliminary work that has suggested that the extended presence of SARS-CoV-2 viral antigen in gut biopsies from participants with COVID-19 may be associated with an improved immune response,21 urges follow-up immunological studies that investigate stool samples. Finally, initiatives such as Researching COVID to Enhance Recovery (RECOVER, NIH) that are poised to elucidate the hitherto elusive phenomenon of PASC should look closely at stool samples as an important factor of SARS-CoV-2 infection with potential long-term impact. Limitations of the study Despite its large size and longitudinal nature, this study has limitations. First, the study is limited in its resolution, having collected only six samples over a 10-month period. Follow-up studies with more frequent sampling, especially in the first 2 months after diagnosis, may help build a more nuanced model of decline of fecal viral RNA concentration over time. This will also allow a closer evaluation of the relative cessation of viral RNA in stool vis-a-vis other respiratory samples such as the OP swab. We were also unable to collect stool samples in a way that would enable recovery of live virus. As this was an outpatient study during the early part of the pandemic, we required participants to collect stool themselves at home and then mail the stool kits to us. For safety and practical purposes, we thus had to provide participants with kits that were rated for virus inactivation. Future studies, which facilitate the careful, consistent collection of stool samples from individuals with COVID-19 in a safe setting, might enhance the likelihood of more accurate measurement of live virus. This would be more direct evidence of SARS-CoV-2 being viable in the gut. Third, we did not obtain direct tissue evidence of infection; to do so would require intestinal biopsies. Of note, recent autopsy-4 and prior biopsy-based21 reports in limited numbers of patients have demonstrated evidence of direct intestinal infection and cytopathic changes. While intestinal biopsies from patients with mild to moderate COVID-19 would be highly informative, to date, these samples have been understandably difficult to obtain. In upcoming large studies, such as the RECOVER study, a subset of patients will be getting such biopsies, and the results of these large-scale studies will be illuminating. Finally, it would be interesting to sequence fecal viral RNA from participants with extended shedding to evaluate the persistence of the original virus variant, evolution of the original variant, and/or potential re-infection by the same or a different SARS-CoV-2 variant. Unfortunately, one of the limitations of current technologies for sequencing variants from complex matrices such as stool is the requirement of an adequate concentration of virus to be able to either amplify or assemble the virus from direct or enriched sequencing. As future technologies are developed for sensitive determination of variant sequences from stool, this type of analysis should be feasible. Of note, this study was carried out prior to the emergence of the strains (Omicron, Delta) that are prevalent today. Different strains may have different relative tropisms to the respiratory versus GI tract and may exhibit differences in clearance rates. This may be the consequence of their inherent biology as well as the immune status of the host due to underlying disorders, prior COVID-19 disease, and natural immunization, or vaccination. Of note, in this study we used samples that were collected as part of a previously published clinical trial.35 The original study reports the enrollment criteria applied to recruit participants. Briefly, the study actively sought to have equal male and female, racially and socioeconomically diverse participants between the ages of 18 and 75 years. The study did not collect information about self-reported gender in recruitment. Participants at risk of current or imminent hospitalization, with a respiratory rate >20 breaths per minute, room air oxygen saturation <94%, history of decompensated liver disease, recent use of interferons, antibiotics, anticoagulants, or other investigational and/or immunomodulatory agents for treatment of COVID-19, and prespecified laboratory abnormalities were excluded. Additionally, pregnant or breastfeeding participants were also excluded. STAR★Methods Key resources table REAGENT or RESOURCE SOURCE IDENTIFIER Biological samples Stool from participants in Peginterferon Lambda-1a (IFN-λ) clinical trial (NCT04331899) Stanford University N/A Oropharyngeal swabs from participants in Peginterferon Lambda-1a (IFN-λ) clinical trial (NCT04331899) Stanford University35 N/A Chemicals, peptides, and recombinant proteins Phosphate buffered saline (PBS) Fisher Scientific BP399-500 0.8 mM Ethylenediaminetetraacetic Acid (EDTA) Fisher Scientific EC200-449-9 Nuclease-free water Ambion AM9937 Tris-HCl pH 8.0 Invitrogen 15567-027 Critical commercial assays QiaAMP Viral RNA Mini kit Qiagen 52906 Custom ddPCR Assay Primer/Probe Mix BioRad 10031277 One-Step RT-ddPCR Advanced Kit for Probes BioRad 1864021 TaqPath 1-Step RT-qPCR Master Mix, CG ThermoFisher A15299 Deposited data A digital repository of all data supporting the findings of this study can be found at Zenodo This study https://zenodo.org/record/6374138 Oligonucleotides Primers for RT-qPCR and ddPCR used in this study, see Data S1 This study N/A Probes for RT-qPCR and ddPCR used in this study, see Data S1 This Study N/A Recombinant DNA Synthetic SARS-CoV-2 RNA ATCC VR-3276SD Zoetis Calf-Guard Bovine Rotavirus-Coronavirus Vaccine Zoetis VLN 190/PCN 1931.20 Software and algorithms Design and Analysis software Thermo Fisher Scientific Version 2.5.1 REDCap Cloud https://projectredcap.org/ Version 1.5 Python https://www.python.org/ Version 3.8.5 Statsmodel package https://www.statsmodels.org/stable/index.html Version 0.12.0 RStudio https://www.rstudio.com/ Version 1.3.959 Other Biomek-FX liquid handler Biomek N/A 12k Flex Applied Biosystems qPCR machine Applied Biosystems N/A QX200 AutoDG Droplet Digital PCR System BioRad N/A BioRad C1000 thermocycler BioRad N/A ddPCR reader BioRad QX200 OMNIGene GUT collection tube DNA Genotek OM-200 Toilet accessory DNA Genotek OM-AC1 DNA/RNA shield fecal collection tube Zymo R1101-E 96-well plates BioRad HSP9601 Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines, see Data S1 Bustin et al. (2009)55 Quantitative Real-Time PCR Experiments (MIQE) guidelines Digital Minimum Information for Publication of Quantitative Real-Time PCR Experiments (dMIQE) guidelines, see Data S1 dMIQE Group & Huggett (2020)56 Digital MIQE guidelines Droplet Digital PCR Applications Guide on QX200 machines BioRad Droplet Digital PCR Applications Guide MicroAmp Optical 384-well plates FisherScientific 43-098-49 Optically clear seal Applied biosystems 4311971 Resource availability Lead contact Correspondence and requests for materials should be addressed to the lead contact, Ami S. Bhatt (269 Campus Dr, CCSR 1155b, Stanford University, Palo Alto, CA 94305. Tel: (650) 498-4438; e-mail: asbhatt@stanford.edu). Materials availability PCR primers sequences are reported in Data S1. Other resources are available upon request of the lead contact. Experimental model and subject details Study design and population A total of 120 adults aged 18 - 71 years who had received a positive SARS-CoV-2 reverse-transcriptase quantitative polymerase chain reaction (RT-qPCR) based respiratory swab test within the past 72 hours were recruited for enrollment in a single-blind, placebo controlled, phase 2 clinical trial of Peginterferon Lambda-1a (IFN-λ) as an intervention for uncomplicated coronavirus disease 2019 (COVID-19). Informed consent was obtained for all participants under Stanford University Institutional Review Board (IRB) approved protocol # 55619 (PIs: Upinder Singh, Prasanna Jagannathan). The primary results of the null study, secondary outcomes, and the full details of study recruitment, inclusion and exclusion criteria were previously reported on and are only briefly summarized here.35 Individuals with study defined lab abnormalities, respiratory rate >20 breaths per minute, room air oxygen saturation levels <94%, pregnancy or breastfeeding, or recent history of hospitalization, uncontrolled liver disease, or use of COVID-19 interventional therapeutics, anticoagulants, antibiotics, and/or antivirals were excluded from the study. Subjects were randomized 1:1 to either the interventional or control study arm to receive a one-time subcutaneous injection of Peginterferon Lambda-1a or saline, respectively, on the first day of enrollment. Randomization was stratified by age (≥50 and <50 years old) and sex. The demographics of study participants are summarized in Table 1. Participant information on sex, age, race and ethnicity was self-reported and was reported in the original clinical manuscript describing this study.35 Information on gender and socioeconomic status was not collected. In addition, healthy adults were recruited to provide stool samples for use as extraction controls under Stanford IRB protocol #42043 (PI: Ami Bhatt). All donors gave informed consent prior to donating stool samples. Information on sex, gender, age, socioeconomic status, race and ethnicity was not collected. Trial registration ClinicalTrials.gov Identifier: NCT04331899 Method details Study samples and data Stool and other data and samples were collected from each set of study participants as outlined below. For the first 28 days following enrollment, participants in the clinical trial completed daily symptom questionnaires administered via REDCap Cloud (version 1.5)36 and self-performed daily measurements of temperature and oxygen saturation using study provided at-home devices. Participants returned to the study site on 1, 3, 5, 7, 10, 14, 21, 28 days (all +/− 1 day) and 120, 210, and 300 days (all +/− 3 weeks) post-enrollment for follow-up visits during which oropharyngeal (OP) swabs were collected, symptoms were queried, and vital signs were recorded. All clinical trial participants were provided a fecal sample collection kit on 0, 5, 21, 28, 120, and 210 days after enrollment and were asked to collect a stool sample in the provided kit, store at room temperature, and drop off for processing at their subsequent study visit or mail back to the study site at the long term follow up time points. We define the following six time points based on when participants returned the stool samples: days 3 (range 0 - 7 days), 14 (8 - 21), 28 (22 - 35), 120 (75 - 165), 210 (166 - 255) and 300 (>255) (Figure 1A). At the start of study enrollment on 25 April 2020, the collection kit consisted of the OMNIGene GUT collection tube (OG), toilet accessory, gloves and Spanish and English translations of manufacturer instructions. Later, starting 14 May 2020, the Zymo DNA/RNA shield fecal collection tube (ZY) was included in the fecal sample collection kit in addition to the OG collection tube. Spanish and English translations of manufacturer instructions specific to the ZY collection tube were also added. Subsequently, all participants were asked to collect a portion of the same stool sample in both of the two kits for each time point. The OG and ZY collection tubes are both marketed to preserve stool samples at ambient temperatures for up to 30 days. This eliminated the burden of sample refrigeration requirements for study participants. Fecal samples were processed within 24 hours of receipt by the lab. Samples collected in the OG and ZY collection tubes were processed similarly, by first vortexing the collection tube for 30 seconds to thoroughly homogenize the sample. Each sample was then aliquoted into 1.8 mL cryovials, labeled with the patient study ID and study time point, and then frozen at −80°C. Healthy control stool samples for use in every batch of RNA extractions were obtained from a healthy individual without prior history of COVID-19 exposure or positive SARS-CoV-2 respiratory test. Healthy stool samples for the limit of blank (LoB) determination were collected in 2018 well prior to the onset of the pandemic. All healthy donors self-collected fecal samples fresh and stored them at 4°C until processing. Within 24 hours of sample collection, samples were aliquoted into cryovials without preservative and frozen immediately at −80°C. Extraction of RNA Stool samples were randomly assigned a sample ID and processed for RNA extraction in batches of 18 following a previously optimized method,10 which is summarized here and in Figure S1. Two positive controls (OG and ZY) were included in each extraction batch for a total of 20 extractions per batch. Positive controls were prepared by adding biopsy punches of stool collected from a healthy individual to OG (4 biopsy punches) and ZY (8 biopsy punches) tubes. Each tube was then spiked with 10 μL of synthetic SARS-CoV-2 RNA at 104 copies/μL, vortexed for 30 seconds for homogenization, transferred in 500 μL aliquots to eppendorf tubes and frozen −80°C. Samples were gradually thawed on ice and vortexed for five seconds to ensure thorough homogenization. 500 μL of the stool-buffer slurry was transferred to an eppendorf tube, spun at 10,000 x g for 2 minutes at room temperature, and 140 μL of the supernatant was transferred to a fresh eppendorf tube for RNA extraction using the QiaAMP Viral RNA Mini kit. RNA extraction was performed as per manufacturer’s protocol and eluted in 100 μL of the elution buffer EB from the kit. Extracted RNA was then transferred to 96 well plates, briefly spun down, sealed and stored at −80 °C until further analysis. Samples collected at the 4, 7 and 10 month timepoints and associated batch controls were additionally spiked with 10 μL of attenuated BCoV vaccine as recommended.10 BCoV was prepared by resuspending one vial of lyophilized Zoetis Calf-Guard Bovine Rotavirus-Coronavirus Vaccine in 3 mL of phosphate buffered saline as per the manufacturer’s instructions. Quantification and statistical analysis RT-qPCR quantification of RNA An RT-qPCR assay to detect and quantify SARS-CoV-2 genomic RNA (gRNA)57, 58 was developed using primer probe sets recommended by the United States Centers for Disease Control and prevention (CDC)59 targeting the Envelope protein (E), Nucleocapsid proteins (N1, N2), and RNA-dependent RNA polymerase protein (RdRP) of the viral genome. To quantify SARS-CoV-2 subgenomic RNA (sgRNA) from stool samples as previously described39 an additional primer probe set targeting the N1 gene with the forward primer annealing to the canonical leader sequence at the 5′ end was included in the assay. All RNA extracts were assayed for all four gRNA targets and the single sgRNA target. Primer and probe sequences are listed in Data S1. Each 20 μL RT-qPCR reaction was composed of 5 μL TaqPath 1-Step RT-qPCR Master Mix, CG, 1.5 μL of primer/probe mix, 8.5 μL of nuclease-free water. The primer/probe mix was prepared with a final concentration of 400 nM of each of the forward and reverse primers and 200 nM of the corresponding probe in 8.5 mM Tris-HCl pH 8.0 and 0.8 mM EDTA. Reactions were prepared in Micro-Amp Optical 384-well plates with 5 μL of stool RNA samples, synthetic RNA standards, or nuclease free water using a Biomek-FX liquid handler. Every assay plate also included standard curves. Standard curves were prepared by serially diluting quantitative synthetic SARS-CoV-2 RNA from 105-10−1 copies per μL. For standard curves in the sgRNA assays, a purified PCR product corresponding to the target gene39 was diluted from 106-10−1 copies per reaction. Nuclease-free water was used as a negative control. RNA extracted from each stool sample was assayed in two technical replicates for each target. Standard curves were run in technical duplicates for all targets on every RT-qPCR assay plate. Eight negative controls were included in each assay plate. Prior to the assay, plates were sealed with an optically clear seal and spun down at room temperature. The samples were assayed in a 12k Flex Applied Biosystems qPCR machine in standard mode using the following cycling conditions: 25°C for 2 minutes, 50°C for 15 minutes, and 95°C for 2 minutes, followed by 45 cycles of 95°C, 3 seconds, and 55°C, 30 seconds. In the RT-qPCR assays, quantification cycle (Cq) value was calculated using the Design and Analysis software. On a plate-by-plate basis, assays with a Cq value greater than the Cq value of the synthetic RNA standard at 1 copy per μL were called undetermined. Cq values for each sample were converted to viral RNA concentration in copies/μL using the linear regression model fit to the standard curve for each plate. We used a statistical model to average over the results of all the technical replicates, and more details about the model are available in the Statistical analysis section. Finally, we calculated the LoB of the assay (additional details are available in the in the following sections of the STAR Methods) and converted all viral RNA concentrations equal to or lower than the LoB to be undetermined, because these were beyond the reliable specificity of the assay. All viral RNA concentrations were expressed on a logarithmic scale by applying the transformation log10(viral RNA concentration+1). SARS-CoV-2 viral RNA concentrations from oropharyngeal swabs were derived from a previously published companion study.35 This study measured the E gene in the SARS-CoV-2 genomic RNA and RNaseP in the human genome in a multiplexed assay. RNaseP was used as an internal control for the extraction of RNA and to monitor the effect of RT-qPCR inhibitors in these samples. Only samples where RNaseP was detected were evaluated. As a requirement for the Stanford FDA Emergency Use Authorization for the SARS-CoV-2 RNA diagnostic test, the sensitivity of the assay for nasopharyngeal swab testing was determined to be 1000 copies/mL. While the FDA did not require the assessment of assay sensitivity for different respiratory tissues, we believe that the assay sensitivity for nasopharyngeal vs. oropharyngeal swabs to be comparable. Similarly, based on previously reported benchmarking and Limit of Detection (LoD) assays, the sensitivity of fecal sample testing for SARS-CoV-2 RNA is 1000 copies/mL.10 Moreover, the assay sensitivity of fecal testing was highly concordant between the tested genes, particularly for the N1, N2, and E genes; RdRP has a slightly lower sensitivity by comparison.10 Therefore, we are confident that the sensitivity of SARS-CoV-2 RNA testing is highly comparable in stool and respiratory biospecimen of the study subjects (1000 copies/mL). ddPCR quantification of RNA Droplet digital PCR (ddPCR) is resilient to PCR inhibitors prevalent in stool, enables absolute quantification without the need for an exhaustive standard curve, and is also more sensitive than traditional qPCR.10 , 37 Therefore, we quantified viral RNA using this orthogonal method as previously described.10 The ddPCR reactions were prepared with the One-Step RT-ddPCR Advanced Kit for Probes. Using a Biomek FX liquid handler, each reaction well was loaded with 5.5 μL of extracted RNA to 5.5 μL Supermix, 2.2 μL reverse transcriptase, 1.1 μL of 300 nM dithiothreitol (DTT), 1.1 μL of 20× Custom ddPCR Assay Primer/Probe Mix and 6.6 μL of nuclease-free water per the manufacturer instructions. For multiplexed reactions, we added 1.1 μL of each of the primer/probe mixes and reduced the amount of nuclease free water to 5.5 μL. We then used a QX200 AutoDG Droplet Digital PCR System to partition reaction samples into droplets of 1 nL using default settings. PCR amplification of the templates was performed on a BioRad T100 thermocycler using the following thermocycling program: 50 °C for 60 min, 95 °C for 10 min, 40 cycles of 94 °C for 30 s and 55 °C for 1 min, followed by 1 cycle of 98 °C for 10 min and 4 °C for 30 min with ramp speed of 1.6 °C/s at each step. Finally, amplified reactions were quantified using a ddPCR reader. The ddPCR analysis was guided by the Droplet Digital PCR Applications Guide on QX200 machines (BioRad)60 and the digital MIQE guidelines.56 We have included the recommended associated checklist in Data S1. We applied a rigorous strategy to threshold the assays and identify true positive reactions as previously described10 and summarized below. Briefly, we analyzed the standards and negative controls in a plate-by-plate fashion and applied a suitable threshold to these samples. This threshold was applied such that the number of positive droplets in the negative control was minimal and the concentration of RNA in the standard matched the theoretical expectation most closely. We then calculated the difference in amplitude between the negative droplets and the threshold in the reactions with the negative control, and applied a threshold to all the other wells such that this same difference in amplitude was maintained. Finally, as with the RT-qPCR reactions, we established an LoB for this assay (additional details are available in the following sections of the STAR Methods), and any sample with viral RNA concentration less than or equal to the LoB was considered to be undetermined. All viral RNA concentrations were expressed on a logarithmic scale by applying the transformation log10(viral RNA concentration+1). Ensuring high specificity in RT-qPCR and ddPCR assays of fecal SARS-CoV-2 RNA In assays to quantify viral RNA, we took a conservative approach at every step to ensure high specificity. First, we adopted a method to determine the limit of blank (LoB) that is based on guidelines set out by the Clinical and Laboratory Standards Institute (CLSI),61 as summarized in the next section. We systematically identified the LoB for stool collected in the OG and ZY kits against each of the four target genes in independent combinations. All samples with an RNA concentration equal to or lower than the corresponding LoB are considered to have an undetermined amount of viral RNA, since this is below a reliable specificity threshold for that assay (example in Data S1). Second, we identified the linear detection range of our assays. A six-point 10-fold dilution series of synthetic SARS-CoV-2 RNA from the American Type Culture Collection (ATCC) starting at 104 log10 copies per μL was used here as previously described.10 Resulting standard curves generated for each of the genes in the genomic RNA measured using RT-qPCR and those measured by ddPCR are shown in Data S1. In assays that detected sgRNA, we used a six-point 10-fold dilution series with pre-quantified sgRNA starting at 106 log10 copies per μL from a previously reported study39 and provide standard curves in Data S1. All samples that yield a viral RNA concentration below the lowest detectable concentration in the linear range of standards are considered to have an undetermined amount of viral RNA. Third, anticipating that few if any stool samples collected beyond the 28 day time point were going to be positive for SARS-CoV-2 RNA, we incorporated a control to guard against false negatives that could result from incomplete or inefficient extraction of RNA, as previously described.10 Briefly, all long-term stool samples were spiked with 10 μL of attenuated Bovine coronavirus (BCoV) prior to RNA extraction. The extracted RNA was then tested for the M gene from BCoV in addition to the regular SARS-CoV-2 based assays. This served to determine if RNA extractions were successful, ensuring we did not falsely report negative SARS-CoV-2 assays as a consequence of ineffective RNA extraction. Out of 239 samples, 237 yielded BCoV RNA, and those that did not were left out of further analysis. Together, these experimental checkpoints increase confident that our reported fecal viral RNA concentrations are accurate. Estimating limits of blanks Understanding the specificity of the assays used in this study to quantify viral RNA is critical to evaluate confidence in results derived thereof. Therefore, we used a strategy based on guidelines set out by the Clinical and Laboratory Standards Institute (CLSI)61 to quantify the limit of blank (LoB) of our stool preservation and detection protocol. To this end, we used stool samples collected from four healthy donors in the Fall of 2018. Since this was from before the emergence of SARS-CoV-2, these samples are confidently negative for SARS-CoV-2 RNA. One stool sample from each of the four donors was aliquoted into separate OG and ZY tubes as per manufacturer instructions. This was performed in independent duplicates by two different operators yielding 16 stool samples. Next, RNA was extracted from each of these samples in duplicate by the two operators resulting in 64 total RNA extracts. The sample preparation protocol is summarized in Data S1. The 64 RNA extracts were assayed for the E, N1, N2 and RdRP genes in the gRNA in duplicate reactions identical to how clinical samples were assayed in this study. Next, these samples were also assayed for the N1 genes in ddPCR assays. Taken together, we calculated the LoB for relevant combinations of stool preservation (OG, ZY), target gene (E, N1, N2 and RdRP), and detection method (RT-qPCR, ddPCR). It was notable that across all targeted genes in both RT-qPCR and ddPCR assays, the LoB measured in the OG kit was higher than that measured in the ZY kit. Specifically, RT-qPCR assays targeting the N1 gene yielded 0.487 log10 copies per μL of viral RNA in samples preserved in OG and 0.237 log10 copies per μL of RNA in those preserved in ZY. These corresponded to 0.429 copies per μL and 0.164 copies per μL of RNA in ddPCR assays targeting the N1 gene. Finally, while targeting the N2 gene via RT-qPCR also yielded low RNA concentrations in these negative controls, E and RdRP were highly specific and yielded no detectable RNA for these targets in the negative controls (Data S1). The RNA concentration derived here is used as the LoB in all further data analysis. Thus, all samples that bear an RNA concentration equal to or lower than the corresponding LoB are considered to have an undetermined amount of viral RNA, since this is below a reliable specificity threshold for that assay (example in Data S1). Guarding against PCR inhibitors for the reliable detection of viral RNA PCR inhibitors are often present in stool. Thus, we wanted to estimate the degree to which our RT-qPCR assays were impacted by PCR inhibition. We posited that diluting the stool RNA extracts prior to assaying for SARS-CoV-2 RNA would dilute any potential PCR inhibitors derived from the stool matrix. Thus, we would expect a higher positivity rate from assaying the diluted extracts. To this end, we assayed 72 clinical samples by RT-qPCR at the concentration they were extracted at (1X), and at a ten-fold dilution of the same samples (0.1X). In aggregate across the 4 RT-qPCR target genes, assaying the samples at 0.1X resulted in a gain of 4 positive samples but a loss of 15 positive samples, likely due to viral RNA concentration falling below the detection limit of the RT-qPCR assay with dilution (Data S1). Thus, the RT-qPCR analysis of the stool RNA extracts likely does not exhibit a high degree of PCR inhibition. Statistical analysis Absolute standardized differences (ASD),62 expressed in units of standard deviations, are displayed in Table 1 to compare the distribution of characteristics in participants reporting GI symptoms at enrollment or not. We interpreted ASDs using Cohen’s guidelines (d: 0.2 = small difference; 0.5 = medium difference; 0.8 = large difference; d < 0.2 = trivial difference).63 Our primary statistical analyses examined associations between participant characteristics and whether the RT-qPCR based detection of SARS-CoV-2 gRNA was positive, focusing only on the stool samples collected during the main study at the first three time points, and including fixed effects to account for the different positivity rates of the four target genes (E, N1, N2 and RdRP) and the two collection kits (OG and ZY). We augmented this with two sensitivity analyses. First, we conducted a subgroup analysis that included samples from all six time-points but that focused on the subset of participants who returned at least one sample during the long-term follow-up; we made decision to focus our primary analysis on the first three time points and to supplement it with this sensitivity analysis to avoid the concern that the decision to join the extended study might correlate with certain patient risk. Second, we conducted subset analyses that focused on individual genes separately. In all cases, we used logistic regression models fit with generalized estimating equations (GEE)64 to account for the correlation between samples and replicates within a participant. To examine whether Peginterferon Lambda-1a (IFN-λ) had an effect on fecal viral RNA shedding, we fit a logistic regression to estimate the odds ratio of fecal shedding in participants receiving the IFN-λ intervention versus those that received a saline placebo. We adjusted the odds ratio by collection kit type (OG and ZY) and gene (E, N1, N2 and RdRP), to account for systematic differences between measurements, and as well as by the patient’s age and sex, because randomization had been stratified by those features.65 We included statistical interaction terms between study arms and indicators for time of collection in the model to estimate the difference between study arms at each time of collection. In addition to the two sensitivity analyses described above, we also used a negative binomial model to assess the association between the IFN-λ intervention and the total viral RNA concentration, whereas before we used GEE to account for correlation within individual patients. In analyses to estimate association between fecal SARS-CoV-2 RNA and symptoms, we regressed the presence of symptoms reported at the time of sample collection on an indicator of the presence of fecal SARS-CoV-2 RNA, adjusted for age, sex, log of the number of days since symptom onset, collection kit type (OG and ZY), and gene (E, N1, N2 and RdRP). We fit a separate logistic regression for each of the symptoms. We additionally fit models including an interaction between fecal SARS-CoV-2 RNA shedding and an indicator of OP shedding to estimate associations among participants with or without an ongoing presence of viral RNA in their OP swabs. All tests were two-sided and conducted at the 0.05 level of significance. Analyses were performed in Python version 3.8.5, using the Statsmodel package, version 0.12.0. IFN-λ does not impact fecal SARS-CoV-2 RNA shedding Exposure to IFN-λ appears to present lower odds of fecal viral RNA shedding at the first time point, around 3 days after receiving the intervention (Figure 3B). However, this association failed to replicate upon closer examination using several sensitivity analyses, as follows.1) We calculated the adjusted odds ratio (aOR) that a person who received the IFN-λ intervention would also be shedding viral RNA in stool at the first three time points, limiting our attention to the subset of individuals who elected to participate in the extended study. Amongst these participants there was no association between the intervention and fecal shedding during any of the six time points (Figures S7A and S7B). 2) We looked at an analysis that was restricted to just individual genes and kits. In this analysis, we find that the association at the first time point is being driven entirely by samples collected in the OG kit, which has previously been shown to have lower sensitivity for fecal SARS-CoV-2 RNA detection10 (Data S1). 3) An analysis that looked at viral RNA concentrations instead of binary test results (positive vs. negative) found no association at any of the three time points (Figure S7C). Supplemental information Document S1. Figures S1–S7 Data S1. Additional data and analysis that informs methods and conclusions in the study, related to Figures 1, 2, and 3, and STAR Methods Document S2. Article plus supplemental information Data and code availability • All data supporting the findings of this study have been deposited at Zenodo (https://zenodo.org/record/6374138) and are publicly available as of the date of publication. • All custom code and mathematical models have been deposited at Zenodo (https://zenodo.org/record/6374138) and are publicly available as of the date of publication. • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments We thank Alexandria Boehm, Marlene Wolfe, and Nasa Sinnott-Armstrong for guidance on processing stool samples and detection of RNA; Angela Rogers for providing stool samples from participants admitted at Stanford Hospital; Rebecca Osbourne, Tiffany Nguyen, and the members of the Stanford Clinical and Translational Research Unit for assistance with stool sample receipt and processing; Elizabeth Ponder for coordinating initial stool sample collection kit distribution to study participants and providing information about funding from Chem-H; Catherine Blish and members of the Blish Lab for receiving and temporary storage of stool samples prior to biobanking; Dean Felsher for access to the QuantStudio 12K Flex qPCR machine; Yvonne Maldonado and Jonathan Altamirano for helping acquire funding to support this work; Said Attiya and Dhananjay Wagh for guidance on applying ddPCR assays; David Solow-Cordero for assistance setting up the Biomek FX and providing access; Luisa Jiminez and Sopheak Sim for assistance in using the Stanford Functional Genomics Facility and High-Throughput Bioscience Center; and Frida Salcedo for help acquiring reagents from Bio-Rad. We are grateful to the Peg-interferon-λ1a clinical trial team for coordinating procurement of stool samples from outpatients enrolled in this trial. Biorender has been a valuable resource for creating schematic illustrations. This work was supported by a ChemH-IMA grant (to A.S.B. and P.J.), the Stanford Dean’s Postdoctoral Fellowship (to A.N.), an AACR Fellowship (to S.Z.), and a NSF Graduate Research Fellowship Program grant (to A.H. and D.T.S.). The laboratory of A.S.B. is supported by 10.13039/100000002 NIH R01 AI148623 and R01 AI143757, and H.H. and the research reported in this publication are supported by the National Center for Advancing Translational Sciences of 10.13039/100000002 NIH award UL1TR003142. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Author contributions A.N., S.Z., E.F.B., and S.E.V. contributed equally to this work. A.N. designed experiments, extracted RNA from stool samples, assayed viral RNA using RT-qPCR and ddPCR, analyzed data, and wrote the manuscript. S.Z. designed experiments, assayed viral RNA using RT-qPCR and ddPCR, analyzed data, generated plots in R, and wrote the manuscript. E.F.B. and S.E.V. designed experiments, biobanked stool samples, extracted RNA from stool samples, and wrote the manuscript. A.D. and H.H. designed experiments, performed statistical analyses, generated plots in Python, and wrote the manuscript. R.M.P. analyzed data. A.H., D.T.S., and R.V. helped design experiments. K.B.J., J.P., H.F.B., U.S., B.A.P., J.A., and P.J. helped collect samples through the Lambda clinical trial and guided data analysis. A.S.B. helped design experiments, analyze data, and write the manuscript. S.Z., A.D., and H.H. performed and replicated the statistical analysis. A.N., E.F.B., S.E.V., and A.S.B. oversaw the statistical analysis. A.N., S.Z., E.F.B., S.E.V., A.D., H.H., and A.S.B. have unrestricted access to all data. A.N., S.Z., E.F.B., S.E.V., A.D., H.H., and A.S.B. prepared the first draft and reviewed and edited the manuscript. All authors read and approved the final manuscript and take responsibility for its content. Declaration of interests The authors declare no competing interests. Inclusion and diversity statement We worked to ensure gender balance in the study arms, recruited participants from diverse ethnic and socioeconomic backgrounds, and provided the study questionnaire and stool collection protocol in Spanish and English. Supplemental information can be found online at https://doi.org/10.1016/j.medj.2022.04.001. ==== Refs References 1 Merola E. Armelao F. de Pretis G. Prevalence of gastrointestinal symptoms in coronavirus disease 2019: a meta-analysis Acta Gastroenterol. Belg. 83 2020 603 615 33321018 2 Cheung K.S. Hung I.F.N. Chan P.P.Y. Lung K.C. Tso E. Liu R. Ng Y.Y. Chu M.Y. Chung T.W.H. Tam A.R. Gastrointestinal manifestations of SARS-CoV-2 infection and virus load in fecal samples from a Hong Kong cohort: systematic review and meta-analysis Gastroenterology 159 2020 81 95 32251668 3 Parasa S. Desai M. Thoguluva Chandrasekar V. Patel H.K. Kennedy K.F. Roesch T. Spadaccini M. Colombo M. Gabbiadini R. Artifon E.L.A. Prevalence of gastrointestinal symptoms and fecal viral shedding in patients with coronavirus disease 2019: a systematic review and meta-analysis JAMA Netw. Open 3 2020 e2011335 32525549 4 Chertow D. Stein S. Ramelli S. Grazioli A. Chung J.-Y. Singh M. SARS-CoV-2 infection and persistence throughout the human body and brain Res. Square 2021 10.21203/rs.3.rs-1139035/v1 5 Mao R. Qiu Y. He J.-S. Tan J.-Y. Li X.-H. Liang J. Shen J. Zhu L.-R. Chen Y. Iacucci M. Manifestations and prognosis of gastrointestinal and liver involvement in patients with COVID-19: a systematic review and meta-analysis. The Lancet Gastroenterol. Hepatol. 5 2020 667 678 6 Sultan S. Altayar O. Siddique S.M. Davitkov P. Feuerstein J.D. Lim J.K. Falck-Ytter Y. El-Serag H.B. AGA Institute AGA Institute rapid review of the gastrointestinal and liver manifestations of COVID-19, meta-analysis of International data, and recommendations for the consultative management of patients with COVID-19 Gastroenterology 159 2020 320 334.e27 32407808 7 Xu C.L.H. Raval M. Schnall J.A. Kwong J.C. Holmes N.E. Duration of respiratory and gastrointestinal viral shedding in children with SARS-CoV-2: a systematic review and synthesis of data Pediatr. Infect. Dis. J. 39 2020 e249 e256 32618932 8 Wang J.-G. Cui H.-R. Tang H.-B. Deng X.-L. Gastrointestinal symptoms and fecal nucleic acid testing of children with 2019 coronavirus disease: a systematic review and meta-analysis Sci. Rep. 10 2020 17846 33082472 9 Brooks E.F. Bhatt A.S. The gut microbiome: a missing link in understanding the gastrointestinal manifestations of COVID-19? Cold Spring Harb. Mol. Case Stud. 7 2021 a006031 33593727 10 Natarajan A. Han A. Zlitni S. Brooks E.F. Vance S.E. Wolfe M. Singh U. Jagannathan P. Pinsky B.A. Boehm A. Standardized preservation, extraction and quantification techniques for detection of fecal SARS-CoV-2 RNA Nat. Commun. 12 2021 5753 34599164 11 Wu Z. McGoogan J.M. Characteristics of and important lessons from the coronavirus disease 2019 (COVID-19) outbreak in China: summary of a report of 72 314 cases from the Chinese center for disease control and prevention JAMA 323 2020 1239 1242 32091533 12 Oran D.P. Topol E.J. The proportion of SARS-CoV-2 infections that are asymptomatic : a systematic review Ann. Intern. Med. 174 2021 655 662 33481642 13 Guo M. Tao W. Flavell R.A. Zhu S. Potential intestinal infection and faecal-oral transmission of SARS-CoV-2 Nat. Rev. Gastroenterol. Hepatol. 18 2021 269 283 33589829 14 Zhang Y. Chen C. Zhu S. Shu C. Wang D. Song J. Song Y. Zhen W. Feng Z. Wu G. Isolation of 2019-nCoV from a stool specimen of a laboratory-confirmed case of the coronavirus disease 2019 (COVID-19) China CDC Wkly. 2 2020 123 124 34594837 15 Xiao F. Sun J. Xu Y. Li F. Huang X. Li H. Zhao J. Huang J. Zhao J. Infectious SARS-CoV-2 in feces of patient with severe COVID-19 Emerg. Infect. Dis. 26 2020 1920 1922 32421494 16 Jeong H.W. Kim S.-M. Kim H.-S. Kim Y.-I. Kim J.H. Cho J.Y. Kim S.-H. Kang H. Kim S.-G. Park S.-J. Viable SARS-CoV-2 in various specimens from COVID-19 patients Clin. Microbiol. Infect. 26 2020 1520 1524 32711057 17 Wang W. Xu Y. Gao R. Lu R. Han K. Wu G. Tan W. Detection of SARS-CoV-2 in different types of clinical specimens JAMA 323 2020 1843 1844 32159775 18 Zhang Y. Chen C. Song Y. Zhu S. Wang D. Zhang H. Han G. Weng Y. Xu J. Xu J. Excretion of SARS-CoV-2 through faecal specimens Emerg. Microbes Infect. 9 2020 2501 2508 33161824 19 Wölfel R. Corman V.M. Guggemos W. Seilmaier M. Zange S. Müller M.A. Niemeyer D. Jones T.C. Vollmar P. Rothe C. Virological assessment of hospitalized patients with COVID-2019 Nature 581 2020 465 469 32235945 20 Albert S. Ruíz A. Pemán J. Salavert M. Domingo-Calap P. Lack of evidence for infectious SARS-CoV-2 in feces and sewage Eur. J. Clin. Microbiol. Infect. Dis. 40 2021 2665 2667 34240259 21 Gaebler C. Wang Z. Lorenzi J.C.C. Muecksch F. Finkin S. Tokuyama M. Cho A. Jankovic M. Schaefer-Babajew D. Oliveira T.Y. Evolution of antibody immunity to SARS-CoV-2 Nature 591 2021 639 644 33461210 22 Lin L. Jiang X. Zhang Z. Huang S. Zhang Z. Fang Z. Gu Z. Gao L. Shi H. Mai L. Gastrointestinal symptoms of 95 cases with SARS-CoV-2 infection Gut 69 2020 997 1001 32241899 23 Bradley B.T. Maioli H. Johnston R. Chaudhry I. Fink S.L. Xu H. Najafian B. Deutsch G. Lacy J.M. Williams T. Histopathology and ultrastructural findings of fatal COVID-19 infections in Washington State: a case series Lancet 396 2020 320 332 32682491 24 Xiao F. Tang M. Zheng X. Liu Y. Li X. Shan H. Evidence for gastrointestinal infection of SARS-CoV-2 Gastroenterology 158 2020 1831 1833.e3 32142773 25 Qian Q. Fan L. Liu W. Li J. Yue J. Wang M. Ke X. Yin Y. Chen Q. Jiang C. Direct evidence of active SARS-CoV-2 replication in the intestine Clin. Infect. Dis. 73 2021 361 366 32638022 26 Britton G.J. Chen-Liaw A. Cossarini F. Livanos A.E. Spindler M.P. Plitt T. Eggers J. Mogno I. Gonzalez-Reiche A.S. Siu S. Limited intestinal inflammation despite diarrhea, fecal viral RNA and SARS-CoV-2-specific IgA in patients with acute COVID-19 Sci. Rep. 11 2021 13308 34172783 27 Cholankeril G. Podboy A. Aivaliotis V.I. Tarlow B. Pham E.A. Spencer S.P. Kim D. Hsing A. Ahmed A. High prevalence of concurrent gastrointestinal manifestations in patients with severe acute respiratory syndrome coronavirus 2: early experience from California Gastroenterology 159 2020 775 777 32283101 28 Effenberger M. Grabherr F. Mayr L. Schwaerzler J. Nairz M. Seifert M. Hilbe R. Seiwald S. Scholl-Buergi S. Fritsche G. Faecal calprotectin indicates intestinal inflammation in COVID-19 Gut 69 2020 1543 1544 32312790 29 Lamers M.M. Beumer J. van der Vaart J. Knoops K. Puschhof J. Breugem T.I. Ravelli R.B.G. Paul van Schayck J. Mykytyn A.Z. Duimel H.Q. SARS-CoV-2 productively infects human gut enterocytes Science 369 2020 50 54 32358202 30 Zang R. Gomez Castro M.F. McCune B.T. Zeng Q. Rothlauf P.W. Sonnek N.M. Liu Z. Brulois K.F. Wang X. Greenberg H.B. TMPRSS2 and TMPRSS4 promote SARS-CoV-2 infection of human small intestinal enterocytes Sci. Immunol. 5 2020 eabc3582 32404436 31 Jang K.K. Kaczmarek M.E. Dallari S. Chen Y.-H. Tada T. Axelrad J. Landau N.R. Stapleford K.A. Cadwell K. Variable susceptibility of intestinal organoid-derived monolayers to SARS-CoV-2 infection Preprint at bioRxiv 2022 10.1101/2021.07.16.452680 32 Hoffmann M. Kleine-Weber H. Schroeder S. Krüger N. Herrler T. Erichsen S. Schiergens T.S. Herrler G. Wu N.-H. Nitsche A. SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor Cell 181 2020 271 280.e8 32142651 33 Saif L.J. Jung K. Comparative pathogenesis of bovine and porcine respiratory coronaviruses in the animal host species and SARS-CoV-2 in humans J. Clin. Microbiol. 58 2020 e01355-20 32522830 34 Nalbandian A. Sehgal K. Gupta A. Madhavan M.V. McGroder C. Stevens J.S. Cook J.R. Nordvig A.S. Shalev D. Sehrawat T.S. Post-acute COVID-19 syndrome Nat. Med. 27 2021 601 615 33753937 35 Jagannathan P. Andrews J.R. Bonilla H. Hedlin H. Jacobson K.B. Balasubramanian V. Purington N. Kamble S. de Vries C.R. Quintero O. Peginterferon Lambda-1a for treatment of outpatients with uncomplicated COVID-19: a randomized placebo-controlled trial Nat. Commun. 12 2021 1967 33785743 36 Lim M.Y. Hong S. Kim B.-M. Ahn Y. Kim H.-J. Nam Y.-D. Changes in microbiome and metabolomic profiles of fecal samples stored with stabilizing solution at room temperature: a pilot study Sci. Rep. 10 2020 1789 32019987 37 Coryell M.P. Iakiviak M. Pereira N. Murugkar P.P. Rippe J. Williams D.B. Heald-Sargent T. Sanchez-Pinto L.N. Chavez J. Hastie J.L. A method for detection of SARS-CoV-2 RNA in healthy human stool: a validation study Lancet Microbe 2 2021 e259 e266 33821247 38 Kuypers J. Jerome K.R. Applications of digital PCR for clinical microbiology J. Clin. Microbiol. 55 2017 1621 1628 28298452 39 Verma R. Kim E. Martínez-Colón G.J. Jagannathan P. Rustagi A. Parsonnet J. Bonilla H. Khosla C. Holubar M. Subramanian A. SARS-CoV-2 subgenomic RNA kinetics in longitudinal clinical samples Open Forum Infect. Dis. 8 2021 ofab310 34295944 40 Saif L.J. Bovine respiratory coronavirus Vet. Clin. North Am. Food Anim. Pract. 26 2010 349 364 20619189 41 Alexandersen, S., Chamings, A., and Bhatta, T.R. SARS-CoV-2 genomic and subgenomic RNAs in diagnostic samples are not an indicator of active replication. Nat. Commun. 11 6059 42 McClary-Gutierrez J.S. Mattioli M.C. Marcenac P. Silverman A.I. Boehm A.B. Bibby K. Balliet M. de Los Reyes F.L. 3rd Gerrity D. Griffith J.F. SARS-CoV-2 wastewater surveillance for public health action Emerg. Infect. Dis. 27 2021 1 8 43 Polo D. Quintela-Baluja M. Corbishley A. Jones D.L. Singer A.C. Graham D.W. Romalde J.L. Making waves: wastewater-based epidemiology for COVID-19 - approaches and challenges for surveillance and prediction Water Res. 186 2020 116404 32942178 44 Medema G. Been F. Heijnen L. Petterson S. Implementation of environmental surveillance for SARS-CoV-2 virus to support public health decisions: opportunities and challenges Curr. Opin. Environ. Sci. Health 17 2020 49 71 33024908 45 Mallapaty S. How sewage could reveal true scale of coronavirus outbreak Nature 580 2020 176 177 46 CDC National Wastewater Surveillance System (NWSS) 2022 Centers for Disease Control and Prevention https://www.cdc.gov/healthywater/surveillance/wastewater-surveillance/wastewater-surveillance.html?CDC_AA_refVal=https%3A%2F%2Fwww.cdc.gov%2Fcoronavirus%2F2019-ncov%2Fcases-updates%2Fwastewater-surveillance.html 47 Mowat A.M. Agace W.W. Regional specialization within the intestinal immune system Nat. Rev. Immunol. 14 2014 667 685 25234148 48 Osikomaiya B. Erinoso O. Wright K.O. Odusola A.O. Thomas B. Adeyemi O. Bowale A. Adejumo O. Falana A. Abdus-Salam I. Long COVID”: persistent COVID-19 symptoms in survivors managed in Lagos State, Nigeria BMC Infect. Dis. 21 2021 304 33765941 49 Taquet M. Dercon Q. Luciano S. Geddes J.R. Husain M. Harrison P.J. Incidence, co-occurrence, and evolution of long-COVID features: a 6-month retrospective cohort study of 273,618 survivors of COVID-19 PLoS Med. 18 2021 e1003773 34582441 50 Ramakrishnan R.K. Kashour T. Hamid Q. Halwani R. Tleyjeh I.M. Unraveling the mystery surrounding post-acute sequelae of COVID-19 Front. Immunol. 12 2021 686029 34276671 51 Weng J. Li Y. Li J. Shen L. Zhu L. Liang Y. Lin X. Jiao N. Cheng S. Huang Y. Gastrointestinal sequelae 90 days after discharge for COVID-19 Lancet Gastroenterol. Hepatol. 6 2021 344 346 33711290 52 Carfì A. Bernabei R. Landi F. Gemelli Against COVID-19 post-acute care study group persistent symptoms in patients after acute COVID-19 JAMA 324 2020 603 605 32644129 53 Kayaaslan B. Eser F. Kalem A.K. Kaya G. Kaplan B. Kacar D. Hasanoglu I. Coskun B. Guner R. Post-COVID syndrome: a single-center questionnaire study on 1007 participants recovered from COVID-19 J. Med. Virol. 93 2021 6566 6574 34255355 54 Al-Aly Z. Xie Y. Bowe B. High-dimensional characterization of post-acute sequelae of COVID-19 Nature 594 2021 259 264 33887749 55 Bustin S.A. Benes V. Garson J.A. Hellemans J. Huggett J. Kubista M. Mueller R. Nolan T. Pfaffl M.W. Shipley G.L. The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments Clin. Chem. 55 2009 611 622 19246619 56 dMIQE GroupHuggett J.F. The digital MIQE guidelines update: minimum information for publication of quantitative digital PCR experiments for 2020 Clin. Chem. 66 2020 1012 1029 32746458 57 Corman V.M. Landt O. Kaiser M. Molenkamp R. Meijer A. Chu D.K. Bleicker T. Brünink S. Schneider J. Schmidt M.L. Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR Euro Surveill. 25 2020 2000045 31992387 58 CDC 2019-Novel Coronavirus (2019-nCoV) Real-Time RT-PCR Diagnostic Panel (2020). (Division of Viral Diseases, Centers for Disease Control and Prevention). 59 Lu X. Wang L. Sakthivel S.K. Whitaker B. Murray J. Kamili S. Lynch B. Malapati L. Burke S.A. Harcourt J. US CDC real-time reverse transcription PCR panel for detection of severe acute respiratory syndrome coronavirus 2 Emerg. Infect. Dis. 26 2020 1654 1665 32396505 60 Droplet Digital PCR Applications Guide https://www.bio-rad.com/webroot/web/pdf/lsr/literature/Bulletin_6407.pdf. 61 Pierson-Perry J.F. Vaks J.E. Vore T.E.K. Durham A.P. Fischer C. Gutenbrunner C. Hillyard D. Kondratovich M.V. Ladwig P. Middleberg R.A. Evaluation of Detection Capability for Clinical Laboratory Measurement Procedures; Approved Guideline 2012 Clinical Laboratory Standards Institute 62 Austin P.C. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research Commun. Stat. - Simulation Comput. 38 2009 1228 1234 63 Cohen J. The Effect Size Index: D. Statistical Power Analysis for the Behavioral Sciences 1988 Routledge Academic 64 Liang K.-Y. Zeger S.L. Longitudinal data analysis using generalized linear models Biometrika 73 1986 13 22 65 Kernan W.N. Viscoli C.M. Makuch R.W. Brass L.M. Horwitz R.I. Stratified randomization for clinical trials J. Clin. Epidemiol. 52 1999 19 26 9973070
PMC009xxxxxx/PMC9005384.txt	==== Front Eur J Intern Med Eur J Intern Med European Journal of Internal Medicine 0953-6205 1879-0828 European Federation of Internal Medicine. Published by Elsevier B.V. S0953-6205(22)00144-3 10.1016/j.ejim.2022.04.009 Clinical Insights Anticoagulation as secondary prevention of massive lung thromboses in hospitalized patients with COVID-19 Sofia Rosaria a Carbone Mattias a Landoni Giovanni ab⁎ Zangrillo Alberto ab Dagna Lorenzo bc⁎ a Department of Anesthesia and Intensive Care, IRCCS San Raffaele Scientific Institute, Milan, Italy b Unit of Immunology, Rheumatology, Allergy and Rare Diseases, IRCCS San Raffaele Scientific Institute, Milan, Italy c Faculty of Medicine, Vita-Salute San Raffaele University, Milan, Italy ⁎ Corresponding authors at: IRCCS Ospedale San Raffaele, Via Olgettina 60, 20132 Milano, Italy. 13 4 2022 13 4 2022 11 2 2022 8 4 2022 11 4 2022 © 2022 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Keywords COVID-19 SARS-CoV-2 Thromboinflammatory syndrome MicroCLOTS Anticoagulants Antithrombotic therapy Abbreviations ARDS, acute respiratory distress syndrome IL-6, interleukin-6 inhibitor MicroCLOTS, microvascular COVID-19 lung vessels obstructive thromboinflammatory syndrome MOF, multiple organ failure PVTs, pulmonary venous thromboses RCT, randomized clinical trials SARS-CoV-2, severe acute respiratory syndrome coronavirus 2 TMA, thrombotic microangiopathy ==== Body pmcText Since February 2020, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection has been responsible for one of the major pandemic we have experienced in the last 100 years [1]. The most severe clinical presentation of COVID-19 is through acute respiratory distress syndrome (ARDS) classifying it as a respiratory illness. The presence of an underlying hypercoagulable state, associated to venous thrombotic events with a prevalence of 30%, was extensively reported worldwide in COVID-19 patients [2], [3], [4]. Moreover, several autopsy studies reported microvascular thromboses in lungs and most organs of deceased patients [5,6]. SARS-CoV-2 infection is thought to be responsible for a specific mechanism of thrombo-inflammation, called the “immunothrombosis model” [7]. The viral-mediated direct cellular damage and the immune response result in the release of proinflammatory cytokines. Cytokines determine the subsequent activation and dysfunction of the endothelium, which contributes to the establishment of an immuno-mediated hypercoagulable state [8]. The pro-thrombotic state is a condition that may precede morbidity and mortality. According to Ciceri et al. [9], this atypical ARDS working hypothesis was named microvascular COVID-19 lung vessels obstructive thromboinflammatory syndrome (MicroCLOTS). This syndrome is thought to be caused by alveolar endothelial damage, followed by progressive endothelial pulmonary involvement. Subsequently, the inflammation and the thrombotic milieu also affect the microcirculation of other organs, eventually leading to multiple organ failure (MOF) [10] and, in certain circumstances, also to a disseminated intravascular coagulation-like state. Among COVID-19 patients with normal angiographic studies, thromboinflammatory markers (D-dimer, C-reactive protein, ferritin, and interleukin 6) are often elevated [11], [12], [13] suggesting the presence of microvascular damage. Nailfold videocapillaroscopy performed on COVID-19 patients showed microvascular abnormalities, resembling acute and post-acute microvascular damage [14]. Furthermore, the COVID-19 radiological pattern is characterized by a unique distribution of pulmonary venous thromboses (PVTs) which overlaps with lung inflamed areas, confirming that in situ thromboses are not embolisms [2]. Several international guidelines recommend heparin-based anticoagulation therapy in all COVID-19 hospitalized patients [15], [16], [17], [18], [19], [20]. This recommendation is based on large observational studies [21,22] which support the efficacy of anticoagulation therapy, while randomized clinical trials (RCTs) comparing the use of heparin versus placebo are lacking. Advantages of heparin include its antithrombotic, anti-inflammatory, and likely antiviral effects [23]. Moreover, heparin has fewer pharmacologic interactions with experimental drugs used in COVID-19 patients, alike the other oral anticoagulants. Despite all of these recommendations, the proper dosage of anticoagulant therapy (prophylactic vs full dose) and the exact time to start anticoagulants remain uncertain [24]. Large RCTs evaluated different anticoagulation strategies in critically ill (ATTAC-ACTIV-4a-REMAP-CAP, HEP-COVID) [25,26] and noncritically ill (ATTAC-ACTIV-4a-REMAP-CAP, RAPID, X-COVID, BEMICOP, HEP-COVID) [26], [27], [28], [29], [30] COVID-19 patients. According to these results, full dose anticoagulation (therapeutic dose) among non-critically ill patients may increase the probability of survival free of organ support, [27] the probability of 28-day survival, [28,30] and it may reduce the probability to developed venous thromboembolism (VTE) [29] with respect to prophylactic dose anticoagulation. However, these findings were not confirmed in patients treated in intensive care units (ICU) [25,26]. Although these RCTs did not include homogeneous populations and the mortality reduction was not confirmed in all studies, it is possible to hypothesize that the efficacy of the anticoagulation strategy may depend on the initiation time of the therapy with respect to the disease course. If COVID-19 MicroCLOTS are similar to the immunothrombosis model, they are probably resistant to classical anticoagulants drugs. In this case, heparin may stop the progression of the coagulation cascade avoiding the increase in thrombi size, but is not able to dissolve clots. As a consequence, it may be reasonable to suggest that the rationale for the use of heparin would not be primary prevention, but secondary prevention and avoidance of thrombi progression and development of multisystemic thrombotic complications. Within the context of mild-to-moderate respiratory illness, hospitalized SARS-CoV-2 infected patients may benefit from full-dose anticoagulation as secondary prevention. On the other hand, critically ill COVID-19 patients have probably already developed extensive lung thrombi. In this case, full-dose anticoagulation may not be able to reverse the established disease process. For these reasons, routine full-dose anticoagulation among ICU critically ill patients while not avoiding thrombotic complications can increase bleeding risk. Thus, anticoagulation therapy for critically ill COVID-19 patients should probably follow the same recommendation that are in place for critically ill non-COVID-19 patients. Even if it reasonable to think that COVID-19 outpatients can benefit from (low dose) anticoagulants, a recent RCT showed no difference in clinical outcomes in patients treated with aspirin, apixaban, or placebo [31]. This might be attributed to the relatively low sample size of the study and/or to the use of drugs different from heparin. An observational large study also suggested that patients on chronic anticoagulants do not have reduced mortality if they develop COVID-19 [32]. As previously highlighted, Sars-CoV-2 exhibits a bidirectional crosstalk between inflammation and thrombosis, or immunothrombosis, and this unique mechanism of inducing coagulopathy paves the way to therapies including antithrombin supplementation, recombinant thrombomodulin, and multiple anti-inflammatory agents. Therefore, monoclonal antibodies targeting pro-inflammatory mediators have been proposed for the treatment of COVID-19 induced microvasculature injury and endothelial damage leading to thrombotic microangiopathy (TMA) [7,8]. Tocilizumab, an interleukin-6 inhibitor (IL-6), may reduce endothelial inflammation, microvascular thrombosis, and mortality [[33], [34], [35]]. Anakinra, an antagonist of interleukin 1 receptor, may dampen systemic inflammation, and reduce mortality [36] in COVID-19 patients, especially when administered early after hospitalization in moderate-to-severe patients outside the ICU. Future studies should investigate the concomitant use of therapeutic dose anticoagulation with anti-inflammatory drugs to prevent the development of critical illness and immunothrombosis. Neutrophils extracellular traps (NETs) play a direct role in the immune-thrombotic process in COVID-19. Some experimental drugs, targeting NET formation, may limit endothelial damage and improve the prognosis [37]. Complement activation, secondary to endothelial injury, suggests the rational use of monoclonal antibodies against C5 and C3, such as Eculizumab and Ravulizumab (ClinicalTrials.gov Identifier: NCT04570397, NCT04288713, NCT04390464), for the treatment of COVID-19 associated thrombotic microangiopathy [38,39]. Summarizing all previous considerations, the hypercoagulable state associated with COVID-19 may be managed firstly by inhibiting the pro-inflammatory state and secondly by establishing anticoagulation at proper dosage, according to the disease course, to avoid the development or worsening of thrombotic complications. In conclusion, our reasoning, which is supported by initial evidence, suggests that full anticoagulation maybe considered in non-ICU patients with COVID-19 at high risk of thrombosis progression and at low risk of bleeding. Other patients (eg ICU patients) might be routinely treated with prophylactic anticoagulants if not otherwise indicated. Further RCTs in homogeneous populations are needed to confirm these observations and to inform guidelines.Fig. 1 Anticoagulation therapy in COVID-19 according to the disease severity. Fig. 1: CRediT authorship contribution statement RS: study conception and design, data interpretation, manuscript drafting. MC: data interpretation, manuscript drafting. GL: study conception and design, critical review of the manuscript. LD: study conception and design, manuscript drafting. AZ: study conception and design, critical review of the manuscript. All Authors read and approved the final version of the manuscript. Declaration of Competing Interest The authors report no conflicts of interest. Acknowledgement Thanks to all the peer reviewers and editors for their opinions and suggestions. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. ==== Refs References 1 World Health Organization Coronavirus Disease 2019 (COVID-19). https://www.who.int/docs/default-source/coronaviruse/situation-reports/20200311-sitrep-51-covid-19.pdf?sfvrsn=1ba62e57_10. 2 De Cobelli F Palumbo D Ciceri F Pulmonary Vascular Thrombosis in COVID-19 Pneumonia J Cardiothorac Vasc Anesth January 1, 2021 10.1053/j.jvca.2021.01.011 Published online 3 Roncon L Zuin M Barco S Incidence of acute pulmonary embolism in COVID-19 patients: Systematic review and meta-analysis Eur J Intern Med 82 2020 29 37 10.1016/j.ejim.2020.09.006 32958372 4 Reichert G Bunel V Dreyfuss D Saker L Khalil A Mal H. Prevalence of proximal deep vein thrombosis in hospitalized COVID-19 patients Eur J Intern Med 89 2021 118 120 10.1016/j.ejim.2021.03.034 33875336 5 Carsana L Sonzogni A Nasr A Pulmonary post-mortem findings in a series of COVID-19 cases from northern Italy: a two-centre descriptive study Lancet Infect Dis 20 10 2020 1135 1140 10.1016/S1473-3099(20)30434-5 32526193 6 Wichmann D Sperhake JP Lütgehetmann M Autopsy Findings and Venous Thromboembolism in Patients With COVID-19 Ann Intern Med 173 4 2020 268 277 10.7326/M20-2003 32374815 7 Abou-Ismail MY Diamond A Kapoor S Arafah Y Nayak L. The hypercoagulable state in COVID-19: Incidence, pathophysiology, and management Thromb Res 194 2020 101 115 10.1016/j.thromres.2020.06.029 32788101 8 Bonaventura A Vecchié A Dagna L Endothelial dysfunction and immunothrombosis as key pathogenic mechanisms in COVID-19 Nat Rev Immunol 21 5 2021 319 329 10.1038/s41577-021-00536-9 33824483 9 Ciceri F Beretta L Scandroglio AM Microvascular COVID-19 lung vessels obstructive thromboinflammatory syndrome (MicroCLOTS): an atypical acute respiratory distress syndrome working hypothesis Crit Care Resusc J Australas Acad Crit Care Med 22 2 2020 95 97 10 Renzi S Landoni G Zangrillo A Ciceri F. MicroCLOTS pathophysiology in COVID 19 Korean J Intern Med September 9, 2020 Published onlineAccessed September 22, 2021 https://www.kjim.org/journal/view.php?doi=10.3904/kjim.2020.336 11 Mahat RK Panda S Rathore V Swain S Yadav L Sah SP. The dynamics of inflammatory markers in coronavirus disease-2019 (COVID-19) patients: A systematic review and meta-analysis Clin Epidemiol Glob Health 11 2021 100727 10.1016/j.cegh.2021.100727 12 Zhou F, Yu T, Du R, et al. Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. The Lancet. 2020;395(10229):1054-1062. doi:10.1016/S0140-6736(20)30566-3. 13 Tang N Li D Wang X Sun Z. Abnormal coagulation parameters are associated with poor prognosis in patients with novel coronavirus pneumonia J Thromb Haemost JTH 18 4 2020 844 847 10.1111/jth.14768 32073213 14 Natalello G De Luca G Gigante L Nailfold capillaroscopy findings in patients with coronavirus disease 2019: Broadening the spectrum of COVID-19 microvascular involvement Microvasc Res 133 2021 104071 10.1016/j.mvr.2020.104071 15 Cuker A Tseng EK Nieuwlaat R American Society of Hematology living guidelines on the use of anticoagulation for thromboprophylaxis in patients with COVID-19: May 2021 update on the use of intermediate intensity anticoagulation in critically ill patients Blood Adv 2021 10.1182/bloodadvances.2021005493 (bloodadvances.2021005493) 16 Spyropoulos AC Levy JH Ageno W Scientific and Standardization Committee communication: Clinical guidance on the diagnosis, prevention, and treatment of venous thromboembolism in hospitalized patients with COVID-19 J Thromb Haemost 18 8 2020 1859 1865 10.1111/jth.14929 32459046 17 Barnes GD Burnett A Allen A Thromboembolism and anticoagulant therapy during the COVID-19 pandemic: interim clinical guidance from the anticoagulation forum J Thromb Thrombolysis 50 1 2020 72 81 10.1007/s11239-020-02138-z 32440883 18 Moores LK Tritschler T Brosnahan S Prevention, Diagnosis, and Treatment of VTE in Patients With Coronavirus Disease 2019 Chest 158 3 2020 1143 1163 10.1016/j.chest.2020.05.559 32502594 19 Bikdeli B Madhavan MV Jimenez D COVID-19 and Thrombotic or Thromboembolic Disease: Implications for Prevention, Antithrombotic Therapy, and Follow-Up J Am Coll Cardiol 75 23 2020 2950 2973 10.1016/j.jacc.2020.04.031 32311448 20 Information on COVID-19 Treatment, Prevention and Research COVID-19 Treatment Guidelines October 5, 2021 Accessed https://www.covid19treatmentguidelines.nih.gov/ 21 Paranjpe I Fuster V Lala A Association of Treatment Dose Anticoagulation With In-Hospital Survival Among Hospitalized Patients With COVID-19 J Am Coll Cardiol 76 1 2020 122 124 10.1016/j.jacc.2020.05.001 32387623 22 Rentsch CT Beckman JA Tomlinson L Early initiation of prophylactic anticoagulation for prevention of coronavirus disease 2019 mortality in patients admitted to hospital in the United States: cohort study BMJ 372 2021 n311 10.1136/bmj.n311 33574135 23 Buijsers B Yanginlar C Maciej-Hulme ML de Mast Q der Vlag J van Beneficial non-anticoagulant mechanisms underlying heparin treatment of COVID-19 patients EBioMedicine 2020 59 10.1016/j.ebiom.2020.102969 24 Giannis D Douketis JD Spyropoulos AC. Anticoagulant therapy for COVID-19: What we have learned and what are the unanswered questions? Eur J Intern Med 96 2022 13 16 10.1016/j.ejim.2021.11.003 34799234 25 REMAP-CAP Investigators, ACTIV-4a Investigators, ATTACC Investigators Therapeutic Anticoagulation with Heparin in Critically Ill Patients with Covid-19 N Engl J Med. 385 9 2021 777 789 10.1056/NEJMoa2103417 34351722 26 Spyropoulos AC Goldin M Giannis D Efficacy and Safety of Therapeutic-Dose Heparin vs Standard Prophylactic or Intermediate-Dose Heparins for Thromboprophylaxis in High-risk Hospitalized Patients With COVID-19: The HEP-COVID Randomized Clinical Trial JAMA Intern Med 181 12 2021 1612 1620 10.1001/jamainternmed.2021.6203 34617959 27 ATTACC Investigators, ACTIV-4a Investigators, REMAP-CAP Investigators Therapeutic Anticoagulation with Heparin in Noncritically Ill Patients with Covid-19 N Engl J Med 385 9 2021 790 802 10.1056/NEJMoa2105911 34351721 28 Sholzberg M Tang GH Rahhal H Effectiveness of therapeutic heparin versus prophylactic heparin on death, mechanical ventilation, or intensive care unit admission in moderately ill patients with covid-19 admitted to hospital: RAPID randomised clinical trial BMJ 375 2021 n2400 10.1136/bmj.n2400 34649864 29 Morici N, Podda G, Birocchi S, et al. Enoxaparin for thromboprophylaxis in hospitalized COVID-19 patients: The X-COVID-19 Randomized Trial. Eur J Clin Invest. n/a(n/a):e13735. doi:10.1111/eci.13735. 30 Marcos-Jubilar M Carmona-Torre F Vidal R Therapeutic versus Prophylactic Bemiparin in Hospitalized Patients with Nonsevere COVID-19 Pneumonia (BEMICOP Study): An Open-Label, Multicenter, Randomized, Controlled Trial Thromb Haemost 12 2021 10.1055/a-1667-7534 Published online October 31 Connors JM Brooks MM Sciurba FC Effect of Antithrombotic Therapy on Clinical Outcomes in Outpatients With Clinically Stable Symptomatic COVID-19: The ACTIV-4B Randomized Clinical Trial JAMA 326 17 2021 1703 1712 10.1001/jama.2021.17272 34633405 32 Montorfano M, Leoni O, Andreassi A, et al. Chronic anticoagulant treatment and risk of mortality in SARS-Cov2 patients: a large population-based study. Minerva Med. Published online February 2022. doi:10.23736/S0026-4806.22.07797-7. 33 Levi M. Tocilizumab in severe COVID-19: A promise fulfilled Eur J Intern Med 95 2022 38 39 10.1016/j.ejim.2021.11.015 34836747 34 Campochiaro C Tomelleri A Matucci-Cerinic M Dagna L. One year later: The case of tocilizumab in COVID-19 Eur J Intern Med 95 2022 5 6 10.1016/j.ejim.2021.10.024 34711474 35 Belletti A, Campochiaro C, Marmiere M, et al. Efficacy and safety of IL-6 inhibitors in patients with COVID-19 pneumonia: a systematic review and meta-analysis of multicentre, randomized trials. Ann Intensive Care. 2021;11(1):152. doi:10.1186/s13613-021-00941-2. 36 Cavalli G De Luca G Campochiaro C Interleukin-1 blockade with high-dose anakinra in patients with COVID-19, acute respiratory distress syndrome, and hyperinflammation: a retrospective cohort study Lancet Rheumatol 2 6 2020 e325 e331 10.1016/S2665-9913(20)30127-2 32501454 37 Barnes BJ Adrover JM Baxter-Stoltzfus A Targeting potential drivers of COVID-19: Neutrophil extracellular trapsNeutrophil extracellular traps in COVID-19 J Exp Med 217 6 2020 e20200652 10.1084/jem.20200652 38 Annane D Heming N Grimaldi-Bensouda L Eculizumab as an emergency treatment for adult patients with severe COVID-19 in the intensive care unit: A proof-of-concept study eClinicalMedicine 28 2020 10.1016/j.eclinm.2020.100590 39 Mastellos DC Pires da Silva BGP Fonseca BAL Complement C3 vs C5 inhibition in severe COVID-19: Early clinical findings reveal differential biological efficacy Clin Immunol Orlando Fla 220 2020 108598 10.1016/j.clim.2020.108598
PMC009xxxxxx/PMC9005438.txt	==== Front J Surg Res J Surg Res The Journal of Surgical Research 0022-4804 1095-8673 Elsevier Inc. S0022-4804(22)00186-X 10.1016/j.jss.2022.04.004 Article The Impact of COVID-19 Pandemic Upon Non-elective Admissions and Surgery at a Safety-Net Hospital A Retrospective Cohort Study Boyev Artem DO Sanjeevi Srinivas MD Estrada Martha M. MD Ko Tien C. MD Wray Curtis J. MD, MS ∗ Department of Surgery, University of Texas Health Science Center at Houston, McGovern Medical School, Houston, Texas ∗ Corresponding author. Department of Surgery, University of Texas Medical School at Houston, 5656 Kelley Street, Suite 30S62008, Houston, TX 77026. Tel.: +1 713 566 5095; fax: +1 713 566 4583. 13 4 2022 10 2022 13 4 2022 278 376385 14 10 2021 13 1 2022 7 4 2022 © 2022 Elsevier Inc. All rights reserved. 2022 Elsevier Inc. Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Introduction In response to the COVID-19 pandemic, hospitals reported decreased admissions for acute surgical diagnoses, but scant data was available to quantify the decrease and its consequences. The objective of this study was to examine the incidence of acute care surgery encounters before and during the COVID-19 pandemic. Materials and Methods A retrospective cohort study was performed at a single, urban, United States safety-net hospital. Emergency room encounters, admissions, non-elective surgical procedures, patient acuity, and surgical complications were compared before and after the start of the COVID-19 pandemic. The primary outcome of the study was the incidence rate (IR) and incidence rate ratios (IRR) for surgical admissions, laparoscopic appendectomy, and urgent laparoscopic cholecystectomy. Results During the COVID-19 (exposure) time period, the number of nonelective procedures was 143 (IR 4.76) which was significantly lower than the control periods (n = 431, IR 7.2), P < 0.001. During the COVID-19 exposure period, there were significantly fewer urgent cholecystectomies performed (1.37 per day versus 2.80-2.93 per day, P < 0.001). There was a trend toward fewer appendectomies performed, but not significant. There was little difference in patient acuity between the exposure and control periods. A higher proportion of patients that underwent urgent cholecystectomy during the COVID time period had been seen in the ED in the prior 30 d (22% versus 5.6%). Conclusions Surgical volume significantly decreased during the COVID-19 pandemic. Management of acute cholecystitis may require re-evaluation as nonsurgical management appears to increase repeat presentations. Keywords Acute care surgery Appendectomy Cholecystectomy COVID-19 Emergency surgery ==== Body pmcIntroduction In 2019, SARS-CoV-2 (Coronavirus) emerged as worldwide viral pathogen. In early 2020 international reports of viral community spread resulted in the COVID-19 pandemic.1 The first case in the United States was diagnosed January 20, 2020, and exponential growth was observed in the following weeks.2 As cases, hospital admissions and deaths began to increase in March 2020, Texas declared a state of emergency on March 13, 2020. Additionally, a restriction was placed on ‘elective’ surgical procedures to conserve personal protective equipment (PPE), intensive care unit beds and other critical hospital resources. Operative volume decreased substantially as low- and intermediate-acuity cases, screening procedures, orthopedic procedures, and even some cancer procedures were postponed.3 , 4 Deferring hospitalization and non-urgent surgeries became a major priority to relieve hospital capacity issues. Surgical management for life-threatening conditions was not restricted. Concurrently many hospitals and insurance companies reported decreased admissions for traditional emergency room diagnoses during the COVID pandemic-related shutdown (e.g., atrial fibrillation, epilepsy/seizure, gastrointestinal bleed, transient ischemic attacks).5 , 6 In parallel, the incidence of acute-care surgical diagnoses, such as abdominal pain, appendicitis and acute cholecystitis, also appeared to decline.7 , 8 The health care challenge for un- or under-insured patients may become more significant. As the concurrent economic shutdown led to increased unemployment, the absolute number of those lacking health insurance is expected to rise.9 This may disproportionately impact safety-net hospitals (SNH) that, by mission or mandate, provide care to a substantial share of vulnerable patients regardless of their ability to pay.10 Prior studies have shown, SNH patients are at-risk for worse outcomes secondary to surgical complications.11 Now that the COVID-19 Delta and Omicron variants are contributing to increases in COVID-related hospitalizations and deaths, the experience of the early-2020 pandemic remain acutely relevant. We sought to examine the incidence of acute care surgery encounters and outcomes in an SNH. Our hypotheses were: (1) fewer patients sought care at the ER during the exposure period and (2) of those who did present to the ER, their acuity level was higher than the control time periods. Materials and Methods Study design and participants Institutional Review Board approval was obtained for a retrospective observational cohort study of emergency room encounters, admissions, and surgical procedures in a safety-net hospital: Lyndon B. Johnson Hospital (LBJ), Houston, Texas. LBJ is a licensed 207 bed acute care safety-net hospital for the Harris Health System affiliated with the McGovern Medical School at the University of Texas Health Sciences Center at Houston. LBJ is the busiest Level III trauma center in Texas, with more than 70,000 emergency patient visits each year.12 The majority of patients at this safety-net hospital are uninsured (54%) with a demographic distribution of 54% Latino, 25% African American, 12% Asian/other and 9% Caucasian.13 Patient encounters in the emergency room (ER) were prospectively identified between March 14, 2020, through April 13, 2020. The COVID-19 exposure time period was defined as 30 d following the March 14, 2020 declaration of Texas State of emergency. Two equivalent control periods were chosen for comparison: March 14, 2018 to April 13, 2018 and March 14, 2019 to April 13, 2019. All emergency room encounters were queried and exported from the electronic medical record (EPIC; Verona, WI), including presenting vital signs (temperature, respiratory rate, pulse, and blood pressure) white blood cell count (WBC), serum bicarbonate and serum creatinine. Emergency Severity Index (ESI) acuity was also recorded for each encounter. ESI is five-level triage algorithm that estimates patients into five groups from one (most serious) to five (least serious).14 All general surgery ER encounters and admissions during the same time periods were recorded. The primary surgical diagnoses of interest for this study were acute appendicitis and acute cholecystitis. Surgical operations were classified as either non-elective (urgent and emergent) or elective. The main surgical procedures queried were open or laparoscopic appendectomy and open or laparoscopic cholecystectomy. The study was approved by the McGovern Medical School Committee for the Protection of Human Subjects (protocol HSC-MS-20-0578). Waiver for informed consent was granted due to the study design and lack of feasibility. Outcomes The primary outcome of the study was the incidence rate (IR) and incidence rate ratios (IRR) for surgical admissions, specifically for acute appendicitis and acute cholecystitis. Secondary outcomes included ER disposition (admission or discharge), time to operative intervention (defined as time from ER admission to time in the operating room), length of stay (LOS), post-operative complications, readmissions, and unplanned ER visits. Findings of perforated versus simple appendicitis and number of cases converted to an open procedure were secondary outcomes for the appendicitis and cholecystitis groups, respectively. As reference data, ER encounters in the 30 d prior and 30 d after each study period were also recorded. These additional date intervals were chosen to determine if a surgical patient had (1) previously sought care in the emergency room within 30 d prior to surgical admission or (2) presented to the ER within 30 d after surgical discharge. Statistical analysis Means and standard deviations or median and interquartile ranges (IQRs) were reported for normally or non-normally distributed continuous variables. Categorical variables were presented as counts and percentages. Standard two-tailed t-tests were used to compare continuous variables and chi-square to compare categorical data associations. Kruskal–Wallis test was performed to compare nonparametric continuous variables. IRR comparing the COVID-19 exposure period to each control period were calculated using Poisson regression to model the number of events per day. Stata 16 (College Station, TX) was used for all statistical analyses. Results During the COVID-19 (exposure) time period, there were a total of 5029 ER encounters (Table 1 ). This was significantly lower than both control periods (n = 7585 and n = 8054). A higher proportion of male patients presented to the ER during COVID (52.7% versus 47.4%, P < 0.001). The ESI Acuity status (ESI 1 or ESI 2) was also slightly higher during COVID (28.3% versus 26.5%). The rate of hospital admission from the ER was higher in the COVID cohort (16.2% versus 14.7%; P = 0.02) and the total time spent in the ER per encounter was significantly lower during the study period (3.5 versus 5.1 h, P < 0.001). However, once a patient was admitted to the hospital the length of stay was significantly longer during COVID (87.2 versus 72.3 h).Table 1 Hospital encounters during COVID-19 and control time periods. Variable COVID-19 (n = 5029) Control (n = 15,639) P-value Age, y (SD) 43.6 (17.1) 43.1 (17.8) 0.14 Gender (%) <0.001 Male 2648 (52.7%) 7419 (47.4%) Female 2381 (47.4%) 8220 (52.6%) ESI acuity (%) <0.001 1 22 (0.4%) 105 (0.7%) 2 1392 (27.9%) 3986 (25.8%) 3 2547 (51.0%) 8422 (54.6%) 4 955 (19.1%) 2687 (17.4%) 5 78 (1.6%) 230 (1.5%) Unknown 35 (0.7%) 211 (1.4%) Pulse, bpm (SD) 81.0 (16.0) 81.2 (16.9) 0.62 Systolic, mmHg (SD) 132.6 (20.8) 131.4 (20.7) <0.01 Diastolic, mmHg (SD) 78.9 (13.1) 78.4 (13.2) 0.04 Respiratory rate, per min (SD) 18.2 (2.7) 18.4 (3.0) <0.01 Temperature, °F (SD) 98.3 (0.6) 98.2 (1.0) <0.01 ER disposition (%) 0.02 Admission 817 (16.2) 2291 (14.7) Discharged 4211 (83.8) 13,348 (85.3) ER duration, h (IQR) 3.5 (4.3) 5.1 (4.6) <0.01 Admission LOS, h (IQR) 87.2 (96.7) 72.3 (83.7) <0.01 A comparison of patient demographics and characteristics in the time period before (Control) and during the COVID-19 pandemic. Unit labels are in column 1. During the exposure period, admissions to the General Surgery service were significantly lower (142 versus 281 admissions/30 d, IRR 0.51, P < 0.001 and 142 versus 293 admissions/30 d, IRR 0.48, P < 0.001) (Table 2 ). The total number of non-elective general surgery procedures was also significantly lower than both controls (143 versus 219 procedures/30 d, IRR 0.65, P < 0.001 and 143 versus 212 procedures/30 d, IRR 0.67, P < 0.001). The frequency of laparoscopic appendectomy was lower albeit not significantly (13 versus 22 procedures/30 d, IRR 0.59, P = 0.13 and 13 versus 23 procedures/30 d, IRR 0.67, P = 0.10). Among patients undergoing laparoscopic appendectomy, there were no differences between groups with regards to vital signs at presentation, WBC, serum creatinine or bicarbonate, length of surgical procedure (Table 3 ). There were also no differences in patients treated non-operatively. However, there was a significant difference in the time to the operating room, 11.5 h during the control time period compared to 8.0 h for the COVID-19 time period (P < 0.01). There was a trend toward shorter length of stay during the COVID-19 exposure time period (24.8 versus 46.6 h, P = 0.16). The percentage of perforated appendicitis was higher in the control periods (26.6% versus 15.3%, P < 0.01). There were no reported complications, unplanned post-operative ER visits, or readmissions during the COVID-19 exposure period, compared with two unplanned post-operative ER visits and two complications during the control period.Table 2 Incidence rates and incidence rate ratios. Variable COVID-19 Control 1 Control 2 All ER encounters (n = 20,668) 5029 7585 8054 Incidence rate (per day) 167.6 252.8 268.5 Incidence rate ratio (95% CI) 0.66 (0.64-0.69) 0.62 (0.60-0.65) P-value <0.001 <0.001 All admissions (n = 3108) 817 1156 1135 Incidence rate (per day) 27.2 38.5 37.8 Incidence rate ratio (95% CI) 0.71 (0.65-0.77) 0.72 (0.66-0.79) P-value <0.001 <0.001 Surgery admissions (n = 716) 142 281 293 Incidence rate (per day) 4.73 9.37 9.77 Incidence rate ratio (95% CI) 0.51 (0.41-0.62) 0.48 (0.39-0.59) P-value <0.001 <0.001 All non-elective add-on cases (n = 574) 143 219 212 Incidence rate (per day) 4.77 7.30 7.07 Incidence rate ratio (95% CI) 0.65 (0.53-0.81) 0.67 (0.54-0.84) P-value <0.001 <0.001 Add-on laparoscopic appendectomy (n = 58) 13 22 23 Incidence rate (per day) 0.43 0.73 0.77 Incidence rate ratio (95% CI) 0.59 (0.27-1.23) 0.57 (0.26-1.16) P-value 0.13 0.10 Add-on laparoscopic cholecystectomy (n = 213) 41 84 88 Incidence rate (per day) 1.37 2.80 2.93 Incidence rate ratio (95% CI) 0.49 (0.33-0.72) 0.47 (0.31-0.68) P-value <0.001 <0.001 A comparison of the daily incidence rate for ER encounters, all admissions, surgical admissions, non-elective add-on cases, appendectomy, and cholecystectomy. This compares the incidence before (Control) and during the COVID-19 pandemic. Table 3 Laparoscopic appendectomy. COVID-19 Patients COVID-19 Controls P-value Non-operative cases (n = 7) (n = 2) (n = 5) Age, y (SD) 33.5 (19.1) 35.4 (16.7) 0.92 Gender 0.81 Male (%) 1 (50%) 2 (40%) Female (%) 1 (50%) 3 (60%) ER heart rate, bpm (SD) 77.5 (24.8) 66.4 (6.1) 0.33 ER temperature, °F (SD) 98.0 (0.3) 98.0 (0.3) 0.94 ER systolic BP, mmHg (SD) 113.5 (7.8) 121.0 (25.6) 0.58 ER diastolic BP, mmHg (SD) 68.5 (9.2) 77.4 (16.6) 0.42 WBC, ×109/L (SD) 14.6 (0.2) 11.5 (3.5) 0.12 CO2, mEq/L (SD) 28.5 (0.7) 25.0 (1.6) 0.01 Creatinine, mg/dL (SD) 0.85 (0.2) 0.58 (0.3) 0.27 Hospital LOS, h (SD) 36.0 (16.8) 62.4 (26.4) 0.21 Perforated (%)/Simple (%) 2 (100%)/0 (0%) 1 (20%)/4 (80%) 0.053 Readmissions (%) 0 (0%) 0 (0%) NA Unplanned ER visits (%) 0 (0%) 0 (0%) NA Complications (%) 0 (0%) 0 (0%) NA Operative cases (n = 58) (n = 13) (n = 45) Age, y (SD) 33.1 (8.2) 34.5 (12.7) 0.71 Gender Male (%) 9 (69%) 28 (62%) 0.64 Female (%) 4 (31%) 17 (38%) ESI acuity 2 (%) 1 (8%) 7 (16%) 0.47 3 (%) 12 (92%) 38 (84%) ER heart rate, bpm (SD) 79.8 (8.0) 77.6 (14.4) 0.61 ER temperature, °F (SD) 98.3 (0.4) 98.3 (0.4) 0.64 ER systolic BP, mmHg (SD) 121.2 (20.1) 121.4 (15.1) 0.97 ER diastolic BP, mmHg (SD) 71.2 (13.6) 73.6 (10.9) 0.52 WBC (SD) 14.7 (5.0) 14.0 (5.0) 0.64 CO2 (SD) 26.4 (3.7) 25.6 (2.7) 0.35 Creatinine (SD) 0.9 (0.3) 0.8 (0.2) 0.07 Duration of procedure (SD) 107.4 (13.1) 109.2 (32.8) 0.84 Time to OR (SD) 8.0 (4.6) 11.5 (3.4) <0.01 Hospital LOS (SD) 24.8 (9.6) 46.6 (54.5) 0.16 Perforated (%)/Simple (%) 2 (15%)/11 (85%) 12 (27%)/33 (73%) <0.01 Readmissions (%) 0 (0%) 0 (0%) NA Unplanned ER visits (%) 0 (0%) 2 (4%) <0.01 Complications (%) 0 (0%) 2 (4%) <0.01 A comparison of appendicitis patient characteristics before and during the COVID-19 pandemic. The top part of the table includes all patients managed non-operatively, while the bottom area includes patients managed with surgery. The units are displayed in column 1. The frequency of laparoscopic cholecystectomy was significantly lower in the exposure period than the control time periods (41 versus 84 procedures/30 d, IRR 0.49 and 41 versus 88 procedures/30 d, IRR 0.47; P < 0.001) (Table 2). For those who presented with acute cholecystitis during COVID, the systolic blood pressure (124.5 versus 119.0) and diastolic blood pressure (76.1 versus 72.7) were significantly higher (P < 0.05) (Table 4 ). There was no difference in WBC, serum creatinine, bicarbonate, duration of case or length of stay between laparoscopic cholecystectomy groups. There was a trend toward a shorter time to operating room during the COVID-19 time period (26.3 versus 32.2 h, P = 0.14). There was also no difference in post-operative ER visits and readmissions. During COVID, there were no conversions to an open cholecystectomy; however, the surgical site infection (SSI) rate was higher during the COVID time period. There were no differences in the other surgical complications. There were also no differences in patients treated non-operatively.Table 4 Laparoscopic cholecystectomy. COVID-19 Patients COVID-19 Controls P-value Non-operative cases (n = 6) (n = 1) (n = 5) Age (SD) 46.0 (NA) 47.4 (8.9) 0.89 Gender 0.01 Male (%) 1 (100%) 0 (0%) Female (%) 0 (0%) 5 (100%) ER heart rate, bpm (SD) 60.0 (NA) 77.6 (17.1) 0.40 ER temperature, °F (SD) 98.0 (NA) 98.1 (0.4) 0.81 ER systolic BP, mmHg (SD) 147.0 (NA) 121.6 (14.2) 0.18 ER diastolic BP, mmHg (SD) 86.0 (NA) 75.6 (11.0) 0.44 WBC, ×109/L (SD) 4.9 (NA) 11.9 (6.2) 0.36 CO2, mEq/L (SD) 25.0 (NA) 26.4 (2.7) 0.66 Creatinine, mg/dL (SD) 0.9 (NA) 0.7 (0.1) 0.19 Hospital LOS, h (SD) 2.0 (NA) 1.8 (1.0) 0.88 Converted to open NA NA Readmissions (%) 0 (0%) 0 (0%) NA Unplanned ER visits (%) 0 (0%) 1 (20%) 0.62 Complications (%) 0 (0%) 0 (0%) NA Operative cases (n = 213) (n = 41) (n = 172) Age (SD) 36.8 (12.1) 39.6 (14.4) 0.25 Gender Male (%) 12 (29%) 33 (19.2%) 0.12 Female (%) 29 (71%) 139 (80.8%) ESI acuity 2 (%) 5 (12%) 16 (9.3%) 0.68 3 (%) 36 (8%) 154 (89.5%) > 3 (%) 0 (0%) 2 (1.2%) ER heart rate, bpm (SD) 76.5 (12.6) 78.6 (12.5) 0.34 ER temperature, °F (SD) 98.3 (0.4) 98.2 (0.4) 0.27 ER systolic BP, mmHg (SD) 124.5 (15.7) 118.7 (15.1) 0.03 ER diastolic BP, mmHg (SD) 76.1 (10.4) 72.5 (9.2) 0.03 WBC, × 109/L (SD) 11.3 (3.4) 10.7 (4.5) 0.42 CO2, mEq/L (SD) 25.6 (3.3) 25.5 (2.5) 0.85 Creatinine, mg/dL (SD) 0.7 (0.2) 0.7 (0.3) 0.80 Duration of procedure, min (SD) 136.4 (41.2) 130.9 (40.1) 0.43 Time to OR (SD) 26.3 (20.2) 32.2 (23.3) 0.14 Hospital LOS (SD) 61.2 (39.4) 62.2 (38.3) 0.87 Converted to open (%) 0 (0%) 2 (1.16%) 0.51 Readmissions (%) 3 (7%) 13 (7.6%) 0.93 Unplanned ER visits (%) 4 (8%) 26 (15.1%) 0.48 Complications (%) 7 (17%) 19 (11.1%) 0.20 SSI (%) 4 (8%) 2 (1.1%) <0.01 Bile Leak (%) 2 (5%) 4 (2.3%) 0.32 Other (%) 1 (2%) 13 (7.6%) 0.28 A comparison of cholecystectomy patient characteristics before and during the COVID-19 pandemic. The top part of the table includes all patients managed non-operatively, while the bottom area includes patients managed with surgery. Units are displayed in column one. As noted above, during COVID significantly fewer patients underwent non-elective laparoscopic cholecystectomy. Of the patients with acute cholecystitis undergoing laparoscopic cholecystectomy during COVID, 22% had been seen and discharged from the ER in the prior 30 d for the same complaint (Table 5 ). This recurrence or relapse rate was significantly lower in the control periods (5.6%) (P = 0.01). There were no recurrences or relapses observed among patients with acute appendicitis, as no patients in either group had previously been seen in the ER.Table 5 Prior ER presentations for same complaint. Patients With Prior ER Presentation Prior ER presentations No prior ER presentation P-value Acute appendicitis Control (%) 0 (0%) 45 (100%) N/A COVID-19 (%) 0 (0%) 13 (100%) Acute cholecystitis Control (%) 12 (7.0) 160 (93.0) 0.01 COVID-19 (%) 9 (22.0) 32 (78.0) This table displays the patients in each category who had previously been evaluated in the emergency room for the same complaint. Of the patients who had appendicitis, no patients had previously been evaluated in the ER in either time period. Of the patients who had cholecystitis, 6.98% of patients during the control time period compared with 22% of patients during the COVID-19 period were return patients after previous discharge from the ER for the same complaint. Discussion Emergency room encounters and surgical admissions at the LBJ safety-net hospital significantly decreased following the state of emergency declaration in Texas. The absolute number of emergency room encounters were 30%-35% lower than the two control time periods in this study. In addition, the frequency of traditional acute-care surgery admissions and non-elective operations were much lower during the COVID-19 pandemic. Palisi et al. reported a significant decrease in overall ER admissions during the COVID-19 pandemic in Italy, although they did not find a significant change in the number of surgical consultations and types of operations performed.15 Our original assumption was the incidence of acute appendicitis and acute cholecystitis would be relatively constant. However, the number of observed non-elective general surgery cases was much lower than expected. Similar to our findings, emergency surgery activity at Spanish hospitals was significantly decreased following the start of the pandemic, and the change was most pronounced for acute cholecystitis and acute appendicitis.16 , 17 Other investigators noted a significant decrease in the incidence and volume of common, urgent medical conditions.15 , 18 , 19 As recently reported, patients may be voluntarily avoiding the ER and even delaying necessary operations such as organ transplantation.20 , 21 Avoiding the ER has led some to fear patients may be staying at home with mild strokes, bowel obstructions and other serious medical conditions.15, 16, 17 , 22 Patients may have also been affected by the decreased availability of public transportation during the lockdown and certain patients at the safety-net hospital may have chosen to stay home and self-treat mildly symptomatic surgical conditions rather than face the longer wait times and infection risk associated with public transit during the pandemic.7 , 23 Interestingly, a higher percentage of male patients presented to the ER during the study period. In addition, higher acuity (ESI-1 or ESI-2) and a higher rate of hospital admission was observed in the COVID cohort. Patients spent less time in the ER which may reflect a more focused approach to triage resulting in quicker discharges of non-acute patient conditions. Yet after patients were admitted during COVID, the hospital length of stay was longer. Reasons behind this observation may be more complex as hospital inpatient efficiency significantly slowed down due to new workflows and SARS-CoV-2 testing requirements. Patient navigation, social work and case management services were also consolidated and reduced during the pandemic. The process of inpatient transfer to rehab hospitals and/or skilled nursing facilities also become more challenging. The sum of these effects likely prolonged the length of stay, which may be viewed as counterproductive during a time when all available resources were needed for COVID related care. We hypothesized that patients seeking emergent care during COVID for appendicitis and cholecystitis would present with more advanced disease. However, the available data did not demonstrate a significant difference between groups. Patients undergoing laparoscopic appendectomy did not show any difference in pre-operative indices of disease severity; furthermore, the percentage of cases with perforated appendicitis was higher during the control time periods. The shorter time from ER admission to the operating room observed during the COVID-19 time period may have contributed to the decrease in cases of perforated appendicitis. The reduced operative volume and decrease in elective cases allowed emergency appendectomies to be performed with less delay. Thus, a “fast track” approach to the management of acute appendicitis may be beneficial, although additional research is required to determine whether decreased time to operation is definitively associated with decreased incidence of perforated appendicitis. The observed trend toward shorter length of stay following laparoscopic appendectomy during the COVID exposure time period can be partly explained by the lower percentage of perforated cases. An Israeli study of similar design also found the weekly incidence of appendicitis decreased 40.7% during the pandemic.7 They also did not observe a significant difference in percentage of complicated versus uncomplicated appendicitis, duration of symptoms prior to presentation, rate of post-operative peritoneal drainage or percentage of serious post-operative complications.7 In contrast, a Turkish study noted a 73% decrease in patients who underwent appendectomy, however, they noted an increased proportion of patients with complicated appendicitis.24 A United States study in Massachusetts also observed a decrease in cases of uncomplicated appendicitis and a corresponding increase in cases of complicated appendicitis, however their sample did include pediatric patients while our sample did not.25 The difference in rates of complicated and uncomplicated appendicitis may be partially explained by differences in access to care and differing attitudes about the pandemic. LBJ hospital, the site of the current study, is part of a county-wide health system, so even the under-privileged and un-insured population can readily access surgical care. Additionally, Texas did not experience the initial COVID-19 surge as acutely as the Northeast, so the attitude toward COVID-19 has been relatively more relaxed. These factors may have contributed to patients in the current study presenting earlier compared to their counterparts in other studies. Patients that underwent laparoscopic cholecystectomy during COVID did have higher systolic and diastolic blood pressure at ER presentation that may be indirect evidence of more severe pain. However, the duration of cholecystectomy and hospital length of stay was not different. It is plausible that patients with mild symptoms may have stayed at home due to fears of COVID exposure if they sought care in the ER.26 Time to operating room also tended to be lower during the COVID-19, again likely a result of empty operating rooms. We suspected the complication rates, post-operative ER presentations, and re-admissions may differ between groups; however, this was not observed. This could be due to the relatively low sample size of surgical cases. Whereas other investigators observed an increase in in-hospital mortality for surgical admissions, this was not observed in our cohort.16 , 24 The one exception is the significantly higher percentage of SSI during the COVID time period, the reasons for which are unclear. A study at an urban, safety-net hospital in Boston also observed a 49% decrease in admissions for cholecystitis.27 However, when stratified by severity, only admissions for Tokyo I mild cholecystitis declined significantly.27 The authors concluded that not all cases of acute cholecystitis progress to more severe disease and some mild cases resolve with outpatient antibiotics or symptomatic treatment.27 Near the start of the COVID-19 timeframe, the American College of Surgeons (ACS) published guidelines for the management of surgical emergencies during the pandemic. These guidelines recommend pain management and delayed surgery for symptomatic cholelithiasis, and laparoscopic cholecystectomy for crescendo symptoms, refractory pain, or acute cholecystitis.28 This study found a significantly higher percentage of cholecystitis patients in the COVID cohort who had previously presented to the ER and were discharge for the same complaint. It seems these patients were managed according to the ACS guidelines. Patients deemed to have mild symptomatic cholelithiasis were discharged from the ED rather than scheduled for surgery. Like other studies, a percentage of these patients re-presented following failed non-operative management or the return of symptoms.29 These returning patients made up a significantly larger percentage of laparoscopic cholecystectomies during the COVID-19 exposure period, a finding that hints at increased prevalence of initial non-operative management. A secondary analysis of this group of patients revealed that all nine patients were Hispanic females, and on average, on initial ER presentation, they exhibited high-normal heart rate (88.9 bpm), alkaline phosphatase (ALP; 83.7 U/L), and white blood cell count (10.8 × 109/L). The demographic makeup of that group may be explained by cultural and culinary preferences. While it was not possible to use the current data to examine all ER patients presenting with abdominal pain and compare the above group to those who were discharged and did not return, the findings are interesting on their own. These findings certainly suggest higher suspicion for acute cholecystitis or refractory/crescendo cholelithiasis requiring surgical intervention in Hispanic females with elevated heart rate, WBC, and ALP. We propose screening criteria of HR ≥ 100 or WBC ≥ 10 or ALP ≥ 90. If this screening criteria is retrospectively applied to the above group, eight out of nine patients would have been initially admitted for cholecystectomy. However, the population served at this hospital is > 50% Hispanic, and symptomatic cholelithiasis is more common in females, so these observations may not be generalizable to hospitals serving a different demographic. Regardless, these observations raise interesting questions for further research. There was little difference in the severity of patient presentation or post-operative complications in the COVID-19 time period compared with the control time period, which suggests that non-operative management of mild right upper quadrant pain without systemic symptoms or systemic markers of inflammation may be a safe approach when necessary to conserve hospital resources. Lastly, the lack of differences in outcome may be explained in part by the makeup of resident/faculty care teams and the increased availability of operative facilities. Due to the need to minimize exposures as well as the need to quarantine infected residents/faculty, the makeup of resident teams was changed dramatically during the COVID-19 time period. Usually, four teams of one or two interns, one or two mid-level residents, and one upper-level resident are at the hospital during the day, and one night intern, one night mid-level resident, and one night upper-level resident are on at night. During the COVID-19 period, one team was on during the day and one team was on at night for a week straight. Usually, a different faculty member is on during the day and another faculty member on at night, and the faculty change daily. During the COVID-19 period, one faculty was on for a week straight during the day and a different faculty for a week straight at night. While the ratio of residents and faculty to patient remained relatively constant, it is very conceivable that this schedule allowed for greater continuity and potentially better care, helping to offset and possibly prevent some complications. Additionally, the increased availability of operative facilities led to decreased ER to OR times, also possibly offsetting and maybe preventing complications associated with increased time to definitive operative intervention. Limitations There are several limitations of this study. This data was not collected for the primary purposes of research which may lead to misclassification bias. It is also possible that patients sought care at other hospitals; however, the majority of patients seen at this safety-net hospital are uninsured and have limited healthcare options. It was not possible to objectively separate patients with symptomatic cholelithiasis from those with acute cholecystitis, the blurring of the two conditions in the data set makes it difficult to provide evidence to definitively support the ACSguidelines. While this is a large dataset of 41,465 emergency room encounters (COVID, controls and reference Groups) and nearly 300 acute care surgery admissions, the actual sample size of appendectomy and cholecystectomy patients is relatively small (approximately 100 patients per time period). Therefore, the study is vulnerable to type II error. Additionally, the study was conducted at a single, urban, safety-net-hospital with the demographics described above, so the findings may not be generalizable to other populations. Future directions This project can be expanded to analyze the rates of appendicitis and cholecystitis over the entire time period since the start of the COVID-19 pandemic. It would be interesting to examine if the lower rates of these disease processes persisted, if there were rebounds after the surges, and if the rates of these conditions returned to the pre-COVID-19 baseline. It would also be interesting to examine whether a shorter time to operative intervention, or a “fast track” appendicitis pathway can result in decreased rates of perforated appendicitis. It would also be possible to expand upon the selection of cholecystitis patients who could safely be managed conservatively without admission or cholecystectomy. It is possible to test the above proposed screening criteria and examine whether it can reduce ER readmissions with persistent or smoldering symptoms of cholecystitis. Conclusions Despite the advances in vaccination over the past 9 mo, the Delta and Omicron variants have put COVID-19 back in the national spotlight with continued surges. Cases, hospitalizations, and deaths are increasing again at an unprecedented rate. New variants and spikes are expected in the future, so the experience of the early-2020 pandemic is very relevant today. Based on the presented data, surgical volume can once again be expected to decrease. While patients with mild disease may resist seeking care, surgical patient acuity is not expected to differ significantly. Moreover, surgical complications are not expected to increase significantly. Additionally, the pandemic experience at different hospitals and in different countries seems to suggest that mild cases of appendicitis and cholecystitis are being managed conservatively without concomitant increases in more severe disease presentation. An important finding of this study is multiple patients with acute cholecystitis had been seen previously in the ER and relapsed during the COVID timeframe. This may reflect bias toward early disposition and discharge of ‘non-COVID’ related illness. During this unprecedented time of healthcare system stress and crisis, improved workflow protocols are needed to prevent multiple emergency room encounters with increased resource consumption for routine surgical diagnoses.30 Preventable emergency room visits should be avoided to reduce potential COVID-19 exposure and conserve healthcare resources. More research is necessary to delineate patients in whom mild appendicitis and cholecystitis may be managed conservatively without increasing the risk of return to the ER or progression of disease. Ongoing efforts to provide safe, surgical care in this era will continue to be a challenge for the foreseeable future as healthcare resources are shifted to combat the COVID-19 pandemic and new emerging faces of this disease. Author Contributions The idea for the study was conceived by Dr Artem Boyev, Dr Curtis Wray, and Dr Srinivas Sanjeevi. Dr Boyev, Dr Wray, Dr Sanjeevi, Dr Estrada and Dr Ko all contributed to data analysis, manuscript writing, and manuscript review and editing. Disclosure Boyev - None, Sanjeevi - None, Estrada - None, Ko - None, Wray -None. The authors report no proprietary or commercial interest in any product mentioned or concept discussed in this article. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. ==== Refs References 1 Cucinotta D. Vanelli M. WHO declares COVID-19 a pandemic Acta Biomed 91 2020 157 160 32191675 2 Holshue M.L. DeBolt C. Lindquist S. First case of 2019 novel coronavirus in the United States N Engl J Med 382 2020 929 936 32004427 3 Fu S.J. George E.L. Maggio P.M. Hawn M. Nazerali R. The consequences of delaying elective surgery: surgical perspective Ann Surg 272 2020 e79 e80 32675504 4 Jain A. Jain P. Aggarwal S. SARS-CoV-2 impact on elective orthopaedic surgery: implications for post-pandemic recovery J Bone Joint Surg Am 102 2020 e68 32618916 5 Cigna Cigna study finds reduced rates of acute non-elective hospitalizations during the COVID-19 pandemic [Cigna Newsroom website]. April 2020. Available at: https://www.cigna.com/about-us/newsroom/studies-and-reports/deferring-care-during-covid-19 6 Garcia S. Albaghdadi M.S. Meraj P.M. Reduction in ST-segment elevation cardiac catheterization laboratory activations in the United States during COVID-19 pandemic J Am Coll Cardiol 75 2020 2871 2872 32283124 7 Tankel J. Keinan A. Blich O. The decreasing incidence of acute appendicitis during COVID-19: a retrospective multi-centre study World J Surg 44 2020 2458 2463 32458019 8 Solomon M.D. McNulty E.J. Rana J.S. The COVID-19 pandemic and the incidence of acute myocardial infarction N Engl J Med 383 2020 691 693 32427432 9 Woolhandler S. Himmelstein D.U. Intersecting U.S. epidemics: COVID-19 and lack of health insurance Ann Intern Med 173 2020 63 64 32259195 10 Gaskin D.J. Hadley J. Population characteristics of markets of safety-net and non-safety-net hospitals J Urban Health 76 1999 351 370 12607901 11 Wakeam E. Hevelone N.D. Maine R. Failure to rescue in safety-net hospitals: availability of hospital resources and differences in performance JAMA Surg 149 2014 229 235 24430015 12 Harris Health System Locations: Lyndon B. Johnson hospital [Harris Health website]. Available at: https://www.harrishealth.org/locations-hh/Pages/lbj.aspx 13 Harris Health System About us: facts and figures [Harris health website]. Available at: https://www.harrishealth.org/about-us-hh/who-we-are/Pages/statistics.aspx 14 Shelton R. The emergency severity index 5-level triage system Dimens Crit Care Nurs 28 2009 9 12 19104244 15 Palisi M. Massucco P. Mineccia M. Celano C. Giovanardi F. Ferrero A. The disappearing of emergency surgery during the COVID 19 pandemic. Fact or fiction? Br J Surg 107 2020 e508 e509 32871023 16 Hessheimer A.J. Morales X. Ginesta C. Where have all the appendicitis gone? patterns of urgent surgical admissions during the COVID19 pandemic Br J Surg 107 2020 e545 e546 32866298 17 Cano-Valderrama O. Morales X. Ferrigni C.J. Reduction in emergency surgery activity during COVID-19 pandemic in three Spanish hospitals Br J Surg 107 2020 e239 10.1002/bjs.11667 32406929 18 De Filippo O. D'Ascenzo F. Angelini F. Reduced rate of hospital admissions for ACS during Covid-19 outbreak in Northern Italy N Engl J Med 383 2020 88 89 32343497 19 Hemingway J.F. Singh N. Starnes B.W. Emerging practice patterns in vascular surgery during the COVID-19 pandemic J Vasc Surg 72 2020 396 402 32361072 20 Hafner K. Fear of COVID-19 leads other patients to decline critical treatment, New York times Published online May 25, 2020. Available at: https://www.nytimes.com/2020/05/25/health/coronavirus-cancer-heart-treatment.html 21 Pereira M.R. Mohan S. Cohen D.J. COVID-19 in solid organ transplant recipients: initial report from the US epicenter Am J Transplant 20 2020 1800 1808 32330343 22 McNamara D. COVID-19: are acute stroke patients avoiding emergency care [Medscape web site] may 31, 2020. Available at: https://www.medscape.com/viewarticle/928337 23 Marrow H. METRO reduces Park & Ride service, closes HOV lanes; two employees test positive for COVID-19 [Community Impact Newspaper Online]. March 30, 2020. Available at: https://communityimpact.com/houston/bellaire-meyerland-west-university/transportation/2020/03/30/metro-reduces-park-ride-service-closes-hov-lanes-two-employees-test-positive-for-covid-19/ 24 Surek A. Ferahman S. Gemici E. Effect of COVID-19 pandemic on general surgical emergencies: are some emergencies really urgent? Level 1 trauma center experience Eur J Trauma Emerg Surg 47 2021 647 652 33136190 25 Orthopoulos G. Santone E. Tirabassi I.F. Increasing incidence of complicated appendicitis during COVID-19 pandemic Am J Surg 221 2021 1056 1060 33012500 26 Rosenbaum L. The untold toll - the pandemic's effects on patients without Covid-19 N Engl J Med 382 2020 2368 2371 32302076 27 Vallès K.F. Neufeld M.Y. Caron E. COVID-19 pandemic and the cholecystitis experience at a major urban safety-net hospital J Surg Res 264 2021 117 123 33812090 28 American College of Surgeons COVID 19 guidelines for triage of emergency general surgery patients [American College of Surgeons web site]. Updated March 25, 2020. Available at: https://www.facs.org/covid-19/clinical-guidance/elective-case 29 Cao A.M. Eslick G.D. Cox M.R. Early laparoscopic cholecystectomy is superior to delayed acute cholecystitis: a meta-analysis of case-control studies Surg Endosc 30 2016 1172 1182 [published correction appears in Surg Endosc. 2016 Mar;30(3):1183] 26139487 30 Gostin L.O. Friedman E.A. Wetter S.A. Responding to Covid-19: how to navigate a public health emergency legally and ethically Hastings Cent Rep 50 2020 8 12
PMC009xxxxxx/PMC9005441.txt	==== Front Resour Policy Resour Policy Resources Policy 0301-4207 1873-7641 Published by Elsevier Ltd. S0301-4207(22)00169-6 10.1016/j.resourpol.2022.102721 102721 Article Does COVID-19 pandemic cause natural resources commodity prices volatility? Empirical evidence from China Guo Shanwen Ph.D a Wang Qibin Ph.D a∗ Hordofa Tolassa Temesgen Ph.D b Kaur Prabjot Dr c Nguyen Ngoc Quynh d Maneengam Apichit e a Institute of Political Economy, Taiwan ChengKung University, Tainan, Taiwan b School of Economics and Finance, Xi'an Jiaotong University, China c Department of Mathematics, Birla Institute of Technology Mesra, Ranchi, Jharkhand, India d Department of Economics, Thuongmai University, Vietnam e Department of Mechanical Engineering Technology, College of Industrial Technology, King Mongkut’s University of Technology North Bangkok, Wongsawang, Bangsue, Bangkok 10800, Thailand ∗ Corresponding author. 13 4 2022 8 2022 13 4 2022 77 102721102721 22 11 2021 13 3 2022 9 4 2022 © 2022 Published by Elsevier Ltd. 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. COVID-19 pandemic caused havoc around the globe in both economic and non-economic sectors. This paper, unlike previous studies, evaluates the role of COVID-19 on the volatility in natural resources. The volatility of natural resources commodity prices has been the center of discussion, especially during the pandemic. Unlike previous studies, this study aims to evaluate the role of the pandemic, i.e., Covid-19 and its possible impact on volatility in natural resources commodity prices for China. China has been the center of this epidemic disease and is considered one of the major economies affected by the Covid-19; therefore, it is better to conduct this study for China. This study uses data from January 2020 till September 2021 to capture the peak time of Covid-19. Moreover, this study employs the novel wavelet power spectrum and wavelet coherence approach to better capture volatility within commodity prices volatility and Covid-19 and evaluate the association between both variables. The empirical results reveal that only natural resources commodity prices are volatile and only short. While Covid-19 positive cases and Covid-19 deaths are not vulnerable during the study period. Moreover, the wavelet coherence conforms that both Covid-19 positive cases and Covid-19 deaths significantly cause volatility in natural resources commodity prices. Although, volatility is found at different periods; still, volatility is observed only in the short-run. The study also provides relevant policy implications to ensure a relevant and timely solution for the existing issue. Moreover, future research guidelines and the study's limitations are also provided. Keywords Commodity prices volatility COVID-19 Pandemic Wavelet approach ==== Body pmc1 Introduction In the shape of COVID-19, the world has experienced another major shock to the global economy after the global financial crisis. The world has changed dramatically during the previous three decades as a result of many economic and non-economic events or crises. Specifically, the Gulf War (1990s), Asian financial crisis (1997), oil price spike (2004), global financial crisis (2007), European sovereign debt crisis (2010–2012), oil supply glut (2015–2016), and the recent Covid-19 pandemic outbreak are among the events attributed for global and regional economic issues (Lyu et al., 2021; Maitra et al., 2021). These events influence industrial/production activities and economic development and create uncertainty in natural resources and their prices (Sun and Wang, 2021; Guan et al., 2021). The Covid-19 pandemic causes severe illness and death, making the people fear this novel disease led economies to recession. In China, this novel pandemic is identified at first, which reached out to most nations. The said economy is affected the most in employment, services, production, tourism, and others. However, the primary sources of economic growth of Chinese economy are considered as the industrial or production sector. Which is severely affected as a result of the lock-down environment in the country. This postponement in the industrial sector reduces production and consequently leads to a decline in the natural resources demand. As a result, the overall supply and demand chain of natural resources like crude oil, coal, natural resources are disturbed. Hence, the prices of such natural resources are not stable in this pandemic period. Therefore, it is important to analyze the association of natural resources volatility and the Covid-19 pandemic in the region affected the most in the world (see Table 1 ).Table-1 Summary of the literature review. Table-1Author (Year) Country (Period) Methodology Findings Ma et al. (2021) China (January 01, 2019–April 01, 2021) Wavelet methods, Frequency domain causality More volatility in natural resources in Covid-19. Bidirectional causality exists between natural resources and economic performance. Sun and Wang (2021) Global Data (January 01, 2019–July 01, 2021) Wavelet methods Natural resources are volatile. No causality between natural resources and economic performance. Gil-Alana and Monge (2020) Global Data (May 04, 2010–May 04, 2020) Long memory techniques Covid-19 has a significant impact on oil prices. Apergis and Apergis (2020) USA (January 21, 2020–April 30, 2020) Mixed Data Sampling modeling Covid-19 and oil prices help reduce the US political polarization. Gupta et al. (2020) China Review The pandemic causes volatility in natural resources. Bildirici et al. (2020) WTI, Brent, Dubai crude oil (May 29, 2006–March 31, 2020) LSTARGARCH Covid-19 and Russia-Saudi conflict is responsible for chaotic behavior of natural resources. Sharif et al. (2020) USA January 21, (2020–March 30, 2020) Wavelet methods Covid-19 significantly affects economic uncertainty. Kartal (2021) Turkey (July 25, 2019–October 30, 2020) Multivariate Adaptive Regression Covid-19 influence oil prices. Narayan (2020) Global sample (February 01, 1995–May 05, 2020) Statistical Analysis Covid-19 promote oil price volatility. Mensi et al. (2020) Global Data (April 23, 2018–April 24, 2020) Asymmetric Multifractal Detrended Fluctuation Analysis Covid-19 causes volatility in natural resources commodity prices. Guan et al. (2021) Natural Resource Dependent Countries (2000–2020) ARDL, PMG Natural resources volatility is harmful for economic growth. Li et al. (2021) Panel (December 16, 2019–December 16, 2020) GARCH(1, 1) Current pandemic declines stock market returns and economic growth. Zhao et al. (2021) Developed countries Panel Regression Covid-19 enhances volatility in stock market returns. Khan et al. (2020) China (1987–2017) Generalized Least Square Natural resources have adverse impact of financial development. Umar et al. (2021a) 12 Oil producing countries (2001Q1–2019Q4) Panel Regression Resource curse surges the possibility of banks default. Umar et al. (2021b) Brent Crude Oil (January 2000–December 2020) Probability-based bubble detection mechanism Bubbles are identified in the oil market in various periods. Goodell and Goutte (2021) December 31, 2019–April 29, 2020 Wavelet Methods In Covid-19, the Bitcoin is a safe haven. Kirikkaleli et al. (2021) United Kingdom (1998–2017) Wavelet Methods Economic growth and nuclear energy consumption is positively correlated. Raza et al. (2018) USA (1979M1–2013M7) Wavelet Methods Oil prices positively affect economic activities. Zhang et al. (2021) China (1965–2019) Granger Causality Economic growth causes carbon emissions Wang et al. (2021) USA (January 4, 2011–July 31, 2020) Econometric techniques Diversification enhances profitability but reduces volatility in portfolio. Umar et al. (2020a) China (1980–2017) FMOLS, DOLS, CCR Economic growth and natural resources are the factors of emissions. Umar et al. (2020b) China (1971–2018) Wavelet methods Enhancement in financial development lower emissions. Since the last three decades, scholars and policy-makers have been involved in a contradictory debate, where the earlier claimed that natural resources abundance weakens economic growth (Gelb, 1988; Sachs and Warner, 2001). However, the latter studies oppose these claims by revealing the conditional positive impact of natural resources on economic growth by enhancing human capital and institutional quality product diversification (Rahim et al., 2021; Joya, 2015; Expo and Nochi Faha, 2020). However, whether a blessing or a curse, natural resources have long been discussed. Currently, volatility in natural resources attracted the attention of policy-makers and the academic world. The Covid-19 epidemic caused damage to the global economy, which also caused fear in financial markets throughout the world (Li et al., 2021). Since the recent outbreak, most economic sectors have been closed down due to the locked-down environment in most parts of the world. Likewise, in China, the spread of the Covid-19 pandemic postponed production, manufacturing, and the pharmaceutical industry, which causes uncertainty in the global supply chain and causes a severe shortage of life saving drugs (Gupta et al., 2020). This postponement in economic and industrial sectors reduces the demand for natural resources such as oil, which dramatically reduces the prices of raw materials and natural resources in China and the rest of the world. Additionally, the Covid-19 pandemic severeness, where the positive cases and death ratio are increases fear in public and consequently increases volatility in natural resources (Devour and Narayan, 2020). Given the significance of crude oil price volatility, it's essential to consider its progress, especially at a time when it reached its lowermost point in history. That is, f or the first time in its history, crude oil prices fell below zero. The price of West Texas Intermediate crude fell to US$37 per barrel on April 20, 2020, representing an exceptional 300 percent decline in price (Devour and Narayan, 2020). The pricing war between Saudi Arabia and Russia and the Covid-19 outbreak might be to blame for this drop. Global economic activity was suspended as a result of the epidemic. Regardless of price recovery, the post-April 2020 era marks one of the most volatile natural resources commodity market periods. The hypothesis of current study is that the outbreak of the Covid-19 pandemic coincided with the oil market's most volatile time, with Covid-19 active or positive cases and deaths contributing to this volatility. We contend that Covid-19 and oil market activity, including price and volatility, are inextricably linked. The rationale is straightforward. The novel Covid-19 pandemic triggered three particular governmental reactions: firstly, the lockdown; secondly, the travel ban, and thirdly, the stimulus package (see Phan and Narayan, 2020). Two of these three governmental reactions (i.e., lockdown and travel restriction) effectively suspended economic activity, including foreign travel, lowering oil demand and consumption. According to the World Trade Organization (WTO), as a consequence of Covid-19, trade in 2020 is predicted to drop by 13–32 percent. In summary, global oil demand has decreased, affecting both the price of oil and its uncertainties, which current study expect to be evident in oil price volatility. The primary objective of this study is to analyze volatility in natural resources commodity prices and Covid-19 pandemic. Specifically, the Covid-19 pandemic is represented by the Covid-19 positive cases and the deaths caused by Covid-19. Nonetheless, many studies in the existing literature examined the influence of Covid-19 pandemic on oil prices. However, these studies ignore volatility in natural resources commodity prices and primarily focus on the economic impact of oil prices. Secondly, this study aims to analyze the causal association Covid-19 positive cases and natural resources commodity prices during the peak period of pandemic. Although the studies of Ma et al. (2021), and Sun and Wang (2021) provide empirical evidence regarding volatility in natural resources commodity prices while considering the Covid-19 pandemic period, yet they ignored the specific causal nexus of mentioned variable and are more oriented to economic performance. Lastly, this study examines the causal association between deaths caused by the Covid-19 pandemic and natural resources commodity prices volatility. The deaths caused by Covid-19 pandemic cause fear among the general public in China, which leads to the postponement of various economic activities. Thus, demand for natural resources has fallen down drastically, which could play a substantial role in natural resources commodity price volatility. Therefore, this study is substantial and could provide relevant policy implications to tackle the issue. This study is novel and contributes to the existing literature in three ways. Firstly, it is one of the pioneering efforts empirically investigating volatility in Covid-19 positive cases, Covid-19 deaths, and natural resources commodity prices. Although, the literature consists of empirical studies for the influence of Covid-19 cases and deaths on oil prices. Still, volatility in the Covid-19 pandemic remained unexplored. However, this is an emerging topic of interest for the policy-makers as it is relevant to a very severe issue that could harm the country's economic system. Secondly, this study provides empirical evidence of the causal linkage between Covid-19 positive cases, Covid-19 deaths, and natural resources commodity prices volatility. There are numerous factors and indicators influencing natural resources commodity prices volatility. However, recently, the Covid-19 pandemic is considered the leading cause of volatility in natural resources for which the literature is not extensive. Therefore, current study provides empirical evidence as contribution to the literature. Moreover, this study used extended dataset that covers Covid-19 pandemic period extensively. Unlike other studies, this study provides comprehensive empirical estimates for volatility and causal nexus using wavelet specifications, which are considered relatively better than the existing methodologies (as discussed in Section-3). Since the Covid-19 pandemic is a threat to human lives and economic systems. Also, China stood the first to experience this fatal pandemic. Therefore, this study's findings and policy implications could help the scholars view and tackle the issue in a more appropriate way. The rest of the paper is organized as following: Section-2 provides relevant literature review; Section-3 presents data and methodology; Section-4 represents results and discussion of the empirical findings; Section-5 provides conclusion and policy implications along with the limitations and future research guidelines. 2 Literature review Since the last three decades, there is a growing literature regarding the influence of natural resources and various economic and non-economic factors and indicators. However, after the emergence of Covid-19 pandemic, the scholars focused more on volatility in natural resources commodity prices. Specifically, Ma et al. (2021) investigated the pre and posit-Covid-19 pandemic periods in case of China by using the wavelet power spectrum, wavelet coherence, and the frequency domain causality tests. The estimated results asserted that natural resources commodity prices are more volatile in Covid-19 pandemic period. Also, the results reveal bidirectional causal association between natural resources commodity price volatility and economic performance. Regarding global economic performance and natural resources commodity price volatility, Sun and Wang (2021) investigated the period of January 01, 2019, to July 01, 2021, by employing the wavelet specifications. The study found that natural resources commodity prices are volatile at different frequencies across different periods, while economic performance is stable. In contrast to Ma et al. (2021), this study found no causal association between the variables. In addition, the recent study of Umar et al. (2021b) used the Probability-based bubble detection mechanism to detect bubbles in the Brent Crude Oil during the period January 2000–December 2020. The study's examined results reveal that bubbles exist in the oil market in various periods. Besides total natural resources volatility, the recent studies are more tending towards volatility in oil prices. Specifically, Gil-Alana and Monge (2020) examined the influence of Covid-19 pandemic on crude oil prices via employing the long memory techniques. The empirical results reveal that the crude oil market was efficient before the outbreak and became efficient after the Covid-19 pandemic. Also, the results unveil that oil prices incorporate mean reverting behavior, indicating that the Covid-19 pandemic shock will be transitory while having long-lasting effect. Using the MIDAS methods, Apergis and Apergis (2020) found that both oil prices and Covid-19 pandemic help in mitigation of the US political polarization. Specifically, the authors claimed that the political leaders minimize their aim for partisan gains particularly in the stressful times. The Covid-19 pandemic, which originated from Wuhan, affects most countries across the globe, reducing global economic growth from 2.9 to 2.4 percent. However, in China, Gupta et al. (2020) illustrate that Covid-19 pandemic spread leads to the close down the production and manufacturing sector, the pharmaceutical industry – disturbs the supply chain leads to shortage of life saving drugs. The authors argued that the reduction in China's production leads to decreased prices of raw materials and natural resources in other parts of the world. Bildirici et al. (2020) investigated West Texas Intermediate (WTI), Brent and Dubai crude daily oil price covering the period from May 29, 2006, to March 31, 2020. The study used Logistic Smooth Transition Autoregressive Generalized Autoregressive Conditional Heteroskedasticity (LSTARGARCH) and the long-short term memory specifications. The empirical findings reveal that the behavior of oil prices is chaotic across the time, where the recent trends, i.e., Covid-19 pandemic, and Russia and Saudi Arabia conflict, are the leading reasons for this behavior in oil prices. Using the wavelet specifications, Sharif et al. (2020) stats that oil is the leading market in US with a lower and higher frequency during the period under observation. Additionally, the authors reveal that Covid-19 pandemic significantly affect the political risk and economic uncertainty of the US relative to their stock market. In addition, Kartal (2021) analyzed Turkish economy from July 25, 2019, to October 30, 2020, using multivariate adaptive regression. The study uncovers that the volatility index has the greatest impact on oil prices irrespective of sample size. Moreover, in Turkey, the COVID-19 outbreak impacts the influence of effective factors on local currency oil prices. In continuation, Narayan (2020) analyzed the impact of Covid-19 and negative oil price news on oil price volatility. The empirical findings suggest that both Covid-19 active cases and negative oil price news significantly increase volatility in oil prices. However, in this panic times, Goodell and Goutte (2021) argued that Bitcoin is the safe haven for investors. Regarding the influence of the Covid-19 pandemic on the price of the most tradable natural resources, i.e., gold and oil prices, Mensi et al. (2020) employed the Asymmetric Multifractal Detrended Fluctuation Analysis approach on 15-min interval intraday data. The empirical findings report that prior to Covid-19 pandemic, the gold market was inefficient due to downward trends, while the inefficiency of gold market was reported during the Covid-19 pandemic. Besides, the oil market is found inefficient by following upward trends before the pandemic and downward trends during the Covid-19 pandemic – which is evidence of the volatility in natural resources commodity prices. Beside volatility in natural resources commodity prices, studies provide empirical evidence regarding the influence of natural resources on various economic and non-economic factors and indicators. Specifically, Guan et al. (2021) analyzed natural resources price volatility and its influence on the economic growth of the natural resources’ dependent economies over 2000–2020. The study used autoregressive distributed lag (ARDL) model and pooled mean group (PMG) estimator for empirical analysis. The results obtained revealed that crude oil had faced a significant downturn, particularly in the 2008 global financial crisis and Covid-19 pandemic. However, gold prices are found less volatile than the crude oil prices during these periods. Moreover, volatility in these natural resources prices deteriorates economic growth in these regions in both short and long run. On the other hand, Raza et al. (2018) claim that oilprices are positively associated to economic activities in the USA during the period 1979M1–2013M7. Due to this, Wang et al. (2021) asserted that diversification is a better option to enhance profitability in the USA. However, economic activities are the leading factor of economic growth across the globe. On the other hand, economic growth and natural resources are the major cause of carbon dioxide emissions Umar et al. (2020a). In addition, the recent study of Zhang et al. (2021) also demonstrates that economic growth causes carbon emissions, which reveals that growth without considering environment is harmful to sustainable development. In addition, Mazur et al. (2021) investigated S&P1500 in order to analyze Covid-19 pandemic and the stock market crash in March 2020. The study found that healthcare, food, natural gas, and software stocks produces high positive returns. However, the equity value in petroleum hospitality, entertainment and real estate sectors fall drastically. The authors argued that the recent pandemic outbreak triggered the stock market crash. Regarding the association between Covid-19 pandemic fear and volatility in the stock market, Li et al. (2021) employed the GARCH(1, 1) specifications and revealed that Covid-19 fear is the driving force for stock market volatility. Specifically, an increase in the Covid-19 cases significantly reduces economic growth and stock market returns. Besides, the authors claimed that the public attention to buying or selling attitude of stocks is greatly dependent on the cases of Covid-19 pandemic. Moreover, Zhao et al. (2021) studied the influence of Covid-19 containment measures on the expected stock price volatility in developed economies. Employing panel regression in the daily data, the study unveils that six-month-ahead volatility indices fell when first or re-imposed lockdowns were announced but did not fall considerably once the lockdowns were eased. For three-month-ahead predicted volatility, such patterns are weaker, and for one-month-ahead expected volatility, they are almost non-existent. To summarize the literature, this study observed from the given literature that natural resources volatility and pandemics are greatly associated. To be more specific, the literature appeals that economic productivity or industrial production has been dropped due to pandemics. Due to this, the demand for natural resources across the globe has slowed down, which causes volatility in natural resources. However, studies have provided a safe haven for future investment instead of natural resources like oil. Such safe havens are gold, and the bitcoins. Moreover, the literature also demonstrates that the pandemic has created fear amongst the industrialists. While this fear of sickness and death leads drop the stock market participation and performance. Nonetheless, the existing literature provides a valuable insight regarding the association of pandemic and natural resources volatility. However, most of the studies have focused only on the economic aspects of the volatility. Relatively, these studies have ignored the primary reason behind natural resources volatility. Moreover, China is the first economy to experience the worst impact of Covid-19 pandemic. Also, the Chinese economy stood among the most affected region across the globe due to postponement of production and industrial sector. In addition, the reduction of production and other economic activities across the globe leads to fluctuations in the prices of natural resources. Therefore, it is important to identify whether Covid-19 is the cause of this volatility in natural resources in China. 3 Data and methodology 3.1 Variables and data Based on the objectives and mentioned literature, this study used a total of three variables to analyze the causal association of Covid-19 pandemic and natural resources commodity price volatility. Specifically, this study used the Covid-19 positive (CP) cases – indicates the infection that spreads across the country, deaths due to Covid-19 positive (CPD) – indicates the number of deaths that is occurred due to the novel pandemic, and oil prices (OP) to represent volatility in natural resources commodity prices. The main reason behind the selection of the variables is that the Covid-19 pandemic creates fear in the economies across the globe. Two factors are responsible for this fear in the general public. Specifically, the fear of Covid-19 sickness, and the death due to this novel disease. Keeping in mind these two factors, the industrial and production sectors postponed production, severely affecting the demand and supply chain of goods and services. Also, temporary closing of the industrial and production sector reduces demand for natural resources, which disturbs imports and exports of natural resources such as oil. Therefore, reducing the demand for natural resources and particularly crude oil, its prices tend to reduce. Also, the trade war between Russia and Saudi Arabia regarding oil exports also affects the demand and supply chain of natural resources. Hence, it is important to analyze whether Covid-19 pandemic is leading in this perspective. In order to discover the results, daily data for the said variables have been obtained from various sources, particularly for China, covering the period from January 2020 to September 2021. Regarding the data sources, the oil prices data is obtained from West Texas Intermediate (WTI),1 while data for Covid-19 active cases and deaths due to Covid-19 positive is extracted from World Health Organization (WHO).2 3.2 Estimation strategy The current study used a wavelet technique to measure the correlation of time series such as Covid-19 active cases and deaths due to Covid-19 on China's natural resources commodity price volatility. Nonetheless, time-varying estimation methodologies such as recursive autoregressive conditional heteroscedasticity (ARCH), generalized ARCH, cointegration, or Diebold and Yilmaz's (2009) rolling-sample spillover index methodology are available. On the other hand, the wavelet technique is regarded as more resilient since it better represents short- and long-run time series trends and causalities. The wavelet technique breaks down a time series into specific time scales instead of other short and long term conventional approaches such as cointegration and error correction. According to Kim Karlsson et al. (2018), wavelet analysis will validate differences in data and vital information, which might be ignored in other time-varying procedures. The wavelet technique, in other words, shows how the time series element evolves over time. This separates the time series under consideration into scaled and shifted mother wavelet time series (Aguiar-Conraria et al., 2008). In addition, wavelet transformations – in which the wavelet's length varies independently over time - allow for natural local analysis of time-series variables. On the other hand, this approach may be compressed into a short wavelet component to examine higher frequency variations and stretched into a long wavelet function to predict the pattern of lower-frequency variations. Short wavelet functions (narrow window) are appropriate for recording fast or quick changes, but the enlarged function can be used to isolate gradual and steady fluctuations (wide window). Also, the wavelet technique leads to efficient estimates by converting non-stationary time series challenges into empirical evidence, which other approaches cannot do (Sifuzzaman et al., 2009). Since there are numerous advantages that the wavelet method offers in empirical research, however, some of the advantages of this time-varying approach is discussed above. As a result of these benefits, the current study used a wavelet approach, which included the wavelet power spectrum and wavelet coherence, to explore vulnerabilities or volatilities and the causal relationship between Covid-19 active cases, deaths caused by Covid-19 pandemic, and the oil prices. Both of these specifications are discussed as following: 3.2.1 Continuous wavelet The current work used a novel wavelet method, in which the wavelet coherence approach reveals time-frequency domain causalities for all variables in the short and long run. In general, a wavelet is an integral squared function having a real value and a mean of zero, denoted by ψ and given as follows:(1) ψT,S(t)=S−1/2ψ(t−TS) From the above equation, The unit variance held by wavelet is shown by S−1/2, which is a normalized constant. Aside from that, the wavelet has two parameters: location/time (T) and frequency (S). Both time and frequency factors are important for detecting the specific wavelet location or position in time, which is mostly related to wavelet fluctuations, and controlling corresponding frequency variations, respectively. The ψ expression will waggle across the t-axis and behave in a wavelike manner. The wavelet utilized here is essentially from the Morlet wavelet family developed by Goupillaud et al. (1984),which may be expressed as(2) ψ(t)=π−141eiω0te−12t2 From the above equation, the factor of normalization is represented by π−14, which indicates the unit energy of wavelet. The Gaussian envelope with a one unit of standard deviation is depicted by e−12t2. Additionally, eiω0t represents complex sinusoid. Moreover, the wavelet in this scenario, exclusively evaluates limited time series data [p(t) = 1, 2, …, T]. The distinction among time and scale localization is uncertain according to Heisenberg's uncertainty principle. As a result, the Morlet wavelet w 0 = 6 is effective for the central frequency, according to Rua and Nunes (2009), due to its performance in optimizing the time and scale localization. 3.2.2 Continuous wavelet transform We can compute the temporal fluctuation of the frequency of time series using the continuous wavelet transformations Wp(T,S). The following equation may be used to express the continuous wavelet transform:(3) Wp(T,S)=∫−∞∞p(t).S−1/2ψ∗(t−T‾S)dt Where the '*' in Eq. (3) denotes complex conjugates, and S specifies whether the wavelet differentiates components of p(t) at a lower or higher scale, which is feasible if the acceptability requirement is met. Furthermore, a higher scale suggests smaller wavelets or fluctuation, whereas a lower scale indicates higher wavelet fluctuations. In this case, wavelet power spectrum (WPS) is appropriate since it delivers more information and amplitude for a given time series. The WPS operated with the following squared form:(4) WPSp(T,S)=\|Wp(T,S)\|2 3.2.3 Wavelet coherence This study used the wavelet coherence approach after evaluating the wavelet power spectrum. Despite the commonalities and contrasts between wavelet coherence and other existing methodologies, this method is unique in that it allows for the detection of correlations between two time variables, p(t) and q(t), in a unified time-frequency domain. The following is the equation for the cross-wavelet transform of both time series:(5) Wpq(T,S)=Wp(T,S)Wq(T,S)‾ From the above equation, Wp(T,S) and Wq(T,S) indicates the cross-wavelet transformation for p(t) and q(t), respectively. In addition, the bar denotes complex conjugates. Wpq (cross-wavelet transform) on the left indicates the covariance of two time series at a certain scale. The wavelet power spectrum records contribution to the series' variance at each time scale, whereas cross-wavelet power measures covariance contribution in time-frequency space. The wavelet coherence, as previously stated, captures the frequency co-movement of two time series variables, p(t) and q. (t). In this regard, Torrence and Compo (1998) suggested the squared form of wavelet coherence, which may be described in Eq. (6) below:(6) R2(T,S)=\|D(S−1Wpq(T,S))\|2D(S−1\|Wp(T,S)\|2)D(S−1\|Wq(T,S)\|2) From the equation mentioned above, the time smoothing process is indicated by D. Regarding the right side of the equation, its values value ranges from 0 to 1, i.e., 0 ≤ R2(T,S) ≤ 1. Specifically, if the R2(T,S) value is approaching to zero, this designates that there is no correlation between the two time series. However, if R2(T,S) value is approaching to one, this reveals that there is a strong correlation between the two time series. Colors ranging from blue to yellow-red might be used to distinguish the connection in a wavelet coherence. The colour blue denotes a lack of or weak connection, but the colour yellow-red denotes a high association between two time variables p(t) and q(t). 3.2.4 Phase There is no definite distinction that indicates whether the relationship between time variables p(t) and q(t) is positive or negative when using the R2(T,S) analytical approach. In this case, the wavelet coherence phase difference is utilized to look for a positive or negative correlation, as well as a lag-lead relation, between the two time series in a combined time-frequency domain. As a result of Torrence and Webster's (1999) research, the wavelet coherence phase difference may be calculated using the following equation:(7) φpq(T,S)=tan−1(L{D(S−1Wpq(T,S))}O{D(S−1Wpq(T,S))}) From the above equation, the real part operator and imaginary part operators of smooth power spectrum is captured by L and O, respectively. Interestingly, φpq(T,S) provides two-dimensional graphical representation, which can be used for the empirical findings regarding wavelet coherence technique. The empirical findings may be interpreted as follows once the results are produced using the wavelet technique. The horizontal axis reflects time, while the vertical axis reflects frequencies in the wavelet power spectrum and wavelet coherence graphical representation. A lower frequency implies a greater size. In the time-frequency domain, wavelet coherence may be used to detect two time series that co-vary. Additionally, the colour yellow-red indicates a strong connection between series, whereas the colour blue indicates a weaker to no connection between the temporal variables under examination. If there is no link between the series, the cooler regions away from the critical region(s) imply time and frequency. Further, the lag and lead phase relationships between the examined variables are shown by arrows in the wavelet graphical displays. At a given scale, the zero phase shift represents the co-movement of two time series. Furthermore, the time series are all in phase or have a positive correlation when the arrows go to the right, but out of phase or have a negative correlation when the arrows travel to the left. Both variables flow in the same direction when two series are in phase, while they flow in the opposite direction when they are anti-phase. Moreover, an arrow moving up, left-up, or right-down on a wavelet coherence schematic graph depicts that the first variable causes the second variable. On the other hand, if the arrows are pointing down, left-down, right up, this leads to the conclusion that the second variable is causing the first variable. After discovering volatility and causal nexus between the variables via the wavelet power spectrum and wavelet coherence, respectively, this study tested the association between these variables via employing the Quantile-on-Quantile (QQ) regression approach. This technique is more powerful in dealing the non-normality or irregularity issue of the data under consideration. Hence, the QQ regression is used as a robustness test. 4 Results and discussion Non-stationary qualities can be seen in a variety of economic and non-economic variables. On the other hand, most estimating methodologies do not allow for empirical investigation of non-stationary series. Furthermore, these non-stationary series might contain significant periodic signals that change amplitude and frequency with time. We used the wavelet power spectrum to capture these shifting trends in Covid-19 positive (CP) cases, deaths due to Covid-19 positive, and natural resource commodity prices from January 2020 to September 2021 in China. In addition, the time-frequency dependency of these variables might raise a few problems, as stated in the study's objectives: First, is there any evidence of time-frequency dependency for these variables? Second, what is the direction of causality between these variables, provided the hypothesis is correct? Finally, whether the relationship occurs in a certain time period (i.e., short-run and long-run) or across the overall time span? The current study used the wavelet coherence technique for empirical analysis of China to answer these questions. As a result, both wavelet methods are used in this work. 4.1 Results of the wavelet power spectrum The graphical depiction of wavelet power spectrum in Fig. 1, Fig. 2, Fig. 3, depicts the zone of influence, which also specifies an edge. However, the wavelet power spectrum produces negligible findings below that edge and cannot be interpreted. Furthermore, Monte Carlo simulation was used to achieve such substantial estimations – demonstrated by the black contour. The black contour represents the empirical findings' at 5% significance level. In addition, the colors of the wavelet power spectrum graph reflect vulnerabilities, with blue (colder) indicating low or no vulnerabilities and red (hot) indicating larger vulnerabilities in time series variables (Kirikkaleli, 2020). Regarding the graphical display of wavelet power spectrum for oil prices (Fig. 1 ), only one significant region is found that indicates the vulnerable oil prices. Specifically, the vulnerable period is between February and May (2020). The scale is found lower, indicating that the frequency is higher in these months, reported as 0–15. Besides, other regions reported showed vulnerability in natural resources commodity prices, yet these regions are insignificant. The vulnerability in natural resources during these months is likely due to the lockdown environment in China (Gupta et al., 2020). Specifically, the lockdown due to Covid-19 pandemic leads to reduced production, trading, and other economic activities, which significantly reduces oil demand and causes fluctuations in oil prices among other natural resources. Besides, the recent conflict of Russia and Saudi Arabia leads to a substantial supply of oil at a lower price for market capturing strategy (Bildirici et al., 2020). This further contributes to the oil market volatility in China. As a result, natural resources commodity prices, i.e., oil prices, in particular, lost prices consistency due to these mentioned issues. These findings are consistent with the earlier empirical findings of Umar et al. (2021b), which demonstrate the existence of bubbles in the oil price market in different periods. Hence, it is concluded that vulnerabilities exist in the oil prices. Concerning Fig. 2 and Fig. 3 , the Covid-19 positive cases and deaths are found stable during the selected period of time. Specifically, a higher fluctuation has not been observed in the Covid-19 positive cases and deaths caused by the Covid-19 pandemic. As depicted by the dark blue color, these variables are not significantly vulnerable, while a weaker vulnerability is observed as depicted by the light blue/green color, which is still insignificant. Hence, it is concluded that in the selected three variables, only natural resources commodity prices are volatile in a shorter period of time.Fig. 1 Wavelet power spectrum for oil prices (OP). Fig. 1 Fig. 2 Wavelet Power Spectrum for Covid-19 positive (CP) Cases. Fig. 2 Fig. 3 Wavelet power spectrum for deaths due to Covid-19 (CPD). Fig. 3 4.2 Results of the wavelet coherence The wavelet coherence graphical display shown in Fig. 4, Fig. 5 shows that the horizontal line represents time, and the vertical line represents frequency. Lower frequency indicates larger scale connection, similar to the wavelet power spectrum, whereas higher frequency indicates lower scale linkage. Wavelet coherence is a more effective method than other existing causality methodologies when considering the time-frequency domain. This method takes two time variables p(t) and q(t) at the same time, as well as their respective co-movements across the selected time period. Focusing on such advantages, the current study used the wavelet coherence approach to investigate the short- and long-term causal links between Covid-19 positive cases and deaths and natural resource commodity price volatility in China across the time period under study. The wavelet coherence graph, like the wavelet power spectrum, uses colors to identify the causal effect, with blue (colder) color denoting weaker or no inter-relationship and red (hot) color denoting high inter-relationship between the variables under investigation (Kirikkaleli, 2020). Furthermore, the black cone-shaped curve denotes a significant zone, and the contour denotes a 5% significant level. Moreover, arrows might be used to indicate the causal linkage: while heading to the right (left), the variables are in phase (antiphase) and traveling in the same (opposite) direction. Further, the first variable is leading when the arrows are pointed up, left-up, or right-down. In contrast, the second variable is leading when the arrows are pointing down, left-down, or right-up. This work's wavelet coherence empirical findings might give vital policy insights for governors, policymakers, and academics. Regarding the wavelet coherence between natural resources commodity prices and Covid-19 positive cases, Fig. 4 reveals four significant regions of causal nexus between the two in different periods. Firstly, in February 2020, there is significant association as depicted by the arrows' direction. In the mentioned period, the arrows are pointing towards the left-up and left-down directions. This indicates that there is a bidirectional causal association between Covid-19 positive cases and oil prices. Secondly, the significant region for causal nexus is found as August–November 2020 period. Particularly, this period is considered as the peak period of Covid-19 pandemic. The arrows found in this period is found traveling to the right-up and left-down. This demonstrates that changes in Covid-19 positive cases significantly cause volatility in natural resources commodity prices. Thirdly, in the period of February 2021 showed significant causal influence of Covid-19 positive cases on oil prices as the arrows are pointing towards the right-up direction. However, this causal nexus is very limited. Lastly, the period of June–July 2021 is observed significant for causal nexus between oil prices and Covid-19 positive cases. Specifically, this causal influence is of higher frequency ranging from 16 to 40. However, the arrows are found traveling towards right-up and right-down, indicating a bidirectional causal association between the discussed variables. However, the influential magnitude of Covid-19 positive cases is found greater. Also, the rightward moving arrows indicate that the association between these two variables is in phase. These findings are in line with the earlier studies of Ma et al. (2021) in case of China and Sun and Wang (2021) in a global perspective – revealing that natural resources are more volatile in the Covid-19 pandemic peak period. The reason behind the phasal movement and causal nexus between Covid-19 pandemic and natural resources commodity prices is China's postponement of production and industrial activities (Gupta et al., 2020). Which further fuel volatility in natural resources commodity prices. However, as the empirical findings reveal that the causal influence is only for a shorter period of time, this is also consistent with the study of Gil-Alana and Monge (2020), which indicates that the Covid-19 shock exhibit mean reverting behavior and transitory.Fig. 4 Wavelet coherence between OP and CP. Fig. 4 Regarding wa velet coherence between oil prices and Covid-19 deaths, Fig. 5 reveals that significant regions hold the causal nexus of these two variables. Specifically, causality is found at various periods during the selected time span. However, these causal nexuses are observed only in the short-run, while missing in the long-run. Specifically, the periods of February 2020, August–November 2020, February 2021, and May–July 2021 are found significant for casual association between Covid-19 deaths and natural resources commodity prices. Besides, the causal association between Covid-19 deaths and oil prices is found only in the shorter run with a different scale and frequencies. From these significant regions, the arrows are pointing to the right, indicating a phasal association between these variables. Moreover, arrows pointing towards up, right-down, right-up, and left-down designate that there is a bidirectional causal association between the said variables. Specifically, the Covid-19 death increases fear in general public (Li et al., 2021). This fear reduces the supply of labor force that offsets production and other economic activities. As a result, demand for natural resources has fallen down, creating volatility in natural resources commodity prices. The findings of this study are consistent with the empirical findings of Narayan (2020), Ma et al. (2021) and Guan et al. (2021), which indicates that natural resources and oil prices, in particular, are volatile during the Covid-19 pandemic, while this volatility in oil prices is driven by Covid-19 pandemic, which could lead to the deterioration of economic growth in both short and long run.Fig. 5 Wavelet coherence between OP and CPD Fig. 5 4.3 Robustness results Nonetheless, the wavelet coherence demonstrates bidirectional causal association between OP –CP, and OP– CPD. However, to validate this association of OP and explanatory variables like CP and CPD, this study uses Quantile-on-Quantile (QQ) regression. The prime advantage of the said approach is that it tackles the irregularity or non-normality issue of data. Fig. 6 provides the empirical estimates of QQ regression for OP and CPD. As per the study of Xu et al. (2021), the darker blue color represents the lowest value of coefficient, and the darker red color indicates the higher coefficient value. While the darker red color indicates higher value of coefficient. Current findings indicate that there is a mixed effect of CP on OP across different quantiles. Specifically, in the earlier quantile (0–0.2) of OP, the influence of CP is observed positive, while in the middle quantiles (0.3–0.6) it is negative but with a lower coefficient value. However, in the higher quantiles (0.8–1) of OP, the influence of CP is found negative with a relatively higher value. This demonstrates that with the increase of Covid-19 pandemic active cases, the oil prices become more unstable and volatility increases. Hence, this study's findings are consistent with the existing study of Sun and Wang (2021) and Ma et al. (2021), that validates the enhancement of volatility in natural resources during the pandemic period.Fig. 6 QQ results for OP and CP Note: The z-axis indicates the coefficient values, the x-axis indicates OP, and the y-axis represents CP. Fig. 6 Similar to the earlier results of OP and CP, the estimated results of the QQ regression for OP and CPD is reported in Fig. 7 . The examined results reveal a mixed association between CPD and OP. That is, the earlier (0–0.3) quantiles of OP are positive across all the quantiles of CPD with a higher coefficient value of 1–1.5. While in the medium quantiles, the coefficient value of CDP decline to the negative with a −2.5 in 0–0.2 quantiles. Whereas the higher quantiles reveal the higher coefficient values. The mixed, i.e., positive and negative values of CPD indicates that volatility of OP dur to fluctuations in the deaths of patients due to Covid-19 pandemic. The current findings of Gil-Alana and Monge (2020) and Gupta et al. (2020) empirically demonstrate that Covid-19 pandemic is the major reason for volatility in natural resources commodity prices in a global and panel data, respectively.Fig. 7 QQ results for OP and CPD Note: The z-axis indicates the coefficient values, the x-axis indicates OP, and the y-axis represents CPD. Fig. 7 4.4 Discussion In recent times, the major issue considered responsible for economic recession and postponement of industrial production is the Covid-19 pandemic. Also, China remained the first country to experience the worst effect of Covid-19 in the shape of industrial postponement. Due to such shock, unemployment rises in the country and similar for the income levels of the investors, industrialists, and households. Consequently, demand for natural resources diminishes while the supply is also surging due to the Russia and Saudi Arabia conflict of capturing higher market proportion by lowering the prices of oil. Simultaneously, the spread of Covid-19 pandemic increases, which causes fear in the general public and leads to the temporary closing of the industries. Keeping this in mind, current study analyzes whether there is any association between the Covid-19 active cases, Covid-19 deaths, and oil prices in case of China. The estimated results of wavelet power spectrum reveals that the oil prices are volatile in the month of May. This indicates that the oil prices are unstable during the Covid-19 pandemic peak period instead of volatile. On the other hand, the wavelet coherence asserted a two-way causality between oil prices and Covid-19 active cases, and similar for the oil prices and Covid-19 deaths. This demonstrates that enhancement in the Covid-19 active cases and death leads the government of China to implement strict policies regarding the health of the general public. In this sense, China found the lockdown a more suitable policy of public health instead of economic sustenance, However, the drawback of this policy is that the industrial production slowed down, which further reduces natural and energy resources demand such as crude oil in the global market. Also, the conflict between Saudi Arabia and Russia further fuels the higher supply chain for economic benefits. Due to this uncertain situation in China caused by Covid-19 pandemic, the oil prices are unstable and volatility persists. Hence, to recover from natural resources volatility, Chinese government must take substantial and appropriate initiatives. 5 Conclusion and policy implications 5.1 Conclusion The recent trend of natural resources commodity price volatility and Covid-19 pandemic nexus motivate scholars to add more to this critical issue. Current study aims to analyze whether there is volatility in natural resources commodity prices and Covid-19? If yes, what is the causal nexus of these variables? In this regard, this study investigated China by using daily data from January 2020 to September 2021. In order to determine volatility, there are various methods available in the literature. However, the wavelet power spectrum is considered the most efficient technique since it considers both the time and frequency domain simultaneously. Besides, the causal nexus could be better analyzed with the novel wavelet coherence specifications by underlining the characteristics mentioned above. Therefore, the current study employed these approaches, which provides short-run and long-run estimates. The empirical findings reveal that oil prices are vulnerable during at only one point. Where it is observed that this vulnerability in oil prices is during the Covid-19 pandemic peak period. Specifically, in the said period, the Covid-19 pandemic was reported as highest in the selected span, which led the country's government to stop industrial production and other economic activities for public health. Also, the causal association between the variables validates the bidirectional causal association between oil prices and Covid-19 active cases, and the oil prices and Covid-19 deaths. Since the emergence of this novel pandemic, China has faced a severe shock in the shape of the lockdown, which triggered the postponement of Chinese industrial sector. Consequently, the prices of natural resources commodity prices fall down and even reach the negative. Nonetheless, the fear of Covid-19 illness and death further leads to closing other economic activities such as trading inside or across the borders. The supply and demand chain is highly disturbed. Hence, the primary reason for fluctuations in natural resource commodity prices is Covid-19 pandemic in the shape of active cases increase and surge in Covid-19 patients deaths. These results are found robust by QQ regression approach. 5.2 Policy implications Based on the empirical findings, this study suggests that immediate actions are required to stabilize natural resources volatility in China. Specifically, there are two appropriate policies that this study suggests: firstly, the price ceiling or price freezing policy could be adopted to regulate the prices of natural resources like crude oil. Nonetheless, the price ceiling or freezing would help the natural resources market stabilize in various crisis conditions regardless of the demand-supply chain disturbance. Secondly, natural resources hedging could be an appropriate policy in these crisis times. The hedging of natural resources will benefit the economy in the short-run and the longer run and tackle the issue of natural resources volatility. Moreover, the Covid-19 is observed as the primary reason for this conflict of natural resources volatility. In this regard, innovative policies in the health sector are required to recover from this novel pandemic for economic and natural resources markets recovery. 5.3 Limitations Although this study provides substantial findings, it is still limited from various perspectives. Specifically, this study utilized only oil prices to represent natural resources commodity prices volatility. However, future researchers could extend this study by investigating other natural resources commodity prices such as coal price, gold price, forest prices, metallic natural resources prices, etc. Also, this study provides empirical results only for China as it is the first country to experience a novel Covid-19 pandemic crisis. However, researchers in the future shall investigate developed, emerging, and developing economies. Moreover, both the pre and post Covid-19 pandemic periods shall be investigated for a comprehensive analysis of Covid-19 pandemic. In addition, the developed and developing economies could be compared regarding the association of Covid-19 pandemic and natural resources volatility in future studies. Credit author statement Shanwen Guo: Supervision, Project administration, Funding acquisition. Qibin Wang: Formal Analysis, Conceptualisation, writeup, empirical analysis. Tolassa Temesgen Hordofa: Review and Discussion. Miss Prabjot Kaur: writeup support and empirical analysis. Soufiyan Bahetta: Supervision, Funding acquisition. Apichit Maneengam: Formal Analysis, Conceptualisation. Data availability The data that has been used is confidential. 1 For oil price data and details, visit: https://www.investing.com/commodities/crude-oil-historical-data. 2 For Covid-19 data and details, visit: https://covid19.who.int/table/. ==== Refs References Aguiar-Conraria L. Azevedo N. Soares M.J. Using wavelets to decompose the time–frequency effects of monetary policy Physica A 387 2008 2863 2878 10.1016/j.physa.2008.01.063 Apergis E. Apergis N. Can the COVID-19 pandemic and oil prices drive the US partisan conflict index? Energy Res. Lett. 2020 10.46557/001c.13144 Bildirici M. Guler Bayazit N. Ucan Y. Analyzing crude oil prices under the impact of COVID-19 by using LSTARGARCHLSTM Energies 13 2020 2980 10.3390/en13112980 Devour N. Narayan P.K. Hourly oil price volatility: the role of COVID-19 Energy Res. Lett. 2020 10.46557/001c.13683 Diebold F.X. Yilmaz K. Measuring financial asset return and volatility spillovers, with application to global equity markets Econ. J. 119 2009 158 171 10.1111/j.1468-0297.2008.02208.x Expo B.N. Nochi Faha D.R. Natural resources, institutional quality, and economic growth: an African tale Eur. J. Dev. Res. 32 2020 99 128 10.1057/s41287-019-00222-6 Gelb A.H. Oil Windfalls: Blessing or Curse? 1988 Oxford university press New York Gil-Alana L.A. Monge M. Crude oil prices and COVID-19: persistence of the shock Energy Res. Lett. 2020 10.46557/001c.13200 Goodell J.W. Goutte S. Co-movement of COVID-19 and Bitcoin: evidence from wavelet coherence analysis Finance Res. Lett. 38 2021 101625 10.1016/j.frl.2020.101625 Goupillaud P. Grossmann A. Morlet J. Cycle-octave and related transforms in seismic signal analysis Geoexploration 23 1 1984 85 102 10.1016/0016-7142(84)90025-5 Guan L. Zhang W.-W. Ahmad F. Naqvi B. The volatility of natural resource prices and its impact on the economic growth for natural resource-dependent economies: a comparison of oil and gold dependent economies Resour. Pol. 72 2021 102125 10.1016/j.resourpol.2021.102125 Gupta M. Abdelmaksoud A. Jafferany M. Lotti T. Sadoughifar R. Goldust M. COVID-19 and economy Dermatol. Ther. 33 2020 e13329 10.1111/dth.13329 32216130 Joya O. Growth and volatility in resource-rich countries: does diversification help? Struct. Change Econ. Dynam. 35 2015 38 55 10.1016/j.strueco.2015.10.001 Kartal M.T. The effect of the COVID-19 pandemic on oil prices: evidence from Turkey Energy Res. Lett. 2021 10.46557/001c.18723 Khan Z. Hussain M. Shahbaz M. Yang S. Jiao Z. Natural resource abundance, technological innovation, and human capital nexus with financial development: a case study of China Resour. Pol. 65 2020 101585 10.1016/j.resourpol.2020.101585 Kim Karlsson H. Li Y. Shukur G. The causal nexus between oil prices, interest rates, and unemployment in Norway using wavelet methods Sustainability 10 2018 2792 10.3390/su10082792 Kirikkaleli D. Does political risk matter for economic and financial risks in Venezuela? J. Econ. Struct. 9 2020 1 10 10.1186/s40008-020-0188-5 Kirikkaleli D. Adedoyin F.F. Bekun F.V. Nuclear energy consumption and economic growth in the UK : evidence from wavelet coherence approach J. Publ. Aff. 21 2021 e2130 10.1002/pa.2130 Li W. Chien F. Kamran H.W. Aldeehani T.M. Sadiq M. Nguyen V.C. Taghizadeh-Hesary F. The nexus between COVID-19 fear and stock market volatility Econ. Res.-Ekon. Istraž. 1–22 2021 10.1080/1331677X.2021.1914125 Lyu Y. Tuo S. Wei Y. Yang M. Time-varying effects of global economic policy uncertainty shocks on crude oil price volatility：New evidence Resour. Pol. 70 2021 101943 10.1016/j.resourpol.2020.101943 Ma Q. Zhang M. Ali S. Kirikkaleli D. Khan Z. Natural resources commodity prices volatility and economic performance: evidence from China pre and post COVID-19 Resour. Pol. 74 2021 102338 10.1016/j.resourpol.2021.102338 Maitra D. Rehman M.U. Dash S.R. Kang S.H. Oil price volatility and the logistics industry: dynamic connectedness with portfolio implications Energy Econ. 102 2021 105499 10.1016/j.eneco.2021.105499 Mazur M. Dang M. Vega M. COVID-19 and the march 2020 stock market crash. Evidence from S&P1500 Finance Res. Lett. 38 2021 101690 10.1016/j.frl.2020.101690 Mensi W. Sensoy A. Vo X.V. Kang S.H. Impact of COVID-19 outbreak on asymmetric multifractality of gold and oil prices Resour. Pol. 69 2020 101829 10.1016/j.resourpol.2020.101829 Narayan P.K. Oil price news and COVID-19—is there any connection? Energy Res. Lett. 2020 10.46557/001c.13176 Phan D.H.B. Narayan P.K. Country responses and the reaction of the stock market to COVID-19—a preliminary exposition Emerg. Mark. Finance Trade 56 2020 2138 2150 10.1080/1540496X.2020.1784719 Rahim S. Murshed M. Umarbeyli S. Kirikkaleli D. Ahmad M. Tufail M. Wahab S. Do natural resources abundance and human capital development promote economic growth? A study on the resource curse hypothesis in Next Eleven countries Resour. Environ. Sustain. 4 2021 100018 10.1016/j.resenv.2021.100018 Raza S.A. Shahbaz M. Amir-ud-Din R. Sbia R. Shah N. Testing for wavelet based time-frequency relationship between oil prices and US economic activity Energy (Oxf.) 154 2018 571 580 10.1016/j.energy.2018.02.037 Rua A. Nunes L.C. International comovement of stock market returns: a wavelet analysis J. Empir. Finance 16 2009 632 639 10.1016/j.jempfin.2009.02.002 Sachs J.D. Warner A.M. The curse of natural resources Eur. Econ. Rev. 45 2001 827 838 10.1016/S0014-2921(01)00125-8 Sharif A. Aloui C. Yarovaya L. COVID-19 pandemic, oil prices, stock market, geopolitical risk and policy uncertainty nexus in the US economy: fresh evidence from the wavelet-based approach Int. Rev. Financ. Anal. 70 2020 101496 10.1016/j.irfa.2020.101496 Sifuzzaman M. Islam M.R. Ali M.Z. Application of Wavelet Transform and its Advantages Compared to Fourier Transform 2009 Sun L. Wang Y. Global economic performance and natural resources commodity prices volatility: evidence from pre and post COVID-19 era Resour. Pol. 74 2021 102393 10.1016/j.resourpol.2021.102393 Torrence C. Compo G.P. A practical guide to wavelet analysis Bull. Am. Meteorol. Soc. 79 1998 61 78 10.1175/1520-0477(1998)079<0061:APGTWA>2.0.CO;2 Torrence C. Webster P.J. Interdecadal changes in the ENSO–monsoon system J. Clim. 12 1999 2679 2690 10.1175/1520-0442(1999)012<2679:ICITEM>2.0.CO;2 Umar M. Ji X. Kirikkaleli D. Shahbaz M. Zhou X. Environmental cost of natural resources utilization and economic growth: can China shift some burden through globalization for sustainable development? Sustain. Dev. 28 2020 1678 1688 10.1002/sd.2116 Umar M. Ji X. Kirikkaleli D. Xu Q. COP21 Roadmap: do innovation, financial development, and transportation infrastructure matter for environmental sustainability in China? J. Environ. Manag. 271 2020 111026 10.1016/j.jenvman.2020.111026 Umar M. Ji X. Mirza N. Rahat B. The impact of resource curse on banking efficiency: evidence from twelve oil producing countries Resour. Pol. 72 2021 102080 10.1016/j.resourpol.2021.102080 Umar M. Su C.-W. Rizvi S.K.A. Lobonţ O.-R. Driven by fundamentals or exploded by emotions: detecting bubbles in oil prices Energy (Oxf.) 231 2021 120873 10.1016/j.energy.2021.120873 Wang L. Ahmad F. Luo G.-L. Umar M. Kirikkaleli D. Portfolio optimization of financial commodities with energy futures Ann. Oper. Res. 1–39 2021 10.1007/s10479-021-04283-x Xu B. Sharif A. Shahbaz M. Dong K. Have electric vehicles effectively addressed CO2 emissions? Analysis of eight leading countries using quantile-on-quantile regression approach Sustain. Prod. Consum. 27 2021 1205 1214 10.1016/j.spc.2021.03.002 Zhang J. Dai Y. Su C.-W. Kirikkaleli D. Umar M. Intertemporal change in the effect of economic growth on carbon emission in China Energy Environ. 32 2021 1207 1225 10.1177/0958305x211008618 Zhao Y. Liu Y. Acharya V. COVID-19 containment measures and expected stock volatility: high-frequency evidence from selected advanced economies IMF Work. Pap 2021 1 10.5089/9781513573502.001 2021
PMC009xxxxxx/PMC9005442.txt	==== Front Biomed Signal Process Control Biomed Signal Process Control Biomedical Signal Processing and Control 1746-8094 1746-8094 Elsevier Ltd. S1746-8094(22)00199-9 10.1016/j.bspc.2022.103677 103677 Article Deep feature fusion classification network (DFFCNet): Towards accurate diagnosis of COVID-19 using chest X-rays images Liu Jingyao ab Sun Wanchun a Zhao Xuehua c Zhao Jiashi a⁎ Jiang Zhengang a⁎ a School of Computer Science and Technology, Changchun University of Science and Technology, Changchun, Jilin 130022, China b School of Computer and Information Engineering, Chuzhou University, Chuzhou 239000, China c School of Digital Media, Shenzhen Institute of Information Technology, Shenzhen 518172, China ⁎ Corresponding authors. 13 4 2022 7 2022 13 4 2022 76 103677103677 31 8 2021 22 3 2022 9 4 2022 © 2022 Elsevier Ltd. All rights reserved. 2022 Elsevier Ltd Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. The widespread of highly infectious disease, i.e., COVID-19, raises serious concerns regarding public health, and poses significant threats to the economy and society. In this study, an efficient method based on deep learning, deep feature fusion classification network (DFFCNet), is proposed to improve the overall diagnosis accuracy of the disease. The method is divided into two modules, deep feature fusion module (DFFM) and multi-disease classification module (MDCM). DFFM combines the advantages of different networks for feature fusion and MDCM uses support vector machine (SVM) as a classifier to improve the classification performance. Meanwhile, the spatial attention (SA) module and the channel attention (CA) module are introduced into the network to improve the feature extraction capability of the network. In addition, the multiple-way data augmentation (MDA) is performed on the images of chest X-ray images (CXRs), to improve the diversity of samples. Similarly, the utilized Grad-CAM++ is to make the features more intuitive, and the deep learning model more interpretable. On testing of a collection of publicly available datasets, results from experimentation reveal that the proposed method achieves 99.89% accuracy in a triple classification of COVID-19, pneumonia, and health X-ray images, there by outperforming the eight state-of-the-art classification techniques. Keywords COVID-19 Deep learning Classification Feature fusion SVM ==== Body pmc1 Introduction COVID-19 has already caused over 5.7 million causalities and infected more than 396 million people, as of February 8, 2022 [1]. Since its appearance in December 2019, it has spread throughout the globe, which has forced countries to take drastic measures, including closing borders, canceling flights and quarantining people in countries with related cases, and containing the virus spread appears to be a challenging task [2]. Owing to the critical health risks associated with it, COVID-19 was declared by the World Health Organization (WHO) as an international public health emergency and pandemic on 30/01/2020 and 11/03/2020, respectively. And new mutated strains continue to emerge. The common symptoms of COVID-19 include fever, cough, shortness of breath, and pneumonia [3], and it affects the human heart, brain, liver and many other organs and requires prompt detection and treatment. It relies primarily on real-time reverse transcription polymerase chain reaction (PCR) for its determination, however, this method takes a longer time to detect. X-ray imaging is a cheaper, faster, and readily available method, where the body gets exposed to a much smaller amount of harmful radiation compared to CT [4]. Chest X-ray imaging (CXRs) is widely used as an assistive diagnostic tool in COVID-19 screening, and it is reported to have high potential prognostic capabilities [5]. However, the diagnosis throughput of human experts is not comparable with that of machines, while early symptoms are difficult to spot and may be overlooked by human experts [6]. Therefore, there exists an urgent need to develop a smarter and more accurate algorithm for assisting to detect diseases automatically (e.g. COVID-19). In recent years, there has been an increasing amount of research on the application of artificial intelligence to disease diagnosis. For instance, A. Esteva et al. [7] trained a CNN on fine-grained skin cancer images, and obtained the results, which were generally consistent with the expert judgment. In another work, M. Guo et al. [8] used a deep learning approach to classify thyroid images with better results, achieving an accuracy of 83.88%. Likewise, Lu et al. [9] introduced wavelet transform and extreme learning machine techniques to predict the healthy or abnormal brain MRI pictures, with an accuracy of 97.04%. Others are the early prognostication of Alzheimer’s disease dementia [10], diagnosis of brain hemorrhage [11], detection of diabetic retinopathy [12], identification of arrhythmia [13], and classification of various types of cancer (e.g. breast [14], brain [15], and prostate [16]). With the wide application of deep learning, its research in the diagnosis of lung diseases is also increasing. For instance, Chen et al. [17] proposed the DualCheXNet which used ResNet and DenseNet to the extract features, and used weighting of multiple classifiers for classification of fourteen diseases of the lung. Gundel et al. [18] proposed a location-aware dense network to improve the accuracy of thoracic disease classification by using high-resolution image data and spatial information of the lesion to accurately find the lesion. Most of the COVID-19 studies use a single network to extract features, but different networks acquire features in different ways and therefore focus on different regions, and fusing multiple networks can make features richer. Therefore, we fuse the features extracted by EfficientNetV2 and ResNet. Deep learning has a strong advantage in feature extraction, while SVM classification is a proven machine learning method; we combine the two to achieve better classification results. Although our network has a strong learning ability, it is poorly interpretable. People cannot perceive small changes in grayscale, but they can perceive color changes better, so we introduce a color visualization method that can present the results of network learning well. The main contributions of this work are as follows:• We proposed a deep feature fusion classification network (DFFCNet), and introduced two modules: deep feature fusion module (DFFM) and multi-disease classification module (MDCM). • EfficientNetV2 was introduced as the backbone network to fuse with the features extracted by ResNet101. The spatial attention (SA) module and the channel attention (CA) module are introduced into the network. • We used four multiple-way data augmentation (MDA) ways to enhance the training set. To simplify the interpretation of proposed deep learning model, a color visualization approach is employed via the Grad-CAM++ technique. • We used the same dataset for our experiments. Compared to the 8 state-of-the-art diagnosis methods for COVID-19, experimental results from this work illustrate very good results achieved through the DFFCNet. The structure of this paper is organized as below. In Section 2, we summarize the current state of research on COVID-19. In Section 3, introduces the dataset, the involved deep learning methods and the proposed new model. In Section 4, we describe the experimental steps and results. In Section 5, finally concludes this paper. 2 Related works The sudden appearance of COVID-19 has led many researchers to propose various artificial intelligence methods to study it. These artificial intelligence methods are divided into three categories. First, deep learning networks such as DenseNet, AlexNet, ResNet and Xception are used for disease diagnosis, and transfer learning can be used to reduce the network parameters. Second, weakly supervised learning or unsupervised methods are used to solve the problem of small labeled samples, and machine learning such as clustering or support vector machines are used as classifiers in order to improve the recognition accuracy. Third, methods such as U-Net are used to segment the lesions. Some researchers use transfer learning for training because it allows higher initial performance of the network, faster training rate, and better convergence of the obtained model, which can reduce the network parameters and make the network small. For instance, Narayan et al. [19] used transfer learning to pre-train Inception (Xception) parameters first on a large dataset and then applied them to the COVID-19 dataset to automatically diagnose diseases. Majeed et al. [20] proposed a new network, named CNN-X, which had fewer parameters and suitabled for smaller datasets. Maghdid et al. [21] used transfer learning to introduce AlexNet and proposed a simple CNN network. Experiments were performed on collected X-ray and CT images with an accuracy of 98%. Katsamenis et al. [22] proposed a simple CNN, which used transfer learning to introduce ResNet-50, changed the last fully connected layer. Pre-trained on ImageNet and achieved better classification results. Montalbo [23] introduced DenseNet as the backbone network through transfer learning, optimized migration learning by freezing some layers and adding a new layer to improve performance, a method called Fused-DenseNet-Tiny, who was able to achieve 97.99% classification accuracy. At the beginning of the outbreak of COVID-19, due to the lack of samples, many researchers proposed many models suitable for small amount of data from this perspective. For example, Aradhya, et al. [24] proposed a new model for one-time learning based on the idea of clustering, introducing two classifiers GRNN (Generalized Regression Neural Network) and PNN (Probabilistic Neural Network). Voulodimos et al. [25] proposed a new online learning model for COVID-19 based on U-Net network, called few-shot driven U-Net. It can learn features of small datasets and accurately segment COVID-19 lesion regions in CT images. Chen et al. [26] developed an end-to-end trainable deep few-shot learning framework in the shortage of annotated COVID-19 CT images in order to save computational costs. It can expand one image into multiple images to accurately diagnose diseases. Yang et al. [27] proposed a new semi-supervised learning network based on less labeled images, which can be applied to new datasets with better generalization performance based on the disease features learned on a limited dataset. In the process of learning disease features, deep learning often learns some features that are not related to the disease, and the obtained model has poor generalization ability. In order to get accurate lesion regions for their study, the segmentation task is necessary, and many researchers have proposed many methods for this problem. Among them, Voulodimos et al. [28] proposed a lightweight segmentation model using U-Net and FCN (Fully Convolutional Neural Networks). The model can be trained without GPU, meaning that it can be run on a PC (personal computer) without parallel computing capabilities. Chen et al. [29] improved U-Net by adding an attention mechanism, and the 10-fold cross-validation results showed a 10% improvement in segmentation performance compared to the traditional U-Net. Saeedizadeh et al. [30] proposed a new model with a new regularization term in U-Net, and the segmentation performance was improved by 2%, and this model is called TV-Unet. Zhou et al. [31] added the spatial attention module and the channel attention module to U-Net, effective feature relations can be obtained. Meanwhile, the dice loss was changed to the focal tversky loss, the obtained model takes only 0.29 s to segment a CT. Chen et al. [32] proposed an unsupervised segmentation network with synthetic data and limited labeled data, which can guide the segmentation network to perform cross-domain learning and improve the segmentation performance. Liu et al. [33] used transfer learning twice for accurate segmentation of COVID-19 lesions, and proposed nCoVSegNet. Due to the small amount of labeled data, the model parameters were first trained on ImageNet for the first time; the pulmonary nodules image lesion features were similar to COVID-19, so the second time was trained on a dataset with labeled pulmonary nodules to further refine the parameters and find similar lesion areas Finally, the CT images of COVID-19 were segmented again, and better results were achieved. As mentioned above, most of the current studies on COVID-19 use single networks for learning, but different networks extract different features, so there is a great need to develop a new method for combining multiple networks for learning. 3 Dataset and methodology For better understanding, Table 5 in Appendix A list the abbreviations. Moreover, the detailed methodology is described below. 3.1 Improvement I: MDA on training set 3.1.1 Original dataset Sait et al. [34] collected 15 publicly available COVID-19 datasets and removed the duplicates to form a new dataset, which is the one used in this work. The dataset contains 1281 COVID-19 X-rays, 1656 viral-pneumonia X-rays, 3270 Normal X-rays, and 3001 bacterial-pneumonia X-rays. We combined viral pneumonia and bacterial pneumonia in a single category. Fig. 1 shows three samples from the dataset. And Fig. 1(a) shows the lesion sites of COVID-19, which we have marked with red arrows. The images in this dataset vary in size and are not labeled for the severity. The main lesion characteristics are described below. The location of infection in COVID-19 is mainly in the bilateral subpleural, whereas in common pneumonia the location of infection is along the trachea, bronchi and blood vessels. The nature of the lesion in COVID-19 is predominantly ground-glass opacities, whereas the main feature of common pneumonia is a solid shadow.Fig. 1 Sample images of CXRs. (a) COVID-19. (b) Normal. (c) Pneumonia. 3.1.2 Dataset preprocessing The dataset contain COVID-19 X-ray, pneumonia X-ray and healthy X-ray images. We resize the set O of original images to a uniform size of 224 × 224, and obtain a new image set R, as shown in Eq. (1).(1) R=Resize(O,[224,224])={o1,o2,o3…,on} The abstraction of images in deep learning networks changes from input to convolutional layer, pooling layer, and the last layer of feature map to fully connected layer. Among them, the feature map can be 3 × 3, 5 × 5, 7 × 7, etc. Among these sizes, if the size is too small, then the information is easily lost, and if the size is too large, the abstraction level of information is not high enough and the computation is more, so the size of 7 × 7 is the most suitable. The input of the image must be 7×(exponential power of 2) and the size of the dataset images are around 300, so 224 = 7 × 32 is the most suitable. 3.1.3 Data augmentation Through random hold-out (RHO) method, the dataset was randomly divided into three subsets: the testing set (X: 20%), the training set (Y: 70%), and the validation set (Z: 10%). The relevant information is listed in Table 1 . Furthermore, to mitigate any potential over fitting, MDA [6] technology is utilized in this work. We used four ways to enhance the training set(2) R→RHO{X,Y,Z} Table 1 Data distribution in the model. Dataset COVID-19 Normal Pneumonia Total Training (70%) 897 2289 3260 6446 Testing (20%) 256 654 931 1841 Validation (10%) 128 327 466 921 Total (100%) 1281 3270 4657 9208 and the relevant sizes related to these subsets satisfy the following equation.(3) \|R\|=\|X\|+\|Y\|+\|Z\|=\|{x1,...,xj}\|+\|{y1,...,yi}\|+\|{z1,...,zq}\| where \|.\| refers to the cardinality of a set, i is the number of training set images, j is the number of testing set images, q is the number of validation set images. Assuming that there are k MDC MDA technique (k MDC = {k1, k2, k3, k4}, in this paper, k MDC including noise injection, rotation, gamma correction and mirror), and n MDA images are generated using each MDA technique, and finally for all MDAs, k MDC × n MDA images are generated. The following four MDA are mainly used in this study: ① Noise injection (N_I) Gaussian noise was injected into all the images of a training set, thereby generating many new noisy images.(4) yk1(i)→=N_I[y(i)]=[y1k1(1),...,ynMDAk1(i)] ② Rotation (Ro) The rotation angle θ Ro = 90° was applied to the images:(5) yk2(i)→=Ro[y(i)]=[y1k2(1,θRo),...,ynMDAk2(i,θRo)] ③ Gamma correction (G_C) The gamma correction factor rG_C = 1.5 was used to produce new images as follows:(6) yk3(i)→=G_C[y(i)]=[y1k3(1,rG_C),...,ynMDAk3(i,rG_C)] ④ Mirror (Mir)(7) yk4(i)→=Mir[y(i)]=[y1k4(1),...,ynMDAk4(i)] (8) yk(i)→=concatyk1(i)→,yk2(i)→,yk3(i)→,yk4(i)→ where yk(i)→ means the data augmentation is concatenation of four MDA results..(9) y(i)→MDAconcaty(i),yk(i)→ where y(i) means the training set consists of the original and augmentation images. As shown in Fig. 2 , we used four ways to enhance the training set. We can observe that one image will become 5 images.Fig. 2 Four multiple-way data augmentation applied to training set. (a) Noise injection. (b) Rotation. (c) Gamma correction. (d) Mirror. 3.2 Improvement II: Backbone network of EfficientNetV2 In classification problems, to achieve better results, methods that increase the network depth, expand the input image size, and increase the network width are commonly used. However, simply increasing the depth of the network limits the accuracy improvement, as it can easily lead to gradient explosion or gradient disappearance. Besides, the storage requirements increase with an increase in network depth. Additionally, if we simply increase the width of the model, this will allow the model to learn more details. However, if the model is not deep enough, deeper features are not easily learned. Moreover, increasing the resolution of an input image enables the model to acquire more features, but increases the computational cost and reduces the training speed. Accordingly, EfficientNet [35] combines the above-mentioned three trade-off cases to achieve the best result, as demonstrated in Fig. 3 . Fig. 3(a) shows the basic network, while Fig. 3(b-d) improves the performance in terms of increasing the network width, depth, and resolution of the input image, respectively. Finally, Fig. 3(e) illustrates the main idea of EfficientNet, which is to integrate the above three elements to improve the network.Fig. 3 Diagrammatic representation of an EfficientNet architecture. In this paper, we used EfficientNetV2 [36] as the backbone network, which is approximately ten times faster than EfficientNet in training, and has a better performance. Likewise, Fused-MBconv corresponds to the key section of EfficentNetV2, which replaces the 1 × 1 boosted convolution and 3 × 3 depth-wise convolution in MBConv with a normal 3 × 3 convolution to improve the training speed, as demonstrated in Fig. 4 . Meanwhile, EfficientNetV2 adopts a progressive learning strategy, where the overall training process is divided into four stages. Each stage has stronger regularization for faster convergence, fewer parameters, and very high accuracy. The corresponding training speed is 5 to 11 times faster for the same computational resources.Fig. 4 Structure of MBConv and Fused-MBConv. 3.3 Improvement III: Network adds Convolutional Block attention module CBAM (Convolutional Block Attention Module) [37] is a lightweight module that has two sub-modules: the channel attention (CA) module and the spatial attention (SA) module. In the CA module, the average pool is used to summarize the feature information, the maximum pool is used to get the information of unique objects, and finally the channel relationship of features is used to find the desired feature description and generate the CA graph. In the SA module, which is mainly a complement to the CA module, it can use the spatial relationship between features to determine the location of information and get the SA map. Finally, the CA is arranged in series with the SA, which can improve the representation capability of CNN. Residual Neural Network (ResNet) [38] was first introduced by K. He et al. Compared to traditional networks, ResNet has fewer parameters (e.g., VGG), better classification, flexible structure. In this paper, we use EfficientNetV2 and ResNet101 to extract features in parallel. The SE module already exists in EfficientNetV2, and we add CBAM to ResNet101 so that both networks can extract features accurately. Fig. 5 shows the exact placement of the modules when integrated into the ResBlock, with the spatial attention module inside the blue border and the CA module inside the red border. We apply CBAM on the convolution output of each block.Fig. 5 Structure of ResBlock + CBAM. 3.4 Improvement IV: Feature fusion The two commonly used feature-level fusion (FLF) methods are concat and add. Add method corresponds to an increase in information amount for the features describing the image; however, the dimensions describing the image do not increase, as show in Fig. 6 (a). On the other hand, concat method refers to a merger of the number of channels, i.e., the number of channels describing the image increases, while relevant information for each feature stays constant. If the dimensions of the two input features × and y are p and q, the dimension of the output feature z is p + q, as show in Fig. 6(b). The relevant mathematical expressions are given in Eq. (5) and Eq. (6). In this study, we used concat for FLF, as show in Fig. 6(a). The number of Fusion(x_y) channels refers to the sum of Feature(x) and Feature(y) channels.(10) Fflf=concat(fE,fR) (11) Fflf=add(fE,fR) where fE is the feature extracted by EfficientnetV2, fR is the feature extracted by ResNet, and Fflf is the fused features set.Fig. 6 Two feature fusion methods (a) Concat. (b) Add. 3.5 Improvement V: SVM as the classifier SVM (Support Vector Machine) [39] majorly solves the data classification problem in pattern recognition, and describes the data as points in space and maps them into one or more hyperplanes, constructed by kernel functions. The core idea is to find the separation interface between different categories so that the samples of two categories fall on both sides of the face and as far away from the separation interface as possible. This assists in separating the two different categories quickly. Eqs.(12), (13) represent formulas for a line or hyper plane, respectively. The traditional SVM only performs binary classification, while the LibSVM [40] program is small, has few parameters, flexible in use, and can perform multi-way classification with a good generalization. The LibSVM is the core of MDCM, as elaborated in Fig. 7 .(12) w→xf→+b=0 (13) yf=mxf+b where w denotes the normal vector of the hyperplane, which determines the direction of the hyperplane. b is the displacement term, which determines the distance between the hyperplane and the origin. xf is the training sample and yf is the output of the training example.Fig. 7 LibSVM implements triple classification. 3.6 Proposed approach In this paper, we propose a deep feature fusion classification network called DFFCNet, which comprises three major stages. For the first stage, the dataset is preprocessed and the training set is enhanced with MDA using four methods. In the second stage, feature learning is performed using EfficientNetV2 and ResNet101, where the CBAM module is added to ResNet101 to enhance its feature extraction capability. The third stage involves the classification of the fused features using SVM, which allows multi-disease efficient classification. Fig. 8 depicts the overall framework. To elaborate further, a pseudo code for DFFCNet algorithm is given in Algorithm 1.Fig. 8 Structure of the proposed DFFCNet. Algorithm 1. Pseudo code of our DFFCNet algorithm.Phase I: Preprocessing X → Z Step 1 Input: Original Image Set O. Step2 Resizing: Resize the image to [224, 224], get dataset R. See Eq. (1). Step 3 YRHO: testing set (X), training set (Y) and validate set (Z). See Eq. (2). Step 4 MDA(Y):N_I、RO、G_C and Mir to augment training set (Y). Phase II: DFFM Step 5 Read one raw Pre-trained model EfficientNetV2 and ResNet. Step 6 Obtaining MBConv and Fused-MBConv Networks from EfficientNetV2 → M1. Step 7 Adding CA and SA to ResNet → ResNet (CBAM) Step 8 Obtaining residual Networks from ResNet (CBAM) → M2. Step 9 Concat (M1, M2). Step12 Generate DFFM. Phase III: MDCM Step13 Get the fusion feature from DFFM → Fflf. See Eq. (10). Step14 Create data labels based on feature values. Step15 Normalize the feature values. Step16 Construct MDCM by radial basis and SVM cross-validation. Step17 Classification results were obtained by MDCM. Step 18 Test confusion matrix, calculate indicators. Step 19 Output: The model DFFCNet and its performances. 4 Experiments and results 4.1 The experiment platform The experiments are performed in a Linux environment, using an NVIDIA DGX Station deep learning workstation with a 32 GB Tesla V100 graphics card to run the experiments. Python language is used to implement the overall code, i.e., data pre-processing and algorithm implementation. Libraries such as Numpy and the deep learning toolbox Pytorch aree used. The learning rate (LR), batch size (BS), epochs, optimizer, and dropout rate (DR) made up the tuned hyper-parameters of the model. The values produced the optimal experimental results: batch size, epochs, DR, and initial LR are set to 8, 30, 0.4 and 0.003. When the loss is reduced, the LR is reduced to the original value of 0.1. The main factor affecting the results is LR, which is usually set about three times higher or lower, so we choose 0.003, 0.001 and 0.01 for the experiment. The trend of Epoch and Loss is shown in Fig. 9 , where the lower the loss, the better the network performance. When LR is 0.01, the loss does not decrease, but increases, as shown in the red line. The 10th epoch has converged when LR is 0.001 and 0.003. The loss is not minimized when LR is 0.001, as shown in the blue line. The loss is minimized when LR is 0.003, as shown in the green line. So neither LR less than 0.003 nor greater than 0.003 can achieve the best results, so we set LR to 0.003.Fig. 9 Relationship between learning rate and loss. 4.2 Experiment to determine the feature-fusion methods The commonly used FLF methods are concat and add. To prove that concat is the best, we conducted an experimental comparison, we fuse the features extracted by EfficientNetV2 and ResNet101 using concat and add respectively, as shown in Fig. 10 . The accuracy of fusion using the add method was 99.40%, and the accuracy of fusion using the concat method reached 99.51%. So we used the concat fusion in this paper.Fig. 10 Comparison of two fusion methods. 4.3 Ablation study of DFFCNet To determine the effect of each improvement, we performed an ablation study, as shown in Table 2 , the experiment shows the results on the testing set. First we use the backbone network EfficientNetV2 for classification and get the accuracy of 97.23%. EfficientNetV2 obtained the accuracy of 99.51% after feature fusion with ResNet101. Next, after adding CBAM to ResNet, the accuracy is 99.73%. Finally, after we replace the classifier with SVM, the accuracy is 99.89%. It can be seen that and every improvement is effect for DFFCNet, especially for feature fusion.Table 2 Ablation study (%). backbone network Feature fusion CBAM SVM Acc (X) √ 97.23 √ √ 99.51 √ √ √ 99.73 √ √ √ √ 99.89 To demonstrate the training process of DFFCNet, we add Fig. 11 . The changes of test accuracy and training loss are shown as the epoch increases, where the horizontal axis is epoch, training loss corresponds to the left vertical axis and test accuracy corresponds to the right vertical axis. As can be seen from Fig. 11, DFFCNet converges around the 10th epoch, while the training loss reaches 0.001 and the accuracy is already close to 100%, which has a good performance.Fig. 11 Test accuracy and training loss of DFFCNet. 4.4 Experimental results 4.4.1 Classification performance In order to evaluate the performance of the proposed DFFCNet method, we used various metrics in the validation set to determine, namely accuracy (Acc(Z)), precision (Pre(Z)), sensitivity (Sen(Z)), specificity (Spe(Z)), recall (Rec(Z)), and F1-score (F1-sc(Z)). The corresponding equations are expressed below.(14) Acc(Z)=TP(Z)+TN(Z)TP(Z)+TN(Z)+FP(Z)+PN(Z) (15) Pre(Z)=TP(Z)TP(Z)+FP(Z) (16) Sen(Z)=TP(Z)TP(Z)+FN(Z) (17) Spe(Z)=TN(Z)TN(Z)+FP(Z) (18) Rec(Z)=TP(Z)TP(Z)+FN(Z) (19) F1-sc(Z)=2∗Pre(Z)∗Rec(Z)Pre(Z)+Rec(Z) Accordingly, Table 3 demonstrates the overall performance of DFFCNet for the validation set of 921 CXRs. The Acc(Z), Pre(Z), Rec(Z), Sen(Z), Spe(Z) and F1-sc(Z) are 99.9%, 100%, 99.2%, 99.2%, 100% and 99.6% for COVID-19 X-ray images. The Acc(Z), Pre(Z), Rec(Z), Sen(Z), Spe(Z) and F1-sc(Z) are 99.9%, 99.7%, 100%, 100%, 99.8% and 99.8% for pneumonia X-ray images. The Acc(Z), Pre(Z), Rec(Z), Sen(Z), Spe(Z) and F1-sc(Z) are 99.8%, 99.8%, 99.8%, 99.8%, 99.8% and 99.8% for normal X-ray images. In conclusion, the performance in terms of accuracy, sensitivity, recall and F1-score on the validation set is good, so the DFFCNet proposed in this paper is effective.Table 3 The classification of DFFCNet networks after two kinds of validation (%). Class Acc (Z) Pre (Z) Rec (Z) Sen (Z) Spe (Z) F1-sc(Z) COVID-19 99.9 100 99.2 99.2 100 99.6 Normal 99.9 99.7 100 100 99.8 99.8 Pneumonia 99.8 99.8 99.8 99.8 99.8 99.8 4.4.2 Confusion matrix To illustrate the classification of data from the validation set, a confusion matrix [41] is employed in this work. For each class C = 1, 2, 3 (1: COVID-19, 2: Pneumonia, 3: Normal), we set that class tag to “positive” and the other two classes to “negative”. Likewise, Fig. 12 presents a schematic confusion matrix for the three categories. The True positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) are used to identify the diagnosis of CXRs by the model. The TP indicates a positive outcome for both real category of the sample and the recognition result. Similarly, the FN highlights a positive real category of the sample, however, it is identified as negative by the model. Alternatively, FP refers to a negative real category of the sample recognized as a positive by the model. Finally, TN indicates a negative category for both real category of the sample and the recognition result. Fig. 13 show the confusion matrix of the DFFCNet model proposed in this paper.Fig. 12 Confusion matrix of multiple class conditions. Fig. 13 Classification results of the DFFCNet visualized with a confusion matrix. 4.4.3 Explainable deep learning using Grad-CAM++ To enhance the intuitive nature of the features and the interpretability of deep learning model, a technique called gradient-weighted class activation mapping plus (Grad-CAM++), proposed by A. Chattopadhay et al. [42], is adopted in this work. Initially, an image is normalized, and the trained neural network model parameters are loaded. Next, the feature map of the target layer is extracted, and the gradient information of the target class on the feature map is recorded. Next, the heat map is obtained through the weighted summation operation of all the feature maps of a target layer. Eventually, using linear interpolation, the heat map is reduced to the size same as that of original image. The obtained heat map is then superimposed on the original image to complete the visualization operation. The visualization results for the features generated after the last convolution layer of DFFCNet, using the Grad-CAM++ method are shown in Fig. 14 . Among them, the COVID-19 and Pneumonia images possess more obvious features whereas the Normal images have no lesion features.Fig. 14 Grad-CAM++ of the DFFCNet. 4.4.4 Comparison with state-of-the-art approaches To demonstrate the effectiveness of the DFFCNet method, we compared it with eight state-of-the-art methods: ECOVNet [43], Fused-DenseNet-Tiny [23], BCNN_SVM [44], COVNet [45], InceptionV3 [46], DTL-V19 [47], ResNet152V2 [48], and VGG16 [49]. All methods used the unified dataset and MDA preprocessing methods, experiments were performed on the testing set. Table 4 illustrates the relevant comparison results. It can be seen that, among all methods, the proposed DFFCNet achieved the best results. Moreover, accuracy achieved 99.89%. The high accuracy was mainly achieved through the feature fusion and attention mechanism coordination. The use of the newly-proposed EfficientNetV2 as the backbone network and SVM as classifier, the effectiveness of which is demonstrated through the experimental results. In addition, the MDA prevents overfitting of the model, thus improving its performance.Table 4 Performance comparison of the proposed DFFCNet with other studies (%). Method Sen (z) Pre (z) F1-sc (z) Acc (z) ECOVNet [43] 97.53 98.15 97.84 97.72 Fused-DenseNet-Tiny [23] 98.15 98.38 98.26 97.99 BCNN_SVM [44] 96.53 98.06 97.26 97.39 COVNet [45] 95.12 94.34 94.65 95.11 InceptionV3 [46] 98.23 98.31 98.26 97.99 DTL-V19 [47] 95.15 95.66 95.40 95.33 ResNet152V2 [48] 98.09 98.25 98.17 97.88 VGG16 [49] 96.94 97.06 96.97 96.58 DFFCNet (this work) 99.60 99.79 99.70 99.89 ECOVNet [43], Fused-DenseNet-Tiny [23], COVNet [45] and DTL-V19 [47] are the proposed methods for COVID-19 diseases. BCNN_SVM [44], InceptionV3 [46], ResNet152V2 [48] and VGG16 [49] are better classification networks proposed in recent years, and these methods are very representative. Compared to the other methods, the strategy proposed in this work is unique. BCNN_SVM [44] used a BCNN bilinear fusion of two deep learning networks, VGG16 and VGG19, to extract the features, and then used an SVM to classify for the presence of COVID-19. Since both fusion networks were VGG, the extracted features were similar and the relevant accuracy was lower than our DFFCNet. Fused-DenseNet-Tiny [23] used transfer learning to introduce DenseNet as a backbone network and optimized transfer learning to improve the performance by freezing some layers and adding new ones. It used the same dataset as in this paper for experiments and the performance was inferior to DFFCNet. Additionally, DTL-V19 [47] used a deep transfer learning of VGG19 for COVID-19 classification. There were fewer training parameters. However, due to the limited network performance, the network had fewer layers and was prone to overfitting. ECOVNet [43] and COVNet [45] were trained for classification using EfficientNetB3 and ResNet50, respectively. These models were simple, lack the fusion of features, and their performance was inferior to our proposed network. Furthermore, we compared the proposed method with the currently popular classification networks, namely InceptionV3 [46], ResNet152V2 [48] and VGG16 [49], and results indicate that these networks were not as effective as DFFCNet. To better display the results, we have added Fig. 15 .Fig. 15 Comparison of our method with 8 state-of-the-art approaches. 5 Conclusion Coping up with the sudden emergence of COVID-19 virus poses a primary challenge for the medical systems. Due to the lack of doctors and testing reagents, it is difficult to timely diagnose all the potential patients. Nevertheless, the application of AI, which can quickly assist to diagnose diseases through CXRs, saves a lot of time. Likewise, this paper proposed a deep feature fusion efficient classification network (DFFCNet). The proposed network enables an accurate diagnosis of COVID-19, health and pneumonia, especially the prediction accuracy of COVID-19 diseases reached 99.89%. To validate the performance of DFFCNet, we compared the experimental results of 8 state-of-the-art methods. DFFCNet achieved good results in terms of accuracy, precision, sensitivity, F1-score. This helps doctors to make faster and more accurate diagnosis of COVID-19, and thus, our method makes a significant contribution to society and hospitals. Moreover, the proposed DFFCNet suffers from two disadvantages: (1) It does not make a judgment about the grade for COVID-19. (2) It cannot handle the datasets constructed via a mixing of CT and CXR. In our future work, we hope to solve the above problems. 6 Data availability The data that support the findings of this study are openly at [https://data.mendeley.com/datasets/9xkhgts2s6/1], reference number. [34]. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Appendix A .Table 5 Abbreviation list. Abbreviation Full name BS batch size CA channel attention CXRs chest X-ray images CBAM Convolutional Block Attention Module DFFCNet deep feature fusion classification network DFFM deep feature fusion module DR dropout rate FLF feature-level fusion FN false negatives FP false positives G_C gamma correction Grad-CAM++ gradient-weighted class activation mapping plus LR learning rate MDA multiple-way data augmentation MDCM multi-disease classification module Mir mirror N_I noise injection ResNet Residual Neural Network Ro rotation RHO random hold-out SA spatial attention SVM Support Vector Machine TN true negatives TP true positives Acknowledgment This work is supported by the 10.13039/501100013061 Jilin Scientific and Technological Development Program (No. 20200401078GX) and the 10.13039/501100003453 Guangdong Natural Science Foundation (2021A1515011994). ==== Refs References 1 World Health Organization, WHO coronavirus disease (COVID-19) Dashboard, Available, https://covid19.who.int/?gclid=Cj0KCQjwtZH7BRDzARIsAGjbK2ZXWRpJROEl97HGmSOx0_ydkVbc02Ka1FlcysGjEI7hnaIeR6xWhr4aAu57EALw_wcB, 2022. 2 Ataguba O.A. Ataguba J.E. Social determinants of health: the role of effective communication in the covid-19 pandemic in developing countries Global Health Action 13 1 2020 1788263 32657669 3 Pormohammad A. Ghorbani S. Khatami A. Farzi R. Baradaran B. Turner D.L. Turner R.J. Bahr N.C. Idrovo J. Comparison of confirmed COVID-19 with SARS and MERS cases-Clinical characteristics, laboratory findings, radiographic signs and outcomes: a systematic review and meta-analysis Rev. Med. Virol. 30 4 2020 e2112 4 Brenner D.J. Hall E.J. Computed tomography—an increasing source of radiation exposure N. Engl. J. Med. 357 22 2007 2277 2284 18046031 5 Shi F. Wang J. Shi J. Wu Z. Wang Q. Tang Z. He K. Shi Y. Shen D. Review of artificial intelligence techniques in imaging data acquisition, segmentation, and diagnosis for COVID-19 IEEE Rev. Biomed. Eng. 14 2020 4 15 6 Wang S.H. Govindaraj V.V. Górriz J.M. Zhang X. Zhang Y.D. Covid-19 classification by FGCNet with deep feature fusion from graph convolutional network and convolutional neural network Inf. Fusion 67 2021 208 229 33052196 7 Esteva A. Kuprel B. Novoa R.A. Ko J. Swetter S.M. Blau H.M. Thrun S. Dermatologist-level classification of skin cancer with deep neural networks Nature 542 7639 2017 115 118 28117445 8 Guo M.H. Du Y.Z. Classification of thyroid ultrasound standard plane images using ResNet-18 networks IEEE 13th Int. Conf. Anti-Counterfeiting 2019 324 328 9 Lu S. Lu Z. Yang J. Yang M. Wang S. A pathological brain detection system based on kernel based ELM Multimed. Tools Appl. 77 3 2018 3715 3728 10 Li H. Habes M. Wolk D.A. Fan Y. Alzheimer's Disease neuroimaging initiative, A deep learning model for early prediction of Alzheimer's disease dementia based on hippocampal magnetic resonance imaging data Alzheimer’s Dementia 15 8 2019 1059 1070 11 Grewal M. Srivastava M.M. Kumar P. Varadarajan S. Radnet: Radiologist level accuracy using deep learning for hemorrhage detection in ct scans 2018 IEEE 15th International Symposium on Biomedical Imaging 2018 281 284 12 Gulshan V. Peng L. Coram M. Stumpe M.C. Wu D. Narayanaswamy A. Venugopalan S. Widner K. Madams T. Cuadros J. Kim R. Raman R. Nelson P.C. Mega J.L. Webster D.R. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs JAMA 316 22 2016 2402 27898976 13 P. Rajpurkar, A. Y. Hannun, M. Haghpanahi, C. Bourn, A. Y. Ng, Cardiologist-level arrhythmia detection with convolutional neural networks, (2017) arXiv preprint arXiv: 1707.01836. 14 Karthik S. Srinivasa Perumal R. Chandra Mouli P.V.S.S.R. Breast cancer classification using deep neural networks Margret Anouncia S. Wiil U.K. Knowledge Computing and Its Applications 2018 Springer Singapore Singapore 227 241 15 Tandel G.S. Biswas M. Kakde O.G. Tiwari A. Suri H.S. Turk M. Laird J. Asare C. Ankrah A.A. Khanna N.N. Madhusudhan B.K. Saba L. Suri J.S. A review on a deep learning perspective in brain cancer classification Cancers 11 1 2019 111 30669406 16 Arvidsson I. Overgaard N.C. Marginean F.E. Krzyzanowska A. Bjartell A. Åström K. Heyden A. Generalization of prostate cancer classification for multiple sites using deep learning 2018 IEEE 15th International Symposium on Biomedical Imaging 2018 191 194 17 Chen B. Li J. Guo X. Lu G. DualCheXNet: dual asymmetric feature learning for thoracic disease classification in chest X-rays Biomed. Signal Process. Control 53 2019 101554 18 S. Guendel, S. Grbic, B. Georgescu, S. Liu, A. Maier, D. Comaniciu, Learning to recognize abnormalities in chest x-rays with location-aware dense networks, inIberoamerican Congress on Pattern Recognition. Springer, 2018, pp. 757–765. 19 Narayan Das N. Kumar N. Kaur M. Kumar V. Singh D. Automated deep transfer learning-based approach for detection of COVID-19 infection in chest X-rays IRBM 2020 20 Majeed T. Rashid R. Ali D. Asaad A. Issues associated with deploying CNN transfer learning to detect COVID-19 from chest X-rays Phys. Eng. Sci. Med. 43 4 2020 1289 1303 33025386 21 Maghdid H.S. Asaad A.T. Ghafoor K.Z. Sadiq A.S. Mirjalili S. Khan M.K. Diagnosing COVID-19 pneumonia from X-ray and CT images using deep learning and transfer learning algorithms Multimodal Image Exploit. Learn. 11734 2021 117340E 22 Katsamenis I. Protopapadakis E. Voulodimos A. Doulamis A. Doulamis N. Transfer learning for COVID-19 pneumonia detection and classification in chest X-ray images 24th Pan-Hellenic Conf. Informatics 2020 170 174 23 Montalbo F.J.P. Diagnosing Covid-19 chest X-Rays with a lightweight truncated DenseNet with partial layer freezing and feature fusion Biomed. Signal Process. Control 68 2021 102583 24 Aradhya V.N. Mahmud M. Guru D.S. Agarwal B. Kaiser M.S. One-shot cluster-based approach for the detection of COVID–19 from chest X–ray images Cogn. Comput. 13 4 2021 873 881 25 Voulodimos A. Protopapadakis E. Katsamenis I. Doulamis A. Doulamis N. A few-shot U-net deep learning model for COVID-19 infected area segmentation in CT images Sensors 21 6 2021 2215 33810066 26 Chen X. Yao L. Zhou T. Dong J. Zhang Y. Momentum contrastive learning for few-shot COVID-19 diagnosis from chest CT images Pattern Recogn. 113 2021 107826 27 Yang D. Xu Z. Li W. Myronenko A. Roth H.R. Harmon S. Xu D. Federated semi-supervised learning for COVID region segmentation in chest CT using multi-national data from China, Italy, Japan Med. Image Analysis 70 2021 101992 28 A. Voulodimos, E. Protopapadakis, I. Katsamenis, A. Doulamis, N. Doulamis, Deep learning models for COVID-19 infected area segmentation in CT images, The 14th PErvasive Technologies Related to Assistive Environments Conference, 2021, pp. 404-411. 29 X. Chen, L. Yao, Y. Zhang, Residual Attention U-Net for Automated Multi-Class Segmentation of COVID-19 Chest CT Images, (2020) arXiv preprint arXiv: 2004.05645. 30 Saeedizadeh N. Minaee S. Kafieh R. Yazdani S. Sonka M. COVID TV-Unet: Segmenting COVID-19 chest CT images using connectivity imposed Unet Comput. Methods Prog. Biomed. Update 1 2021 100007 31 Zhou T. Canu S. Ruan S. Automatic COVID-19 CT segmentation using U-Net integrated spatial and channel attention mechanism Int. J. Imaging Syst. Technol. 31 1 2021 16 27 33362345 32 Chen H. Jiang Y. Loew M. Ko H. Unsupervised domain adaptation based COVID-19 CT infection segmentation network Appl. Intell. 2021 1 14 33 Liu J. Dong B. Wang S. Cui H. Fan D.P. Ma J. Chen G. COVID-19 lung infection segmentation with a novel two-stage cross-domain transfer learning framework Med. Image Anal. 74 2021 102205 34 U. Sait, K. G. Lal, S. Prajapati, R. Bhaumik, T. Kumar, S. Sanjana, K. Bhalla. Curated Dataset for COVID-19 Posterior-Anterior Chest Radiography Images (X-Rays), Mendeley Data, 1, 2020. 35 Tan M. Le Q. Efficientnet: Rethinking model scaling for convolutional neural networks Int. Conf. Mach. Learn., PMLR 2019 6105 6114 36 M. Tan, Q. Le, Efficientnetv2: Smaller models and faster training, (2021) arXiv preprint arXiv: 2104.00298. 37 Woo S. Park J. Lee J.Y. Kweon I.S. Cbam: Convolutional block attention module Proceedings of the European conference on computer vision (ECCV) 2018 3 19 38 He K. Zhang X. Ren S. Sun J. Deep residual learning for image recognition 2016 IEEE conference on computer vision and pattern recognition(CVPR) 2016 770 778 39 Cortes C. Vapnik V. Support-vector networks Mach. Learn. 20 3 1995 273 297 40 Chang C.C. Lin C.J. LIBSVM: a library for support vector machines ACM Trans. Intell. Syst. Technol. (TIST) 2 3 2011 1 27 41 Chicco D. Tötsch N. Jurman G. The Matthews correlation coefficient (MCC) is more reliable than balanced accuracy, bookmaker informedness, and markedness in two-class confusion matrix evaluation BioData Min. 14 1 2021 1 22 33430939 42 A. Chattopadhay, A. Sarkar, P. Howlader, V. N. Balasubramanian, Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks, 2018 IEEE winter conference on applications of computer vision (WACV), 2018, pp. 839–847. 43 Garg A. Salehi S. La Rocca M. Garner R. Duncan D. Efficient and visualizable convolutional neural networks for COVID-19 classification using chest CT Expert Syst. Appl. 195 2022 116540 44 Mastouri R. Khlifa N. Neji H. Hantous-Zannad S. A bilinear convolutional neural network for lung nodules classification on CT images Int. J. Comput. Assist. Radiol. Surg. 16 1 2021 91 101 33140257 45 Ko H. COVID-19 pneumonia diagnosis using a simple 2D deep learning framework with a single chest CT image: model development and validation Journal of medical Internet research 22 6 2020 e19569 46 Szegedy C. Vanhoucke V. Ioffe S. Shlens J. Wojna Z. Rethinking the inception architecture for computer vision 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas 2016 2818 2826 47 Panwar H. Gupta P.K. Siddiqui M.K. Morales-Menendez R. Bhardwaj P. Singh V. A deep learning and grad-CAM based color visualization approach for fast detection of COVID-19 cases using chest X-ray and CT-Scan images Chaos, Solitons Fractals 140 2020 110190 48 K. He, X. Zhang, S. Ren, J. Sun, Identity mappings in deep residual networks, European conference on computer vision, Springer, Cham, 2016, pp. 630-645. 49 K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, (2014) arXiv preprint arXiv: 1409.1556.
PMC009xxxxxx/PMC9005573.txt	==== Front 0100714 6400 Pediatr Res Pediatr Res Pediatric research 0031-3998 1530-0447 34645953 10.1038/s41390-021-01729-7 nihpa1736148 Article Genetic Determinants of Metabolic Biomarkers and their Associations with Cardiometabolic Traits in Hispanic/Latino Adolescents Kim Daeeun 1 Justice Anne E. 2 Chittoor Geetha 2 Blanco Estela 34 Burrows Raquel 5 Graff Mariaelisa 1 Howard Annie Green 6 Wang Yujie 1 Rohde Rebecca 1 Buchanan Victoria L. 1 Voruganti V. Saroja 7 Almeida Marcio 8 Peralta Juan 8 Lehman Donna M. 9 Curran Joanne E. 8 Comuzzie Anthony G. 10 Duggirala Ravindranath 8 Blangero John 8 Albala Cecilia 5 Santos José L. 11 Angel Bárbara 5 Lozoff Betsy 12 Gahagan Sheila 3 North Kari E. 1 1. Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC 2. Department of Population Health Sciences, Geisinger, Danville, PA 3. Division of Academic General Pediatrics, Child Development and Community Health at the Center for Community Health, University of California at San Diego, San Diego, CA 4. Department of Public Health, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile 5. Department of Public Health Nutrition, Institute of Nutrition and Food Technology (INTA), University of Chile, Santiago, Chile 6. Department of Biostatistics, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC 7. Department of Nutrition and Nutrition Research Institute, University of North Carolina at Chapel Hill, Kannapolis NC 8. Department of Human Genetics and South Texas Diabetes and Obesity Institute, University of Texas Rio Grande Valley, Brownsville, TX 9. Departments of Medicine and Epidemiology and Biostatistics, University of Texas Health San Antonio, San Antonio, TX 10. The Obesity Society, Silver Spring, MD 11. Department of Nutrition, Diabetes and Metabolism, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile 12. Department of Pediatrics, University of Michigan, Ann Arbor, MI Author Contribution D. K. and K. E. N. designed the study and drafted the initial manuscript; E. B. and R. B. collected the data; A. E. J., G. C., M. G., and Y. W., carried out genetic data cleaning; D. K., M. G., A.G.H., Y. W., R. R., and V. L. B. conducted statistical analysis; M. A., J. P., D. M. L., J. E. C., A. G. C., R. D., and J. B. involved in the validation study; D. K., K. E. N., M. G., A. G. H., A. E. J., and G. C. involved in interpretation of the results; All authors revised the manuscript and contributed to the content, and approved the submission and publication of the paper. Corresponding Author: Professor Kari E. North, Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA. kari_north@unc.edu, Postal address: 123 W. Franklin Street, Building C, Suite 421, CB#8050, Chapel Hill, NC 27599-8050, Phone 919-966-2148; Fax 919-966-9800 3 9 2021 8 2022 13 10 2021 04 10 2022 92 2 563571 http://www.nature.com/authors/editorial_policies/license.html#terms Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms Background Metabolic regulation plays a significant role in energy homeostasis, and adolescence is a crucial life-stage for the development of cardiometabolic disease (CMD). Objectives This study aims to investigate the genetic determinants of metabolic biomarkers—adiponectin, leptin, ghrelin, and orexin—and their associations with CMD risk factors. Methods We characterized the genetic determinants of the biomarkers among Hispanic/Latino adolescents of the Santiago Longitudinal Study (SLS) and identified the cumulative effects of genetic variants on adiponectin and leptin using biomarker polygenic risk scores (PRS). We further investigated the direct and indirect effect of the biomarker PRS on downstream body fat percent (BF%) and glycemic traits using structural equation modeling. Results We identified putatively novel genetic variants associated with the metabolic biomarkers. A substantial amount of biomarker variance was explained by SLS-specific PRS, and the prediction was improved by including the putatively novel loci. Fasting blood insulin and insulin resistance were associated with PRS for adiponectin, leptin, and ghrelin, and BF% was associated with PRS for adiponectin and leptin. We found evidence of substantial mediation of these associations by the biomarker levels. Conclusion The genetic underpinnings of metabolic biomarkers can affect the early development of CMD, partly mediated by the biomarkers. ==== Body pmcINTRODUCTION Obesity in early life and subsequent CMD is a major public health concern. In 2015-2016 in the US, one out of five adolescents aged 12-19 years were affected by obesity1 with a prevalence of 25.8% among self-identified Hispanic/Latino (H/L) youth (aged 2-19 years) compared to 14.1% among non-Hispanic white youth.1 Such alarming data are not restricted to the US H/L population, as many self-identified H/L populations in South and Central America are also suffering. For example, Chilean citizens also suffer from an increasing burden of obesity due to the epidemiological and nutritional transition that began in the 1980s.2,3 Energy homeostasis is particularly critical in obesity biology4 since its related hormones play significant roles in balancing energy expenditure and energy intake by exchanging physiological information between the central nervous system and other parts of the body.4,5 In this regard, appetite-related metabolic hormones function as signaling molecules, and malfunction of this system may interrupt energy homeostasis and consequently contribute to the development of obesity. Four such metabolic hormones – adiponectin, leptin, ghrelin, and orexin –are involved in diverse cardiometabolic pathways. Adiponectin is well-known for its protective roles in appetite regulation, energy metabolism, diabetes, and inflammation.6 Leptin shows anorexigenic and proinflammatory roles; individuals with obesity tend to have higher levels of circulating leptin together with leptin resistance.7 Ghrelin is an appetite stimulating gut hormone that has been associated with glucose metabolism, which is closely linked to diabetes.8–11 Orexin is a hypothalamic neuropeptide that regulates feeding behaviors and arousal status12–14 and is ubiquitously expressed in different body sites15, including plasma.16 There is a paucity of information on the genetic determinants of metabolic biomarkers, particularly among adolescents and in H/L. Such work is critical as H/L have continental ancestry and admixture that is different than that found in European populations. Moreover, the interrelations among the genetic underpinnings of metabolic biomarkers, phenotypic variability, and downstream CMD risk factors (e.g., body fat percent (BF%) and insulin resistance (IR)) among adolescents have also been understudied. Previous studies have revealed that some obesity-associated genetic factors are associated with appetite regulation17,18 and mendelian randomization studies have suggested causal relationships between metabolic biomarkers and CMD risk factors. However, studies focusing on the mediating roles of metabolic biomarkers in the relationship between the genetic underpinnings of metabolic biomarkers and CMD risk factors are lacking. We sought to address these important gaps in the literature by first, describing the known and novel large-effect genetic determinants of the four metabolic biomarkers among H/L adolescents (Figure S1). Second, assessing the influence of aggregated genetic contributions to each biomarker by constructing polygenic risk scores (PRS) based on both known and novel genetic loci. Lastly, we aimed to investigate the relationship between the aggregated effects of biomarker-influencing genetic factors on obesity and glycemic traits, to estimate the degree to which this association is mediated by these biomarker levels. Investigating these relationships among adolescents with a high burden of obesity could provide critical insight into the biological mechanisms of obesity and downstream CMD during adolescence. METHODS Study population The Santiago Longitudinal Study (SLS) originally began as a preventive trial for infancy iron deficiency anemia (IDA), funded by National Institutes of Health (NIH-R01 HD014122).19 From 1991 to 1996, a total of 1,792 infants were recruited at community clinics in lower/middle-class neighborhoods, in Santiago, Chile. Inclusion criteria for the infancy study were term singleton birth, vaginal delivery, birth weight ≥3kg, and absence of major perinatal health problems.19 Follow up studies occurred at 5, 10, and 16 years and included anthropometric measurement, psychosocial data, and developmental assessments.20,21 At 16 years, 679 adolescents completed a study of risk for obesity and CMD that targeted participants with the most comprehensive data in childhood. Data collected included anthropometric measures, cardiovascular risk and metabolic biomarkers from fasting blood samples, and self-reported health-related behaviors. Complete phenotype and genotype data were available for 543 (80%) of these participants. The study was reviewed and approved or found to be exempt by Institutional Review Boards of the University of California at San Diego, University of Michigan, University of North Carolina at Chapel Hill, and the Institute of Nutrition and Food Technology, University of Chile. Genetic data and quality control Participants were genotyped on the Illumina Multiethnic Genotyping Array with imputation to the 1000 Genomes Phase III AMR reference panel. Quality control included individual call rate of > 90%, assessment of sex mismatch, relatedness, and ancestry outliers. Single nucleotide polymorphisms (SNPs) with effect allele frequency (EAF)<0.05, indels, and imputation quality score<0.5 were excluded, resulting in ~6 million SNPs that were assessed for their association with the biomarkers. Measurement Metabolic biomarkers. Fasting blood samples were obtained, stored at −80°C, and analyzed. Enzyme-linked immunosorbent assay (ELISA) was used to measure adiponectin and leptin levels (R&D Systems, Minneapolis, MN and DRG International, Inc., New Jersey, NJ, respectively). The radioimmunoassay (RIA) technique was utilized to quantify ghrelin and orexin-A levels (Phoenix Pharmaceuticals, Inc., Burlingame CA). BF%. Dual energy x-ray absorptiometry (DEXA) scan was used to measure body fat mass (Lunar Prodigy Corp. Software: Lunar iDXA ENCORE 2011, Version 13.60.033, Madison, WI). BF% was calculated as [100(%) × fat mass(kg)/total body mass (kg)]. Glycemic traits. From overnight fasting blood, glucose and insulin levels were quantified with an enzymatic colorimetric assay (QCA S.A., Amposta, Spain) and radioimmunoassay (RIA DCP Diagnostic Products Corporation, LA), respectively. In addition, IR was assessed based on continuous measure of homeostatic model assessment of insulin resistance (HOMA-IR; calculated as [(glucose (mg/dL) × insulin (μUI/dL))/405]; a value of HOMA-IR ≥ 2.6 can be diagnosed as insulin resistance.).22 Analytical approach Genome-wide association tests. We regressed four metabolic biomarkers on SNPs using SUGEN 23, assuming an additive genetic model and adjusting for sex and population substructure using the first 5 principal components (PCs). PCs were constructed by EIGENSTRAT24 using genetic information of the participants. Serum adiponectin, ghrelin, and orexin levels were natural log-transformed before regression analyses. Due to a detection limit issue in measuring leptin levels, rank-based normalized residuals of a Tobit regression model adjusting for sex were used in the genome-wide association analyses for leptin levels. To correct for multiple testing, we considered SNPs with p-value <5 × 10−8 as demonstrating genome-wide significance, and those with a p-value <5 × 10−6 as demonstrating suggestive significance. We identified putative novel loci and reported the lead SNP from all 1 MB regions of the genome-wide significant and/or suggestive significant associations where no previous GWAS signal had been reported. Validation of the GWAS findings. For adiponectin and leptin levels, we investigated if the SNPs identified from SLS demonstrated similar associations in a separate Mexican American validation data set (see Supplementary Material for more information). We considered a signal to be validated if the same SNP was associated with the metabolic biomarker level at a Bonferroni-corrected significance level (p <0.05/the number of SNPs tested for validation for each biomarker) with directional consistency. Functional interrogation of putative novel signals. Among validated SNPs, we interrogated the potential candidate genes in each locus, selected based on 1) the evidence of functional link to a genetic locus (SNP) and 2) the previous reports on plausible gene functions. To examine the functional connection between a SNP and a gene, we queried the expression quantitative trait loci (eQTLs) from NIH Genotype-Tissue Expression (GTEx) project25 and biologically plausible gene functions using databases such as PubMeda, Online Mendelian Inheritance in Manb, and GeneCardsc. Transferability of known associations. We described the transferability of known associations for adiponectin and leptin in SLS, defined when the association was reported with the exact SNP at nominal statistical significance (p <0.05) and directional consistency. As of 12/15/2020, there were 48 and 15 previously reported SNPs associated with adiponectin and leptin, respectively, at the genome-wide significance level. There were no previously reported SNPs associated with ghrelin or orexin levels at genome-wide significance. Constructing PRS of adiponectin and leptin based on known and putative novel loci. Given our small sample size and our desire to assess the evidence of mediated effects of genetic variants for these biomarkers on obesity and glycemic traits, we assessed the aggregated effects of previously reported genetic loci for adiponectin and leptin levels by constructing PRS. Since the previously reported SNPs were tag SNPs primarily identified in studies of Europeans, we considered three different approaches to PRS construction: 1) PRS constructed only by the previously reported SNPs (PRSReported), 2) PRS constructed by the SLS-specific tagging SNPs of the known loci (i.e., the lowest p-value within +/− 500kb of the previously reported SNPs) (PRSSLS), and 3) PRS constructed by both SLS-specific previous signals and putatively novel signals from the current study (PRSSLS+Novel). PRSReported was calculated by summing the number of biomarker-increasing alleles of the previously reported SNPs. If there were more than two SNPs in linkage disequilibrium (LD) (AMR r2>0.2), we only included one tag SNP that was directionally consistent and had the lowest p-value among directionally consistent SNPs. If any of them were not directionally consistent, we included the SNP with the lowest p-value. PRSSLS was defined as the number of biomarker-increasing alleles of the SLS-specific SNPs with the lowest p-values from the current study among variants within +/− 500kb of the known SNPs. When we observed correlated SLS-specific SNPs (AMR r2>0.2), we only included the SNP with the lower p-value for the association with the biomarker. PRSSLS+Novel was derived by summing up the PRSSLS and the number of biomarker-increasing alleles of the SNPs with genome-wide significance or suggestive significance from the current study. Investigating the associations between the adiponectin and leptin levels and the constructed PRS for the biomarker in the H/L population. We evaluated the aggregated genetic effects on adiponectin and leptin levels among SLS by regressing each biomarker (for adiponectin, ghrelin, and orexin, natural log-transformed values; for leptin, rank-based inverse normalized residuals of a Tobit regression adjusting for sex) on each PRS assuming a linear relationship and assessing prediction model by comparing R2 (or adjusted R2 for the models with covariates). Specifically, we compared four separate models for each biomarker as follows: a model only including PRS of the biomarker (Model 1), a model additionally including sex as a covariate (Model 2), a model additionally including the first five PCs as covariates (Model 3), and a model additionally including the BMI Z-score (Model 4). Investigating the relationship between the aggregated effects of biomarker influencing genetic factors on BF% and glycemic traits. Adjusting for sex and the first five PCs, we estimated the total effect of the metabolic biomarker PRS on BF% and glycemic traits (fasting blood glucose (FBG), fasting blood insulin (FBI), and IR) and the percent mediated through the biomarker levels using causal mediation analysis (PROC CAUSALMED) in SAS 9.4 (SAS Institute Inc., Cary, NC).26,27 The proposed pathways, both direct and indirect, from PRS to BF% and glycemic traits are shown in Figure S2. For adiponectin and leptin levels, all three types of PRS were investigated; for ghrelin and orexin levels only those suggestive signals identified in the GWAS herein were used. All pathways, including those to the biomarker and those to the cardiometabolic trait, were adjusted for sex and the first 5 PCs. To further account for the potential heterogeneities by sex, sex-stratified sensitivity analyses were conducted. In each stratified analysis, age and the first five PCs were included as covariates. For leptin, rank-based inverse normalized residuals of leptin from a Tobit regression adjusting for age were included as mediator. And for the other biomarkers, natural log-transformed biomarker levels were used as mediators. Statistical significance was determined using Bonferroni-corrected p-values. The results of sex-stratified analyses were presented in the supplementary tables (Table S12 – Table S15). RESULTS Descriptive statistics. Table 1 reports the descriptive statistics from the 543 SLS participants (259 females) included in the current analyses. The average levels of adiponectin, leptin, ghrelin, and orexin were 11.3μg/mL (SD: 5.3), 11.7ng/mL (SD:13.5), 239.6pg/mL (SD: 151.0), and 16.7pg/mL (SD: 4.2), respectively. The descriptive characteristics of the validation sample are shown in Table S1. Identification of novel signals. No SNP-metabolic biomarker associations displayed genome-wide statistically significant evidence of association. However, 8, 5, 13, and 8 loci displayed suggestive associations with adiponectin, leptin, ghrelin, and orexin, respectively. The lead SNPs, nearest genes, and corresponding p-values for adiponectin, leptin, ghrelin, and orexin are reported in Table 2. Validation of putative novel signal. Ten of 13 adiponectin- or leptin-associated SNPs were available in the validation set. Of the examined genetic markers for validation (6 for adiponectin and 4 for leptin), only the marker rs12066716 was found to be significantly associated (Bonferroni-corrected p <0.00625 (=0.05/8)) with adiponectin with directional consistency [β(±SE) in the discovery set: −0.181 ± 0.039, β(±SE) in the validation set: −0.191 ± 0.065], The other 9 signals for adiponectin and leptin level were directionally consistent but not statistically significant (Table S2). Functional interrogation of putative novel signals. For our validated adiponectin-associated SNP (rs12066716), we identified a candidate gene demonstrating both functional links to an index SNP and plausible biological functions in determining adiponectin level. The index SNP (rs12066716) is an eQTL for C1QTNF12 (C1q/TNF-related protein 12, CTRP12) in visceral adipose tissue (p for eQTL = 8.1 × 10−5)d. CTRPs, including CTRP12 (the product of C1QTNF12), are a family of adiponectin paralogs,28–30 thus the association between the gene and adiponectin is highly feasible. Transferability of known associations. For adiponectin, a total of 48 different SNPs have been reported for association with adiponectin from 11 studies31–41, 44 of which were available in the current study (Table S3). Thirty-five of these 44 SNPs displayed directionally consistent associations with adiponectin levels, 8 of which displayed nominal statistical significance (p<0.05). Fifteen variants have been associated with leptin levels with genome-wide significance from 3 studies42–44; 13 of these variants were available in the current study (Table S4). Six of these 13 associations were directionally consistent, but none were nominally significant (p<0.05). Aggregated genetic influence on adiponectin and leptin levels Adiponectin. As some of the 44 available known SNPs are in LD with one another, we selected 29 independent SNPs (AMR LD r2<0.2) and constructed PRSReported for adiponectin (Table S5). We also identified 24 independent SLS-specific tagging SNPs in the known loci by selecting the SNPs with lowest p-value for the association with adiponectin levels within +/− 500kb of the reported SNPs (Table S7). All the 24 SLS-specific tagging SNPs were independent of each known SNP (AMR LD r2<0.2). Using these 24 SLS-specific tagging SNPs for the known loci, we generated PRSSLS. Lastly, we additionally summed up the number of adiponectin-increasing alleles of the 8 putative novel variants from the PRSSLS and generated adiponectin-PRSSLS+Novel. The prediction results of the natural log-transformed adiponectin levels by PRS were reported in Table 3. While the proportion of variance explained by PRSReported was 0.035, the proportion of variance explained by PRSSLS was 0.253 (Table 3). We further estimated the proportion of variance explained by PRSSLS+Novel from the current study as 0.367. While the estimated variance explained is likely to be biased upward because of winner’s curse, it is obvious that the SLS-specific tagging SNPs better characterized the genetic effects of these loci on this biomarker (also below for leptin) for our SLS population. Leptin. Among the 13 available known SNPs, we selected 7 independent SNPs (AMR LD r2 <0.2) and constructed PRSReported for leptin levels (Table S6). We also identified 8 SLS-specific tagging SNPs of the known loci – i.e., SNPs with the lowest p-values within +/− 500kb of the known variant – and constructed PRSSLS for leptin levels (Table S8). All the 8 SLS-specific tagging SNPs are independent of the known SNPs (AMR LD r2<0.2). Then, we summed up the number of leptin-increasing alleles of the 5 putative novel variants from the PRSSLS and calculated PRSSLS+Novel. We predicted the rank-based normalized residuals of Tobit regressed leptin level (sex as an explanatory variable) by using the three PRS (Table 3). Whereas the variance explained by PRSReported was 0.001 (and the regression coefficient was not significant with a significance level of 0.05), the variance explained by PRSSLS was 0.082 and by PRSSLS+Novel was 0.206 (Table 3). It is possible that the R2 for leptin was lower than for adiponectin partly due to fewer previous GWAS of leptin than adiponectin. Effects of the biomarker-associated genetic determinants on downstream BF% and glycemic traits Adiponectin. FBI (p=0.0037), IR (p=0.0064) and BF% (p<0.0001) were significantly associated with PRSSLS+Novel (Table 4). For FBI and IR, significant indirect effects (i.e., mediated through adiponectin levels) were noted. The mediated effects of adiponectin PRSSLS+Novel on FBI and IR accounted for 79.0% (95% CL: 2.4, 155.5; p=0.04) and 83.2% (95% CL: −0.5, 166.8; p=0.05) of the overall effects, respectively. For BF%, we also found evidence of mediation, but the proportion of mediated effects over the total effects was only 46.7% (95% CL: 2.4, 90.9; p=0.04). Given the wide confidence interval of these estimates, point estimate should be interpreted with caution. In terms of direction of effects, both directly and indirectly the adiponectin-increasing effects appeared to be metabolically protective (decrease in FBI, IR, and BF%). For PRSSLS, the total effects on FBI and BF% were estimated as being significant as for PRSSLS+Novel, but for PRSReported, none of them was significantly affected by the PRS (Table S9). Leptin. FBI (p=0.0014), IR (p=0.0043), and BF% (p<0.0001) were significantly affected by the leptin PRSSLS+Novel (Table 4). The total effect of leptin-increasing PRS was associated with poorer metabolic outcomes (i.e., increase in FBI, IR, and BF%). For FBI and IR, the estimate of the direct effect of PRS on leptin was negative and the estimate of the indirect effect of PRS on the CMD health outcome through leptin was positive. Therefore, the indirect pathway through leptin accounted for most of this association between FBI and IR. For BF%, where both direct and indirect associations were positive, the proportion of mediated effects was 96.9% (95% CL: 65.2, 128.5; p<0.01). For PRSSLS, only the total effect on BF% was significant (also associated with worse metabolic outcomes); none of the hypothetical downstream BF% and glycemic traits were significantly associated with PRSReported (Table S10). According to the results from sex-stratified analyses, the total effects of leptin PRSSLS+Novel on FBI and IR are largely driven by females, however, the indirect effects mediated through leptin are comparable–in terms of direction, magnitude, and significance – among females and males (Table S13). Ghrelin. FBI (p<0.0001) and IR (p<0.0001) were significantly affected by the ghrelin PRS based on the putative novel variants from the current study (Table 4) with ghrelin-increasing PRS associated with overall decreases in IR. As with leptin, we found differences in the direction of the association between direct and indirect effects. Higher PRS were indirectly related with lower FBG (−0.223 [95% CI: (−0.286, −0.161)]) and IR (−0.232 [95% CI: (−0.295, −0.169)]), and the total association was similar in magnitude even after accounting for the direct effects. Orexin. None of the cardiometabolic traits were affected by the orexin-increasing PRS (Table 4). DISCUSSION The identification of genetic mechanisms influencing metabolic biomarkers has the potential to identify important pathways for obesity and its downstream consequences. Yet, the bulk of obesity research has focused on how genes relate to BMI and waist circumference, instead of on relationships with eating behaviors and/or mechanistic biomarkers which may have early influences on the development of obesity, through appetite for example. Further still, most extant data has primarily focused on homogeneous middle-aged adults, with very few genetic studies on culturally and ancestrally diverse adolescents. The current study identified 34 suggestive novel genetic signals associated with four metabolic biomarkers and validated one novel signal associated with adiponectin level. In addition, we revealed that associations between biomarkers and aggregated genetic effects based on known variants were improved when study-specific tag SNPs for known loci and novel variants were included in the PRS. This implicates the need for ancestry-specific studies to validly capture aggregated genetic effects across populations. Lastly, we demonstrated that metabolic biomarker-influencing genetic factors were also associated with some CMD risk factors, especially with FBI and IR (for adiponectin, leptin, and ghrelin) and BF% (for adiponectin and leptin), implicating causal roles of these biomarkers in obesity and IR as early as adolescence. Although our study did not demonstrate novel genome-wide significant signals for appetite markers, likely due to insufficient power, 34 putative novel signals with suggestive evidence for association with metabolic biomarkers were mapped, with the rs12066716 adiponectin association validated in a H/L adult population (p<0.00625) despite the age differences between the two study populations. While the SNP is intronic to TTL10, it is associated with gene expression in several other nearby genes in adipose tissue. Notably, rs12066716 is an eQTL for C1QTNF12, which encodes CTRP12 (also called adipolin), a member of the family of adiponectin paralogs.28,30,45,46 CTRP12 improves insulin sensitivity29 and has been shown to correlate with adiponectin levels (r=0.34) 47 In addition, CTRP12 levels are currently being evaluated as a diagnostic biomarker for T2D48 Finally, other studies have demonstrated that CTRP12 levels were lower among patients with coronary artery disease.47 We found that little to none of the variance in adiponectin and leptin levels were explained by PRS based on the exact reported SNPs from the literature, likely due to the ancestral difference between our population (i.e., mainly European vs. H/L) and published studies. Once we substituted SLS-specific tagging SNPs of the known loci for the reported SNPs, the prediction performance of the PRS improved substantially (R2 0.035 to 0.253 for adiponectin and 0.001 to 0.082 for leptin). We conducted permutation testing to place our adiponectin-PRSSLS findings in context. We randomly selected 24 SNPs among all SNPs that displayed significant association (p<0.05) with adiponectin in our GWAS. We then constructed the PRS (the number of risk alleles) from each these SNPs and calculated the R-square values from each model. Repeating this simulation 1000 times, we observed the distribution of R-square values (maximum: 0.220; range: 0.125 – 0.220; Figure S4). All R-square values were of smaller magnitude than our adiponectin-PRSSLS estimate (0.253). Thus, our results demonstrate that the proportion of variation explained by adiponectin-PRSSLS was greater than chance, likely an overestimation as well. This finding demonstrates the need for valid ancestry-specific tagging SNPs for PRS. Furthermore, since we identified additional improvement of PRS’ performance after incorporating the putative novel signals, large-scale studies for diverse populations to discover additional genetic loci associated with metabolic biomarker are warranted. However, it should be noted that the predictions by PRSSLS and PRSSLS+Novel were likely to be overestimated because of the overlap between the discovery set of the associations and the validation set for the association. Nonetheless, as illustrated in our previous study49, the current findings also underscore the importance of considering genetic diversity across different cohorts, and the influence of this diversity on unique underpinnings to disease in future application of PRS to diverse ethnic groups.50 For adiponectin, leptin, and ghrelin, PRS of biomarkers were simultaneously associated with FBI, IR and BF% (except for ghrelin) among H/L adolescents. These common associations may suggest shared genetic pathways, at least to a certain degree, between metabolic biomarkers and CMD as early as adolescence. Many of the significant overall effects included significant mediated effect by metabolic biomarkers (FBI, IR, and BF% for leptin and adiponectin, and FBI and IR for ghrelin), and this implies that each metabolic biomarker might be somewhere on the causal pathways from the biomarker-determining genetic factors to the CMD risk factors—i.e., the metabolic biomarkers are causally associated with the CMD risk factors—if our assumption on the hypothetical causal relationship (see Figure S2) is valid. Notably, FBI and IR are closely associated with PRS for adiponectin, leptin, and ghrelin through indirect effects. Since IR is considered a root cause of various CMD 51, it is crucial to elucidate the roles of these metabolic biomarkers in developing IR. For the significant mediated effects on BF% for adiponectin and leptin, our findings also suggest an early causal effect of adiponectin and leptin on body fat, suggesting an importance of these biomarkers for the development of obesity at a very young age. Such findings may have relevance for the timing of planning interventions. The observed significant indirect effects on FBI, IR, and BF% support the protective roles of adiponectin in cardiometabolic health. While inverse associations of plasma adiponectin with IR and adiposity have been well established 52,53, the findings from mendelian randomization studies to assess the causal relationships between adiponectin and IR (or insulin sensitivity) and between adiponectin and BMI have been contradictory 54–57 or unsupported 58. Such findings may be related to weak genetic instruments, population stratification bias, genetic pleiotropy, and LD. Although our analysis has not formally assessed causal relationships between these biomarkers, our data support a potential influence of adiponectin on FBI, IR, and BF% by the presence of significant indirect effects of adiponectin-associated genetic factors mediated by adiponectin levels. In addition, for BF%, the effect sizes of the direct (β = −0.072 despite not meeting the significance criteria; p = 0.084) and indirect effect (β = −0.063, p = 0.018) were comparable. This implies the presence of other pathways from the adiponectin-associated genes to BF% independent of adiponectin. Further studies investigating potential common pathways shared between BF% and adiponectin levels are needed. For the downstream effect of leptin-influencing genetic variants on adiposity, a previous study reported that a leptin-decreasing allele (rs17151919-A) was associated with higher BMI during early childhood (under 8-year) 44 However, our results demonstrated a leptin-increasing PRS was associated with higher BF%. Such results may be understood in the context of distinct leptin effects across the life-course and in the context of poor cardiometabolic health. Although leptin typically inhibits appetite and increases energy expenditure59, individuals with poor cardiometabolic health tend to have higher circulating leptin levels due to low leptin sensitivity60 – i.e., leptin resistance. Since the previous study and our results focused on early childhood and late adolescence, respectively, the opposite direction of the association between leptin-influencing genetic variants and adiposity might have been driven by the development of leptin resistance. For ghrelin, similar patterns with adiponectin were demonstrated except for BF%. This implicates the potential protective roles of circulating ghrelin in cardiometabolic traits especially for insulin biology. In line with this finding, many previous studies including a meta-analysis61 on the association between ghrelin and IR in obesity reported the negative correlation among people with obesity. Thus, our study adds to the body of literature that supports ghrelin’s role in glucose metabolism, but the mechanisms by which circulating ghrelin levels play roles in regulating insulin sensitivity need to be further studied. Overall, our results contribute to the mechanistic evidence of obesity-risk genes influencing adiposity via the appetite regulatory system, as previously shown for monogenic obesity disorders, which, without exception, involve disturbances of appetite leading to severe early-onset obesity.62 A substantial evidence base of prospective studies links impaired satiety mechanisms to excessive weight gain63–65, and bivariate twin analyses are consistent with common genetic pathways underlying satiety responsiveness and weight in infancy66. This suggests that genetically susceptible individuals are particularly vulnerable to the abundance of highly palatable food in the modern obesogenic environment. The major strength of this study is the availability of four metabolic biomarkers in an adolescent H/L population at high risk for obesity and downstream CMD. In particular, genome-wide studies of ghrelin and orexin are rare. For adiponectin and leptin, we also leveraged the known information from the previous studies to assess the aggregated genetic effects. Lastly, we extensively investigated cross-trait associations of biomarker-influencing genetic loci with other CMD risk factors. There are notable limitations to our study, including the modest sample size and the lack of availability of studies to replicate our findings. Nine out of 10 putative novel signals for adiponectin and leptin level were not replicated from the validation data set, which could be related to the distinct admixture patterns across Mexican Americans and Chileans, the differences in age across the data sets, or small sample sizes. In addition, PRSSLS and PRSSLS+Novel may have overestimated the aggregated genetic effects and cross-trait associations with BF%, FBI, and IR. Furthermore, as analyses were conducted cross-sectionally, it is difficult to determine temporality among traits, e.g., metabolic biomarkers and cardiometabolic traits, and the possibilities of reverse causation between biomarker level and cardiometabolic traits. In conclusion, our study identified several putatively novel genetic variants associated with the metabolic biomarkers with substantial phenotypic variance explained by SLS-specific PRS. We also demonstrated that some of the aggregated genetic factors may be directly linked to BF%, FBI, and IR or mediated through metabolic biomarkers. Our findings reinforce a need for longitudinal analyses to confirm the genetic determinants regulating metabolic homeostasis and their further influence on cardiometabolic disorder development. From a public health standpoint, such findings are critical, as once cardiometabolic health is established in adolescence, it is very difficult to reverse. Thus, findings from this study yielded important information on biological mechanisms and candidates for prevention efforts, especially relevant in this high-risk ancestrally diverse population. Supplementary Material 1 Statement of Financial Support This work was funded in part by University of North Carolina Nutrition Research Institute internal pilot grant, AHA grant 15GRNT25880008 and NIH award K99/R00HL130580-02. We thank the participants and their family members from the Santiago Longitudinal Study (SLS) (R01 HL088530, R01 HD33487). Dr. North is additionally supported by R01HL151152 and R01 DK122503. Work for the validation study was supported in part by National Institutes of Health (NIH) grants P01 HL045522, R01 DK047482, DK053889, R01 HL113323, R37 MH059490, and T2D-GENES Consortium grants (U01 DK085524, U01 DK085584, U01 DK085501, U01 DK085526, and U01 DK085545). We thank the participants of the San Antonio Family Heart Study and the San Antonio Family Diabetes/Gallbladder Study for their continued cooperation and participation in our research programs. Table 1. Distributions of variables among 16-year follow-up of the Santiago Longitudinal Study (SLS) Variable Total (N=543)a Mean SD Age (years) 16.8 0.3 Anthropometric variables Body Mass Index (BMI) (kg/m2) 23.8 4.6 BMI Z-scores 0.5 1.0 Metabolic biomarkers Adiponectin (μg/mL) 11.3 5.3 Leptin (ng/mL)b 11.7 13.5 Ghrelin (pg/mL) 239.6 151.0 (N=542; 1 missing female) Orexin (pg/mL) 16.7 4.2 Glycemic traits Fasting blood glucose (mg/dL) 88.4 9.8 Fasting blood insulin (μUI/dL) 8.1 5.6 Insulin resistance [HOMA-IRc (glucose × insulin/405)] 1.8 1.3 Body fat percent (%) (N=537; 5 missing females and 1 missing male) 28.9 10.7 a 259 females (47.7%) and 284 males (52.3%) b Leptin levels in females: 18.8 ng/mL (SD: 14.7); leptin levels in males: 5.2 ng/mL (SD: 8.0) c HOMA-IR: homeostatic model assessment of insulin resistance Table 2. Lead SNPs with suggestive significance (p < 5.00E-06) associated with four appetite/metabolic biomarkers from the GWA analyses in participants of SLS GWAS Traits Nearest gene Lead SNP CHR BP EA/OA EAF BETA SE p-value Adiponectin TTLL10 rs12066716a 1 1123434 A/T 0.17 −0.181 0.039 3.04E-06 AK5 rs59559185 1 77825925 A/G 0.18 0.182 0.039 2.55E-06 LRRC8C rs17130858 1 90167923 G/A 0.25 0.158 0.033 1.38E-06 C3orf31 rs407484 3 11944316 A/T 0.10 −0.271 0.053 4.00E-07 VCAN rs33601 5 82773175 A/G 0.66 0.144 0.031 4.94E-06 CLDN10 rs2992893 13 96044769 C/T 0.52 −0.148 0.030 7.02E-07 FLJ22447 rs17098985 14 62092745 T/C 0.13 −0.236 0.045 1.29E-07 PNMAL1 rs11083829 19 46967169 C/A 0.74 −0.174 0.036 1.48E-06 Leptin ZNF804B rs34056816 7 89190522 C/T 0.06 0.748 0.161 3.47E-06 SQRDL (SQOR) rs8026541 15 46695491 T/C 0.72 0.367 0.069 1.14E-07 CDH5 rs233521 16 66282923 A/G 0.80 0.350 0.076 4.29E-06 LOC100506172 rs77137714 16 73378933 T/C 0.14 0.409 0.088 3.70E-06 MIR4739 rs62063332 17 77635587 G/A 0.09 −0.567 0.124 4.64E-06 Ghrelin LOC100129138 rs12144587 1 106005238 G/T 0.56 −0.154 0.033 4.07E-06 ADAM30 rs76491617 1 120434349 C/T 0.15 −0.239 0.051 2.80E-06 PSEN2 rs67899970 1 227092656 T/C 0.14 0.228 0.049 4.05E-06 CNTNAP5 rs181846169 2 124237624 T/C 0.05 0.419 0.087 1.24E-06 KCNH8 rs4535245 3 19541306 A/G 0.14 −0.222 0.048 3.86E-06 NXPH1 rs6966968 7 8840378 G/A 0.10 0.295 0.059 5.28E-07 R4B11FIP2 rs34988394 10 119738984 T/C 0.16 0.234 0.050 2.83E-06 CADM1 rs4938201 11 115218713 G/A 0.39 −0.162 0.035 3.31E-06 RPSAP52 rs11175889 12 66117678 G/T 0.32 0.173 0.036 1.86E-06 CDH11 rs17385192 16 63892792 G/T 0.24 0.184 0.039 2.05E-06 FERMT1 rs6076964 20 6210743 T/C 0.06 0.341 0.075 4.65E-06 CRYAA rs11701620 21 44601373 G/C 0.06 0.346 0.073 1.84E-06 LOC100271722 rs12159453 22 46433442 A/G 0.08 −0.305 0.064 1.58E-06 Orexin PPAP2B rs944844 1 56891592 G/A 0.16 −0.099 0.021 2.30E-06 SLC4A5 rs3771728 2 74503190 C/T 0.08 −0.126 0.027 3.62E-06 ABCA11P rs60941356 4 407189 T/C 0.08 −0.132 0.029 4.22E-06 GABRB3 rs8036016 15 26977581 T/C 0.17 0.108 0.020 6.83E-08 FLJ45079 rs56201168 17 75734152 C/T 0.06 −0.162 0.033 7.10E-07 RNF152 rs34264102 18 59468629 G/A 0.07 0.148 0.031 2.46E-06 SIK1 rs11702068 21 44768217 T/C 0.19 −0.093 0.020 2.27E-06 MIAT rs9608521 22 27199237 T/C 0.18 −0.089 0.019 3.18E-06 SLS, Santiago Longitudinal Study; CHR, chromosome; BP, base pair; EA, effect allele; OA, other allele; EAF, effect allele frequency; SE, standard error a Validated signal (In the validation set: β = −0.191 ± 0.065, P = 0.003, EA = A, and EAF = 9%) Table 3. Associations of adiponectin and leptin levels with PRSReported, PRSSLS, and PRSSLS+Novela in SLS participants Biomarkerb Modelc PRSReported PRSSLS PRSSLS+Novel Adiponectin Betad SEd Pd R2 Adj. R2 Beta SE P R2 Adj. R2 Beta SE P R2 Adj. R2 1 0.027 0.006 <0.0001 0.035 0.034 0.089 0.007 <0.0001 0.253 0.252 0.080 0.005 <0.0001 0.367 0.366 2 0.028 0.006 <0.0001 0.092 0.089 0.088 0.006 <0.0001 0.305 0.302 0.080 0.004 <0.0001 0.421 0.419 3 0.028 0.006 <0.0001 0.101 0.089 0.090 0.006 <0.0001 0.324 0.316 0.082 0.004 <0.0001 0.443 0.436 4 0.029 0.006 <0.0001 0.167 0.155 0.085 0.006 <0.0001 0.352 0.342 0.079 0.004 <0.0001 0.463 0.455 Leptin 2 −0.017 0.025 0.4919 0.001 −0.001 0.190 0.027 <0.0001 0.082 0.080 0.231 0.020 <0.0001 0.206 0.204 3 −0.015 0.025 0.5387 0.018 0.007 0.189 0.027 <0.0001 0.098 0.088 0.230 0.019 <0.0001 0.220 0.212 4 −0.020 0.022 0.3744 0.222 0.212 0.160 0.025 <0.0001 0.278 0.268 0.194 0.018 <0.0001 0.361 0.352 a PRSReported: polygenic risk score (PRS) constructed by the previously reported SNPs; PRSSLS: PRS constructed by SLS-specific tagging SNPs (with the lowest p-value) within +/− 500kb region of the previously reported SNP; PRSSLS+Novel: PRS constructed by additional putative novel SNPs from the current study in addition to the PRSSLS b Adiponectin: natural log-transformed adiponectin level; leptin: rank-based normalized values of residuals from Tobit regression (sex-adjusted) for leptin levels. c Models: Adiponectin: [Model 1] log (Adiponectin levels) ~ PRS; [Model 2] log (Adiponectin levels) ~ PRS + Sex; [Model 3] log (Adiponectin levels) ~ PRS + Sex + PC1 + PC2 + PC3 + PC4 + PC5; [Model 4] log (Adiponectin levels) ~ PRS + Sex + PC1 + PC2 + PC3 + PC4 + PC5 + BMI Z-scores Leptin: [Model 2] Rank-based normalized values of residuals from Tobit model (sex-adjusted) ~ PRS; [Model 3] Rank-based normalized values of residuals from Tobit model (sex-adjusted) ~ PRS + PC1 + PC2 + PC3 + PC4 + PC5; [Model 4] Rank-based normalized values of residuals from Tobit model (sex-adjusted) ~ PRS + PC1 + PC2 + PC3 + PC4 + PC5 + BMI Z-score d Beta: a regression coefficient for PRS; SE: standard error of the regression coefficient; P: p-value Table 4. The effects of biomarker PRS on downstream body fat percent (BF%) and glycemic traits (fasting blood glucose (FBG), fasting blood insulin (FBI), and insulin resistance (IR)) in SLS participants. Total effecta Direct effecta Indirect effecta Beta 95% CL P Beta 95% CL Beta 95% CL Adiponectin PRS Glycemic Traits FBG 0.001 −0.081 0.082 0.9831 0.026 −0.080 0.131 −0.025 −0.092 0.042 FBI −0.124 −0.208 −0.040 0.0037b −0.026 −0.134 0.082 −0.098 −0.167 −0.029 HOMA-IR −0.117 −0.201 −0.033 0.0064 b −0.020 −0.128 0.088 −0.097 −0.167 −0.028 BF% −0.135 −0.198 −0.071 <0.0001b −0.072 −0.153 0.010 −0.063 −0.115 −0.011 Leptin PRS Glycemic Traits FBG −0.032 −0.114 0.050 0.4406 −0.064 −0.154 0.027 0.032 −0.008 0.071 FBI 0.137 0.053 0.221 0.0014 b −0.009 −0.098 0.079 0.146 0.100 0.192 HOMA-IR 0.123 0.039 0.207 0.0043 b −0.021 −0.110 0.068 0.144 0.098 0.190 BF% 0.185 0.123 0.248 <0.0001 b 0.006 −0.054 0.066 0.180 0.139 0.221 Ghrelin PRS Glycemic Traits FBG −0.061 −0.142 0.020 0.1381 0.057 −0.042 0.156 −0.118 −0.178 −0.058 FBI −0.217 −0.299 −0.134 <0.0001 b 0.007 −0.089 0.103 −0.223 −0.286 −0.161 HOMA-IR −0.215 −0.298 −0.133 <0.0001 b 0.017 −0.080 0.113 −0.232 −0.295 −0.169 BF% −0.064 −0.127 0.000 0.0499 0.029 −0.049 0.106 −0.092 −0.139 −0.046 Orexin PRS Glycemic Traits FBG −0.069 −0.150 0.012 0.0939 −0.059 −0.145 0.026 −0.010 −0.038 0.018 FBI −0.001 −0.085 0.083 0.9872 0.032 −0.056 0.121 −0.033 −0.063 −0.003 HOMA-IR −0.013 −0.097 0.071 0.7588 0.019 −0.069 0.108 −0.033 −0.063 −0.003 BF% 0.011 −0.053 0.075 0.7320 0.031 −0.037 0.099 −0.020 −0.042 0.003 Note: HOMA-IR (homeostatic model assessment of insulin resistance); 95% CL (Wald 95% Confidence Limits) a Total effect: The overall effects of the biomarker PRS (constructed by putative novel SNPs in addition to SLS-specific tagging SNPs) on BF% or glycemic traits; direct effect: the effect of the biomarker PRS on BF% or glycemic traits after adjusting for the biomarker levels; indirect effect: the effect of the biomarker PRS on BF% or glycemic traits mediated through biomarker levels. All effects were estimated after adjusting for sex and 5 principal components of participants’ genetic composition. Adiponectin levels, ghrelin levels, orexin levels, FBG, FBI, and HOMA-IR were natural log-transformed. For leptin, rank-based inverse normalized residuals of a Tobit regression model adjusting for sex were used to account for the limit of detection issue. All the exposures, mediators, and outcomes were standardized. Model: Standardized Natural log transformed glycemic traits or BF% ~ Standardized biomarker-PRS + standardized natural log-transformed biomarker levels (or rank-based normalized values of residuals from Tobit model (sex-adjusted) for leptin levels) + Sex + PC1 + PC2 + PC3 + PC4 + PC5 Mediator: standardized natural log-transformed biomarker levels (or rank-based normalized values of residuals from Tobit model (sex-adjusted) for leptin levels) Covariates: Sex, PC1, PC2, PC3, PC4, PC5 b Significant total effects of biomarker PRS on downstream BF% and glycemic traits (after correction for multiple testing p < 0.0125 (=0.05/4) Impact This study characterized the genetic underpinnings of four metabolic hormones and investigated their potential influence on adiposity and insulin biology among Hispanic/Latino adolescents. Fasting blood insulin and insulin resistance were associated with polygenic risk score (PRS) for adiponectin, leptin, and ghrelin, with evidence of some degree of mediation by the biomarker levels. Body fat percent (BF%) was also associated with PRS for adiponectin and leptin. This provides important insight on biological mechanisms underlying early metabolic dysfunction and reveals candidates for prevention efforts. Our findings also highlight the importance of ancestrally diverse populations to facilitate valid studies of the genetic architecture of metabolic biomarker levels. Disclosure Statement: The authors have nothing to disclose. Category of Study: Population study Consent Statement: The research conducted herein constitutes an analysis of de-identified existing data and therefore is not considered human subject research. a https://pubmed.ncbi.nlm.nih.gov/ b https://www.ncbi.nlm.nih.gov/omim c https://www.genecards.org/ d https://www.gtexportal.org/home/snp/rs12066716 ==== Refs REFERENCES 1 Hales CM , Carroll MD , Fryar CD , & Ogden CL , Prevalence of Obesity Among Adults and Youth: United States, 2015-2016. NCHS data brief, 1–8 (2017). 2 Albala C , Vio F , Kain J , & Uauy R , Nutrition transition in Chile: determinants and consequences. Public Health Nutr 5 , 123–128, doi:10.1079/PHN2001283 (2002).12027274 3 Muzzo S , Burrows R , Cordero J , & Ramirez I , Trends in nutritional status and stature among school-age children in Chile. Nutrition 20 , 867–872, doi:10.1016/j.nut.2004.06.007 (2004).15474874 4 Ghanemi A , Yoshioka M , & St-Amand J , Broken Energy Homeostasis and Obesity Pathogenesis: The Surrounding Concepts. Journal of Clinical Medicine 7 , 453–453, doi:10.3390/jcm7110453 (2018). 5 Richard D , Cognitive and autonomic determinants of energy homeostasis in obesity. Nat Rev Endocrinol 11 , 489–501, doi:10.1038/nrendo.2015.103 (2015).26122319 6 Nigro E New insight into adiponectin role in obesity and obesity-related diseases. Biomed Res Int 2014 , 658913, doi:10.1155/2014/658913 (2014).25110685 7 Abella V Leptin in the interplay of inflammation, metabolism and immune system disorders. Nat Rev Rheumatol 13 , 100–109, doi:10.1038/nrrheum.2016.209 (2017).28053336 8 Chabot F , Caron A , Laplante M , & St-Pierre DH , Interrelationships between ghrelin, insulin and glucose homeostasis: Physiological relevance. World J Diabetes 5 , 328–341, doi:10.4239/wjd.v5.i3.328 (2014).24936254 9 Gray SM , Page LC , & Tong J , Ghrelin regulation of glucose metabolism. J Neuroendocrinol 31 , e12705, doi:10.1111/jne.12705 (2019).30849212 10 Klok MD , Jakobsdottir S , & Drent ML , The role of leptin and ghrelin in the regulation of food intake and body weight in humans: a review. Obesity reviews : an official journal of the International Association for the Study of Obesity 8 , 21–34, doi:10.1111/j.1467-789X.2006.00270.x (2007).17212793 11 Poher AL , Tschöp MH , & Müller TD , Ghrelin regulation of glucose metabolism. Peptides 100 , 236–242, doi:10.1016/j.peptides.2017.12.015 (2018).29412824 12 Ebrahim IO , Howard RS , Kopelman MD , Sharief MK , & Williams AJ , The hypocretin/orexin system. J R Soc Med 95 , 227–230, doi:10.1258/jrsm.95.5.227 (2002).11983761 13 Lin L The sleep disorder canine narcolepsy is caused by a mutation in the hypocretin (orexin) receptor 2 gene. Cell 98 , 365–376, doi:10.1016/S0092-8674(00)81965-0 (1999).10458611 14 Sakurai T Orexins and orexin receptors: A family of hypothalamic neuropeptides and G protein-coupled receptors that regulate feeding behavior. Cell 92 , 573–585, doi:10.1016/S0092-8674(00)80949-6 (1998).9491897 15 Heinonen MV , Purhonen AK , Makela KA , & Herzig KH , Functions of orexins in peripheral tissues. Acta Physiol (Oxf) 192 , 471–485, doi:10.1111/j.1748-1716.2008.01836.x (2008).18294339 16 Arihara Z Immunoreactive orexin-A in human plasma. Peptides 22 , 139–142, doi:10.1016/S0196-9781(00)00369-7 (2001).11179609 17 Cecil JE , Tavendale R , Watt P , Hetherington MM , & Palmer CN , An obesity-associated FTO gene variant and increased energy intake in children. N Engl J Med 359 , 2558–2566, doi:10.1056/NEJMoa0803839 (2008).19073975 18 Llewellyn CH , Trzaskowski M , van Jaarsveld CHM , Plomin R , & Wardle J , Satiety mechanisms in genetic risk of obesity. JAMA Pediatr 168 , 338–344, doi:10.1001/jamapediatrics.2013.4944 (2014).24535189 19 Lozoff B Behavioral and developmental effects of preventing iron-deficiency anemia in healthy full-term infants. Pediatrics 112 , 846–854 (2003).14523176 20 East P Infant iron deficiency, child affect, and maternal unresponsiveness: Testing the long-term effects of functional isolation. Developmental psychology 53 , 2233–2244, doi:10.1037/dev0000385 (2017).28933876 21 Pacheco LS Early onset obesity and risk of metabolic syndrome among Chilean adolescents. Preventing Chronic Disease 14 , doi:10.5888/pcd14.170132 (2017). 22 Burrows R Healthy Chilean Adolescents with HOMA-IR >= 2.6 Have Increased Cardiometabolic Risk: Association with Genetic, Biological, and Environmental Factors. Journal of Diabetes Research 2015 , doi:Artn 783296 10.1155/2015/783296 (2015). 23 Lin DY Genetic association analysis under complex survey sampling: the Hispanic Community Health Study/Study of Latinos. Am J Hum Genet 95 , 675–688, doi:10.1016/j.ajhg.2014.11.005 (2014).25480034 24 Price AL Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 38 , 904–909, doi:10.1038/ng1847 (2006).16862161 25 Consortium GT , The Genotype-Tissue Expression (GTEx) project. Nat Genet 45 , 580–585, doi:10.1038/ng.2653 (2013).23715323 26 VanderWeele TJ Explanation in causal inference : methods for mediation and interaction. (Oxford University Press, 2015). 27 VanderWeele TJ , & Vansteelandt S , Conceptual issues concerning mediation, interventions and composition. Stat Interface 2 , 457–468 (2009). 28 Enomoto T Adipolin/C1qdc2/CTRP12 protein functions as an adipokine that improves glucose metabolism. Journal of Biological Chemistry 286 , 34552–34558, doi:10.1074/jbc.M111.277319 (2011).21849507 29 Wei Z C1q/TNF-related protein-12 (CTRP12), a novel adipokine that improves insulin sensitivity and glycemic control in mouse models of obesity and diabetes. Journal of Biological Chemistry 287 , 10301–10315, doi:10.1074/jbc.M111.303651 (2012).22275362 30 Wong GW , Wang J , Hug C , Tsao TS , & Lodish HF , A family of Acrp30/adiponectin structural and functional paralogs. Proceedings of the National Academy of Sciences of the United States of America 101 , 10302–10307, doi:10.1073/pnas.0403760101 (2004).15231994 31 Chung CM A genome-wide association study reveals a quantitative trait locus of adiponectin on CDH13 that predicts cardiometabolic outcomes. Diabetes 60 , 2417–2423, doi:10.2337/db10-1321 (2011).21771975 32 Heid IM Clear detection of ADIPOQ locus as the major gene for plasma adiponectin: results of genome-wide association analyses including 4659 European individuals. Atherosclerosis 208 , 412–420, doi:10.1016/j.atherosclerosis.2009.11.035 (2010).20018283 33 Morisaki H CDH13 gene coding T-cadherin influences variations in plasma adiponectin levels in the Japanese population. Hum Mutat 33 , 402–410, doi:10.1002/humu.21652 (2012).22065538 34 Spracklen CN Exome-Derived Adiponectin-Associated Variants Implicate Obesity and Lipid Biology. Am J Hum Genet 105 , 15–28, doi:10.1016/j.ajhg.2019.05.002 (2019).31178129 35 Wu Y A meta-analysis of genome-wide association studies for adiponectin levels in East Asians identifies a novel locus near WDR11-FGFR2. Hum Mol Genet 23 , 1108–1119, doi:10.1093/hmg/ddt488 (2014).24105470 36 Dastani Z Novel loci for adiponectin levels and their influence on type 2 diabetes and metabolic traits: a multi-ethnic meta-analysis of 45,891 individuals. PLoS Genet 8 , e1002607, doi:10.1371/journal.pgen.1002607 (2012).22479202 37 Qi L Novel locus FER is associated with serum HMW adiponectin levels. Diabetes 60 , 2197–2201, doi:10.2337/db10-1645 (2011).21700879 38 Jee SH Adiponectin concentrations: a genome-wide association study. Am J Hum Genet 87 , 545–552, doi:10.1016/j.ajhg.2010.09.004 (2010).20887962 39 Wu Y Genome-wide association study for adiponectin levels in Filipino women identifies CDH13 and a novel uncommon haplotype at KNG1-ADIPOQ. Hum Mol Genet 19 , 4955–4964, doi:10.1093/hmg/ddq423 (2010).20876611 40 Richards JB A genome-wide association study reveals variants in ARL15 that influence adiponectin levels. PLoS Genet 5 , e1000768, doi:10.1371/journal.pgen.1000768 (2009).20011104 41 Ling H Genome-wide linkage and association analyses to identify genes influencing adiponectin levels: the GEMS Study. Obesity (Silver Spring) 17 , 737–744, doi:10.1038/oby.2008.625 (2009).19165155 42 Folkersen L Genomic and drug target evaluation of 90 cardiovascular proteins in 30,931 individuals. Nat Metab 2 , 1135–1148, doi:10.1038/s42255-020-00287-2 (2020).33067605 43 Kilpelainen TO Genome-wide meta-analysis uncovers novel loci influencing circulating leptin levels. Nat Commun 7 , 10494, doi:10.1038/ncomms10494 (2016).26833098 44 Yaghootkar H Genetic Studies of Leptin Concentrations Implicate Leptin in the Regulation of Early Adiposity. Diabetes 69 , 2806–2818, doi:10.2337/db20-0070 (2020).32917775 45 Sargolzaei J , Chamani E , Kazemi T , Fallah S , & Soori H , The role of adiponectin and adipolin as anti-inflammatory adipokines in the formation of macrophage foam cells and their association with cardiovascular diseases. Clin Biochem 54 , 1–10, doi:10.1016/j.clinbiochem.2018.02.008 (2018).29452073 46 Schaffler A , & Buechler C , CTRP family: linking immunity to metabolism. Trends Endocrinol Metab 23 , 194–204, doi:10.1016/j.tem.2011.12.003 (2012).22261190 47 Fadaei R Decreased serum levels of CTRP12/adipolin in patients with coronary artery disease in relation to inflammatory cytokines and insulin resistance. Cytokine 113 , 326–331, doi:10.1016/j.cyto.2018.09.019 (2019).30337217 48 Bai B Circulating C1q complement/TNF-related protein (CTRP) 1, CTRP9, CTRP12 and CTRP13 concentrations in Type 2 diabetes mellitus: In vivo regulation by glucose. PLoS ONE 12 , doi:10.1371/journal.pone.0172271 (2017). 49 Justice AE Genetic determinants of BMI from early childhood to adolescence: the Santiago Longitudinal Study. Pediatr Obes 14 , e12479, doi:10.1111/ijpo.12479 (2019).30515969 50 Carlson CS Generalization and dilution of association results from European GWAS in populations of non-European ancestry: the PAGE study. PLoS Biol 11 , e1001661, doi:10.1371/journal.pbio.1001661 (2013).24068893 51 Ginsberg HN , Insulin resistance and cardiovascular disease. J Clin Invest 106 , 453–458, doi:10.1172/JCI10762 (2000).10953019 52 Arita Y Paradoxical decrease of an adipose-specific protein, adiponectin, in obesity. 1999. Biochem Biophys Res Commun 425 , 560–564, doi:10.1016/j.bbrc.2012.08.024 (2012).22925674 53 Tilg H , & Moschen AR , Adipocytokines: mediators linking adipose tissue, inflammation and immunity. Nat Rev Immunol 6 , 772–783, doi:10.1038/nri1937 (2006).16998510 54 Borges MC Metabolic Profiling of Adiponectin Levels in Adults: Mendelian Randomization Analysis. Circ Cardiovasc Genet 10 , doi:10.1161/CIRCGENETICS.117.001837 (2017). 55 Yaghootkar H Mendelian randomization studies do not support a causal role for reduced circulating adiponectin levels in insulin resistance and type 2 diabetes. Diabetes 62 , 3589–3598, doi:10.2337/db13-0128 (2013).23835345 56 Mente A Causal relationship between adiponectin and metabolic traits: a Mendelian randomization study in a multiethnic population. PLoS One 8 , e66808, doi:10.1371/journal.pone.0066808 (2013).23826141 57 Gao H Evidence of a causal relationship between adiponectin levels and insulin sensitivity: a Mendelian randomization study. Diabetes 62 , 1338–1344, doi:10.2337/db12-0935 (2013).23274890 58 Nielsen MB , Colak Y , Benn M , & Nordestgaard BG , Causal Relationship between Plasma Adiponectin and Body Mass Index: One- and Two-Sample Bidirectional Mendelian Randomization Analyses in 460 397 Individuals. Clin Chem, doi:10.1093/clinchem/hvaa227 (2020). 59 Farooqi IS Effects of Recombinant Leptin Therapy in a Child with Congenital Leptin Deficiency. New England Journal of Medicine 341 , 879–884, doi:10.1056/NEJM199909163411204 (1999).10486419 60 Myers MG , Cowley MA , & Munzberg H , Mechanisms of leptin action and leptin resistance. Annu Rev Physiol 70 , 537–556, doi:10.1146/annurev.physiol.70.113006.100707 (2008).17937601 61 Zhang CS The Correlation Between Circulating Ghrelin and Insulin Resistance in Obesity: A Meta-Analysis. Frontiers in Physiology 9 , doi:ARTN 1308 10.3389/fphys.2018.01308 (2018). 62 Montague CT Congenital leptin deficiency is associated with severe early-onset obesity in humans. Nature 387 , 903–908, doi:10.1038/43185 (1997).9202122 63 van Jaarsveld CH , Boniface D , Llewellyn CH , & Wardle J , Appetite and growth: a longitudinal sibling analysis. JAMA Pediatr 168 , 345–350, doi:10.1001/jamapediatrics.2013.4951 (2014).24535222 64 van Jaarsveld CH , Llewellyn CH , Johnson L , & Wardle J , Prospective associations between appetitive traits and weight gain in infancy. Am J Clin Nutr 94 , 1562–1567, doi:10.3945/ajcn.111.015818 (2011).22071702 65 Parkinson KN , Drewett RF , Le Couteur AS , Adamson AJ , & Gateshead Milennium Study Core, T. Do maternal ratings of appetite in infants predict later Child Eating Behaviour Questionnaire scores and body mass index? Appetite 54 , 186–190, doi:10.1016/j.appet.2009.10.007 (2010).19887093 66 Llewellyn CH , van Jaarsveld CH , Plomin R , Fisher A , & Wardle J , Inherited behavioral susceptibility to adiposity in infancy: a multivariate genetic analysis of appetite and weight in the Gemini birth cohort. Am J Clin Nutr 95 , 633–639, doi:10.3945/ajcn.111.023671 (2012).22277555
PMC009xxxxxx/PMC9005588.txt	==== Front 9815755 30139 Prostate Cancer Prostatic Dis Prostate Cancer Prostatic Dis Prostate cancer and prostatic diseases 1365-7852 1476-5608 34645983 10.1038/s41391-021-00460-y nihpa1740908 Article Characterization of a Castrate Resistant Prostate Cancer Xenograft Derived from a Patient of West African Ancestry Patierno Brendon M. BS 1 Foo Wen-Chi MD 23 Allen Tyler PhD 2 Somarelli Jason A. PhD 12 Ware Kathryn E. PhD 12 Gupta Santosh PhD 12 Wise Sandra PhD 4 Wise John P. Sr. 4 Qin Xiaodi MS 2 Zhang Dadong PhD 2 Xu Lingfan PhD 3 Li Yanjing PhD 3 Chen Xufeng PhD 3 Inman Brant A. MD 25 McCall Shannon J. MD 23 Huang Jiaoti MD, PhD 23 Kittles Rick A. PhD 6 Owzar Kouros PhD 27 Gregory Simon PhD 128 Armstrong Andrew J. MD 1259 George Daniel J. MD 125 Patierno Steven R. PhD 12 Hsu David S. MD, PhD 128# Freedman Jennifer A. PhD 12#* 1 Department of Medicine, Division of Medical Oncology, Duke University School of Medicine, Durham, North Carolina, 27710, United States of America 2 Duke Cancer Institute, Duke University School of Medicine, Durham, North Carolina, 27710, United States of America 3 Department of Pathology, Duke University School of Medicine, Durham, North Carolina, 27710, United States of America 4 Department of Pharmacology and Toxicology, University of Louisville School of Medicine, Louisville, Kentucky, 40202, United States of America 5 Department of Surgery, Duke University School of Medicine, Durham, North Carolina, 27710, United States of America 6 Division of Health Equities, Department of Population Sciences, City of Hope, Duarte, California, 91010, United States of America 7 Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, North Carolina, 27710, United States of America 8 Center for Genomics and Computational Biology, Duke University, Durham, North Carolina, 27710, United States of America 9 Department of Pharmacology and Cancer Biology, Duke University, Durham NC USA # These authors contributed equally to this work. Author Contributions: BMP: Conceptualization, Investigation, Formal Analysis, Writing, Visualization; WCF: Investigation, Formal Analysis, Writing, Visualization; TA: Formal Analysis, Writing, Visualization; JAS: Methodology, Writing; KEW: Methodology, Writing; SG: Investigation, Formal Analysis, Writing, Visualization; SW: Investigation, Formal Analysis, Writing, Visualization; JPW: Methodology, Writing; XQ: Formal Analysis, Writing, Visualization; DZ: Formal Analysis, Writing, Visualization; LX: Methodology; YL: Methodology; XC: Methodology; BAI: Conceptualization, Resources, Writing; SJM: Resources, Writing; JH: Conceptualization, Writing; RAK: Investigation, Formal Analysis, Writing, Visualization; KO: Conceptualization, Validation, Writing; SG: Conceptualization, Validation, Writing; AJA: Conceptualization, Writing; DJG: Conceptualization, Writing; SRP: Conceptualization, Writing, Supervision, Funding Acquisition; DSH: Conceptualization, Writing, Supervision, Project Administration, Funding Acquisition; JAF: Conceptualization, Writing, Supervision, Project Administration, Funding Acquisition * Corresponding author: Jennifer A. Freedman, 905 South LaSalle Street, Durham, NC 27710, Phone: 919-684-6354, Fax: 919-660-0178, jennifer.freedman@duke.edu 24 9 2021 13 10 2021 13 4 2023 10.1038/s41391-021-00460-yUsers may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use: https://www.springernature.com/gp/open-research/policies/accepted-manuscript-terms Background: Prostate cancer is a clinically and molecularly heterogeneous disease, with highest incidence and mortality among men of African ancestry. To date, prostate cancer patient-derived xenograft (PCPDX) models to study this disease have been difficult to establish because of limited specimen availability and poor uptake rates in immunodeficient mice. Ancestrally diverse PCPDXs are even more rare, and only six PCPDXs from self-identified African American patients from one institution were recently made available. Methods: In the present study, we established a PCPDX from prostate cancer tissue from a patient of estimated 90% West African ancestry with metastatic castration resistant disease, and characterized this model’s pathology, karyotype, hotspot mutations, copy number, gene fusions, gene expression, growth rate in normal and castrated mice, therapeutic response, and experimental metastasis. Results: This PCPDX has a mutation in TP53 and loss of PTEN and RB1. We have documented a 100% take rate in mice after thawing the PCPDX tumor from frozen stock. The PCPDX is castrate- and docetaxel-resistant and cisplatin-sensitive, and has gene expression patterns associated with such drug responses. After tail vein injection, the PCPDX tumor cells can colonize the lungs of mice. Conclusion: This PCPDX, along with others that are established and characterized, will be useful pre-clinically for studying the heterogeneity of prostate cancer biology and testing new therapeutics in models expected to be reflective of the clinical setting. ==== Body pmcIntroduction Significant disparities in prostate cancer (PCa) incidence and mortality exist among racial/ethnic groups, as the age-adjusted incidence and mortality rates for PCa among African American men are 1.7- and 2.3-fold greater, respectively, compared with white men (1,2,3). Such racial disparities likely result from a complex interplay among social, lifestyle/environmental, health system, and biological determinants of health (6). To model race-related aggressive PCa and develop therapeutic agents against it, preclinical models derived from PCas from racially/ethnically diverse patients are required. To date, the MDA PCa 2b cell line is the only non-virally transformed PCa cell line derived from a patient of West African ancestry (7), and only recently were the first six patient-derived xenograft (PDX) models of PCa from self-identified African American patients established (8). Researchers have begun to use patient-derived models of PCa to study the molecular basis of PCa subtypes and develop new therapeutics (11–25). PCa PDX (PCPDX) models have been reported; for example, Lawrence et al. reported ten PCPDXs established largely from metastases from patients who had undergone treatment with enzalutamide, and showed efficacy of ribosome targeting therapeutics against these PCPDXs (23). Other studies have shown that explant and xenograft models of androgen independent PCa closely recapitulate immunostaining and RNA-sequencing profiles of clinical CRPC specimens (7, 24). PCPDXs have been difficult to establish, as they do not readily grow in the flank of immunodeficient mice, and mouse prostates have a limited capacity to house and grow tumors (13). For example, in a recent study by the Novartis Institute for Biomedical Research, which presented the generation of 1,075 PDXs representing a range of different tumor types, no PCPDXs were reported (12). Attempts to improve PCa uptake in mice have included using tissue slice grafts to maximize tumor composition, supplementation with testosterone, and implantation under the renal capsule to maximize vascularization (13,14,17). However, these studies have still generated limited numbers of PCPDXs. Other methods to develop preclinical models of PCa include a recent study using metastatic foci from tissue acquisition necropsy to generate 21 PCPDXs, designated the LuCAP series (22). These represent PCas from a spectrum of metastatic sites and are derived from patients of unknown ancestry (22). The LuCAP series have been continually passaged in hosts rather than stored as frozen stock, and models in this series have a range of take rates when grafted into a new host, with a reported average take rate of 10%. As aforementioned, Palanisamy et al. recently reported establishment of 154 PDXs derived from 99 PCa patients, with take rates between 20% to 30%, including 47 that they have characterized and that can be expanded as cell lines (MDA PCa 2a and 2b) or PDXs (8). The PCPDX models are from 88 self-identified Caucasian patients, 6 self-identified African American patients, and 5 self-identified Hispanic patients. Estimated ancestry was not reported. Herein, we established a PCPDX model using malignant prostate tissue from a patient of estimated 90% West African ancestry. We demonstrated successful engraftment of this PCa back into mice after liquid nitrogen freeze with a 100% take rate. In addition, we characterized this PCPDX, including pathology, karyotype, hotspot mutations, copy number, gene fusions, gene expression, growth rate in normal and castrated mice, therapeutic response, and ability to colonize the lungs in vivo after tail vein injection. Materials and Methods Human PCa Specimens Between 2014 and 2017, 230 men undergoing surgery for PCa consented to have some of their prostate tissue harvested for research by the Duke BioRepository & Precision Pathology Center (BRPC) under Duke IRB protocol 00035974. Tissue from 189 of these patients was transferred and used for research including PCPDX development under Duke IRB protocol 00079652. Transferred samples were de-identified (all 18 HIPAA identifiers removed), and only a limited set of clinical information (age, gender, self-identified race, ethnicity, pathologic diagnosis, surgical procedure performed, prior treatment history) was provided with the samples. PCPDX Generation All animal studies were performed under a Duke University Animal Care and Use Committee (IACUC) approved protocol and PDXs were developed, as previously described (18, 19). Briefly, surgical pathology created frozen Optimal Cutting Temperatures (OCTs) from prostate samples, which were then cut, mounted on slides, and stained. Following a pathological review of the slides, areas of tissue having the highest tumor content were released in minimal media for PCPDX generation. This process ensured higher tumor content in the samples used and ensured samples were implanted approximately 95 minutes or less after termination of blood flow. Tissues were then washed with phosphate buffered saline (PBS) and then minced into pieces approximately 2mm in size and injected into the flanks of 8–10-week-old JAX NOD.CB17-PrkdcSCID-J (SCID) or NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice obtained from the Duke University Rodent Genetic and Breeding Core to develop subcutaneous tumors. For the development of sub-renal capsule tumors, 8–10-week-old JAX NOD.CB17-PrkdcSCID-J or NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ mice were placed in a left lateral decubitus position and the fur clipped at the site above the left kidney. The fur-clipped surgical site was prepared with a disinfectant scrub followed by an alcohol derivative three times. A small incision was made in the left flank and extended through the fascia to enter the peritoneal cavity. The kidney was displaced outside the peritoneum and the capsule was then lifted up above the kidney using forceps to create a small space between the kidney and capsule. An approximately 2mm3 specimen was then inserted into the space using a plastic 10μL Gilson pipette tip. The specimen was pushed far enough so that it remained stably within the capsule and the hole in the capsule could be cauterized without damaging the specimen. The kidney was placed back into the peritoneum and the wound in the peritoneum was then sutured closed. Lymphomas have been known to occur in these animals around approximately six months; therefore, mice were checked daily for initial growth and observed for a total of four months. If growth was observed beneath the implantation site, the hair surrounding the tumor was cut with clippers and the skin was sterilized with ethanol so that tumors could be carefully measured with OSHLUN electronic calipers. Mice were sacrificed once tumors exceeded 10 × 10 mm and tissue was harvested at 150mg/mL in PBS for serial passaging and for testing therapeutic response. To test therapeutic response, tissue was dissociated with the MACS GentleMACS Dissociater and 200μL of the suspension, approximately 108 cells, was injected into each new mouse with a 19-guage needle. Histology Surgical specimens were collected and formalin-fixed, paraffin-embedded (FFPE) for standard histological evaluation by Hematoxylin and Eosin (H&E). Additional 4 μM sections were used for immunohistochemistry (IHC) to confirm the diagnosis and characterize the PCa. 4 μM sections were also cut from FFPE blocks of the PDX specimen for standard histological evaluation by H&E and characterization by IHC. The antibodies used for immunohistochemical characterization were CK8/18 (Dako, 1:200), Pin-4 (combination of p63 (Dako, 1:100), CK34BE12 (Dako, 1:25), and P504S (Dako, 1:50)), PSA (Dako, RTU), PAP (Leica, 1:200), Synaptophysin (Leica, RTU), Chromogranin (Leica 1:200), NSE (Leica, RTU), Ki-67 (Leica, 1:50), CK34BE12 (Dako, RTU), and Derm CK (combination of AE1/3 (Leica, 1:800), Cam 5.2 (Leica, 1:100), and MNF116 (Leica, 1:400)). Karyotype Karyotype analysis was performed using logarithmically growing cells, arrested at metaphase using colchicine, and cells were harvested using standard methods for chromosomal analysis (26). Briefly, cells were collected by trypsinization, pelleted, and re-suspended in 0.075M KCl followed by fixing with methanol:acetic acid (3:1). Cells were dropped onto clean wet slides and aged for 45 min in a 90°C oven. Metaphases were Giemsa-banded, and 30 metaphases were karyotyped. Ancestry Estimate Global genetic ancestry analysis was performed as previously described (27). Briefly, DNA was genotyped for 100 ancestry informative markers (AIMs) using the Sequenom MassARRAY iPLEX platform. The AIMs panel consisted of carefully selected autosomal markers previously identified and validated for estimating continental ancestry information in admixed populations. Individual single-nucleotide polymorphism (SNP) genotype calls were generated using Sequenom TYPER software. Individual admixture estimates for the patient were calculated using a model-based clustering method as implemented in the program STRUCTURE v2.3. Given uncertainty regarding the ancestry of the sample, the admixture model was used to determine which estimation of K (number of sub populations) was the best fit for the data. K was set from 2 to 5 and 100 iterations were run. K = 3 was determined to have the best fit and was used in generating estimates. Copy Number Alterations To investigate genome-wide copy number alterations, an array comparative genomic hybridization (CGH) experiment was conducted on PCPDX DNA samples in duplicates. Array CGH was a two-color based method, where PCPDX DNA was labeled with Cyanine-3 (Cy5)-dUTP and reference DNA with Cyanine-3(Cy3)-dUTP, and hybridized together. Male genomic DNA from a healthy individual was used as a reference control for aCGH hybridization (SureTag DNA Labeling Kit, Cat#5190-3400). Agilent SurePrint G3 Human CGH array (60-mers, 4X180K, Cat#G4449A, Agilent Technologies, CA, USA) containing about 170,334 Distinct Biological Features spotted on the array chip with an estimated 13Kb overall median probe arrangement (hg19), and an approximate 5–10 Mb copy-neutral loss of heterozygosity resolutions was used. After 24 hours of hybridization and washing (Cat#5188-5226, Agilent OligoCHG Wash Buffers), slides were scanned in an Agilent Scanner C. Raw data was uploaded in Agilent CytoGenomics Software and analyzed for the detection of genomic copy gains or losses. Further, to reduce false-positives, stringent filtering criteria were applied by using a minimum three probes to call a copy gain or loss, and an average absolute log2 ratio for gain ≥ 0.25 and loss ≤ −0.25 was used. Additionally, all genomically altered genes were judged manually based on probe distribution within chromosomal aberration regions analyzed in the Agilent CytoGenomics software (28). Hotspot Mutations The Ion AmpliSeq™ Cancer Hotspot Panel v2 (CHP2) (Thermo Fisher) was used to identify hotspot mutations. This is a next generation sequencing (NGS)-based assay consisting of 207 amplicons in a single primer pool that target various cancer “hotspot” variants across approximately 2 800 Catalogue of Somatic Mutations in Cancer (COSMIC) mutations from 50 oncogenes and tumor suppressor genes using an amplicon-based AmpliSeq library. For the CHP2 assay, sequencing was performed on the Ion Torrent Personal Genome Machine (Thermo Fisher). To identify additional gene mutations, NGS was also performed using the VariantPlex Myeloid Kit (ArcherDX), which has 75 gene targets that are frequently mutated in myeloid malignancies. The gene content partially overlaps the gene content of the CHP2 assay. Downstream sequencing was performed on the high-throughput Illumina NextSeq 500 (Illumina). Gene Expression and Fusions Total RNA was extracted from mouse tumors with the High Pure RNA Isolation Kit (Roche Life Science). The concentration and quality was assessed on a Qubit 2.0 (ThermoFisher Scientific) and a 2100 Bioanalyzer (Agilent Technologies), respectively. Illumina Truseq Stranded total RNA-seq Kit combined with ribozome (ribo-zero) and globin depletion was used to prepare total RNA-seq libraries. Total RNA was first depleted of the rRNA and Globin mRNA using biotinylated probes that selectively bind rRNA and globin species. In this protocol, strand specificity information was determined with the dUTP “marking” method. After reverse transcription, Illumina sequencing adapters were ligated to the dscDNA fragments and amplified to produce the final RNA-seq library. This protocol did not amplify the strand marked with dUTP, allowing strand-specific sequencing. Libraries were indexed using a dual indexing approach, allowing pooling of multiple libraries and sequencing on the same sequencing flow cell. Before pooling and sequencing, fragment length distribution and library quality were assessed on a 2100 Bioanalyzer using the High Sensitivity DNA Kit (Agilent Technologies). All libraries were pooled in equimolar ratio and sequenced. Libraries were sequenced on an Illumina NovaSeq 6000 sequencer. Multiplexing 24 libraries on an Illumina S2 NovaSeq flow cell yielded between 200–300 million 151bp paired-end sequences per sample after combining two runs. Once generated, sequence data was de-multiplexed and Fastq files generated using Illumina Bcl2Fastq2 conversion software. FastQC (v0.11.5) (29) and MultiQC (30) were used to assess the initial quality of the sequences stored in fastq files pre- and post-quality control. Low quality sequences and adapters were assessed and removed using Trimmomatic (v0.36-3) (31). Sequencing reads that passed quality control were aligned to reference human genome GRCh38 using STAR aligner (v2.6.1d) (32). The aligned reads were annotated to genomic features using HTSeq (33) implemented in the STAR program. The reference genome sequence and GTF file were obtained from GENCODE project (Release 31). Mapping rates to rRNA and intron regions were assessed using the CollectRnaSeqMetrics tool from Picard v2.13.2 accessed 10-3-2017 (http://broadinstitute.github.io/picard). Potential tissue contamination from mice was assessed using several strategies based on five selected samples (including two of the PDXs from mice treated with enzalutamide, one PDX from a mouse treated with the vehicle (treatment control), as well as two primary human cell populations, which served as negative controls with no mouse tissue present). The assessment strategies included: different resources of reference genome assemblies: GENCODE GRCh38 primary assembly and UCSC GRCh38 assembly; different versions of STAR aligner (32): v2.5.4b and v2.6.1d; using human and mouse combined reference genomes: GENCODE GRCh38 primary human genome + GENCODE GRCm38 primary mouse genome; UCSC GRCh38 human genome and UCSC GRCm38 mouse genome; different settings for STAR aligner arguments: outFilterMatchNminOverLread and outFilterScoreMinOverLread with 0.33, 0.5 or 0.66 (default). Gene differential expression was performed within the framework of a negative binomial model using the R (v3.5.3) (32) and its extension package DESeq2 (v1.20.0) (33). All statistical analyses were adjusted for multiple testing within the framework of control of the false-discovery rate, unless stated otherwise (34, 35, 36). Gene fusion candidates were analyzed using STAR-Fusion (v1.6.0) (37). The gene fusion reference library was obtained from the Broad Institute pre-built CTAT library: GRCh38_gencode_v29_CTAT_lib_Mar272019.plug-n-play.tar.gz accessed 04-02-2019 (https://data.broadinstitute.org/Trinity/). The fusion results were visualized using R package chimeraviz v1.10.0 (38). Statistical analyses were mainly scripted using the R statistical environment (39) along with its extension packages from the comprehensive R archive network (CRAN; https://cran.r-project.org/) and the Bioconductor project (40). The analyses were carried out with adherence to the principles of reproducible analysis using the knitr package (41) for generation of dynamic reports. All analyses were programmed and documented on 04/02/2019 using mercurial (https://www.mercurial-scm.org/) for source code management and are available through a non-commercial source code repository (https://gitlab.oit.duke.edu/dcibioinformatics/pubs/hsu-pc-pdx-rnaseq) (Gitlab registered account required). Therapeutic Response Drug sensitivity studies on the PCPDX were performed as follows. Mice from the second passage of the original PDX were sacrificed once tumors exceeded 10 × 10 mm and tissue from multiple tumors were harvested at 150mg/mL in PBS for serial passaging and for testing therapeutic response. To test therapeutic response, tissue from multiple tumors was dissociated with the MACS GentleMACS Dissociater, and 200μL of the resulting suspension, approximately 108 cells, was injected into the flanks of each JAX NOD.CB17-PrkdcSCID-J mouse. Growth was monitored until the tumors measured ~100 −150mm3. Mice were then randomized into 6 groups: enzalutamide (n=5), control 1 (DMSO) (n=5), castration surgery (n=5), control 2 (DPBS) (n=5), docetaxel (n=5) and cisplatin (n=5). All of the mice in each group were used to assess tumor growth in each condition. To perform blinded assessments for tumor measurements, a second technician was given cages of mice without seeing cage number or marking and this technician measured the tumor volume. For RNA sequencing analysis, RNA was isolated from the tumors of 3 mice from each drug treatment and respective vehicle control groups. RNA sequencing data from the castration surgery group is not shown. Enzalutamide, docetaxel and cisplatin were acquired from the Duke University Pharmacy stockroom. Enzalutamide was diluted in 1:20 DMSO in ultrapure distilled water and delivered at 10mg/kg by oral gavage five times per week. Docetaxel was diluted in 1:20 DMSO and cisplatin was diluted in DPBS and administered by intraperitoneal injection twice per week at 8mg/kg and 3.5mg/kg, respectively as standard doses (25). Experimental Metastasis To establish an experimental metastasis model, tumor samples were homogenized and filtered through a 30 μM and then 10 μM filter, suspended in PBS, and injected into the tail veins of NSG mice at a density of 5×105/200 μL using a 30G1/2 needle and a 1-mL syringe. A total of 10 weeks after injection the animals were sacrificed. The lungs were removed, rinsed in PBS, and fixed in 10% neutral buffered formalin. Whole lungs were sectioned and stained with hematoxylin and eosin (H&E) according to routine protocols, and a genitourinary pathologist performed histological evaluation. Data Availability All RNA sequencing data are available from GEO (GSE146402). Results Patient Clinical History Clinical history of the patient showed that he was initially diagnosed with Gleason pattern 5 + 5 = 10 prostate adenocarcinoma. He was then treated initially with external beam radiation therapy and did well until seven years later when his PSA began to rise (Figure 1). Following this biochemical recurrence, he was started on combined androgen blockade with bicalutamide (non-steroidal androgen receptor inhibitor) and goserelin (luteinizing-hormone releasing hormone (LH-RH) agonist). Six years later, his PCa became castrate-resistant and his PSA rose to 3.1 ng/ml. Imaging demonstrated bone metastases in the ribs and pelvis and he began treatment with abiraterone and prednisone. Within two months, his PSA dropped to 0.83 ng/mL and soon thereafter, he was started on denosumab as bone-directed therapy for his bone metastases. Within another month, it was determined this high-grade PCa required a pelvic exenteration. At pelvic exenteration, the tumor had entirely replaced the prostate, and demonstrated invasion into the seminal vesicles, bladder and rectal wall. The rectal mucosa was uninvolved. There was also evidence of angioinvasion and perineural invasion and 3/17 regional lymph nodes contained tumor. The patient’s disease stage was pT4N1Mx. The sample used to establish the PCPDX was taken from this procedure from malignant prostate tissue. Generation of the PCPDX From 2014–2017, 189 patients undergoing surgery for their PCa were identified and, after the surgical procedure, prostate tissue cores were collected for the establishment of PCPDXs. The majority of samples were from patients undergoing radical prostatectomy, and we collected data on race and PSA levels for these patients (Supplementary Table 1). We also recorded the attempts at PDX formation, including mouse line, number of mice used, and site of implantation (Supplementary Table 1). The vast majority of these cases were Gleason Grade 6–7 adenocarcinomas, and only one patient had high-grade disease. A total of 41% of the prostate tissue cores were grafted into NOD.SCID mice and 59% were grafted into NSG mice. From the 189 samples injected, we generated one PCPDX for an overall take rate of 0.52%. This PCPDX was generated from a specimen grafted into a NSG mouse under the renal capsule, without testosterone supplementation. Our standard operating procedure (SOP) for generation of PDXs required freezing a portion of each PDX tumor for the first three passages to generate fresh frozen vials of PDX for future PDX generation. Per our SOP, we tested the viability of the fresh frozen stock starting at the third passage to re-generate PDX and for this PDX, the rate of regeneration from fresh frozen stock was 100% (10/10). Growth can be detected in these PDXs after just two weeks, and tumors will reach 1000mm3 in the following 15–20 days. Two subsequent generations were implanted, and we observed growth rates in the same range as well as homogenous histology between the generations. Patient and PCPDX Histology Histological evaluation of the patient’s tumor revealed a poorly differentiated malignancy arranged in solid sheets without obvious glandular formation. Tumor cells exhibited pleomorphic nuclei, prominent nucleoli, and increased mitoses. Large areas of geographic necrosis were readily identified (Figure 2). A wide panel of immunohistochemical stains showed the tumor to be consistent with a highly de-differentiated PCa (Table 1). Epithelial stains (cytokeratin cocktail, CK8/18, CK34bE12) were only focally positive while many prostate lineage specific markers (NKX3.1, PSA, PAP, and P501S) were negative. Racemase, which is expressed in the majority of PCas, was focally positive in this tumor. Stains for neuroendocrine differentiation (synaptophysin, chromogranin, NSE) and androgen receptor were also negative. The tumor showed a moderately high proliferation index by Ki-67. The tumor stained negative for p63, excluding urothelial carcinoma. Similar to the patient tumor, the PCPDX showed pleomorphic tumor cells with prominent nucleoli and areas of geographic necrosis (Figure 2). Tumor cells were arranged predominantly in nests, but scattered luminal structures were present, consistent with glandular differentiation. Immunohistochemical staining showed strong and more broadly positive staining for epithelial markers by cytokeratin cocktail, CK8/18 and CK34bE12 in the PCPDX, but prostatic lineage specific markers such as NKX3.1, PSA, PAP and P501S were negative, consistent with what we observed in the original patient’s tumor (Figure 2 and Table 1). Stains for racemase, p63, AR, and neuroendocrine markers synaptophysin, chromogranin and NSE were also all negative. The Ki-67 stain was diffusely positive indicating a high level of proliferation. Patient Ancestry and Karyotype of the PCPDX Ancestral genotyping of DNA from the patient revealed estimated 90% West African, 7% Native American, and 2% European ancestry for this self-identified African American patient. Notably, after three passages, the PCPDX had similar estimated ancestry, with estimated 88% West African, 9% Native American, and 3% European ancestry (Supplementary Table 2). Karyotype analysis on a total of 30 PCPDX cells showed an extra Y chromosome, additional material on chromosome 8, and deletion of material on chromosome 9 (arrows indicate abnormal chromosomes) (Figure 3A). Two cells also had deletion of material on chromosome 20 (data not shown). ISCN designation: 47, XYY, add (8) (p23), del (9) (p22) [30]. Additional non-clonal abnormalities were observed in four cells; one cell had additional material on chromosome 9q, one cell had a deletion in chromosome 11q, one cell had a deletion in chromosome 13q, one cell had additional material on chromosome 4, a deletion in chromosome 6p, and a deletion in chromosome 7p (Supplementary Figure 1). Genomic Characterization of the PCPDX We first performed next generation sequencing using two hotspot mutation assays to screen the PCPDX sample for mutations shown previously to be drivers of PCa (Table 2). Both screens detected a frameshift mutation in TP53 (NM_000546.5:c.1005_1008delTGAG). Neither assay detected reads for PTEN or RB1, which is highly suggestive of deletions involving these genes. We next performed copy number analysis of the PCPDX (Table 3). Genome wide analysis revealed 660 genes with copy number alterations (Supplementary Table 3). Pathway analysis of this gene list using Gene Annotation Tool to Help Explain Relationships (GATHER) showed the three most significant pathways affected to be “response to virus, response to biotic stimulus, and defense response” (Supplementary Table 4) (42). Consistently, Ingenuity Pathway Analysis showed almost every significant pathway alteration related to immune cell development, communication, and trafficking (Supplementary Table 5 and Supplementary Table 6) (43). The PCPDX showed several important copy number alterations, including loss of UDP Glucuronosyltransferase Family 2 Member B17 (UGT2B17) and Phosphodiesterase 4D Interacting Protein (PDE4DIP). Loss of UDP glucuronosyltransferase family 2 members, including UGT2B17, have been reported as sufficient to restore free dihydrotestosterone, sustained androgen signaling, and development of castration resistance (44, 45). A recent study reported PDE4DIP loss appears often in circulating tumor cells taken from men with mCRPC (28). Phosphatase and tensin homolog (PTEN), Retinoblastoma-Associated Protein (RB1), and Friend Leukemia Integration 1 Transcription Factor (FLI1) loss have all been shown to be important in PCa progression (46–48). Gain of G Protein Nucleolar 3 (GNL3) has been correlated with PCa metastasis, and SRY-Box Transcription Factor 2 (SOX2) is often upregulated in PCa leading to metastasis and chemotherapy resistance (47–49). Finally, we used RNA-seq data to identify gene fusions (Figure 3B). Using this method, we did not detect the TMPRSS2/ERG fusion, found in approximately 50% of PCas (2). Studies have reported that this fusion is less prevalent in patients of African ancestry and all of the PCPDXs derived from self-identified African American patients were reported to be ERG negative (8, 50). We did detect snoRNA and ribonuclease-associated RNA (RMRP and RPPH1) fusions, which have been determined to play a role in PCa progression (51). Therapeutic Response of the PCPDX To compare drug efficacy in the PCPDX and patient, we treated the PCPDX by physical castration and with enzalutamide to determine if the PCPDX, like the patient tumor, was castrate-resistant and resistant to secondary hormonal therapy. We also sought to gain insight from RNA Seq data from enzalutamide treated PDXs regarding gene expression and enzalutamide resistance. As shown in Figure 4A, the PCPDX demonstrated primary resistance to both physical castration and enzalutamide. As a control, we treated a group of mouse tumors established from LNCaP cells with the same dose of enzalutamide, and tumor growth was completely inhibited in these experiments (data not shown). Figure 4C shows the response of individual tumors in each group. Docetaxel remains a standard chemotherapy for treatment of advanced PCa, and this patient previously received this treatment. Therefore, we evaluated the PCPDX response to docetaxel (8 mg/kg) and saw a trend of tumor growth inhibition, but not a significant decrease in tumor volume (Figure 4B). Finally, to determine sensitivity of refractory disease, we treated the PCPDX with cisplatin. Cisplatin is an alkylating agent that binds to DNA bases causing cross-links and breaks in DNA strands, thus interfering with DNA replication (52), and is widely used clinically in castrate-resistant PCa. The PCPDX treated with cisplatin revealed significant tumor growth inhibition (Figure 4), suggesting efficacy of cisplatin in this castrate-resistant, docetaxel resistant PCa. Mice treated with cisplatin maintained a healthy weight and appearance, and by the end of the treatment, tumor growth in some mice was not only inhibited, but tumor volume was beginning to decrease (Figure 4D). Genomic Characterization of the PCPDX Treated with Enzalutamide, Docetaxel or Cisplatin As the PCPDX was resistant to enzalutamide and docetaxel, and sensitive to cisplatin, we performed RNA sequencing from three tumors (biological replicates) from the following treatment groups: 1) cisplatin, 2) docetaxel, 3) enzalutamide, and 4) control to begin to understand the mechanism of response to these drugs. Initial analysis focused on gene expression changes between treated and corresponding control groups, and specifically examined genes previously identified to play important roles in PCa progression to advanced and metastatic disease (2, 53). PCPDX tumors treated with cisplatin compared with the corresponding control group revealed expression changes in 3 996 genes with an unadjusted p-value < 0.01 (Supplementary Table 7), including downregulation of Enhancer Of Zeste 2 Polycomb Repressive Complex 2 Subunit (EZH2), Cyclin Dependent Kinase Inhibitor 1B (CDKN1B), and Ras Homolog Family Member A (RHOA), inhibition of which have all been linked previously to cisplatin resistance (54–58) (Figure 5). Ingenuity Pathway Analysis of the top 1 000 genes with significant expression changes showed that the nucleotide excision repair pathway and DNA damage checkpoint regulation were significantly altered (Supplementary Table 8 and Supplementary Figure 2). PCPDX tumors treated with enzalutamide compared with the corresponding control group showed expression changes in 1 066 genes with an unadjusted p-value < 0.01 (Supplementary Table 9). Gene expression changes included upregulation of NDRG2, which has been associated with resistance to abiraterone and enzalutamide in PCa (59), as well as PAX6, which has been linked to AR transactivation and therapeutic resistance (60) (Figure 5). In addition, Kallikrein Related Peptidase 2 (KLK2) and R-Spondin 2 (RSPO2) have been shown to be upregulated in advanced PCa, and both have increased expression greater than two-fold in our model in response to treatment with enzalutamide (61, 62). Ingenuity Pathway Analysis of the top 1 000 genes with significant expression changes revealed that the top pathways altered had significant effects on gastrointestinal and reproductive disease progression (Supplementary Table 10 and Supplementary Figure 3). PCPDX tumors treated with docetaxel interestingly did not show as many changes in gene expression when compared with the corresponding control group, with only 201 genes with unadjusted p-value < 0.01 (Supplementary Table 11). Despite this, we identified upregulation of RUNX2, and downregulation of NR4A1, which both play a role in docetaxel resistance (63–66) (Figure 5). We did not perform Ingenuity Pathway Analysis using these data, as there were not enough expression changes for robust analysis of altered pathways. Experimental Metastasis of the PCPDX To measure the experimental metastasis (67) of the PCPDX, we injected PCPDX tumor cells into the tail veins of four NSG mice. After 10 weeks, mice were sacrificed and lungs were harvested. As shown in Figure 6, the injection of PCPDX tumor cells resulted in metastases to the lung. For each mouse, a biopsy of the lung was obtained and two H&E slides were prepared. The whole lung was also made into a FFPE block and two H&E slides were prepared. In the lungs of four mice undergoing tail vein injections, 100% of the mice had metastatic disease (Table 4). Discussion Herein, we established a PCPDX derived from malignant prostate tissue from a patient of estimated 90% West African ancestry. To our knowledge, this is the first PCPDX derived from a patient of estimated 90% West African ancestry and joins only six other PCPDXs derived from self-identified African American patients available to the PCa and the PCa disparities research community. This PCPDX is a model of castration resistant and highly treatment refractory disease and with the capacity to colonize the lungs. From the present study and the literature, it is clear that establishing PCPDXs is extremely challenging. We found that taking samples from standard 12-hole punches of biopsies resulted in only 10% of samples containing viable tumor for implantation into immunodeficient mice, likely because PCas do not form grossly identifiable tumors and these tumors are often comprised of heterogeneous cell populations, including stroma and other normal cells. In addition, we found that higher Gleason score (8+) samples would often grow to form a nodule in the primary inoculation mice, but would not survive passaging into another mouse. Interestingly, histopathological examination of these masses revealed proliferative basal cells without evidence of invasive cancer. Furthermore, many PCPDXs established to date have low take rates and slow initial growth rates. For example, the LuCaP series of mCRPC PDXs have self-reported take rates ranging from 10% to 80%, and take 6–36 weeks to detect initial growth (22). The PCPDX established herein is unique in that it has a 100% take rate and growth can be detected just two weeks. Immunohistochemical staining of the patient tumor specimen confirmed a highly dedifferentiated tumor, with rare cells expressing cytokeratin. The PCPDX; however, stained highly positive for cytokeratin, confirming an epithelial lineage from the patient tumor. In addition, the PCPDX showed much stronger and more diffusely positive staining for CK34bE12 as well as Ki67. These differences between the patient tumor specimen and PCPDX staining are indicative of the tumor cell clone that was able to most effectively invade and colonize after implantation into the mouse. While further characterization of the patient tumor would have been informative, there was a limited amount of this sample available, and the available sample was exhausted during these studies. The PCPDX model represents the epithelial and highly proliferative cancer cells from the patient’s heterogenous and aggressive disease. In addition, the PCPDX stained negative for AR, unlike the other six PCPDXs derived from self-identified African American patients (8). This PCPDX therefore provides a model to study loss of AR expression in the context of therapeutic resistance and disease progression in a patient of West African ancestry. Genomic profiling and therapeutic response of the PCPDX established herein revealed insight into the biology and subtype of this preclinical model. The PCPDX has a mutation in TP53 as well as deletions of PTEN and RB1. TP53 mutation and deletion of PTEN and RB1 have been linked to poor prognosis in many cancers, including PCa, and have been shown to indicate a more aggressive PCa phenotype (38,41,43). These alterations were not necessarily present in the patient’s primary tumor and some/all of them were likely selected for in response to treatment. While it is possible to determine exact copy number gain/loss via aCGH, we were unable to definitively count the amount of copy number gains or losses within these samples given the dynamic range of the assay. Therefore, we were unable to conclude whether deletions were hemizygous or homozygous. Deletion of PTEN was also reported in two of the PCPDXs derived from self-identified African American patients (8). Importantly, the MDA PCa 2b cell line does not have mutated TP53 or deletions of PTEN and RB1 (7). MDA PCA 2b cells are also androgen dependent and express AR as well as PSA. These differences from our model highlight the heterogeneity of PCa among patients of all races and ethnicities, and the necessity of generating models that reflect the disease at different stages and in men of all races and ethnicities. In addition, we also noted gain of SOX2 and loss of UGT2B17 in the PCPDX established herein, and these alterations are consistent with an advanced disease that has become resistant to secondary hormonal therapy (39,42,43). Loss of UGT2B17 was also reported in one of the PCPDXs derived from a self-identified African American patient (8). Regarding additional copy number alterations, as in the case of this PCPDX, gain of CYP11B1 and loss of PHLPP1 were also each reported in one of the PCPDXs derived from self-identified African American patients (8). Copy number alterations involving CD24 and MAPK14 were also identified in this PCPDX and each reported in one of the PCPDXs derived from self-identified African American patients; however, in this PCPDX both involved gain and in each of the PCPDXs derived from self-identified African American patients both involved loss (8). Finally, we noted a number of gene fusions in this PCPDX, which have also been shown to correlate with advanced disease. Interestingly, in contrast to the lack of response to enzalutamide and docetaxel, the tumor volume of PCPDXs treated with cisplatin significantly decreased compared with the corresponding control group. It is important to note that the PCPDX treatments were done using the same passage of the PCPDX to seed tumors, but a different piece of frozen tumor stock was used for the castration experiment and the chemotherapy experiment. Because of the difference this may have caused in the growth rates of the control groups and the small number of animals tested, we cannot make a definitive conclusion regarding the responsiveness of the PCPDX to the tested therapies. However, we observed a significant decrease in growth after treatment with cisplatin, and the corresponding RNAseq data revealed changes in genetic pathways related to DNA damage. These data suggest the potential efficacy for treatment of this tumor with cisplatin. RNA sequencing of tumors post-treatment compared with respective vehicle control groups revealed changes in gene expression that were consistent with pathways previously reported to be altered in response to the tested therapeutics. Treatment with cisplatin resulted in decreased expression of a number of genes involved in DNA repair and cell cycle checkpoint regulation. We also observed increased expression of PAX6, KLK2, and RSPO2 in response to treatment with enzalutamide, each of which have been shown to be upregulated in advanced PCa (44, 61, 62). IPA analysis of the genes with expression changes after treatment with enzalutamide confirmed many genes were involved in advanced reproductive disease. These data suggest that the PCPDX recapitulates advanced CRPC disease, having developed the genetic networks necessary to resist treatment with enzalutamide. The PCPDX model established and characterized herein represents late stage disease and recapitulates the biology often seen in advanced CRPC, thus providing a model of highly proliferative and drug resistant PCa with experimental metastatic potential from a patient of West African ancestry. Generation of a larger panel of PCPDXs from ancestrally diverse patients at various stages of disease will aid in development of biomarkers and therapeutic agents for aggressive PCa. Ultimately, such precision medicine interventions will reduce PCa disparities and improve outcomes for all men with aggressive PCa. Supplementary Material Supplementary Figure 1 Supplementary Figure 2 Supplementary Figure 3 Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Supplementary Table 4 Supplementary Table 5 Supplementary Table 6 Supplementary Table 7 Supplementary Table 8 Supplementary Table 9 Supplementary Table 10 Supplementary Table 11 Acknowledgements We acknowledge the BioRepository & Precision Pathology Center (BRPC), a shared resource of the Duke University School of Medicine and Duke Cancer Institute, for providing access to the human biospecimens used under Institutional Review Board oversight in this work, the assistance of the Duke University Health System Clinical Molecular Diagnostics Laboratory, Duke Sequencing and Genomic Technologies Shared Resource, and the Duke Cancer Institute Bioinformatics Shared Resource, and Bonnie LaCroix, laboratory manager. Support: P30 Cancer Center Support Grant (P30 CA014236), NIH Basic Research in Cancer Health Disparities R01 Award R01CA220314 to SRP PI, JAF and DSH Co-I, DJG and KO Collaborator, WCF Pathologist. Figure 1. Patient clinical history. Patient PSA during years from diagnosis and treatment history. The patient initially had an elevated PSA, which dropped and remained < 3.1 after radiation therapy. After a period of response to radiation therapy, his PSA increased and his prostate cancer progressed to advanced CRPC. Before attainment of the surgical specimen used to generate the PCPDX, the patient had failed radiation therapy, androgen deprivation therapy, secondary hormonal therapy, denosumab, alendronate, and chemotherapy. Figure 2. Patient and PDX immunohistochemical profiles. H&E staining at 20x magnification of the patient’s tumor and the established PDX both demonstrate pleomorphic nuclei with prominent nucleoli, as well as areas of geographic necrosis (). The patient tumor cells are arranged in solid sheets with no obvious glandular differentiation, whereas the PDX cells are arranged predominantly in nests with scattered luminal structures (arrows). Immunohistochemical staining shows patient and PDX tumor cells to be negative for NXK3.1, PSA and PAP. A small number of cells in the patient tumor specimen stained positive for cytokeratin, with intermittent cells staining positive for Ki-67. The PCPDX model stained much more diffusely positive for cytokeratin, with many cells staining strongly for Ki-67. The patient tumor specimen shows tumor cells to be focally positive for racemase (cytoplasmic stain, red chromogen) (arrows), with rare cells expressing CK34bE12. The PCPDX stains negative for racemase, and stains diffusively positive for CK34bE12. Figure 3. PCPDX DNA and RNA analysis. A. Karyotype analysis on a total of 30 PCPDX cells revealed a dominant karyotype including an extra Y chromosome, additional material on chromosome 8, and deletion of material on chromosome 9 (abnormalities indicated with red arrows). Additional karyotypes are included in Supplementary Figure 1. B. Gene fusions in the PCPDX visualized as a circus plot, notably lacking the TMPRSS2/ERG fusion seen in 50% of prostate cancers. Figure 4. Response of the PCPDX and individual PCPDX tumors to clinically relevant doses of standard-of-care prostate cancer therapies. Treatment was begun at the second measured time-point, at which point tumors had reached ~100mm3. An asterisk () indicates a statistically significant data point defined by the Holm-Sidak t-test (PRISM) where ** P < .01). A. Average tumor volume over time in mice treated with enzalutamide and castration surgery versus control. B. Average tumor volume over time of mice treated with docetaxel and cisplatin versus control. C. Response of individual PCPDX tumors in mice from panel A. D. Response of individual PCPDX tumors in mice from panel B. Figure 5. PCPDX gene expression post-treatment with enzalutamide, docetaxel, or cisplatin. Heat map (left) and volcano plots (right) of fold changes in gene expression in genes of interest post-treatment with enzalutamide, docetaxel or cisplatin compared with respective controls. For the heat map, columns represent treatment groups (three mice), p-value < 0.01 indicated. For the volcano plots, fold change 1.5, positive (red), negative (blue), p-value < 0.01 indicated. Figure 6. H&E of mouse lung tissue with metastatic tumor foci. H&E slides were assessed at 4x magnification by a genitourinary pathologist for tumor foci, indicated by an asterisk. Table 1. Patient Tumor and PCPDX Immunohistochemical Staining. A panel of immunohistochemical stains were performed on the clinical and PCPDX specimens. The first column indicates the IHC stain being used, and the second and third column indicate the staining results in either the PDX or patient sample respectively. Cytokeratin stains confirm the epithelial lineage of the tumor. Prostatic lineage markers such as NKX 3.1, PSA and PAP negative staining indicate a highly de-differentiated cancer. A cocktail of antibodies performed on the clinical specimen showed tumor cells to be focally positive for racemase (cytoplasmic stain, red chromogen) and CK34bE12 (cytoplasmic stain, brown chromogen). The PCPDX has a similar IHC profile as well as shares specific features from certain cells in the patient tumor specimen, despite colonization and propagation in mice. Stain PDX Patient Cytokeratin cocktail Strong and Diffusely Positive Focally Positive CK8/18 Strong and Diffusely Positive Focally Positive PIN4 Diffusely CK34be12 Positive Focally Positive for racemase and CK34bE12 PSA Negative Negative NKX 3.1 Negative Negative PAP Negative Negative P501S Negative Negative AR Negative Negative GR Negative Positive Synaptophysin Negative Negative Chromogranin Negative Negative NSE Negative Negative Ki-67 High Moderately high Table 2. Hotspot Mutations in the PCPDX. A next generation sequencing based assay was used to screen the PCPDX sample for mutations shown previously to be drivers of prostate cancer. The first column indicates the gene and the second column indicates whether the PCPDX has a mutation in the indicated gene and, if so, the type of mutation. Hotspot Mutations Gene Panel Mutation AKT1 None ATM None BRAF None HRAS None NOTCH1 None NRAS None IDH1 None PIK3CA None PTEN Deletion RB1 Deletion TP53 Frameshift Table 3. Copy Number Alterations in the PCPDX. aCGH data was used to determine copy number alterations. Gain/loss in genes of interest are indicated, with genes indicated in the first column and gain/loss in the indicated gene given in the second column. Genes PDX Copy Number Alterations NRAS Gain CYP11B1 Gain UGT2B17 Loss PTEN Loss PHLPP1 Loss RB1 Loss KMT2D Loss KDM5D Gain FANCA Gain FLI1 Loss NDRG1 Gain EGFR Loss ETS1 Loss CHEK2 Gain SOX2 Gain CD24 Gain CD44 Loss GNL3 Gain ZMYM5 Gain MAPK14 Gain Table 4. Tumor foci in mouse lungs after tail vein injections. The number of, size of, and presence of necrosis in tumor foci in each sample indicated in the first column was assessed by a pathologist and indicated in the second, third, and fourth column, respectively. The whole lung was accessed as well as 2mm3 biopsies, which could be used for further assessment if marked positive. Sample type Approximate # of foci Size range (mm) Necrosis? Biopsy of Lung 1 2 <1 – 2 Y Whole Lung 1 10 <1 – 1 N Biopsy of Lung 2 2 <1 N Whole Lung 2 1 <1 N Biopsy of Lung 3 0 NA NA Whole Lung 3 2 <1 N Biopsy of Lung 4 1 1 N Whole Lung 4 2 <1 N The authors declare no potential conflicts of interest. ==== Refs References 1. Siegel RL , Miller KD , Jemal A : Cancer statistics, 2019. Cancer J Clin 2019, 69 (1 ): 7–34 2. Cancer Genome Atlas Research Network. The Molecular Taxonomy of Primary Prostate Cancer. Cell. 2015;163 (4 ):1011–1025.26544944 3. Irshad S , Bansal M , Castillo-Martin M , Zheng T , Aytes A , Wenske S , A molecular signature predictive of indolent prostate cancer [published correction appears in Sci Transl Med. 2013 Sep 18;5(203):203er9]. Sci Transl Med. 2013;5 (202 ):202ra122. 4. Robbins AS , Whittemore AS , Thom DH . Differences in socioeconomic status and survival among white and black men with prostate cancer. American journal of epidemiology. 2000;151 (4 ):409–16.10695600 5. Pietro GD , Chornokur G , Kumar NB , Davis C , Park JY . Racial Differences in the Diagnosis and Treatment of Prostate Cancer. Int Neurourol J. 2016;20 (Suppl 2 ):S112–S119. doi:10.5213/inj.1632722.361 27915474 6. Polite BN , Adams-Campbell LL , Brawley OW , Bickell N , Carethers JM , Flowers CR , Charting the future of cancer health disparities research: A position statement from the American Association for Cancer Research, the American Cancer Society, the American Society of Clinical Oncology, and the National Cancer Institute. CA Cancer J Clin. 2017;67 :353–361 28738442 7. Navone NM , van Weerden WM , Vessella RL , Williams ED , Wang Y , Isaacs JT , Movember gap1 pdx project: An international collection of serially transplantable prostate cancer patient-derived xenograft (pdx) models. Prostate. 2018;78 :1262–1282.30073676 8. Palanisamy N , Yang J , Shepherd PDA , Li-Ning-Tapia EM , Labanca E , Manyam GC , “The MD Anderson Prostate Cancer Patient-Derived Xenograft Series (MDA PCa PDX) Captures the Molecular Landscape of Prostate Cancer and Facilitates Marker-Driven Therapy Development.” Clinical Cancer Research, 2020, doi:10.1158/1078-0432.ccr-20-0479. 9. Barbieri CE , Bangma CH , Bjartell A , Catto JW , Culig Z , Gronberg H , The mutational landscape of prostate cancer. Eur Urol. 2013;64 (4 ):567–576. doi:10.1016/j.eururo.2013.05.029 23759327 10. Spratt DE , Chan T , Waldron L , Speers C , Feng FY , Ogunwobi OO , Racial/Ethnic Disparities in Genomic Sequencing. JAMA Oncol. 2016;2 (8 ):1070–1074.27366979 11. Bluemn EG , Coleman IM , Lucas JM , Coleman RT , Hernandez-Lopez S , Tharakan R , Androgen receptor pathway-independent prostate cancer is sustained through FGF signaling. Cancer Cell. 2017;32 :474–89. e6.29017058 12. Gao H , Korn JM , Ferretti S , Monahan JE , Wang Y , Singh M , High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response. Nat Med. 2015; 21 :1318–25. 10.1038/nm.3954.26479923 13. Russell PJ , Russell P , Rudduck C , Tse BW , Williams ED , Raghavan D . Establishing prostate cancer patient derived xenografts: lessons learned from older studies. Prostate. 2015;75 (6 ):628–636. doi:10.1002/pros.22946 25560784 14. Zhao H , Nolley R , Chen Z , Peehl DM . Tissue slice grafts: an in vivo model of human prostate androgen signaling. Am J Pathol. 2010;177 (1 ):229–239. doi:10.2353/ajpath.2010.090821 20472887 15. Zhao H , Thong A , Nolley R , Reese SW , Santos J , Ingles A , Patient-derived tissue slice grafts accurately depict response of high-risk primary prostate cancer to androgen deprivation therapy. J Transl Med. 2013;11 :199. Published 2013 Aug 28. doi:10.1186/1479-5876-11-199 23985008 16. Wang Y , Revelo MP , Sudilovsky D , Cao M , Chen WG , Goetz L , Development and characterization of efficient xenograft models for benign and malignant human prostate tissue. 2005. Prostate 64 : 149–159 15678503 17. Priolo C , Agostini M , Vena N , Ligon AH , Fiorentino M , Shin E , Establishment and genomic characterization of mouse xenografts of human primary prostate tumors. Am J Pathol. 2010;176 (4 ):1901–1913.20167861 18. Lin D , Wyatt AW , Xue H , Wang Y , Dong X , Haegart A , High fidelity patient-derived xenografts for accelerating prostate cancer discovery and drug development. Cancer Res 2014;74 :1272–83. 10.1158/0008-5472.CAN-13-2921-T 24356420 19. Goldstein AS , Drake JM , Burnes DL , Finley DS , Zhang H , Reiter RE , Purification and direct transformation of epithelial progenitor cells from primary human prostate. Nat Protoc. 2011;6 (5 ):656–667.21527922 20. Uronis JM , Osada T , McCall S , Yang XY , Mantyh C , Morse MA , Histological and molecular evaluation of patient-derived colorectal cancer explants. PLoS One. 2012;7 (6 ):e38422.22675560 21. Kim MK , Osada T , Barry WT , Yang XY , Freedman JA , Tsamis KA , Characterization of an oxaliplatin sensitivity predictor in a preclinical murine model of colorectal cancer. Mol Cancer Ther. 2012;11 (7 ):1500–1509.22351745 22. Nguyen HM , Vessella RL , Morrissey C , Brown LG , Coleman IM , Higano CS , LuCaP Prostate Cancer Patient-Derived Xenografts Reflect the Molecular Heterogeneity of Advanced Disease an--d Serve as Models for Evaluating Cancer Therapeutics. Prostate. 2017;77 (6 ):654–671.28156002 23. Lawrence MG , Obinata D , Sandhu S . Patient-derived Models of Abiraterone- and Enzalutamide-resistant Prostate Cancer Reveal Sensitivity to Ribosome-directed Therapy. Eur Urol. 2018;74 (5 ):562–572. doi:10.1016/j.eururo.2018.06.020 30049486 24. Li Q , Deng Q , Chao HP , Liu X , Lu Y , Lin K , Linking prostate cancer cell AR heterogeneity to distinct castration and enzalutamide responses. Nat Commun. 2018;9 (1 ):3600. Published 2018 Sep 6. 30190514 25. Pienta KJ , Abate-Shen C , Agus DB , Attar RM , Chung LWK , Greenberg NM , The current state of preclinical prostate cancer animal models. Prostate. 2008;68 (6 ):629–639.18213636 26. Wise JP Sr , Wise SS , Little JE . The cytotoxicity and genotoxicity of particulate and soluble hexavalent chromium in human lung cells. Mutat Res 2002;517 (1–2 ):221–9.12034323 27. Al-Alem U , Rauscher G , Shah E , Batai K , Mahmoud A , Beisner E , Association of genetic ancestry with breast cancer in ethnically diverse women from Chicago. PLoS One 2014;9 :e112916.25423363 28. Gupta S , Li J , Kemeny G , Bitting RL , Beaver J , Somarelli JA , Whole Genomic Copy Number Alterations in Circulating Tumor Cells from Men with Abiraterone or Enzalutamide-Resistant Metastatic Castration-Resistant Prostate Cancer. Clin Cancer Res. 2017;23 :1346–1357. doi: 10.1158/1078-0432.CCR-16-1211.27601596 29. Andrews S , FastQC A Quality Control tool for High Throughput Sequence Data. 2014. 30. Ewels P , Magnusson M , Lundin S , Kaller M . MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics, 2016. 32 (19 ): p. 3047–8.27312411 31. Bolger AM , Lohse M , and Usadel B , Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics, 2014. 30 (15 ): p. 2114–20.24695404 32. Dobin A , Davis CA , Schlesinger F , Drenkow J , Zaleski Z , Jha S , STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 2013. 29 (1 ): p. 15–21.23104886 33. Anders S , Pyl PT , and Huber W , HTSeq--a Python framework to work with high-throughput sequencing data. Bioinformatics, 2015. 31 (2 ): p. 166–9.25260700 34. Anders S and Huber W , Differential expression analysis for sequence count data. Genome Biology, 2010. 11: p. R106.20979621 35. Benjamini Y and Hochberg Y , Controlling the False Discovery Rate - a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological, 1995. 57 (1 ): p. 289–300. 36. Bass A , Storey J . qvalue: Q-value estimation for false discovery rate control. R (published bioinformatics package) package version 2.8.0, http://github.com/jdstorey/qvalue, 2015. 37. Haas B , Dobin A , Stransky N , Li B , Yang X , Tickle T , STAR-Fusion: Fast and Accurate Fusion Transcript Detection from RNA-Seq. bioRxiv, 2017. 10.1101/120295 38. Lagstad S , chimeraviz: Visualization tools for gene fusions. R package version 1.10.0, 2019. 39. R Core Team, R: A Language and Environment for Statistical Computing. 2019: Vienna, Austria. 40. Gentlema R , Carey VJ , Bates DM , Bolstad B , Dettling M , Dudoit S , Bioconductor: open software development for computational biology and bioinformatics. Genome Biology, 2004. 5 (10 ): p. R80.15461798 41. Xie Y , Dynamic documents with R and knitr. Second edition. ed. 2015, Boca Raton: CRC Press/Taylor and Francis. xxvii , 266. 42. Chang JT , Nevins JR . GATHER: a systems approach to interpreting genomic signatures. Bioinformatics. 2006;22 :2926–2933. doi: 10.1093/bioinformatics/btl483.17000751 43. Krämer A , Green J , Pollard J Jr , Tugendreich S Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics. 2014;30 :523–530 24336805 44. Robinson D , Van Allen EM , Wu YM , Schultz N , Lonigro RJ , Mosquera JM , Integrative clinical genomics of advanced prostate cancer. Cell. 2015;161 :1215–1228.26000489 45. Zhu Z , Chung YM , Sergeeva O , Kepe V , Berk M , Li J , Loss of dihydrotestosterone-inactivation activity promotes prostate cancer castration resistance detectable by functional imagingJ. Biol. Chem 2018 293 : 17829 46. Leinonen KA , Saramäki OR , Furusato B , Kimura T , Takahashi H , Egawa S , Loss of PTEN is associated with aggressive behavior in ERG-positive prostate cancer. Cancer Epidemiol Biomarkers Prev. 2013; 22 : 2333–2344.24083995 47. Kregel S , Kiriluk KJ , Rosen AM , Cai Y , Reyes EE , Otto KB , Sox2 is an androgen receptor-repressed gene that promotes castration-resistant prostate cancer. PLoS One. 2013;8 :e53701.23326489 48. Mu P Zhang Z , Benelli M , Karthaus W , Hoover E , Chen C , Wongvipat J , “SOX2 promotes lineage plasticity and antiandrogen resistance in TP53- and RB1-deficient prostate cancer.” Science (New York, N.Y.) vol. 355 ,6320 (2017): 84–88. doi:10.1126/science.aah4307 49. Lee M , Williams KA , Hu Y , Andreas J , Patel SJ , Zhang S , GNL3 and SKA3 are novel prostate cancer metastasis susceptibility genes. Clin Exp Metastasis. 2015;32 (8 ):769–782.26429724 50. Magi-Galluzzi C , Tsusuki T , Elson P , Simmerman K , LaFargue C , Esgueva R , TMPRSS2-ERG gene fusion prevalence and class are significantly different in prostate cancer of Caucasian, African-American and Japanese patients. Prostate, 2011. 71 (5 ): p. 489–97 20878952 51. Crea F , Quagliata L , Michael A , Liu H , Frumento P , Azad AA , “Integrated analysis of the prostate cancer small-nucleolar transcriptome reveals SNORA55 as a driver of prostate cancer progression.” Molecular oncology vol. 10 ,5 (2016): 693–703. doi:10.1016/j.molonc.2015.12.010 26809501 52. Rocha CRR , Silva MM , Quinet A , Cabral-Neto JB , Menck CFM . DNA repair pathways and cisplatin resistance: an intimate relationship. Clinics (Sao Paulo). 2018;73 (suppl 1 ):e478s. Published 2018 Sep 6. doi:10.6061/clinics/2018/e478s 30208165 53. Abida W “Genomic correlates of clinical outcome in advanced prostate cancer.” Proceedings of the National Academy of Sciences of the United States of America vol. 116 ,23 (2019): 11428–11436. doi:10.1073/pnas.1902651116 31061129 54. Xu L , Tang H , Wang K , Zheng Y , Feng J , Dong H , “Pharmacological inhibition of EZH2 combined with DNA-damaging agents interferes with the DNA damage response in MM cells.” Molecular medicine reports vol. 19 ,5 (2019): 4249–4255.30942459 55. Zhan J , Wang P , Li S , Song J , He H , Wang Y , “HOXB13 networking with ABCG1/EZH2/Slug mediates metastasis and confers resistance to cisplatin in lung adenocarcinoma patients.” Theranostics vol. 9 ,7 2084–2099. 6 Apr. 2019, doi:10.7150/thno.29463 31037158 56. Qin X , Guo H , Wang X , Zhu X , Yan M , Wang X , “Exosomal miR-196a derived from cancer-associated fibroblasts confers cisplatin resistance in head and neck cancer through targeting CDKN1B and ING5.” Genome biology vol. 20 ,1 12. 14 Jan. 2019, doi:10.1186/s13059-018-1604-0 30642385 57. Shen DW , Pouliot LM , Gillet JP , Ma W , Johnson AC , Hall MD , “The transcription factor GCF2 is an upstream repressor of the small GTPAse RhoA, regulating membrane protein trafficking, sensitivity to doxorubicin, and resistance to cisplatin.” Molecular pharmaceutics vol. 9 ,6 (2012): 1822–33. doi:10.1021/mp300153z 22571463 58. Street CA , Routhier AA , Spencer C , Perkins AL , Masterjohn K , Hackathorn A , “Pharmacological inhibition of Rho-kinase (ROCK) signaling enhances cisplatin resistance in neuroblastoma cells.” International journal of oncology vol. 37 ,5 (2010): 1297–305. doi:10.3892/ijo_00000781 20878077 59. Yu C , Wu G , Li R , Gao L , Yang F , Zhao Y , “NDRG2 acts as a negative regulator downstream of androgen receptor and inhibits the growth of androgen-dependent and castration-resistant prostate cancer.” Cancer biology & therapy vol. 16 ,2 (): 287–96. doi:10.1080/15384047.2014.1002348 60. Elvenes J , Thomassen EI , Johnsen SS , Kaino K , Sjøttem E , Johansen T . “Pax6 represses androgen receptor-mediated transactivation by inhibiting recruitment of the coactivator SPBP.” PloS one vol. 6 ,9 e24659. 15 Sep. 2011, doi:10.1371/journal.pone.0024659 21935435 61. Shang Z , Niu Y , Cai Q , Chen J , Tian J , Yeh S , Human kallikrein 2 (KLK2) promotes prostate cancer cell growth via function as a modulator to promote the ARA70-enhanced androgen receptor transactivation. Tumor Biol. 35 , 1881–1890 (2014). 10.1007/s13277-013-1253-6 62. Patel Rachana , Brzezinska Elspeth Rachana A. , Repiscak Peter , Ahmad Imran , Mui Ernest , Gao Meiling , Activation of β-Catenin Cooperates with Loss of Pten to Drive AR-Independent Castration-Resistant Prostate Cancer. Cancer Res. February 1 2020 (80) (3) 576–590; DOI: 10.1158/0008-5472.CAN-19-1684 31719098 63. Nesbitt H , Browne G , O’Donovan KM , Byrne NM , Worthington J , McKeown SR , (2016). Nitric oxide up-regulates RUNX2 in LNCaP prostate tumours: implications for tumour growth in vitro and in vivo. J. Cell Physiol. 23 , 473–482. 10.1002/jcp.25093 64. Yang Y , Bai Y , He Y , Zhao Y , Chen J , Ma L , “PTEN Loss Promotes Intratumoral Androgen Synthesis and Tumor Microenvironment Remodeling via Aberrant Activation of RUNX2 in Castration-Resistant Prostate Cancer.” Clinical cancer research : an official journal of the American Association for Cancer Research vol. 24 ,4 (2018): 834–846. doi:10.1158/1078-0432.CCR-17-2006 29167276 65. Ozaki T , Yu M , Yin D , Sun D , Zhu Y , Bu Y , “Impact of RUNX2 on drug-resistant human pancreatic cancer cells with p53 mutations.” BMC cancer vol. 18 ,1 309. 20 Mar. 2018, doi:10.1186/s12885-018-4217-9 29558908 66. Yu L , Su YS , Zhao J , Wang H , Li W . “Repression of NR4A1 by a chromatin modifier promotes docetaxel resistance in PC-3 human prostate cancer cells.” FEBS Lett, 587 (2013), pp. 2542–2551 23831020 67. Elkin M & Vlodavsky I Tail vein assay of cancer metastasis. Curr Protoc Cell Biol Chapter 19, Unit19.12, (2001). DOI: 10.1002/0471143030.cb1902s12
PMC009xxxxxx/PMC9005614.txt	==== Front Medizinrecht Medizinrecht Medizinrecht 0723-8886 1433-8629 Springer Berlin Heidelberg Berlin/Heidelberg 6167 10.1007/s00350-022-6167-0 Jahresregister Jahresregister 2021 39. Jahrgang 2021 13 4 2022 2022 39 Suppl 1 132 © Springer-Verlag 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. issue-copyright-statement© Springer-Verlag GmbH Deutschland, ein Teil von Springer Nature 2022 ==== Body pmc
PMC009xxxxxx/PMC9005617.txt	==== Front Radiol Med Radiol Med La Radiologia Medica 0033-8362 1826-6983 Springer Milan Milan 1487 10.1007/s11547-022-01487-4 Chest Radiology Prognostic significance of peripheral consolidations at chest x-ray in severe COVID-19 pneumonia http://orcid.org/0000-0003-3669-8837 Novelli Federica fede.novelli78@gmail.com 1 Pinelli Valentina valentina.pinelli@asl5.liguria.it 1 Chiaffi Luigi luigi.chiaffi@asl5.liguria.it 1 Carletti Anna Maria annamaria.carletti@asl5.liguria.it 1 Sivori Massimiliano massimilianbo.sivori@asl5.liguria.it 1 Giannoni Ugo ugo.giannoni@asl5.liguria.it 2 Chiesa Fabio fabio.chiesa@asl5.liguria.it 2 Celi Alessandro alessandro.celi@unipi.it 3 1 Division of Pulmonology, ASL5 Spezzino, La Spezia, Italy 2 Division of Radiology, ASL5 Spezzino, La Spezia, Italy 3 grid.5395.a 0000 0004 1757 3729 Department of Surgery, Medicine, Molecular Biology and Critical Care, University of Pisa, Via Paradisa, 2, 56124 Pisa, Italy 13 4 2022 17 24 6 2021 23 3 2022 © Italian Society of Medical Radiology 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. To evaluate the possible prognostic significance of the development of peripheral consolidations at chest x-ray in COVID-19 pneumonia, we retrospectively studied 92 patients with severe respiratory failure (PaO2/FiO2 ratio < 200 mmHg) that underwent at least two chest x-ray examinations (baseline and within 10 days of admission). Patients were divided in two groups based on the evolution of chest x-ray toward the appearance of peripheral consolidations or toward a greater extension of the lung abnormalities but without peripheral consolidations. Patients who developed lung abnormalities without peripheral consolidations as well as patients who developed peripheral consolidations showed, at follow-up, a significant worsening of the PaO2/FiO2 ratio but a significantly lower mortality and intubation rate was observed in patients with peripheral consolidations at chest x-ray. The progression of chest x-ray toward peripheral consolidations is an independent prognostic factor associated with lower intubation rate and mortality. Keywords COVID-19 Chest x-ray Consolidation Prognosis ==== Body pmcIntroduction Portable chest x-ray (CXR) is widely used for the follow-up of lung abnormalities in coronavirus disease-19 (COVID-19). The most common CXR lung abnormalities in COVID-19 are lung consolidations, ground glass opacities (GGO) and reticular opacities [1–6]. One particular feature of COVID-19 pneumonia is the high frequency of peripheral lung involvement, often mimicking other inflammatory processes such as organizing pneumonia or eosinophilic pneumonia [1, 7, 8]. Chest computed tomography (CT) studies have shown peripheral involvement in 33–86% of patients, characterized by GGO in the early stages and greater prevalence of consolidations in more advanced stages [9–11]. Such peripheral lung involvement can be readily identified in CXR, especially when characterized by consolidations. CXR studies have reported 40–63% of exclusive involvement of the peripheral zone [3–5]. Although some studies have evaluated the prognostic significance of chest CT or CXR findings, they mainly took into consideration the total extent of the disease rather than a specific characteristic [6, 12, 13]. In this study, we evaluated the characteristics of patients who developed CXR peripheral consolidation and the possible prognostic significance of such finding. Patients and methods Patients This was a retrospective study of patients admitted to our sub-intensive respiratory unit from September to December 2020 with SARS-CoV-2 infection confirmed by real-time RT-PCR on nasopharyngeal swab. The inclusion criteria were as follows: (1) patients with severe disease (defined by a PaO2/FiO2 ratio < 200 mmHg at nadir); (2) patients who underwent at least two CXR’s (one at admission and the second within 10 days of admission); (3) patients with lung abnormalities involving > 50% of the lung parenchyma or with peripheral consolidations in at least one of the two CXR’s. The study was conducted in compliance with the Declaration of Helsinki, with the approval of the local ethic committee. We divided the patients based on the presence or absence of peripheral consolidations at baseline or follow-up CXR. Among the patients admitted in the study period, we arbitrarily selected the first 46 patients of the two groups among those that fulfilled the inclusion criteria. No patient was excluded for any reason after being considered in the first batch of 46 per group. To investigate the prognostic significance of the appearance of peripheral consolidations on CXR in patients hospitalized for severe respiratory failure due to bilateral SARS-CoV-2 pneumonia we analyzed two clinically relevant outcomes, death within the hospitalization and the need for endotracheal intubation (ETI). Chest x-ray All patients underwent digital anteroposterior bedside CXR at full inspiration using a portable radiography unit at admission and within 10 days of admission. In an initial training set, 25 CXR’s were reviewed independently by two experienced thoracic radiologists (FC and UG) and also observed by two clinical pulmonologists (FN and VP). The remaining CXR’s were independently evaluated by the pulmonologists; results were compared, and when disagreement was found, final decisions were determined by consensus. The finding considered in the evaluation of CXR were the presence and extension of reticular alterations, GGO and consolidations. CXR alterations were defined according to the Fleischner Society’s nomenclature [14]: reticular alterations, as innumerable small linear opacities; GGO, as areas of increased lung opacity, within which the margins of pulmonary vessels may be indistinct; consolidation, as an increase in lung opacity that obscures the margins of the vessels and airway walls. We have defined the extension of the lung abnormalities using a cut-off of 50% of lung fields and we have assigned a score of 1 if the lung abnormalities were extended less than 50% of the lung fields and a score of 2 if they were extended more than 50% of the lung fields. We have defined peripheral zones as areas beyond 2 cm from the lobar bronchial structures as far as visible [15]. We divided the patients on the basis of the presence or absence of peripheral consolidations at baseline or follow-up CXR (Fig. 1a–d).Fig. 1 Classification of chest x-ray. a Score 1, no peripheral consolidations: reticulations and GGO extended less than 50% of the lung fields, without peripheral consolidations. b Score 2, no peripheral consolidations: reticulations and GGO extended more than 50% of the lung fields, without peripheral consolidations. c Score 1 + peripheral consolidations: reticulations, GGO and consolidations extended less than 50% of the lung fields, presence of peripheral consolidations. d lung abnormalities (reticulations, GGO and consolidations) extended more than 50% of the lung fields, presence of peripheral consolidations. GGO ground-glass opacities Statistical analysis Data analysis was performed with SPSS software, version 25. Data were expressed as mean ± standard deviation or median [interquartile range] for normally distributed and non-normally distributed variables, respectively. Comparison between patient groups were performed with unpaired t-test and Mann–Whitney test as appropriate. Paired data were compared with paired t-test and paired Wilcoxon test. Multivariate logistic analysis was used to identify significant variables in predicting mortality. Two-side p values below 0.05 were considered statistically significant. Results A total of 92 patients was considered, mean age 70 ± 10 years; 78% were males. Tables 1 and 2 describe the patients characteristics and the chest x-ray findings at hospital admission. The median time between onset of symptoms and the hospital admission was 6 days. Patients had a severe respiratory failure per inclusion criteria, lymphopenia with normal white blood cells, increased value of C-reactive protein (CRP) and procalcitonin (PCT). In 80% of cases PCT was < 0.55 ng/mL, commonly considered the threshold to suspect bacterial superinfection in COVID-19 [16]. Table 2 describes the chest x-ray findings at hospital admission. According to the onset of symptoms, we subdivided the chest x-ray in the 4 stages described in the literature [17, 18]: stage 1 (0–4 days), stage 2 (5–8 days), stage 3 (9–13 days) and stage 4 (≥ 14 days). Most patients (54.3%) were admitted to the hospital in stage 2 and had little extension of the lung abnormalities (70.7% had a score 1 of chest x-ray). Only 12% had peripheral consolidations, none of them in stage 1.Table 1 Baseline characteristics of the patients Patients characteristics Baseline Time from symptoms onset (days) 6.0 [4.0] Arterial gas analysis PaO2/FiO2, mmHg (median [interquartile range]) 213 [114] Laboratory exams White blood cell count (× 103/μL) (median [interquartile range]) 7.7 [4.5] Neutrophil count (× 103/μL) (median [interquartile range]) 6.16 [4.47] Neutrophils, % (median [interquartile range]) 82.6 [9.7] Lymphocyte count (× 103/μL) (median [interquartile range]) 0.74 [0.58] Lymphocytes, % (median [interquartile range]) 10.7 [7.0] NLR (median [interquartile range]) 7.5 [7.6] CRP (mg/dL) (median [interquartile range]) 7.7 [8.8] PCT (ng/mL) (median [interquartile range]) (only 74 pts) 0.17 [0.37] PCT > 0.55 ng/mL (number, %) (only 74 pts) 15 (20.3%) CRP C-reactive protein, PCT procalcitonin p < 0.05 §p < 0.01 Table 2 Baseline chest x-ray findings according to stage Stage of illness from the onset Number of patients 92 Score 1, no PC 65 (70.7%) Score 1 + PC 11 (12%) Score 2, no PC 16 (17.3%) Stage 1 (0–4 days) 22 (23.9%) 17 (77.3) 0 5 (22.7) Stage 2 (5–8 days) 50 (54.3%) 36 (72) 6 (12) 8 (16) Stage 3 (9–13 days) 15 (16.3%) 8 (53.3) 5 (33) 2 (13.3) Stage 4 (≥ 14 days) 5 (5.4%) 4 (80) 0 1 (20) PC peripheral consolidations Patients repeated chest x-ray after a median time of 6.2 days. Table 3 shows demographic and laboratory findings at baseline and follow-up and the rate of negative outcome of patients subdivided in two groups according the presence of peripheral consolidations at follow-up.Table 3 Comparing between patient without and with peripheral consolidations Peripheral consolidations n = 46 No peripheral consolidations n = 46 Age 67.2 ± 10.4 72.9 ± 9.7 Sex M/F, % 78.3/21.7 78.3/21.7 PaO2/FiO2, mmHg Baseline 219.5 [79.0] 190.0 [118.0] FU 101.0 [28.0] 97.5 [30.0] Δ% − 50.3 [22.3] − 47.8 [35.2] White blood cell count (× 103/μL) Baseline 8.15 [4.52] 7.15 [4.77] FU 9.55 [4.52] 10.6 [8.02] Δ% 11.3 [70.6] 56.6 [112.7]* Neutrophils count (× 103/μL) Baseline 6.50 [3.82] 5.73 [4.92] FU 8.01 [4.16] 9.21 [8.25]* 78.8 [84.5] 77.8 [115.4]* Lymphocyte count (× 103/μL) Baseline 0.81 [0.57] 0.68 [0.54] FU 0.79 [0.47] 0.58 [0.47]§ Δ% 2.8 [89.3] − 24.0 [51.5]§ CRP (mg/dL) Baseline 7.4 [5.5] 10.3 [10.4]* FU 3.7 [6.7] 11.3 [11.0]§ Δ% − 45.5 [95.6] 6.8 [137.3]§ PCT (ng/mL) Baseline 0.11 [0.25] 0.28 [0.58]§ FU 0.05 [0.12] 0.23 [0.67]§ Δ% − 50.0 [56.2] 0 [241.7] PCT > 0.55 ng/mL Baseline (74 pts) 6 (14.3%) 9 (28.1%) FU (86 pts) 4 (9.1%) 13 (31%)§ ETI n, % 3 (6.5) 16 (34.8)§ Death n,% 6 (13) 30 (65.2)§ CRP C-reactive protein, PCT procalcitonine, FU follow-up, ETI endotracheal intubation p < 0.05 §p < 0.01 Patients with peripheral consolidations were younger. Both groups had severe baseline respiratory failure that further worsened at follow-up, but with no significant difference between them. Both groups had similar neutrophil and lymphocyte count at baseline but at follow-up patients without peripheral consolidations had an increase in neutrophils and a strong reduction of lymphocytes. Furthermore, while the patients with peripheral consolidations had a reduction of CRP and PCT value at follow-up, those without peripheral consolidations showed a persistence of high values of these inflammatory markers. In particular, 30% of patients in the second group had a PCT value > 0.55 ng/mL, that is suspect for bacterial super-infection, albeit non diagnostic. The corresponding figure for patients in the first group was 6%. Finally, we observed a significantly lower mortality and intubation rate in patients with peripheral consolidations at CXR (Table 3). Table 4 Multivariate logistic analysis for death/need for endotracheal intubation Variable Odds ratio 95% CI Age 1.04 0.98–1.10 Baseline CRP 1.02 0.93–1.13 Δ% CRP 1.004 1.001–1.009 Δ% Lymphocytes (%) 0.99 0.98–1.01 PCT > 0.55 ng/mL at FU 3.73 0.78–17.80 Peripheral consolidations at CXR 0.08 0.02–0.27§ CRP C-reactive protein, PCT procalcitonine, CXR chest x-ray, FU follow-up *p < 0.05 §p < 0.01 We performed a multivariate logistic analysis to investigate a composite outcome that combines mortality and the need for ETI and using age, baseline CRP, CRP Δ, lymphocytes Δ, PCT > 0.55 ng/mL at follow-up and the presence of peripheral consolidations at CXR as covariates. The presence of peripheral consolidations at CXR was an independent predictor of better prognosis (Table 4). Discussion The main purpose of this study was to investigate the prognostic significance of the appearance of peripheral consolidations on CXR in patients hospitalized for severe respiratory failure due to bilateral SARS-CoV-2 pneumonia. Our data show that in severe patients this type of radiographic response is associated with a better prognosis, with a significant reduction in the rate of intubation and mortality compared to patients who do not develop peripheral consolidations but go toward a greater extension of the lung abnormalities. While the development of peripheral consolidations is associated with clinical/laboratory characteristics suggestive of a better prognosis (lower age, higher lymphocyte count and lower CRP and PCT) it remained an independent predictor of good prognosis after controlling for such variables. This result is very important, especially in light of the fact that both groups of patients have the same baseline level of severity and the same worsening of respiratory failure within the first 10 days of hospitalization. Therefore, observing in such critical patients a radiographic evolution toward peripheral consolidations allows physicians to obtain important prognostic information. At the present time, CXR is the main radiologic tool for monitoring the progression of lung abnormalities in COVID-19, especially in critical patients admitted to sub-intensive and intensive care unit [19] and some studies have utilized a CXR scores to quantify the pulmonary involvement in COVID-19 and to predict the prognosis [2, 5, 6, 20, 21]. However, most such studies used scores that quantified the total extent of radiographic changes, without analyzing the prognostic significance of specific changes. An exception is represented by the study by Giraudo et al., which assessed the contribution of GGO and consolidations separately, attributing a worse prognosis to the latter. However, the consolidations described by Giraudo include all consolidations regardless of their localization [21]. We hypothesize that the same severity of respiratory failure in the two groups in our study is an expression of a greater impact on gas exchange of the consolidations (possibly due to the shunt effect), while the better prognosis in these patients is the expression of a less extensive disease. Further studies with chest CT scans associated with lung ultrasound or V/Q scan are warranted to further investigate this hypothesis. Indeed, as described in the work by Parra Gordo et al., which analyzed both CXR and computed tomography, the radiologic stage of peripheral consolidations would match with a moderate lung involvement with the pattern of organizing pneumonia, while the CXR stage characterized by diffuse pulmonary opacities would match with a severe lung involvement with radiologic pattern of diffuse alveolar damage [7]. In our study, both patient groups have a severity of respiratory failure corresponding to moderate-severe ARDS, and various studies show that ARDS with diffuse alveolar damage is associated with higher mortality than ARDS without diffuse alveolar damage [22, 23]. We found that the CXR evolution toward peripheral consolidations is associated with a lower age, a lower reduction of lymphocytes, a lower CRP and PCT value, features that have been previously shown to be associated with a better prognosis [12, 24–28]. On the contrary, we observed persistence of elevated value of CRP and PCT in the group of patients with lung abnormalities involving > 50% of the lung parenchyma but without peripheral consolidations. This observation might suggest that bacterial superinfection is responsible for the worst prognosis of patients without peripheral consolidations. However, some data in the literature show that PCT may be an indicator of disease severity in COVID-19 independent of bacterial superinfection [29]. Furthermore, all patients were treated with antibiotics per standard practice at the time, and in all patients with PCT > 0.55 antibiotic therapy was scaled up; therefore, it is not likely that the superinfection, when present, had a major impact in the outcome. This study has limitations. First, it was retrospective. Second, we did not perform a correlation between CXR and CT findings, since in our center CXR was the main radiologic tool for monitoring the progression of lung abnormalities in COVID-19. Third, only 6 patients had a negative outcome (IOT or death) in the group of peripheral consolidations. Finally, another limitation of the present study is the use of portable CXR, that has lower quality compared to posteroanterior standard radiographs and in which evaluation of the left lower lobe is limited. However, this method has the advantages of limiting the movement of severe patients and to reduce environmental contamination. In conclusion, according to our data, in severe COVID-19 patients the progression of CRX toward peripheral consolidations is an independent prognostic factor associated with lower intubation rate and mortality even in the presence of worsening respiratory failure; further studies are needed to confirm the significance of this finding and to understand the underlying pathophysiological and histopathological features. Authors' contributions All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Novelli Federica and Pinelli Valentina. The radiographic score was discussed and agreed with the radiologists Chiesa Fabio e Giannoni Ugo. The first draft of the manuscript was written by Novelli Federica and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding No funds were received for this research project. Availability of data and materials Data available on request from the authors. Declarations Conflict of interest The authors have no conflicts of interest to declare that are relevant to the content of this article. Ethics approval The study was conducted in compliance with the Declaration of Helsinki, with the approval of the local ethic committee. Consent to participate Informed consent was obtained from all individual participants included in the study. Consent for publication Patients signed informed consent regarding publishing their data. The submission does not include images that may identify the person. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Jacobi A Chung M Bernheim A Eber C Portable chest X-ray in coronavirus disease-19 (COVID-19): a pictorial review Clin Imaging 2020 64 35 42 10.1016/j.clinimag.2020.04.001 32302927 2. Yasin R Gouda W Chest X-ray findings monitoring COVID-19 disease course and severity Egypt J Radiol Nucl Med 2020 10.1186/s43055-020-00296-x 3. Vancheri SG Savietto G Ballati F Maggi A Canino C Bortolotto C Valentini A Dore R Stella GM Corsico AG Iotti GA Mojoli F Perlini S Bruno R Preda L Radiographic findings in 240 patients with COVID-19 pneumonia: time-dependence after the onset of symptoms Eur Radiol 2020 30 11 6161 6169 10.1007/s00330-020-06967-7 32474630 4. Ippolito D Maino C Pecorelli A Allegranza P Cangiotti C Capodaglio C Mariani I Giandola T Gandola D Bianco I Ragusi M Franzesi CT Corso R Sironi S Chest X-ray features of SARS-CoV-2 in the emergency department: a multicenter experience from northern Italian hospitals Respir Med 2020 170 106036 10.1016/j.rmed.2020.106036 32469732 5. Wong HYF Lam HYS Fong AHT Leung ST Chin TWY Lo CSY Lui MMS Lee JCY Chiu KWH Chung TWH Lee EYP Wan EYF Hung IFN Lam TPW Kuo MD Ng MY Frequency and distribution of chest radiographic findings in patients positive for COVID-19 Radiology 2020 296 2 E72 E78 10.1148/radiol.2020201160 32216717 6. Borghesi A Maroldi R COVID-19 outbreak in Italy: experimental chest X-ray scoring system for quantifying and monitoring disease progression Radiol Med (Torino) 2020 125 5 509 513 10.1007/s11547-020-01200-3 32358689 7. Parra Gordo ML Buitrago Weiland G GrauGarcía M ArenazaChoperena G Aspectos radiológicos de la neumonía COVID-19: evolución y complicaciones torácicas Radiologia 2021 63 1 74 88 10.1016/j.rx.2020.11.002 33334590 8. Smith DL Grenier JP Batte C Spieler B A characteristic chest radiographic pattern in the setting of the COVID-19 pandemic Radiol: Cardiothorac Imaging 2020 2 5 e200280 10.1148/ryct.2020200280 33778626 9. Chung M Bernheim A Mei X Zhang N Huang M Zeng X Cui J Xu W Yang Y Fayad ZA Jacobi A Li K Li S Shan H CT imaging features of 2019 novel coronavirus (2019-nCoV) Radiology 2020 295 1 202 207 10.1148/radiol.2020200230 32017661 10. Ng MY Lee EYP Yang J Yang F Li X Wang H Lui MM Lo CS Leung B Khong PL Hui CK Yuen KY Kuo MD Imaging profile of the COVID-19 infection: radiologic findings and literature review Radiol Cardiothorac Imaging 2020 2 1 e200034 10.1148/ryct.2020200034 33778547 11. Salehi S Abedi A Balakrishnan S Gholamrezanezhad A Coronavirus disease 2019 (COVID-19): a systematic review of imaging findings in 919 patients Am J Roentgenol 2020 215 1 87 93 10.2214/ajr.20.23034 32174129 12. Feng Z Yu Q Yao S Luo L Zhou W Mao X Li J Duan J Yan Z Yang M Tan H Ma M Li T Yi D Mi z, Zhao H, Jiang Y, He Z, Li H, Nie W, Liu Y, Zhao J, Luo M, Liu X, Rong P, Wang W, Early prediction of disease progression in COVID-19 pneumonia patients with chest CT and clinical characteristics Nat Commun 2020 11 1 4968 10.1038/s41467-020-18786-x 33009413 13. Liang W Liang H Ou L Chen B Chen A Li C Li Y Guan W, Sang L, Lu J, Xu Y, Chen G, Guo H, Guo J, Chen Z, Zhao Y, Li S, Zhang N, Zhong N, He J, Development and validation of a clinical risk score to predict the occurrence of critical illness in hospitalized patients With COVID-19 JAMA Int Med 2020 180 8 1081 1089 10.1001/jamainternmed.2020.2033 14. Hansell DM Bankier AA MacMahon H McLoud TC Müller NL Remy J Fleischner society: glossary of terms for thoracic imaging Radiology 2008 246 3 697 722 10.1148/radiol.2462070712 18195376 15. Timmerman R McGarry R Yiannoutsos C Papiez L Tudor K DeLuca J Ewing M Abdulrahman R DesRosiers C Williams M Fletcher J Excessive toxicity when treating central tumors in a phase II study of stereotactic body radiation therapy for medically inoperable early-stage lung cancer J Clin Oncol 2006 24 30 4833 4839 10.1200/jco.2006.07.5937 17050868 16. Pink I Raupach D Fuge J Vonberg RP Hoeper MM Welte T Rademacher J C-reactive protein and procalcitonin for antimicrobial stewardship in COVID-19 Infection 2021 49 935 943 10.1007/s15010-021-01615-8 34021897 17. Pan F Ye T Sun P Gui S Liang B Li L Zheng D Wang J Hesketh RL Yang L Zheng C Time course of lung changes at chest CT during recovery from coronavirus disease 2019 (COVID-19) Radiology 2020 295 715 721 10.1148/radiol.2020200370 32053470 18. Mazzei MA Guerrini S Zanoni M Franchi F Valente S Cusi MG Frediani B Volterrani L Novel coronavirus (COVID-19) pneumonia: portable chest X-ray or computed tomography? An Italian perspective Lung India 2021 38 S72 S73 10.4103/lungindia.lungindia_453_20 33686985 19. American College of Radiology. ACR recommendations for the use of chest radiography and computed tomography (CT) for suspected COVID-19 infection. https://www.acr.org/Advocacy-and-Economics/ACR-Position-Statements/Recommendations-for-Chest-Radiography-and-CT-for-Suspected-COVID19-Infection 20. Toussie D Voutsinas N Finkelstein M Cedillo MA Manna S Maron SZ Jacobi A Chung M Bernheim A Eber C Coincepcion J Fayad ZA Gupta YS Clinical and chest radiography features determine patient outcomes in young and middle-aged adults with COVID-19 Radiology 2020 297 1 E197 E206 10.1148/radiol.2020201754 32407255 21. Giraudo C Cavaliere A Fichera G Weber M Motta R Pelloso M Tosato F Lupi A Calabrese F Carretta G Cattelan AM De Conti G Cianci V Navalesi P Plebani M Rea F Vettor r, Vianello A, Stramare R, Validation of a composed COVID-19 chest radiography score: the CARE project ERJ Open Res 2020 6 4 00359 02020 10.1183/23120541.00359-2020 22. Cardinal-Fernández P Bajwa EK Dominguez-Calvo A Menéndez JM Papazian L Thompson BT the presence of diffuse alveolar damage on open lung biopsy is associated with mortality in patients with acute respiratory distress syndrome Chest 2016 149 5 1155 1164 10.1016/j.chest.2016.02.635 26896701 23. Cardinal-Fernández P, Lorente JA, Ballén-Barragán A, Matute-Bello G (2017) Acute respiratory distress syndrome and diffuse alveolar damage. new insights on a complex relationship. Ann Am Thorac Soc 14(6):844–850. 10.1513/AnnalsATS.201609-728P 24. Zhou F Yu T Du R Fan G Liu Y Liu Z Xiang J Wang Y Song B Gu X Guan L Wei Y Li H Wu X Xu J Tu S Zhang Y Chen H Cao B Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study The Lancet 2020 395 10229 1054 1062 10.1016/s0140-6736(20)30566-3 25. Liu Y Yang Y Zhang C Huang F Wang F Yuan J Wang Z Li J Feng C Zhang Z Wang L Peng L Chen L Qin Y Zhao D Tan S Yin L Xu J Zhou C Jiang C Liu L Clinical and biochemical indexes from 2019-nCoV infected patients linked to viral loads and lung injury Sci China Life Sci 2020 63 3 364 374 10.1007/s11427-020-1643-8 32048163 26. Wu C Chen X Cai Y Xia J Zhou X Xu S Huang H Xhang L Zhou X Du C Zhang Y Song J Wang S Chao Y Yang Z Xu J Zhou X Chen D Xiong W Xu L Zhou F Jiang J Bai C Zheng J Song Y Risk factors associated with acute respiratory distress syndrome and death in patients with coronavirus disease 2019 pneumonia in Wuhan, China JAMA Int Med 2020 180 7 934 943 10.1001/jamainternmed.2020.0994 27. Liu J, Liu Y, Xiang P, Pu L, Xiong H, Li C, Zhang M, Tan j, Xu Y, Song R, Song M, Wang L, Zhang W, Han B, Tang L, Wang X, Zhou G, Zhang T, Li b, Wang Y, Chen Z, Wang X et al (2020) Neutrophil-to-lymphocyte ratio predicts critical illness patients with 2019 coronavirus disease in the early stage. J Transl Med. 10.1186/s12967-020-02374-0 28. Rui HU Chaofei H Shiyao P Mingzhu Y Xiang C Procalcitonin levels in COVID-19 patients Int J Antimicrob Agents 2020 56 106051 10.1016/j.ijantimicag.2020.106051 32534186 29. Heer RS Mandal AK Kho J Szawarski P Csabi P Grenshaw D Walker IA Missouris CG Elevated procalcitonin concentrations in severe COVID-19 may not reflect bacterial co-infection Ann Clin Biochem 2021 58 5 520 527 10.1177/00045632211022380 34018843
PMC009xxxxxx/PMC9005618.txt	==== Front J Geol Soc India Journal of the Geological Society of India 0974-6889 Geological Society of India New Delhi 2002 10.1007/s12594-022-2002-5 Original Article Partitioning of Rare Earth Elements (REEs) from Coal to Coal Fly Ash in Different Thermal Power Stations (TPSs) of India Maity Sudip sudip_maity@yahoo.com sudipmaity@cimfr.nic.in 12 Singh Choudhary Akshay K. dynamicakshay007@gmail.com 12 Kumar Santosh kumarsantoshhzb1993@gmail.com 12 Gupta Pavan K. pkchehit@gmail.com 1 1 grid.505934.e CSIR-Central Institute of Mining and Fuel Research (Digwadih), PO: FRI, Dhanbad, 828 108 India 2 grid.469887.c 0000 0004 7744 2771 Academy of Scientific and Innovative Research (AcSIR), Ghaziabad, 201 002 India 13 4 2022 2022 98 4 460466 31 3 2021 6 8 2021 © Geological Society of India, Bengaluru, India 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Rare earth elements (REEs) have been a topic of profound interest for several decades especially in the present age of electronic and digital revolution. India has the world’s richest beach sands with REEs, yet it imports some strategic REEs to fulfil its demand. It’s high time to explore alternative sources to meet its demand and coal ash from Thermal Power Stations (TPS) can be a very good alternative resource. In the present study, coal and coal fly ash (CFA) from seven Indian TPSs have been evaluated for estimation of REEs and variations in minerals compositions. Mineralogy of the samples is estimated using X-Ray diffraction (XRD) technique. Coal samples mostly consist of quartz and kaolinite however phase transformations of minerals occurred due to high temperature treatment during combustion. CFA mostly contains quartz and mullite. REEs have been determined by Inductively Coupled Plasma-Atomic Emission Spectrometry (ICP-AES) and considerable occurrence of any specific REE is not observed. Among the studied TPSs, Pr has the highest concentration among REEs in ash, reaching up to 63 ppm. The Outlook Coefficient (Cout) of REEs is in the range of 0.3–4.5 and 0.1–1.2 for coal and CFA respectively. In this research paper, Enrichment Coefficient (Δe) has been introduced to see the enrichment of REE in CFA with respect to the mother coal and a graph of Δe vs glassy phase has been plotted to observe the partitioning of REEs. Occurrence of Light REEs is more prominent than Heavy REEs. issue-copyright-statement© Geological Society of India 2022 ==== Body pmcAcknowledgements SM likes to thank Science and Engineering Research Board, Govt. of India for Research Grant (No. EMR/2017/000856). We thankfully acknowledge all the Power Plants for allowing us to collect the samples. Authors thank SAIF, IIT, Bombay for providing the ICP - AES analyses and RQA Research Group, CSIR - CIMFR, Dhanbad for providing the proximate analyses of coals. We also thankful to Director, CSIR- Central Institute of Mining and Fuel Research, Dhanbad for allowing publishing this manuscript. ==== Refs References Balaram V Rare earth elements: A review of applications, occurrence, exploration, analysis, recycling, and environmental impact Geosci. Front. 2019 10 1285 1303 10.1016/j.gsf.2018.12.005 Blissett RS Smalley N Rowson NA An investigation into six coal fly ashes from the United Kingdom and Poland to evaluate rare earth element content Fuel 2014 119 236 239 10.1016/j.fuel.2013.11.053 Central Electricity Authority (2019) Annual Report on Fly Ash Generation and its utilization at coal/lignite based Thermal Power Stations in the country for the year 2018–19. Dey S Saini MK Narayan JP Chaudhury N Prediction of gross calorific value of Indian non-caking coals on the basis of ash and moisture Jour. Mines Met. Fuels 2012 60 1&2 31 38 Dong Y Sun X Wang Y Huang C Zhao Z The sustainable and efficient ionic liquid-type saponification strategy for rare earth separation processing ACS Sustain. Chem. Eng 2016 4 1573 1580 10.1021/acssuschemeng.5b01499 Firman, Haya, A. (2021) Study on the Potential Rare Earth Elements in Coal Combustion Product from Banjarsari Power Plant, South Sumatera. IOP Conf. Ser.: Mater. Sci. Eng., v.1125. doi:10.1088/1757-899X/1125/1/012003 Hassas BV Rezaee M Zhou C Pisupati SV Precipitation of rare earth elements from acid mine drainage by CO2 mineralization process Chem. Eng. Jour. 2020 399 125716 10.1016/j.cej.2020.125716 Kambekar A Haldive SA Experimental Study on Combined Effect of Fly Ash and Pond Ash on Strength and Durability of Concrete IJSER 2013 4 5 81 86 Ketris MP Yudovich YE Estimations of Clarkes for Carbonaceous biolithes: World averages for trace element contents in black shales and coals Internat. Jour. Coal Geol. 2009 78 135 148 10.1016/j.coal.2009.01.002 Key World Energy StatisticsInternational Energy AgencyKolker A Scott C Hower JC Vazquez JA Lopano CL Dai S Distribution of rare earth elements in coal combustion fly ash, determined by SHRIMP-RG ion microprobe Internat. Jour. Coal Geol. 2019 184 1 10 Krishnamurthy P Rare Metal (RM) and Rare Earth Element (REE) Resources: World Scenario with Special Reference to India Jour. Geol. Soc. India 2020 95 465 474 10.1007/s12594-020-1463-7 Kumari P Singh AK Wood DA Hazra B Predictions of Gross Calorific Value of Indian Coals from their Moisture and Ash Content Jour. Geol. Soc. India 2019 93 437 442 10.1007/s12594-019-1198-5 Lin R Howard B Roth E Bank T Granite E Soong Y Enrichment of Rare Earth Elements from Coal and Coal By-Products by Physical Separations Fuel 2017 200 506 520 10.1016/j.fuel.2017.03.096 Maity S Coal Fly Ash (CFA) as a source of Rare Earth Elements (REE): an alternative route of value added utilization of Indian Coal Fly Ash ENCO — 2019, 20 – 22 Feb, 2019, Vigyan Bhavan, New Delhi 2019 1 474 479 Pan J Hassas BV Rezaee M Zhou C Pisupati SV Recovery of rare earth elements from coal fly ash through sequential chemical roasting, water leaching, and acid leaching processes Jour. Clean. Prod. 2020 284 124725 10.1016/j.jclepro.2020.124725 Rybak A Rybak A Characteristics of Some Selected Methods of Rare Earth Elements Recovery from Coal Fly Ashes Metals 2021 11 1 42 10.3390/met11010142 Seredin VV A new method for primary evaluation of the outlook for rare earth element ores Geol. Ore Depos. 2010 52 5 428 433 10.1134/S1075701510050077 Seredin VV Dai S Coal deposits as potential alternative sources for lanthanides and yttrium Internat. Jour. Coal Geol. 2012 94 67 93 10.1016/j.coal.2011.11.001 Stuckman MY Lopano CL Granite EJ Distribution and speciation of rare earth elements in coal combustion by-products via synchrotron microscopy and spectroscopy Internat. Jour. Coal Geol. 2018 195 125 138 10.1016/j.coal.2018.06.001 U.S. Geological Survey (2020) Mineral Commodity Summaries 2020; U.S. Geol. Journal Surv. doi:10.3133/mcs2020 Wang N Sun X Zhao Q Yang Y Wang P Leachability and adverse effects of coal fly ash: A review Jour. Hazard. Mater. 2020 396 122725 10.1016/j.jhazmat.2020.122725 32353729 Yuan Z Chengyi W Leiming X Yunxiang N The Distribution of Trace Elements in Granitoids in the Nanling Region of China Chin. Jour. Geochem. 1993 12 30 193 205 Zhang W Rezaee M Bhagavatula A Li Y Groppo J Honaker R A review of the occurrence and promising recovery methods of rare earth elements from coal and coal by products Internat. Jour. Coal Prep. Util. 2015 35 295 330 10.1080/19392699.2015.1033097
PMC009xxxxxx/PMC9005623.txt	==== Front J Geol Soc India Journal of the Geological Society of India 0974-6889 Geological Society of India New Delhi 2004 10.1007/s12594-022-2004-3 Original Article Seismic Site Characterization Using Ambient Noise and Earthquake HVSR in the Easternmost Part of Shillong Plateau, India Vijayan Athira Agrawal Mohit mohit@iitism.ac.in Gupta Ravindra K. grid.417984.7 0000 0001 2184 3953 Department of Applied Geophysics, Indian Institute of Technology (Indian School of Mines), Dhanbad, 826 004 India 13 4 2022 2022 98 4 471478 3 6 2021 4 10 2021 © Geological Society of India, Bengaluru, India 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. This study provides site characterization results using passive seismic techniques for the easternmost part of Shillong plateau of Northeastern India. The prime objective of this research is to generate the Horizontal-to-Vertical Spectral Ratios (HVSR) from earthquakes and ambient microtremor data for generation of 1-D shear wave velocity profiles to reveal the basement depths and fundamental resonance frequencies underneath three stations of Shillong plateau. The earthquake waveforms are recorded for a period of eighteen months (i.e. December 2018 to June 2020) using broadband seismometer; while the ambient microtremor data is collected from a three component highly sensitive velocity meter. The earthquake HVSRs and noise HVSRs are consistent with each other and range from ∼3.6 Hz to ∼14.5 Hz. The Rayleigh wave ellipticity are generated from the corresponding HVSR curves for inversion to determine compressional and shear wave velocity structure. We report that the compressional and shear wave velocity profiles match fairly well for both earthquakes and ambient microtremor data. Overall, Vs30 lies in the range of 480 m/s to 2600 m/s at all locations in the easternmost part of Shillong plateau; while the sedimentary layer velocity lies between 450 m/s and 1100 m/s. The thickness of the sedimentary layer is assessed from the velocity profiles and using an empirical formula makes a good match and varies from 11.3 m to 31.2 m. issue-copyright-statement© Geological Society of India 2022 ==== Body pmcAcknowledgement The authors acknowledge the support of the Head of the Department of Applied Geophysics for providing necessary infrastructure to successfully complete this study. Authors are grateful to associate editor, Dr. Prantik Mandal, and two anonymous reviewers for providing constructive comments which have eventually improved the quality of manuscript. The first author is also thankful to Prof. S.K. Pal who provided the support for ‘Tromino’. All the departmental faculty members are thankfully acknowledged for their timely support during Covid-19 pandemic, which helped us significantly learn the complex concepts associated with this research. This research was funded by the Indian Department of Science and Technology (Project No. ECR/2016/001271 and SR/FST/ES-I/2017/12). ==== Refs References Baro O Kumar A A review on the tectonic setting and seismic activity of the Shillong plateau in the light of past studies Disaster Advances 2015 8 7 34 45 Baro O Kumar A Seismic source characterization for the Shillong Plateau in Northeast India Jour. Seismol. 2017 21 5 1229 1249 10.1007/s10950-017-9664-2 Baro O Kumar A Ismail-Zadeh A Seismic hazard assessment of the Shillong Plateau, India Geomatics, Natural Hazards and Risk 2018 9 1 841 861 10.1080/19475705.2018.1494043 Bignardi S The uncertainty of estimating the thickness of soft sediments with the HVSR method: A computational point of view on weak lateral variations Jour. Appl. Geophys. 2017 145 28 38 10.1016/j.jappgeo.2017.07.017 Bilham R Tom La Touche and the great Assam earthquake of 12 June 1897: Letters from the epicenter Seismol. Res. Lett. 2008 79 3 426 437 10.1785/gssrl.79.3.426 BIS (Bureau of Indian Standards). (2016). Indian standard criteria for earthquake resistant design of structures. IS 1893-16, New Delhi, India. Biswas, R., Baruah, S. and Bora, D.K. (2015). Mapping sediment thickness in Shillong City of Northeast India through empirical relationship. Jour. Earthquakes, v.2015, Article ID 572619. Biswas R Baruah S Bora N Assessing shear wave velocity profiles using multiple passive techniques of Shillong region of northeast India Natural Hazards 2018 94 3 1023 1041 10.1007/s11069-018-3453-2 Bora DK Baruah S Depth of mid-crustal discontinuity from reflected seismic waves on local earthquake seismograms recorded at Shillong Plateau, Northeast India Geomatics, Natural Hazards and Risk 2012 3 4 355 364 10.1080/19475705.2012.668564 Bora DK Baruah S Mapping the crustal thickness in Shillong-Mikir Hills Plateau and its adjoining region of northeastern India using Moho reflected waves Jour. Asian Earth Sci. 2012 48 83 92 10.1016/j.jseaes.2011.12.007 Bora DK Hazarika D Borah K Rai SS Baruah S Crustal shear-wave velocity structure beneath northeast India from teleseismic receiver function analysis Jour. Asian Earth Sci. 2014 90 1 14 10.1016/j.jseaes.2014.04.005 Borah K Bora DK Goyal A Kumar R Crustal structure beneath northeast India inferred from receiver function modeling Phys. Earth Plane. Inter. 2016 258 15 27 10.1016/j.pepi.2016.07.005 Coudurier-Curveur A Tapponnier P Okal E Van der Woerd J Kali E Choudhury S Karaka Ç A composite rupture model for the great 1950 Assam earthquake across the cusp of the East Himalayan Syntaxis Earth Planet. Sci. Lett. 2020 531 115928 10.1016/j.epsl.2019.115928 Devi E U Rao N P Kumar M R Modelling of sPn phases for reliable estimation of focal depths in northeastern India Curr. Sci. 2009 96 1251 1255 Duarah BP Phukan S Understanding the tectonic behaviour of the Shillong plateau, India using remote sensing data Jour. Geol. Soc. India 2011 77 2 105 112 10.1007/s12594-011-0013-8 Fäh D Kind F Giardini D A theoretical investigation of average H/V ratios Geophys. Jour. Internat. 2001 145 2 535 549 10.1046/j.0956-540x.2001.01406.x Fäh D Kind F Giardini D Inversion of local S-wave velocity structures from average H/V ratios, and their use for the estimation of site-effects Jour.f Seismol. 2003 7 4 449 467 10.1023/B:JOSE.0000005712.86058.42 Fäh, D., Wathelet, M., Kristekova, M., Havenith, H., Endrun, B., Stamm, G., … and Cornou, C. (2009). Using ellipticity information for site characterisation. NERIES JRA4 “Geotechnical Site Characterisation”. task B, 2. Gupta, R.K., Agrawal, M., Pal, S.K. and Das, M. K. (2021). Seismic site characterization and site response study of Nirsa (India). Natural Hazards, doi:10.1007/s11069-021-04767-w. Gupta RK Agrawal M Pal SK Kumar R Srivastava S Site characterization through combined analysis of seismic and electrical resistivity data at a site of Dhanbad, Jharkhand, India Environ. Earth Sci. 2019 78 226 10.1007/s12665-019-8231-2 Ibs-von Seht M Wohlenberg J Microtremor measurements used to map thickness of soft sediments Bull. Seismol. Soc. Amer. 1999 89 1 250 259 10.1785/BSSA0890010250 Kalita BC Ground water prospects of Shillong Urban Aglomerate 1998 Meghalaya Central Ground Water Board Kayal JR De R Microseismicity and tectonics in northeast India Bull. Seismol. Soc. Amer. 1991 81 1 131 138 10.1785/BSSA0810010131 Khan P K Chakraborty P P The seismic b-value and its correlation with Bouguer gravity anomaly over the Shillong Plateau area: tectonic implications Jour. Asian Earth Sci. 2007 29 1 136 147 10.1016/j.jseaes.2006.02.007 Kumar D Reddy DV Pandey AK Paleoseismic investigations in the Kopili fault zone of North East India: Evidences from liquefaction chronology Tectonophysics 2016 674 65 75 10.1016/j.tecto.2016.02.016 Mahajan AK Kumar P Site characterisation in Kangra Valley (NW Himalaya, India) by inversion of H/V spectral ratio from ambient noise measurements and its validation by multichannel analysis of surface waves technique Near Surface Geophysics 2018 16 3 314 327 10.3997/1873-0604.2018008 Mahesh P Catherine JK Gahalaut VK Kundu B Ambikapathy A Bansal A Kalita S Rigid Indian plate: constraints from GPS measurements Gondwana Res. 2012 22 3–4 1068 1072 10.1016/j.gr.2012.01.011 Martorana R Capizzi P D’Alessandro A Luzio D Di Stefano P Renda P Zarcone G Contribution of HVSR measures for seismic microzonation studies Annals of Geophysics 2018 61 2 SE225 10.4401/ag-7786 Mukul M Jade S Bhattacharyya AK Bhusan K Crustal shortening in convergent orogens: Insights from global positioning system (GPS) measurements in northeast India Jour. Geol. Soc. India 2010 75 1 302 312 10.1007/s12594-010-0017-9 Nakamura, Y. (1989) A method for dynamic characteristics estimation of subsurface using microtremor on the ground surface. Railway Technical Research Institute, Quarterly Reports, 30(1). Nath SK Thingbaijam KKS Raj A Earthquake hazard in Northeast India-A seismic microzonation approach with typical case studies from Sikkim Himalaya and Guwahati city Jour. Earth System Sci. 2008 117 2 809 831 10.1007/s12040-008-0070-6 Nath, S. K., Vyas, M., Pal, I. and Sengupta, P. (2005) A seismic hazard scenario in the Sikkim Himalaya from seismotectonics, spectral amplification, source parameterization, and spectral attenuation laws using strong motion seismometry. Jour. Geophys. Res.: Solid Earth, v.110(B1). Nayak GK Rao VK Rambabu HV Kayal JR Pop?up tectonics of the Shillong Plateau in the great 1897 earthquake (Ms 8.7): Insights from the gravity in conjunction with the recent seismological results Tectonics 2008 27 1 TC(1018) 10.1029/2006TC002027 Nogoshi M Igarashi T On the propagation characteristics of microtremors Jour. Seism. Soc. Japan 1970 23 264 280 Odum, J.K., Williams, R.A., Stephenson, W.J., Worley, D.M., von Hillebrandt-Andrade, C., Asencio, E., Irizarry, H. and Cameron, A., (2007) Near-surface shear wave velocity versus depth profiles, Vs 30, and NEHRP classifications for 27 sites in Puerto Rico. U. S. Geological Survey. Oldham T A catalogue of Indian earthquakes from the earliest time to the end of A.D. 1869 Mem. Geol. Surv. India 1882 19 163 215 Olsen KB Site amplification in the Los Angeles basin from three-dimensional modeling of ground motion Bull. Seismol. Soc. Amer. 2000 90 6B S77 S94 10.1785/0120000506 Pandey AK Roy PNS Baidya PR Gupta AK Estimation of current seismic hazard using Nakamura technique for the Northeast India Natural Hazards 2018 93 2 1013 1027 10.1007/s11069-018-3338-4 Parolai S Bormann P Milkereit C New relationships between Vs, thickness of sediments, and resonance frequency calculated by the H/V ratio of seismic noise for the Cologne area (Germany) Bull. Seismol. Soc. Amer. 2002 92 6 2521 2527 10.1785/0120010248 Raghukanth STG Nadh Somala S Modeling of strong-motion data in northeastern India: Q, stress drop, and site amplification Bull. Seismol. Soc. Amer. 2009 99 2A 705 725 10.1785/0120080025 Ramesh, D.S., Ravi Kumar, M., Uma Devi, E., Solomon Raju, P., and Yuan, X. (2005) Moho geometry and upper mantle images of northeast India. Geophys. Res. Lett., v.32(14). Raoof J Mukhopadhyay S Koulakov I Kayal J R 3?D seismic tomography of the lithosphere and its geodynamic implications beneath the northeast India region Tectonics 2017 36 5 962 980 10.1002/2016TC004375 Reddy C Sunil P S Prajapati S Ponraj M Amrithraj S Geodynamics of the NE Indian lithosphere: geodetic and seismotectonic perspective Mem. Geol. Soc. India 2011 77 241 250 Sambridge M Geophysical inversion with a neighbourhood algorithm-I. Searching a parameter space. Geophys Jour. Internat. 1999 138 2 479 494 Sambridge M Geophysical inversion with a neighbourhood algorithm-II. Appraising the ensemble Geophys. Jour. Internat. 1999 138 3 727 746 10.1046/j.1365-246x.1999.00900.x Sandhu, M., Sharma, B., Mittal, H., & Chingtham, P. (2020). Analysis of the site effects in the North East region of India using the recorded strong ground motions from moderate earthquakes. Jour. Earthquake Engg., pp.1–20. SESAME. (2004) Guidelines for the implementation of the H/V spectral ratio technique on ambient vibrations-measurements, processing and interpretations, SESAME European research project EVG1-CT-2000-00026, deliverable D23.12. SESAME: Site EffectS Assessment Using AMbient Excitations, March, pp.1–62. Sharma B Chopra S Chingtham P Kumar V A study of characteristics of ground motion response spectra from earthquakes recorded in NE Himalayan region: an active plate boundary Natural Hazards 2016 84 3 2195 2210 10.1007/s11069-016-2543-2 Singh AP Purnachandra Rao N Ravi Kumar M Hsieh M C Zhao L Role of the Kopili Fault in Deformation Tectonics of the Indo-Burmese Arc Inferred from the Rupture Process of the 3 January 2016 M w 6.7 Imphal Earthquake Bull. Seismol. Soc. Amer. 2017 107 2 1041 1047 10.1785/0120160276 Srinivasan P Sen S Bandopadhaya PC Study of variation of Paleocene-Eocene sediments in the shield areas of Shillong Plateau Rec. Geol. Surv. India 1996 129 77 78 Sukhija, B.S., Reddy, D.V., Kumar, D. and Nagabhushanam, P. (2006) Comment on “Interpreting the style of faulting and paleoseismicity associated with the 1897 Shillong, northeast India, earthquake: Implications for regional tectonism” by CP Rajendran et al. Tectonics, v.25(2). Tillotson E The great Assam earthquake of August 15, 1950 Nature 1951 167 4239 128 130 10.1038/167128a0 Tün M Pekkan E Özel O Guney Y An investigation into the bedrock depth in the Eskisehir Quaternary Basin (Turkey) using the microtremor method Geophys. Jour. Internat. 2016 207 1 589 607 10.1093/gji/ggw294 Vorobieva I Mandal P Gorshkov A Block-and-fault dynamics modelling of the Himalayan frontal arc: Implications for seismic cycle, slip deficit, and great earthquakes Jour. Asian Earth Sci. 2017 148 131 141 10.1016/j.jseaes.2017.08.033 Wathelet M Chatelain J L Cornou C Giulio G D Guillier B Ohrnberger M Savvaidis A Geopsy: A user-friendly open-source tool set for ambient vibration processing Seismol. Res. Lett. 2020 91 3 1878 1889 10.1785/0220190360
PMC009xxxxxx/PMC9005628.txt	==== Front Int J Mach Learn Cybern Int J Mach Learn Cybern International Journal of Machine Learning and Cybernetics 1868-8071 1868-808X Springer Berlin Heidelberg Berlin/Heidelberg 35432624 1555 10.1007/s13042-022-01555-1 Original Article Tuning of data augmentation hyperparameters in deep learning to building construction image classification with small datasets http://orcid.org/0000-0003-2136-9870 Ottoni André Luiz C. andre.ottoni@ufrb.edu.br 12 de Amorim Raphael M. amorimba@gmail.com 2 Novo Marcela S. marcela.novo@ufba.br 3 Costa Dayana B. dayanabcosta@ufba.br 4 1 grid.440585.8 0000 0004 0388 1982 Technologic and Exact Center, Federal University of Recôncavo da Bahia, Cruz das Almas, Brazil 2 grid.8399.b 0000 0004 0372 8259 Electrical Engineering Graduate Program, Federal University of Bahia, Salvador, Brazil 3 grid.8399.b 0000 0004 0372 8259 Department of Electrical and Computer Engineering, Federal University of Bahia, Salvador, Brazil 4 grid.8399.b 0000 0004 0372 8259 Department of Structural and Construction Engineering, Federal University of Bahia, Salvador, Brazil 13 4 2022 2023 14 1 171186 24 5 2021 22 3 2022 © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Deep Learning methods have important applications in the building construction image classification field. One challenge of this application is Convolutional Neural Networks adoption in a small datasets. This paper proposes a rigorous methodology for tuning of Data Augmentation hyperparameters in Deep Learning to building construction image classification, especially to vegetation recognition in facades and roofs structure analysis. In order to do that, Logistic Regression models were used to analyze the performance of Convolutional Neural Networks trained from 128 combinations of transformations in the images. Experiments were carried out with three architectures of Deep Learning from the literature using the Keras library. The results show that the recommended configuration (Height Shift Range = 0.2; Width Shift Range = 0.2; Zoom Range =0.2) reached an accuracy of 95.6% in the test step of first case study. In addition, the hyperparameters recommended by proposed method also achieved the best test results for second case study: 93.3%. Keywords Deep learning Convolutional neural networks Hyperparameter tuning Data augmentation Building construction image classification issue-copyright-statement© Springer-Verlag GmbH Germany, part of Springer Nature 2023 ==== Body pmcIntroduction Deep Learning methods have important applications in the Digital Image Processing field [4, 23, 42, 47]. In this sense, a possible application of Deep Learning is building construction area [8, 10, 14, 41, 53]. In recent literature, there are several applications in this research field, such as: crack detection [8, 54], road crack classification [53], safety guardrail detection [22], structural damage recognition [11], detecting safety helmet [41], safety harness detection [10], classification of rock fragments [50], damage detection of a steel bridge [1], tunnel lining defects [49] and facade defects classification [14]. Deep Learning methods can also be applied in recognition of vegetation in building facades images [32]. In fact, the growth of biological manifestations on building facades may indicate the deterioration and degradation of constructions [2, 24]. In addition, the detection of this pathology in inspection images can assist in the conservation of historic buildings [7, 21, 24, 37]. In this sense, [32] proposes a Deep Learning approach for recognizing vegetation in buildings. Another possibility is to use Deep Learning analysis of roof structures [33]. In the literature, there are several examples of works that investigated the efficient of roofs structure [5, 12, 33, 43]. For example, in a recent study, [33] proposes a methodology to tuning of two hyperparameters (learning rate and optimizer) of Neural Networks in the building roof image classification. It is also worth noting that, one of the relevant factors on [32] and [33] was the experiments with Data Augmentation. [32] verified the improvement in validation accuracy when using Data Augmentation to increase the training database. In fact, Data Augmentation techniques play an important role in the application of Machine Learning in small datasets [4, 9, 42, 51, 52]. This is because, the generation of artificial images directly contributes to increase the capacity for the generalization of the Deep Learning model and thus decrease the chance of overfitting [4, 9]. In this respect, one of the challenges of using Data Augmentation is the definition of which transformations (such as zoom, rotation, flip) will be applied to the images [6, 28, 34, 44, 48]. In terms of Machine Learning, this problem can be treated as in the area of Hyperparameter Tuning [19, 20, 27, 30, 31, 39, 40]. In the literature, some studies have analyzed the influence of Data Augmentation hyperparameter combinations in different applications, such as: plant classification [34], transmission line inspection [44] and covid-19 diagnostic process in chest X-ray radiological imaging [28]. In [48], different types of Data Augmentation methods were analyzed for crack detection in constructions. However, the literature lacks proposals to optimize the combinations of Data Augmentation hyperparameters for the application of Deep Learning in building construction image classification, especially in the recognition of vegetation on building facades and roofs defects classification. The objective of this paper is to propose a rigorous methodology for tuning of Data Augmentation hyperparameters in Deep Learning to building construction image classification with small data sets. For this, two case studies are observed: vegetation recognition in facades [32] and roofs structures analysis [33]. In order to do that, Logistic Regression models [16] will be used to analyze the performance of Convolutional Neural Networks (CNN) [4, 9] trained from 128 combinations of transformations in the images. For comparison purposes, three CNN architectures from the literature will also be adopted: MobileNet [17], DenseNet-121 [18] e CNN8 [32]. This paper is organized into five sections. Section 2 presents theoretical concepts of CNNs and Logistic Regression. Section 3 presents the proposed methodology. Sections 4 and 5 describe the results and conclusions, respectively. Theoretical foundation Convolutional neural networks Convolutional neural networks (CNNs) are Deep Learning methods with several researches in computer vision field [4, 9, 15, 23]. One of the main factors that make CNNs a relevant Machine Learning technique is the ability to automatically extract features from processed images [9]. In addition, another important point is the use of layers and elements with different functionalities in the network architecture, such as [4, 9]:Input layer: receives input signals (e.g .: image). Weights: adjusted during the training process (trainable parameters). Convolutional filters (kernels): have a set of weights, according to their size. For example, if the kernel size is 3×3, then the filter contains 9 trainable weights. Activation function: transforms a signal into a limited output. Some examples are the functions ReLu, softmax and sigmoid. Convolutional Layer: applies the convolution operation between the filters and the input matrix in the layer. As an output, new matrices are generated (feature map), according to the number of kernels in the layer. Pooling: applies a transformation to decrease the input matrix dimensions. For this, statistical functions can be used: maximum (max) or average (avg). Flatten: transforms the matrices resulting from convolutional operations into a single vector. Dropout: randomly disconnects a set of neurons at each training epoch. Fully connected layers: similar to the structures of traditional Artificial Neural Networks, in which, all neurons and layers are connected. Output layer: shows the output of CNN, such as a binary classifier neuron. Thus, in view of the complexity of the CNN architectures, an important factor is the use of tools for efficient implementation [4, 9]. In this line, it is worth mentioning the Keras library [4]. Keras is available on R interface1 and Python language for development of Deep Learning applications. In addition, it allows execution on CPU or GPU. Another relevant factor is the simplicity to use Data Augmentation methods. In this sense, the Keras library was adopted in this work, as described in Sect. 3. Logistic regression The methods based on Linear Regression (simple and multiple) aim to model a continuous output from one or more independent variables [29]. On the other hand, Logistic Regression is a technique addressed for the analysis of categorical data [13, 16]. Moreover, in a logistic function, the variable response is binary or dichotomous. The Logistic Regression model can be represented by Eq. (1) [13]:1 p(x)=exp(β0+∑k=1mβkxk)1+exp(β0+∑k=1mβkxk) where, β0 to βk are the coefficients of the regression model; x are the independent variables; and p(x) is the probability of success. Thus, if p(x) is the probability of an event occurring, then the expression 1-p(x) represents the probability of an event not occurring. The ratio between p(x) and 1-p(x) is called chance (Eq. (2)) [13]:2 chance=p(x)1-p(x) In this line, the neperian logarithm of chance provides a linear model, according to Eq. (3) [13]:3 lnp(x)1-p(x)=β0+∑k=1mβkxk where, this equation is called logit and is a simplification of the Logistic Regression model. Methodology Database of the first case study In this study, the database presented by [32] was used for training, validation and testing of Convolutional Neural Networks in the recognition of vegetation on building facades. The dataset has 390 images, divided into two classes: Class 0: without vegetation on the building’s facade. Class 1: with vegetation on the building’s facade. According to [32], the images of the training and validation datasets were defined from photographs adapted from The Zurich Urban Micro Aerial Vehicle Dataset [26]. These images were recorded in 2015 by an vant from the urban streets of Zurich (Switzerland). On the other hand, the images of the test dataset were selected from the website Pixabay.2 The dataset analyzed during the current study are available from the corresponding author on request or in the web link:drive.google.com/file/d/1l6KA80mZdKqxlpfpenIH57mCYESL3uyq/ Figures 1 and 2 present examples of images from the database of classes 0 and 1, respectively.Fig. 1 Examples of images of the class 0 - without vegetation on the facade Fig. 2 Examples of images of the class 1 - with vegetation on the facade The database (390 images) was partitioned following the same structure proposed by [32]:Training (250 images): 125 images in class 0 (without vegetation) and 125 images in class 1 (with vegetation). Validation (50 images): 25 images in class 0 (without vegetation) and 25 images in class 1 (with vegetation). Test (90 images): 45 images in class 0 (without vegetation) and 45 images in class 1 (with vegetation). Data augmentation Data preprocessing is an important stage in the Machine Learning field [9]. This is because, this step can decrease the learning complexity and improve the accuracy results [9]. In this line, Data Augmentation techniques can be used in the training of CNNs [4, 9, 42, 51, 52]. Data Augmentation is an approach applied mainly to small data learning [4, 42]. In this sense, Data Augmentation methods generate more data for training from the existing images [4]. The aim is to increase the CNN model’s ability to generalize and avoid overfitting [4, 9]. For this, artificial images are created from random transformations in the original data [4]. Zoom, rotation and flip are some examples of possible transformations for the generation of augmented images [9]. In this paper, Data Augmentation methods were applied from Keras library in R software [4]. For this, a image_data_generator() method was adopted [4]. The function image_data_generator() generates batches of data with new modified images from the original data. In this regard, the following random transformations3 were used to increase training data [4]:Rotation range: an integer number that defines the degree range for random rotations. Rotation is a circular movement around a fixed point. The processing images will have random rotations on a predefined range of degrees according with data entrance. Horizontal flip: if this input is“true”them the images will be randomly mirrored in the horizontal direction (left-right). Vertical flip: if this input is“true”them the images will be randomly mirrored in the vertical direction (up-bottom). Shear range: distort the image along an axis to create or rectify the perception angles. There are two shear transformation, X-Shear that shift X coordinates values and Y-Shear that’s shift Y coordinates values. Width shift range: shifts the image randomly to the left or to the right (horizontal shifts). If the value is float and ≤ 1 it will take the percentage of total width as range. For example, in an image that width is 100 pixels and if width_shift_range = 1.0 then it will shift image randomly between -100% to 100% or -100px to 100px. Positive values will shift the image to the right side and negative values will shift the image to the left side. Height shift range: shifts the image randomly to up or down (vertically shifts). If the value is float and ≤ 1 it will take the percentage of total height as range. For example, in an image that height is 100 pixels and if height_shift_range = 1.0 then it will shift image randomly between -100% to 100% or -100px to 100px. Positive values will shift the image to the upside and negative values will shift the image to underside. Zoom range: it will do a randomly augmentation of the image adding new pixels values. It can be specified with the percentage of the zoom as single float or a range as an array. For example, if zoom_range = 0.4 the range will be [0.6, 1.4] between 60% (zoom in) or 140% (zoom out). Figure 3 presents examples of images generated by Data Augmentation with the Keras library.Fig. 3 Examples of images generated by Keras data augmentation: a original image; b–d rotation range (R=40); e horizontal flip (H=TRUE); f vertical flip (V=TRUE); g and h height shift range (He=0.2); l shear range (S=0.2); j–l width shift range (W=0.2); m–p zoom range (Z=0.2) The number of artificially generated images depends on training settings: batch_size, steps_per_epoch and epoch. For example, in this study, these parameters were defined in first phase of experiments as: batch_size = 32; steps_per_epoch = 100; and epoch = 10. Thus, for each simulation were generated randomly around 32,000 new images for training. This value is more than 100 times greater than the number of original photographs for training (250). Neural network architectures In this paper, three CNNs architectures were adopted: CNN-8 [32], DenseNet-121 [18] and MobileNet [17]. Recently, these structures (or variations) are discussed in some papers in research field of building construction image processing with deep learning, such as, in the tasks: crack detection (DenseNet) [46], structural health monitoring (FC-DenseNet) [38], detecting safety helmet (SSD MobileNet) [41], road damage detection (SSD MobileNet) [25] and recognition of vegetation in buildings (CNN-8) [32]. In this study, CNN architectures were used for binary classification, for example, class 0 (without vegetation on the building’s facade) and class 1 (with vegetation on the building’s facade) [32]. For this, the keras_model _sequential() method in the Keras library was used [4], as described below:CNN-8: CNN architecture used by [32] to vegetation image recognition in buildings. The structure has 8 layers and 3,985,345 trainable parameters. In addition, this architecture is based on a model proposed by [4], originally with 12 layers. DenseNet-121: Dense Convolutional Network is an architecture proposed by [18]. This structure is characterized by connecting each layer to all other layers (dense connection). Moreover, it has 7,479,169 trainable parameters. To use this architecture, the application_densenet121() method in the Keras library was adopted. MobileNet: CNN architecture proposed by [17] for mobile and embedded vision applications. The structure uses deptwise separable convolutions (factorized convolutions). In addition, it has 28 layers and 3,732,289 trainable parameters. To use this architecture, the application_mobilenet() method in the Keras library was adopted. In all experiments, CNN architectures were trained with an adagrad optimizer and a learning rate of 0.01. In addition, the dimensions 50×50×3 were standardized as input to the neural network: input_shape = c(50, 50, 3). It is also noteworthy that all three architectures were configured with the last two layers as fully connected. In the last layer having the binary classifier neuron with sigmoid activation function [4]. Hyperparameter tuning Design of experiments In this section, the design of experiments for tuning of data augmentation hyperparameters with Logistic Regression [16] is described. The simulations of the Convolutional Neural Network models were conducted in the R software [35] with the Keras library [4]. For this, an Intel Core i7-8565 (CPU) and NVIDIA GeForce MX110 (GPU) were used. For the experiments evaluation were used three metrics: accuracy in validation or testing (Acc), number of images correctly classified (C) and number of images incorrectly classified, that is, errors (E). Equations (4) to (6) present these formulas.4 Acc=TP+TNTP+TN+FP+FN, 5 C=TP+TN, 6 E=FP+FN, where,TP: true positives, that is, correct classifications in class 1 (facade with vegetation). FN: false negatives, that is, incorrect classifications in class 1 (facade with vegetation). TN: true negatives, that is, correct classifications in class 0 (facade without vegetation). FP: false positives, that is, incorrect classifications in class 0 (facade without vegetation). The experiments were conducted in three stages: Data Augmentation Hyperparameters. Data Augmentation and CNN Architectures. Test Experiments. In the first phase, seven hyperparameters were defined for adjustment, each with two levels of treatments (0 - without transformation and 1 - with transformation):Rotation Range (R): 0 or 40. Horinzontal Flip (H): FALSE or TRUE. Vertical Flip (V): FALSE or TRUE. Height Shift Range (He): 0 or 0.2. Shear Range (S): 0 or 0.2. Width Shift Range (W): 0 or 0.2. Zoom Range (Z): 0 or 0.2. Thus, a total of 128 (27) combinations of data augmentation hyperparameters were analyzed in the first stage. For each configuration, five CNN models (repetitions) were trained in 10 epochs with 100 steps (steps por epoch) adopting the MobileNet architecture [17]. The metrics observed in this phase were the total number of images correctly classified (C) and errors (E) in the validation dataset, used to adjust a Logistic Regression model. In the second stage of experiments, the best combinations of the first phase were used. In addition, three CNN architectures from the literature were adopted: MobileNet [17], DenseNet-121 [18] e CNN8 [32]. For each combination (configuration of data augmentation × architecture), 5 repetitions were performed with 20 epochs. The total correct classifications (C) and errors (E) were observed in the validation dataset. Furthermore, the accuracy (Acc) in the validation step was also analyzed. Finally, in the third phase of experiments, the hyperparameter combinations performance were analyzed in the test dataset. In this sense, new trainings were carried out with the data augmentation configurations selected in 5 repetitions with 30 epochs. For each of the CNN models trained in this phase, the accuracy in the classification of the test database was analyzed. Logistic regression method In this paper, the method for hyperparameter tuning uses Logistic Regression [16]. The objective is to evaluate the probability of hits and errors in the building construction image classification, according to the settings of data augmentation. For this, the response variable (y) is binary and was modeled as follows:y=1:correctimageclassification.0:incorrectimageclassification. For the first step, the explanatory variables (x1, x2, x3, x4, x5, x6 and x7) refer to the seven hyperparameters analyzed:x1:Rotation Range (R)x2:Horizontal Flip (H)x3:Vertical Flip (V)x4:Height Shift Range (He)x5:Shear Range (S)x6:Width Shift Range (W)x7:Zoom Range (Z) The Equation (7), in turn, presents the Logistic Regression model (logito format) proposed for the recommendation of hyperparameters:7 lnp(x)1-p(x)=β0+β1x1+β2x2+β3x3+……+β4x4+β5x5+β6x6+β7x7 The coefficients (β) of Eq. (7) can be obtained by the maximum likelihood method [13]. Then, the regression coefficients hypothesis test must be performed. In this sense, the significance of the effects of each variable present in the model is analyzed in two hypotheses:H0:βk=0,Ha:βk≠0. When the initial hypothesis (H0) is accepted (p>0.05), the variable xk (associated with the βk coefficient) does not have statistical significance in the model. On the other hand, if the alternative hypothesis is accepted (p<0.05), the hyperparameter (k) has significance in the Logistic Regression model. The adjusted coefficients (β) also may perform the calculations of the odds associated with each hyperparameter configuration. In this aspect, the OR metric represents the odds ratio of correct classification between the two levels of a hyperparameter. For example, the odds ratio for hyperparameter 1 (Rotation Range) is given by Eq. (8):8 OR1=exp(β1), where OR1 is the odds ratio of level 1 (R=40) in relation to level 0 (R=0) of hyperparameter 1 (Rotation Range). Thus, if OR1>1 the chance of success in adopting R=40 is greater than R=0. Otherwise (OR1<1), then the chance of CNN correctly classifying an image is greater if trained without the rotation transformation. A similar analysis can be made after calculating the odds ratio of the other analyzed hyperparameters. Thus, Eq. () presents the general formulation for the odds ratio:9 ORk=exp(βk). Therefore, odds ratio indices are used to define hyperparameter configurations for the sequence of experiments, as will be described in the next subsection (HPtuningLogReg Algorithm). In the sequence, logistic regression models are also used to analyze the results of the second stage of experiments. In this case, the objective is to evaluate the influence of the selected hyperparameter combinations for the three CNN architectures adopted (CNN8 [32], DenseNet-121 [18] and MobileNet [17]). For this, three logistic regression models are adjusted (one per architecture) and the odds ratio indices for the hyperparameter configurations are observed. HPtuningLogReg algorithm Algorithm 1 presents the method proposed in R language for tuning of data augmentation hyperparameters with Logistic Regression: HPtuningLogReg Algorithm. The code has been divided into four steps: data input, adjustment of the logistic regression model, hyperparameter tuning and summary. In the first phase (lines 1 to 14), the results of the experiments are read and prepared for the method sequence. The“correct”and “error”vectors store the number of correct and incorrect classifications, respectively, for each of the observed hyperparameter configurations. Then, in line 16, the function glm of the R language is used to adjust the Logistic Regression model. In addition, the anova method (line 17) is also used to perform statistical tests of analysis of variance and to calculate the p-value (“paov”). In sequence, from the adjusted coefficients (“modelglm$coefficients”), the odds ratio measures are calculated (line 18). In phase 3 (lines 19 to 45), the Logistic Regression model is adopted to hyperparameter tuning. For this, a repetition loop is performed by varying the hyperparameter index (data augmentation transformation type): 1 - R, 2 - H, 3 - V, 4 - He, 5 - S, 6 - W and 7 - Z. In line 21, the statistical significance of the variable present in the model is analyzed. If the alternative hypothesis (H1) is accepted (“paov$‘Pr(>Chi)‘[i+1]<0.05”) there is significance for the coefficient βk, that is, there is a statistical difference between the two treatments of the hyperparameter k. In this case, the value of the odds ratio is presented and then recommended hyperparameter level with the greatest chance of success in the validation dataset image classification (lines 22 to 35). On the other hand, if initial hypothesis (H0) is accepted (lines 36 to 45), there is no statistically significant difference between the two values of the hyperparameter k (p>0.05). Finally, default treatment (“H[i,2]”) is recommended, that is, the transformation k in the data augmentation process should not be applied. In step 4, a summary of the recommended hyperparameter values is presented. In this case, only the hyperparameters whose decision variables received level 1 (C[i] == 1) in step 3 are shown. Hyperparameter tuning to second small dataset In this case study, another type of problem in buildings was analyzed: gutter integrity and cleanliness pathology in roofs [43]. For this, images were used from the database presented and described by [33, 43, 45] and made available by the Research Group in Construction Technology and Management (School of Engineering - UFBA).4 These images were captured from roof inspections with an unmanned aerial vehicle. In a previous study, [33] used this dataset in experiments to tuning of two CNN hyperparameters (learning rate and optimizer). For this, the images were divided into two classes: (0) roofs with clean gutters and (1) roofs with dirty gutters. Thus, the database adopted by [33] has 220 images, separated for the training, validation and test phases:Training (160 images): 80 images in class 0 and 80 images in class 1. Validation (30 images): 15 images in class 0 and 15 images in class 1. Test (30 images): 15 images in class 0 and 15 images in class 1. Figures 4 and 5 present examples of images (two classes) of the second small dataset.Fig. 4 Examples of images of the class 0 (roofs with clean gutters) in second small dataset Fig. 5 Examples of images of the class 1 (roofs with dirty gutters) in second small dataset In this sense, the Hyperparameter Tuning methodology (Section 3.4) was adopted in new experiments with this second case study. Then, HPtuningLogReg was applied to tuning of Data Augmentation hyperparameters. The dataset analyzed during the second case study is available from the corresponding author on request or in the web link:https://abre.ai/dataset2 Results Results of first small dataset This section presents the results for the first case study: recognition of vegetation on building facades. Stage 1: Data augmentation hyperparameters In stage 1, HPtuningLogReg algorithm was adopted to adjust the logistic regression model and tuning of Data Augmentation hyperparameters. For this, results of 128 hyperparameters combinations analyzed were used. The Equation (10) presents the adjusted linear model (logito).10 y=β0+β1x1+β2x2+β3x3+··+β4x4+β5x5+β6x6+β7x7=1.524-0.005x1-0.118x2-0.130x3+··+0.179x4-0.0085x5+0.114x6+0.111x7 Table 1 shows the results of the test statistic (p), recommended values and odds ratio (OR) per hyperparameter.Table 1 Results of Data Augmentation hyperparameter tuning with logistic regression in stage 1. Hyperparameter β p Value x OR Rotation R. (R) – 0.005 0.86 0 0 0.995 Hor. Flip (H) – 0.118 0.00 False 0 0.888 Vertical Flip (V) – 0.130 0.00 False 0 0.878 Height S. R. (He) 0.179 0.00 0.2 1 1.196 Shear Range (S) – 0.008 0.79 0 0 0.992 Width S. R. (W) 0.114 0.00 0.2 1 1.121 Zoom Range (Z) 0.111 0.00 0.2 1 1.118 Bold values indicate the hyperparameters with OR > 1 and p < 0.05 The results of Table 1 shows that the effects related to transformations of Rotation Range and Shear Range have no statistical effect (p>0.05). Thus, the algorithm recommended the use of the default value for these hyperparameters (R=0 and S=0). On the other hand, the effects of the variables referring to Horizontal Flip and Vertical Flip showed statistical significance (p<0.05). However, the odds ratio values for these hyperparameters were less than 1 (OR<1). Thus, HPtuningLogReg algorithm recommended the use of H=0 and V=0. Table 1 also presents the results for the other three transformations: Height Shift Range, Width Shift and Zoom. The effects of these variables were statistically significant (p<0.05). The recommended value of Height Shift method was 0.2 with an estimated odds ratio of 1.196. In this respect, the adjusted model reveals that adopting the level xHe=1 has around 20% more chances of success in the image classification, in relation to not adopting this transformation in the training base. The adjusted values for Width Shift and Zoom were also 0.2, with odds ratios of 1.121 and 1.118, respectively. Thus, it is estimated that the chance of correct image classification when using W=0.2 or Z=0.2 is around 12% greater than performing the training without these Data Augmentation effects. Thus, from the Logistic Regression results, the HPtuningLogReg algorithm recommended three transformations in the images for the training process: Height Shift Range, Width Shift Range e Zoom Range. These hyperparameters analyzed at two levels each, result in eight combinations of Data Augmentation transformations (2×2×2). Table 2 presents these combinations set and their respective levels of decision variablesTable 2 Hyperparameter combinations of data augmentation selected in stage 1 Comb. He W Z xHe xW xZ 1 0 0 0 0 0 0 2 0 0 0.2 0 0 1 3 0 0.2 0 0 1 0 4 0 0.2 0.2 0 1 1 5 0.2 0 0 1 0 0 6 0.2 0 0.2 1 0 1 7 0.2 0.2 0 1 1 0 8 0.2 0.2 0.2 1 1 1 The hyperparameter combinations presented in Table 2 were used in the next stage of experiments, as shown in the following section. Stage 2: Data augmentation and CNN architectures In stage 2, the hyperparameter combinations defined in the previous phase were evaluated in conjunction with three architectures in the literature: CNN8 [32], DenseNet-121 [18] and MobileNet [17]. Table 3 presents the results of accuracy in the validation step and the statistical metrics of the logistic regression models (OR and p).Table 3 Results of validation accuracy (%) and statistical metrics for each method (CNN architecture + data augmentation combination) Arch. Comb. 1 2 3 4 5 Mean OR p 1 78.0 78.0 78.0 76.0 82.0 78.4 1.000 – 2 86.0 82.0 78.0 78.0 76.0 80.0 1.102 0.66 3 86.0 82.0 86.0 86.0 84.0 84.8 1.537 0.07 CNN8 4 92.0 88.0 86.0 84.0 92.0 88.4 2.099 0.00 5 84.0 90.0 80.0 86.0 86.0 85.2 1.586 0.05 6 84.0 84.0 84.0 86.0 82.0 84.0 1.445 0.11 7 90.0 92.0 88.0 90.0 88.0 89.6 2.374 0.00 8 88.0 88.0 86.0 90.0 86.0 87.6 1.946 0.01 1 88.0 84.0 90.0 88.0 84.0 86.8 1.000 – 2 90.0 86.0 94.0 92.0 84.0 89.2 1.256 0.41 3 92.0 92.0 88.0 90.0 86.0 89.6 1.310 0.33 DenseNet-121 4 90.0 88.0 86.0 90.0 92.0 89.2 1.256 0.41 5 86.0 90.0 94.0 88.0 86.0 88.8 1.206 0.49 6 90.0 88.0 90.0 88.0 92.0 89.6 1.310 0.33 7 90.0 94.0 92.0 92.0 90.0 91.6 1.658 0.09 8 92.0 96.0 96.0 92.0 94.0 94.0 2.382 0.01 1 76.0 74.0 76.0 78.0 76.0 76.0 1.000 – 2 84.0 88.0 88.0 88.0 90.0 87.6 2.231 0.00 3 82.0 84.0 90.0 82.0 78.0 83.2 1.564 0.05 MobileNet 4 92.0 88.0 90.0 88.0 90.0 89.6 2.720 0.00 5 82.0 80.0 78.0 84.0 82.0 81.2 1.364 0.16 6 84.0 86.0 86.0 86.0 90.0 86.4 2.006 0.00 7 88.0 92.0 88.0 86.0 94.0 89.6 2.721 0.00 8 92.0 90.0 90.0 92.0 90.0 90.8 3.117 0.00 Bold values indicate the data augmentation combinations with p < 0.05 From the Table 3 it is possible to observe that the highest accuracy average (87.6%) for the CNN8 architecture was achieved by adopting the combination 7 (He=0.2; W=0.2; Z=0). In this case, adopting configuration 7 has approximately 2 times more chances of success (OR=2.374) in the classification in relation to the reference combination (He=0; W=0; Z=0). It is also noteworthy that the four combinations (4, 5, 7 and 8) showed statistical significance (p≤0.05). On the other hand, when analyzing the results of DenseNet-121 in the Table 3, the highest mean accuracy (94.0%) was obtained by the combination 8. In addition, only this configuration (He=0.2; W=0.2; Z=0.2) was statistically significant (level of 5%). The experiments with the MobileNet architecture, revealed that adopt the configuration 8 has around 3 times more chances of success (OR=3.117) in image classification. Moreover, five combinations (2, 3, 5, 6, 7 and 8) are statistically different (p≤0.05). In this sense, configuration 8 (He=0.2; W=0.2; Z=0.2) was the only to present statistical significance for the three architectures. In addition, the odds ratio for this combination was over 1.9 for CNN8, DenseNet-121 and MobileNet. Thus, combination 8 was selected for the sequence of experiments in the test stage. To illustrate, Figs. 6 and 7 present samples of images generated by Data Augmentation, adopting the combination 8: He=0.2; W=0.2; Z=0.2.Fig. 6 Examples of images generated by data augmentation (He=0.2; W=0.2; Z=0.2) for class 0 (without vegetation on the building facade) Fig. 7 Examples of images generated by data augmentation (He=0.2; W=0.2; Z=0.2) for class 1 (with vegetation on the building facade) Tests results In this step, simulations were performed out on the test dataset adopting three architectures (CNN8, DenseNet-121 and MobileNet) and two data augmentation configurations:Proposed in this paper (P) - defined from steps 1 and 2 (Hyperparameter Tuning): (He=0.2; W=0.2; Z=0.2). Literature (L) - presented in [4] and used by [32] for the same dataset of this study: (R=40; H=TRUE; He=0.2; S=0.2; W=0.2; Z=0.2). Table 4 presents the accuracy in test step for each analyzed configurations.Table 4 Maximum accuracy in the test step in each of the repetitions and respective mean accuracy (M) for first small dataset Arch. C. 1 2 3 4 5 M. CNN8 P 95.6 87.8 92.2 93.3 91.1 92.0 CNN8 L 91.1 80.0 93.3 91.1 93.3 89.8 DenseNet P 77.8 77.8 73.3 83.3 65.6 75.6 DenseNet L 87.8 77.8 83.3 71.1 70.0 78.0 MobileNet P 71.1 65.6 81.1 74.4 63.3 71.1 MobileNet L 54.4 87.8 58.9 55.6 53.3 62.0 Bold values indicate the best result of test accuracy and mean accuracy Comparison between methods recommended by the data augmentation configurations (Proposed (P) and Literature (L)) and architectures From the Table 4 it is possible to observe that the highest accuracy average (92.0%) was achieved by the CNN8 architecture when adopting the proposed data augmentation combination. Moreover, this configuration (CNN8 + P) also resulted in the highest accuracy value in one repetition: 95.6%. This value is equivalent to the correct classification of 86 images out of a total of 90 photographs on the test dataset. In this sense, Table 5 presents the confusion matrix for the adoption of CNN8 + P (Repetition 1).Table 5 Confusion matrix with the best results for the test step (first small dataset) TP = 43 FN=2 FP=2 TN = 43 Bold values indicate the number of true positives (TP) and true negatives (TN) It can be seen in Table 5 that the CNN model correctly classified 43 images in the positive class and 43 images in the negative class (accuracy of 95.6%). Thus, for each class, CNN only missed 2 images in the test dataset (error around 4.44%). It is also worth noting that, in the study of [32], the maximum accuracy achieved was 90% for the same test images. Thus, indicating that the careful adjustment of the Data Augmentation hyperparameters can increase the results in the classification. Table 6 summarizes the recommended hyperparameters for the analyzed database.Table 6 Selected hyperparameters for first small dataset Hyperparameter Recommendation Architecture CNN8 Height Shift Range (He) 0.2 Width Shfit Rang (W) 0.2 Zoom Range (Z) 0.2 Results of second small dataset This section report the results of applying the proposed methodology in a second small dataset: gutter integrity in roofs structures. In this sense, the Equation 11 presents the linear model adjusted for this case study:11 y=β0+β1x1+β2x2+β3x3+··+β4x4+β5x5+β6x6+β7x7=1.750-0.245x1+0.071x2-0.027x3+··-0.008x4+0.089x5-0.024x6-0.091x7 Table 7 shows the results of the test statistic (p), recommended values and odds ratio (OR) per hyperparameter (second small dataset).Table 7 Results of data augmentation hyperparameter tuning with logistic regression (second small dataset) Hyperparameter β p Value x OR Rotation R. (R) – 0.245 0.00 0 0 0.783 Hor. Flip (H) 0.071 0.07 False 0 1.074 Vertical Flip (V) – 0.027 0.48 False 0 0.973 Height S. R. (He) – 0.008 0.85 0 0 0.992 Shear Range (S) 0.089 0.02 0.2 1 1.094 Width S. R. (W) – 0.024 0.53 0 0 0.976 Zoom Range (Z) – 0.091 0.02 0 0 0.913 Bold value indicates the hyperparameters OR > 1 and p < 0.05 Table 7 shows that the only hyperparameter recommended by the HPtuningLogReg algorithm for the second case study was Shear Range (S), because p<0.05 and OR>1. On the other hand, four transformations did not reach statistical significance (p>0.05): Horizontal Flip (H), Vertical Flip (V), Height Shift Range (He) and Width Shift Range (W). In addition, two hyperparameters achieved statistical effect (p<0.05), but obtained OR<1: Rotation Range (R) and Zoom Range (Z). In this regard, the test stage was carried out with two Data Augmentation configurations:Proposed in this paper (P) - Hyperparameter Tuning for second small dataset: S=0.2. Literature (L) - presented in [4] and used by [33]: R=40; H=TRUE; He=0.2; S=0.2; W=0.2; Z=0.2. Table 8 presents the accuracy results of test step to second small dataset.Table 8 Maximum accuracy in the test step in each of the repetitions and respective mean accuracy (M) for second small dataset Arch. C. 1 2 3 4 5 M. CNN8 P 83.3 93.3 93.3 86.7 83.3 88.0 CNN8 L 56.7 70.0 66.7 56.7 73.3 64.7 DenseNet P 70.0 83.3 90.0 90.0 70.0 80.7 DenseNet L 86.7 86.7 90.0 80.0 86.7 86.0 MobileNet P 70.0 70.0 60.0 70.0 63.3 66.7 MobileNet L 70.0 80.0 73.3 83.3 70.0 75.3 Bold values indicate the best result of test accuracy and mean accuracy Comparison between methods recommended by the data augmentation configurations (Proposed (P) and Literature (L)) and architectures Table 8 shows that the highest average accuracy (88.0%) for second case study was achieved by the CNN8 architecture with the proposed configuration. Moreover, this Data Augmentation combination (S=0.2) achieved the maximum accuracy value in one iteration: 93.3%. Comparison with other studies In this section, a comparative study is carried out between the present proposal and other recent works in the literature: I [32], II [3], III [48], IV [36] and V [54]. For this, five features were observed: CNN application (classification, detection or segmentation), type of problem, analyzed hyperparameters and tuning of Data Augmentation methods. All analyzed paper applied Deep Learning models for image processing of building construction. Table 9 presents the comparison results.Table 9 Comparison of this proposal with different papers that applied CNNs in the image processing of building construction: I [32], II [3], III [48] , IV [36] and V [54]. Proposed I II III IV V [32] [3] [48] [36] [54] CNN application Classification ✓ ✓ – – – – Detection – – ✓ ✓ ✓ – Segmentation – – – – ✓ ✓ Problem Crack detection – – – ✓ ✓ ✓ Bridge inspection – – ✓ – – – Roofs defects classification ✓ – – – – – Vegetation in facades ✓ ✓ – – – – Analyzed Hyperparameters Rotation Range ✓ ✓ ✓ ✓ ✓ ✓ Horinzontal Flip ✓ ✓ ✓ – – ✓ Vertical Flip ✓ – – – – ✓ Height Shift Range ✓ ✓ – – – – Shear Range ✓ ✓ – – ✓ – Width Shift Range ✓ ✓ – – – – Zoom Range ✓ ✓ – – – – Others – – – ✓ ✓ – Tuning of Data Augmentation Yes ✓ – ✓ ✓ – – No – ✓ – – ✓ ✓ From the Table 9 it is confirmed that the main contribution of this paper is the proposal of a methodology for tuning of data augmentation hyperparameters to building construction image classification, especially in vegetation recognition and roofs defects classification. In another way, it should be noted that most of the other studies in this area are dedicated to the problem of crack detection (or segmentation). Furthermore, this proposal innovates by analyzing 128 (27) combinations of hyperparameters from seven Data Augmentation transformations: rotation range, horizontal flip, vertical flip, height shift range, shear range, width shift range and zoom range. In general, other papers analyze less combinations and transformations of Data Augmentation in the building construction image processing field. Another important contribution is the proposal for the application of logistic regression models for hyperparameter tuning. The papers by [3] and [48] also present methodologies for recommending Data Augmentation hyperparameters. However, these studies are applied to other problems (crack detection or bridge inspection) and do not adopt logistic regression methods. Conclusion The objective of this paper was to propose a rigorous methodology for tuning of Data Augmentation hyperparameters in Deep Learning to small datasets. In this sense, the main contributions of this study are:Careful analysis of Data Augmentation transformations in the application of Deep Learning in building image classification, especially in the recognition of vegetation on facades and roofs defects classification. Design of experiments with 128 combinations of Data Augmentation using the Keras library and the R software. Proposal of the HPtuningLogReg method using Logistic Regression to tuning of Data Augmentation hyperparameters. Comparison of Data Augmentation configurations by adopting three Convolutional Neural Network architectures from the literature. Regarding the results, in the first stage of experiments were recommended three Data Augmentation transformations for first case study: Height Shift Range (He), Width Shift Rang (W) and Zoom Range (Z). According to the Logistic Regression model, adopting He=0.2 guarantees an increase of around 20% of success in the correct image classification. On the other hand, by adopting W=0.2 or Z=0.2 the chance of success is increased by approximately 12%. Moreover, from the second stage of experiments, the configuration (He=0.2; W=0.2; Z=0.2) was the only one to present statistical significance (p≤0.05) for the three CNN architectures analyzed. Finally, in the testing stage, the selected Data Augmentation configuration reached the highest average accuracy (92%) when adopting the CNN8 architecture. In addition, this combination also resulted in the greatest accuracy in one repetition: 95.6%. This value is equivalent to the correct classification of 86 images out of a total of 90 photographs on the test dataset of first case study. For the second case study, the logistic regression model recommended the Shear Range transformation for Data Augmentation. In this sense, the hyperparameters selected for this application also achieved the best results in the test phase: 93.3%. In future work, it is expected to analyze other Data Augmentation transformations. In addition, it is also suggested to test more levels for specific hyperparameters, for example, Zoom Range ranging from 10% to 50%. It is also worth highlighting the importance of investigating possible limitations of the logistic regression model, for example, the proposed approach did not accounted for the interactions among different predictor variables. With interaction effects, each predictor (hyperparameter) influences others and could result others solutions of tuning. Another important point will be the adoption of the HPtuningRegLog method in tuning of Data Augmentation settings in other applications with small dataset of building construction image classification. Acknowledgements The authors are grateful to Research Group in Construction Technology and Management (GETEC) - School of Engineering (UFBA) - for providing the second image dataset, the Robotics & Perception Group (University of Zurich) for providing “The Zurich Urban Micro Aerial Vehicle Dataset”, UFBA and UFRB. Declarations Conflict of interest The Authors listed in this article declare that they have no conflict of interest. Ethical approval This article does not contain any studies with human participants or animals performed by any of the authors. 1 https://keras.rstudio.com/. 2 www.pixabay.com. 3 https://keras.rstudio.com/reference/. 4 getec.eng.ufba.br. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Ali R Cha Y-J Subsurface damage detection of a steel bridge using deep learning and uncooled micro-bolometer Constr Build Mater 2019 226 376 387 10.1016/j.conbuildmat.2019.07.293 2. Barberousse H Lombardo RJ Tell G Couté A Factors involved in the colonisation of building facades by algae and cyanobacteria in france Biofouling 2006 22 02 69 77 10.1080/08927010600564712 16581671 3. Bianchi E Abbott AL Tokekar P Hebdon M Coco-bridge: Structural detail data set for bridge inspections J Comput Civ Eng 2021 35 3 04021003 10.1061/(ASCE)CP.1943-5487.0000949 4. Chollet F, Allaire JJ (2018) Deep learning with R. Manning Publications 5. Conceição J Poça B De Brito J Flores-Colen I Castelo A Inspection, diagnosis, and rehabilitation system for flat roofs J Perform Constr Facil 2017 31 6 04017100 10.1061/(ASCE)CF.1943-5509.0001094 6. Cubuk ED, Zoph B, Shlens J, Le QV (2020) Randaugment: Practical automated data augmentation with a reduced search space. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pp 702–703 7. da Silva GR, Valões DC, Nascimento CF, SNA A., Candeia MA, Santiago H, Oliveira DV, Everton G, Lima JC, Souza JM (2 021) Elaboration of a damage map the facades of a public building in the city of triunfo/pe. Int J Adv Eng Res Sci 8:237–244 8. Dung CV Autonomous concrete crack detection using deep fully convolutional neural network Autom Constr 2019 99 52 58 10.1016/j.autcon.2018.11.028 9. Elgendy M (2020) Deep learning for vision systems. Manning Publications 10. Fang W Ding L Luo H Love PE Falls from heights: a computer vision-based approach for safety harness detection Autom Constr 2018 91 53 61 10.1016/j.autcon.2018.02.018 11. Gao Y Mosalam KM Deep transfer learning for image-based structural damage recognition Comput-Aid Civ Infrastruct Eng 2018 33 9 748 768 10.1111/mice.12363 12. Garcez N Lopes N de Brito J Silvestre J System of inspection, diagnosis and repair of external claddings of pitched roofs Constr Build Mater 2012 35 1034 1044 10.1016/j.conbuildmat.2012.06.047 13. Giolo S. R (2017). Introduction to categorical data analysis with applications (in portuguese). Editora Blucher 14. Guo J, Wang Q, Li Y, Liu P. (2020). Façade defects classification from imbalanced dataset using meta learning-based convolutional neural network. Computer-Aided Civil and Infrastructure Engineering 15. He K, Zhang X, Ren S, Sun J (2016). Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778 16. Hosmer DW Jr, Lemeshow S, Sturdivant RX (2013) Applied logistic regression, vol 398. John Wiley & Sons 17. Howard A. G, Zhu M, Chen B, Kalenichenko D, Wang W, Weyand T, Andreetto M, Adam H. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861 18. Huang G, Liu Z, Van Der Maaten L, Weinberger K. Q (2017). Densely connected convolutional networks. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2261–2269 19. Hutter F, Hoos H, Leyton-Brown K. (2014). An efficient approach for assessing hyperparameter importance. In Proceedings of International Conference on Machine Learning 2014 (ICML 2014), pages 754–762 20. Hutter F, Kotthoff L, Vanschoren J, editors (2019). Automated Machine Learning: Methods, Systems, Challenges. Springer. In press, available at http://automl.org/book 21. Kaamin M, Ahmad N, Razali S, Mokhtar M, Ngadiman N, Masri D, Hussin I, Asri L. (2020). Visual inspection of heritage mosques using unmanned aerial vehicle (uav) and condition survey protocol (csp) 1 matrix: A case study of tengkera mosque and kampung kling mosque, melaka. volume 1529 22. Kolar Z Chen H Luo X Transfer learning and deep convolutional neural networks for safety guardrail detection in 2d images Autom Constr 2018 89 58 70 10.1016/j.autcon.2018.01.003 23. Lecun Y Bengio Y Hinton G Deep learning Nature 2015 521 7553 436 444 10.1038/nature14539 26017442 24. Loukma M Stefanidou M Causes of deterioration of ottoman mosques WIT Transactions on The Built Environment 2018 177 173 180 10.2495/IHA180141 25. Maeda H Sekimoto Y Seto T Kashiyama T Omata H Road damage detection and classification using deep neural networks with smartphone images Computer-Aided Civil and Infrastructure Engineering 2018 33 12 1127 1141 10.1111/mice.12387 26. Majdik AL Till C Scaramuzza D The zurich urban micro aerial vehicle dataset The International Journal of Robotics Research 2017 36 3 269 273 10.1177/0278364917702237 27. Mantovani RG Rossi AL Alcobaça E Vanschoren J de Carvalho AC A meta-learning recommender system for hyperparameter tuning: Predicting when tuning improves svm classifiers Inf Sci 2019 501 193 221 10.1016/j.ins.2019.06.005 28. Monshi MMA, Poon J, Chung V, Monshi FM (2021) Covidxraynet: Optimizing data augmentation and cnn hyperparameters for improved covid-19 detection from cxr. Computers in Biology and Medicine 133:104375 29. Myers RH, Montgomery DC, Anderson-Cook CM (2016) Response surface methodology: process and product optimization using designed experiments. John Wiley & Sons 30. Neary P. (2018). Automatic hyperparameter tuning in deep convolutional neural networks using asynchronous reinforcement learning. In 2018 IEEE International Conference on Cognitive Computing (ICCC), pages 73–77 31. Ottoni ALC Nepomuceno EG de Oliveira MS de Oliveira DCR Tuning of reinforcement learning parameters applied to sop using the scott-knott method Soft Comput 2020 24 4441 4453 10.1007/s00500-019-04206-w 32. Ottoni ALC Novo MS A deep learning approach to vegetation images recognition in buildings: a hyperparameter tuning case study IEEE Lat Am Trans 2021 19 12 2062 2070 10.1109/TLA.2021.9480148 33. Ottoni A. L. C, Novo M. S, Costa D. B. (2021). Hyperparameter tuning of convolutional neural networks for building construction image classication. The Visual Computer 34. Pawara P, Okafor E, Schomaker L, Wiering M. (2017). Data augmentation for plant classification. In International Conference on Advanced Concepts for Intelligent Vision Systems, pages 615–626. Springer 35. R Core Team (2020) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria 36. Ren Y, Huang J, Hong Z, Lu W, Yin J, Zou L, Shen X (2020) Image-based concrete crack detection in tunnels using deep fully convolutional networks. Construction and Building Materials 234:117367 37. Rocha E Macedo J Correia P Monteiro E Adaptation of a damage map to historical buildings with pathological problems: Case study at the church of carmo in olinda, pernambuco Revista ALCONPAT 2018 8 1 51 63 10.21041/ra.v8i1.198 38. Sajedi SO Liang X Uncertainty-assisted deep vision structural health monitoring Computer-Aided Civil and Infrastructure Engineering 2021 36 2 126 142 10.1111/mice.12580 39. Schratz P Muenchow J Iturritxa E Richter J Brenning A Hyperparameter tuning and performance assessment of statistical and machine-learning algorithms using spatial data Ecol Model 2019 406 109 120 10.1016/j.ecolmodel.2019.06.002 40. Shankar K Zhang Y Liu Y Wu L Chen C-H Hyperparameter tuning deep learning for diabetic retinopathy fundus image classification IEEE Access 2020 8 118164 118173 10.1109/ACCESS.2020.3005152 41. Shen J Xiong X Li Y He W Li P Zheng X Detecting safety helmet wearing on construction sites with bounding-box regression and deep transfer learning Computer-Aided Civil and Infrastructure Engineering 2021 36 2 180 196 10.1111/mice.12579 42. Shorten C, Khoshgoftaar T. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, 6(1). cited By 456 43. Silveira, B., Melo, R., and Costa, D. B. (2021). Using uas for roofs structure inspections at post-occupational residential buildings. In Toledo Santos, E. and Scheer, S., editors, Proceedings of the 18th International Conference on Computing in Civil and Building Engineering, pages 1055–1068, Cham. Springer International Publishing 44. Song C, Xu W, Wang Z, Yu S, Zeng P, Ju Z. (2020). Analysis on the impact of data augmentation on target recognition for uav-based transmission line inspection. Complexity, 2020 45. Staffa L. B, Sa L. S. V, Lima M. I. S. C, Costa D. B. (2020). Use of image processing techniques for inspection of building roof structures for technical assistance purposes (in portuguese). ENTAC - National Meeting of the Built Environment Technology 46. Wang J.-J, Liu Y.-F, Nie X, Mo Y. (2022). Deep convolutional neural networks for semantic segmentation of cracks. Structural Control and Health Monitoring, 29(1). cited By 0 47. Wang X Zhao Y Pourpanah F Recent advances in deep learning Int J Mach Learn Cybern 2020 11 747 750 10.1007/s13042-020-01096-5 48. Wang Z Yang J Jiang H Fan X Cnn training with twenty samples for crack detection via data augmentation Sensors 2020 20 17 4849 10.3390/s20174849 32867223 49. Xue Y Li Y A fast detection method via region-based fully convolutional neural networks for shield tunnel lining defects Computer-Aided Civil and Infrastructure Engineering 2018 33 8 638 654 10.1111/mice.12367 50. Yang Z, He B, Liu Y, Wang D, Zhu G (2021) Classification of rock fragments produced by tunnel boring machine using convolutional neural networks. Automation in Construction 125:103612 51. Younis MC, Keedwell E (2019) Semantic segmentation on small datasets of satellite images using convolutional neural networks. Journal of Applied Remote Sensing 13(4):046510 52. Zeng S Zhang B Zhang Y Gou J Dual sparse learning via data augmentation for robust facial image classification Int J Mach Learn Cybern 2020 11 8 1717 1734 10.1007/s13042-020-01067-w 53. Zhou S, Song W. (2020). Deep learning-based roadway crack classification using laser-scanned range images: A comparative study on hyperparameter selection. Automation in Construction, 114 54. Zhou S, Song W (2021) Crack segmentation through deep convolutional neural networks and heterogeneous image fusion. Automation in Construction 125:103605
PMC009xxxxxx/PMC9005648.txt	==== Front Lancet Glob Health Lancet Glob Health The Lancet. Global Health 2214-109X World Health Organization. Published by Elsevier Ltd S2214-109X(22)00177-2 10.1016/S2214-109X(22)00177-2 Comment An assertive, practical, and substantive agenda to catalyse meaningful change Kutzin Joseph a Dalil Suraya a Barroy Helene a Barkley Shannon a Dkhimi Fahdi a Jowett Matthew a Marten Robert a Mathauer Inke a Meessen Bruno a Sparkes Susan a Xu Ke a a World Health Organization, Geneva 1211, Switzerland 4 4 2022 5 2022 4 4 2022 10 5 e606e608 © 2022 World Health Organization 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcThe Lancet Global Health Commission on financing primary health care1 combines a shared vision with practical guidance on how to align health financing with overall reform strategies that place primary care service delivery at the core. The Commission reinforces key messages that WHO has put forward on health financing reforms to enable progress towards universal health coverage (UHC). It then extends these by application to primary care as a critical service delivery element for the progressive realisation of UHC. WHO's guidance on health financing is crystallised into a framework for regular country assessment to inform policy dialogue.2 The alignment of the Commission with this guidance is clear, as reflected in the table . While the decision to limit the operational definition of primary health care (PHC) to service delivery platforms was made for the purposes of the Commission, certain key financing issues merit further attention. We point to these towards the end of this Comment.Table Alignment of the Lancet Global Health Commission recommendations with WHO guidance on health financing WHO progress matrix desirable attribute Commission recommendation Health expenditure is based predominantly on public funding sources Public resources should provide the core of PHC funding, with minimal reliance on direct payments when services are accessed Benefit design includes explicit limits on user charges and protects access for vulnerable groups Public resources should provide the core of PHC funding, with minimal reliance on direct payments when services are accessed Pooling structure and mechanisms across the health system enhances the potential to redistribute available prepaid funds Reduce fragmentation, thereby creating an enabling environment for more equitable cross-subsidies between healthy and ill as well as rich and poor, more efficient integration between levels of care, and better coordination with services in (often donor-funded) disease or intervention-specific programmes Health system and financing functions are integrated or coordinated across schemes and programmes Reduce fragmentation, thereby creating an enabling environment for more equitable cross-subsidies between healthy and ill as well as rich and poor, more efficient integration between levels of care, and better coordination with services in (often donor-funded) disease or intervention-specific programmes Resource allocation to providers reflects a combination of population health needs and provider performance Payment methods should assign resources based on people's health needs and align incentives with people-centred services Purchasing arrangements are tailored in support of service delivery objectives Payment methods should assign resources based on people's health needs and align incentives with people-centred services A set of priority health service benefits within a unified framework is implemented for the entire population Pooled funds should cover PHC and enable all people to receive it free at the point of use Health budget formulation and structure support flexible spending and are aligned with sector priorities Public financial management systems must be flexible and straightforward, enabling managers to respond to the changing needs of patients, families, and communities Providers can directly receive revenues, flexibly manage them, and report on spending and outputs Funds flow to and are managed by frontline providers (autonomy) PHC=primary health care. From a health financing perspective, perhaps the most far-reaching and potentially influential recommendations in the Commission are (a) to move towards a coherent mixed-provider payment model for PHC with capitation at the core, and (b) to universalise PHC coverage while eliminating or greatly reducing out-of-pocket payments for these services. We agree; here we identify some key issues of policy and implementation alignment that are implied by these directions. Universality requires that the funding base, particularly but not only in low-income and middle-income countries (LMICs), would rely predominantly on general government budget revenues, regardless of whether they flow directly to providers or via a service purchasing agency such as a health insurance fund. For any payment system paid from government budget revenues to be effective, budget formulation, and execution—the fundamentals of public financial management—will need to function sufficiently well to enable providers to receive a steady, predictable flow of funds3 with the ability to manage these flexibly.4 This can be facilitated with the design of a programme-based budget that is defined in a way (eg, as access to PHC services) that aligns with the capitation strategy and shifts control and monitoring from inputs to predefined performance indicators that are progressively refined over time. In reality, a country's starting point for reorienting provider payment to a more coherent set of incentives for “people-centred care” is more complex than the extremes of either line-item budgets or unmanaged fee-for-service, given the fragmentation of revenues flowing from different schemes and programmes that exists in most countries. Thus, a diagnostic of initial payment arrangements5 will be essential to create a realistic transition process. As the people-centred payment model of capitation becomes more sophisticated, it should move beyond one level of care and be designed in a way to explicitly encourage provision of services and tasks at lower-level facilities and provider types that are increasingly close to the population they serve. But financial incentives alone are not enough; from early in the reform process, it will be essential to identify service delivery strategies tailored to local needs, including settings challenged by inadequate levels or mix of health workforce, which address underlying performance issues, and then to align the specifics of purchasing with these strategies. Universalising coverage for PHC services means, in effect, that entitlement to services would have to be entirely or predominantly non-contributory in nature, or done so de facto by relying on general revenues to fund coverage for the uninsured in countries relying mainly on social health insurance. This mechanism aligns with a growing consensus6, 7 about the weakness of contributory-based entitlement (typically in the form of social health insurance) in contexts of limited labour formality. As noted by the Commission, most out-of-pocket spending in PHC is for medicines. To eliminate or greatly reduce such payments, it is essential that prescribed medicines are made available either without co-payment or with low and explicit limits that are fixed in absolute rather than percentage terms.8 This has obvious implications for prioritising generic medicine, management of prescribing patterns, strong price negotiation processes, and where relevant, contracting with private pharmacies. Universalising PHC combined with the approach of progressing towards a purposively aligned mixed-provider payment system centred on capitation will require, concurrently, taking steps towards a unified or interoperable population (and eventually health service use) database, regardless of affiliation to specific health programme or coverage schemes. This is perhaps a “hidden agenda” for UHC, as unified data systems are a critical input for learning, adaptable (ie, resilient) health systems, reflected in the strategic pathway shown in figure 11 of the report. Finally, choosing capitation is a political act, because it requires an explicit, up-front decision on the share or amount of the budget that will be allocated to this purpose. Thus, the Commission's recommendations to have applied political economy analysis9 as an integral part of policy development is particularly relevant here. From the perspective of expenditure tracking, interpreting PHC as a service delivery system or platform is appealing because it is measurable. Because PHC is not a category in the System of Health Accounts, estimates have to be constructed based on choices about which expenditure categories to assign to it. A “global” measure of PHC is inherently challenging because countries differ in how they organise PHC, and thus no global measure will be equally relevant to all countries. The global measure used by WHO10 was the product of extensive consultation. However, as more countries engage in monitoring PHC spending, and as new service delivery models emerge following the pandemic, ensuring policy relevance and cross-country comparability requires that we periodically revisit the measure; the Commission's recommendations are very helpful for this agenda. As the Commissioners have recognised, what is most important is to improve the quality and increase the frequency with which countries produce their own health accounts. Then, based on how PHC is organised in the country, they can cross-tabulate the “provider” and “function” classifications to assign expenditures to PHC in a way that is most policy-relevant in the national context. For those of us who work on health financing, considering primary health care only in terms of service delivery is familiar, comfortable ground. The concept of PHC, however, is broader,11 and a next generation of work on “financing for health” is needed to reflect this, for example by taking on the challenge of multisectoral budgeting12 aimed at addressing cross-cutting public health functions and health determinants emanating from outside as well as inside the health system. As the Commission authors note, we need to adapt our financing levers to population-based essential public health functions across service delivery platforms. In addition, enabling people and communities to take a more proactive role in their health likely requires some rethinking of health financing instruments from a more “demand-side” perspective. The Commission highlights the importance of these issues, but country evidence is limited, and there is great scope for further policy development. Similarly, the design of financing instruments to support health in all policies and an “economy for health”13 warrant further attention to lay out an actionable agenda going forward. There is no global blueprint for how to organise financing to support PHC as the means for the progressive realisation of progress towards UHC. But the fact that we do not know everything does not mean that we do not know anything. Some ways of doing things are better than others.14 The Commission continues and deepens this more assertive approach to health financing. The Commission starts from the position that stronger PHC is the best (more equitable, more efficient) approach for the progressive realization of UHC, and then applies what is known about health financing to provide concrete guidance and clear directionality to governments and international agencies. The Commission's specific emphasis on the political economy of PHC-oriented health financing reforms is important to address the question of why, despite all the years that have passed since 1978, these reforms have not gained traction in so many countries. By clearly stating concrete technical approaches that align to PHC and explicitly recognising the inherently political nature of these processes, the Commission will, we hope, catalyse practical approaches that move beyond rhetoric and agenda-setting to actual implementation. We declare no competing interests. ==== Refs References 1 Hanson K Brikci N Erlangga D The Lancet Global Health Commission on financing primary health care: putting people at the centre Lancet Glob Health 2022 published online April 4. 10.1016/S2214-109X(22)00005-5 2 Jowett M Kutzin J Kwon S Hsu J Sallaku J Salano JG Assessing country health financing systems: the health financing progress matrix 2020 World Health Organization Geneva 3 Piatti-Funfkirchen M Barroy H Pivodic F Margini F Budget execution in health: concepts, trends and policy issues 2021 World Bank Washington, DC 4 O'Dougherty S Mtei G Kutzin J Barroy H Piatti-Funfkirchen M Direct facility financing: concept and role for UHC 2022 World Health Organization Geneva 5 Mathauer I Dkhimi F Analytical guide to assess a mixed provider payment system 2018 World Health Organization Geneva https://www.who.int/publications/i/item/978-92-4-151533-7 6 Kutzin J Yip W Cashin C Alternative financing strategies for universal health coverage Scheffler RM Handbook of global health economics and public policy. Vol 1, Ch 5, 267–309 – The Economics of Health and Health Systems 2016 World Scientific Publishing Singapre https://www.worldscientific.com/doi/abs/10.1142/9789813140493_0005 7 Yazbeck AS Savedoff WD Hsiao WC The case against labor-tax-financed social health insurance for low- and low-middle-income countries Health Affairs 39 2020 892 897 32364862 8 Thomson S Cylus J Evetovits T Can people afford to pay for health care? New evidence on financial protection in Europe 2019 WHO Regional Office for Europe Copenhagen 9 Sparkes SP Bump JB Őzçelik EA Kutzin J Reich MR Political economy analysis for health financing reform Health Systems Reform 5 2019 183 194 31369319 10 Eigo N McDonald L Hernandez P Rivas L Dupuy J Xu K Measuring primary health care expenditure under SHA 2011 2021 World Health Organization Geneva 11 Ghebreyesus TA Strengthening our resolve for primary health care Bull World Health Organ 98 2020 726 26A 33177764 12 McGuire F Vijayasingham L Vassall A Financing intersectoral action for health: a systematic review of co-financing models Globalization Health 15 2019 86 31849335 13 WHO Council on the Economics of Health for All Financing Health for All: increase, transform, and redirect. Council Brief No. 2 2021 World Health Organization Geneva https://www.who.int/publications/m/item/council-brief-no-2 14 Kutzin J Anything goes on the path to universal health coverage? No Bull World Health Organ 90 2012 867 868 23226900
PMC009xxxxxx/PMC9005653.txt	==== Front Lancet Glob Health Lancet Glob Health The Lancet. Global Health 2214-109X The Author(s). Published by Elsevier Ltd. S2214-109X(22)00005-5 10.1016/S2214-109X(22)00005-5 The Lancet Global Health Commissions The Lancet Global Health Commission on financing primary health care: putting people at the centre Hanson Kara Prof SD a* Brikci Nouria MSc a Erlangga Darius PhD a Alebachew Abebe MSc c De Allegri Manuela Prof PhD d Balabanova Dina Prof PhD a Blecher Mark PhD e Cashin Cheryl PhD f Esperato Alexo PhD g Hipgrave David PhD h Kalisa Ina MSc i Kurowski Christoph MD j Meng Qingyue Prof PhD k Morgan David BSc l Mtei Gemini PhD m Nolte Ellen Prof PhD b Onoka Chima Prof PhD n Powell-Jackson Timothy Prof PhD a Roland Martin Prof DM o Sadanandan Rajeev MPhil p Stenberg Karin MSc q Vega Morales Jeanette PhD r Wang Hong PhD s Wurie Haja PhD t a Department of Global Health and Development, London School of Hygiene and Tropical Medicine, London, UK b Department of Health Services Research and Policy, London School of Hygiene and Tropical Medicine, London, UK c Breakthrough International Consultancy, Addis Ababa, Ethiopia d Heidelberg Institute of Global Health, University Hospital and Faculty of Medicine, University of Heidelberg, Heidelberg, Germany e National Treasury, Pretoria, South Africa f Results for Development, Washington DC, USA g Bill & Melinda Gates Foundation, New Delhi, India h UNICEF, Iraq Country Office, Baghdad, Iraq i World Health Organization, Kigali, Rwanda j World Bank, Washington DC, USA k China Center for Health Development Studies, Peking University, Beijing, China l Health Division, The Organisation for Economic Co-operation and Development, Paris, France m Abt Associates, Dar es Salaam, Tanzania n Department of Community Medicine, University of Nigeria, Enugu, Nigeria o Department of Public Health and Primary Care, University of Cambridge, UK p Health Systems Transformation Platform, New Delhi, India q World Health Organization, Geneva, Switzerland r Pronova Technologies, Santiago, Chile s Bill & Melinda Gates Foundation, Seattle, WA, USA t College of Medicine and Allied Health Sciences, University of Sierra Leone, Freetown, Sierra Leone * Correspondence to: Prof Kara Hanson, Department of Global Health and Development, London School of Hygiene & Tropical Medicine, London, WC1H 9SH, UK 4 4 2022 5 2022 4 4 2022 10 5 e715e772 © 2022 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license 2022 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. ==== Body pmcExecutive summary The COVID-19 pandemic has brought the need for well-functioning primary health care (PHC) into sharp focus. PHC is the best platform for providing basic health interventions (including effective management of non-communicable diseases) and essential public health functions. PHC is widely recognised as a key component of all high-performing health systems and is an essential foundation of universal health coverage. PHC was famously set as a global priority in the 1978 Alma-Ata Declaration. More recently, the 2018 Astana Declaration on PHC made a similar call for universal coverage of basic health care across the life cycle, as well as essential public health functions, community engagement, and a multisectoral approach to health. Yet in most low-income and middle-income countries (LMICs), PHC is not delivering on the promises of these declarations. In many places across the globe, PHC does not meet the needs of the people—including both users and providers—who should be at its centre. Public funding for PHC is insufficient, access to PHC services remains inequitable, and patients often have to pay out of pocket to use them. A vicious cycle has undermined PHC: underfunded services are unreliable, of poor quality, and not accountable to users. Therefore, many people bypass primary health-care facilities to seek out higher-level specialist care. This action deprives PHC of funding, and the lack of resources further exacerbates the problems that have driven patients elsewhere. Focus on financing Health systems are fuelled by their financing arrangements. These arrangements include the amount of funding the system receives, the ways funds are moved through the system to frontline providers, and the incentives created by the mechanisms used to pay providers. Establishing the right financing arrangements is one crucially important way to support the development of people-centred PHC. Improving financing arrangements can drive improvements in how PHC is delivered and equip the system to respond effectively to evolving population health needs. Thus attention should be paid simultaneously to both financing and service delivery arrangements. In this report, the Lancet Global Health Commission on financing PHC argues that all countries need to both invest more and invest better in PHC by designing their health financing arrangements—mobilising additional pooled public funding, allocating and protecting sufficient funds for PHC, and incentivising providers to maintain the health of the populations they serve—in ways that place people at the centre and by addressing inequities first. Financing is political Answering the question of how to make these changes goes far beyond technical considerations. Fundamentally shifting a health system's priorities—away from specialist-based and hospital-based services and towards PHC—involves political choices and creates numerous political challenges. Successfully reorienting a system towards PHC requires savvy political leadership and long-term commitment, as well as proactive, adaptable strategies to engage with stakeholders at all levels that account for the social and economic contexts. Therefore, this report addresses both technical and political economy considerations involved in strengthening financing for PHC. Spending more and spending better on PHC Despite broad recognition of the importance of PHC, there is no global consensus on what exactly constitutes PHC. This makes it challenging to measure and report on levels of expenditure on PHC. In this Commission we define PHC as a service delivery system or platform, together with the human and other resources needed for it to function effectively. We found that LMICs spend far too little on PHC to provide equitable access to essential services and that much of the (significant) variation in PHC spending levels across countries is explained by national income levels, although there is variation in the amount of government resources allocated to PHC at any given level of economic development. Furthermore, at every level of PHC spending, there is substantial variation in performance, suggesting that we need to spend better as well as spending more. In this Commission, we analysed provider payment methods and found that the sources of PHC expenditure remain fragmented and overly reliant on out-of-pocket payments. Population-based provider payment mechanisms, such as capitation, should be the cornerstone of financing for people-centred PHC. However, these mechanisms are rare in LMICs, where input-based budgets are standard practice. Furthermore, many features of primary health-care organisation that are necessary for population-based payment strategies (such as empanelment, registration, and gatekeeping) are absent in LMICs. Redressing these limitations to improving financing PHC is urgent, as new challenges continue to arise. As in other parts of the health sector, PHC will continue to become more integrated, digitally-driven, and pluralistic; therefore, PHC financing arrangements also need to evolve to support, drive, and guide these changes to better meet human needs. The Commission takes the position that progressive universalism should drive every aspect of PHC. That means putting the rights and needs of the poorest and most vulnerable segments of a population first. This requires unwavering ethical, political, and technical commitment and focus. Together with this overarching principle, we identified four key attributes of people-centred financing arrangements that support PHC. (1) Public resources should provide the core of primary health-care funding. Revenue-raising mechanisms should be defined based on the ability to pay and be progressive. Out-of-pocket payments must be reduced to levels where they are no longer a financial barrier to accessing needed care, impoverish households, or push households deeper into poverty. In most LMICs, this level of public funding for PHC can only be generated through increased allocations to PHC from general tax revenue, and therefore requires an expansion of countries' taxation capacities. In low-income countries, more development assistance will be needed to expand the resource envelope for PHC. (2) Pooled funds should be used to allow all people to receive PHC that is provided free at the point of use. Only once universal coverage with PHC is achieved should pooled resources be extended to cover other entitlements. In this way, PHC can help fulfil the promise of universal health coverage. (3) Resources for PHC should be allocated equitably (across levels of service delivery and geographic areas) and protected as they flow through the system to frontline providers. Countries should deploy a set of strategic resource allocation tools (including a needs-based per-capita resource allocation formula and effective public financial management tools) to match primary health-care funding with population needs and ensure these resources reach the frontline, and prioritise the poorest and most vulnerable people. (4) Payment mechanisms for primary health-care providers should support allocation of resources based on people's health needs, create incentive environments that promote PHC that is people-centred, and foster continuity and quality of care. To achieve these goals, a so-called blended provider payment mechanism with capitation at its core is the best approach to paying for PHC. Capitation should form the core of the primary health-care financing system because it directly links the population with services. Combining capitation with other payment mechanisms, such as performance-based payments for specific activities, enables additional objectives to be achieved. Each country is at a different point along its path towards the goal of effective financing for PHC. The four attributes outlined both represent goals and present a guide for working towards those goals. This Commission recognises that, depending on the context, the evolution of an effective primary health-care financing system in some countries might occur through incremental changes, whereas others can implement comprehensive reforms. Improving PHC financing can occur in response to bottom-up advocacy, top-down policy or, most likely, through a combination of grassroots and technocratic approaches. Political, social, and economic factors are therefore as important as technical design elements when it comes to enacting efficient and equitable primary health-care financing reform. Changing the ways in which PHC is financed requires support from a wide range of stakeholders, and deliberate political strategies, to determine and then stay the course. The change also requires good information about PHC resource levels and flows so that this reorientation can be effectively managed and monitored. In this Commission, we provide five recommendations. (1) People-centred financing arrangements for PHC should have public resources provide the bulk of primary health-care funding; pooled funds cover primary-health care, enabling all people to receive PHC that is provided free at the point of service use; resources for PHC are allocated equitably across levels of service delivery and geographic areas, and are protected so that sufficient resources reach frontline primary health-care service providers and patients; and primary health-care provider payment mechanisms support the allocation of resources based on people's health needs, create incentive environments that promote PHC that is people centred, foster continuity and quality of care, and remain flexible enough to support rapidly changing service delivery models. (2) Spending more and spending better on PHC requires a whole-of-government approach involving all ministries whose remit interacts with health and requires the support of civil society. Key actors and stakeholders should be involved in designing and implementing financing arrangements for PHC that are people-centred. Although the specifics will vary depending on the national context, there are important roles and responsibilities for ministries of health, ministries of finance, local government authorities, communities and civil society groups, health-care providers and organisations, donors, and technical agencies. (3) Each country should plot out a strategic pathway towards people-centred financing for PHC that reflects the attributes outlined above, including investments in supporting basic health system functions. Technical strategies should be underpinned from the outset by analysis of the political economy. (4) Global technical agencies should reform the way primary health-care expenditure data are collected, classified, and reported to enable longitudinal and cross-country analyses of achievement of key primary health-care financing goals. (5) Academic researchers, technical experts, and policy makers, among others, should pursue a robust research agenda on financing arrangements for PHC that place people at the centre to support achievement of key primary health-care financing goals. Introduction Primary health care (PHC) is a key component of all high-performing health systems,1 an essential foundation for universal health coverage (UHC), and a prerequisite for meeting the Sustainable Development Goals. It is a pathway to achieving good health at low cost2 by providing essential and cost-effective health interventions, including health promotion; maternal, newborn, and child health care; immunisations; and treatment for common illnesses across the life course. As the global burden of non-communicable diseases increases, PHC is emerging as the locus of both prevention and the coordination of life-long management of chronic conditions. PHC also has an important role in providing essential public health functions, including responding to epidemic diseases such as the COVID-19 pandemic. When successfully delivered, PHC serves as a key vehicle for fulfilling governmental and societal commitments. For example, primary health-care expansion improves equity when its services reach vulnerable segments of the population.3 Because primary health-care services are provided where people live and work,4 and because PHC focuses on population health, it can address many determinants of health that underpin various sources of vulnerability.5 PHC can protect households' financial wellbeing by fostering good health and reducing the risks of disease among breadwinners, caregivers, and other family members, and by averting the need for expensive secondary and tertiary health care.6 In fragile states and conflict-affected settings, primary health-care services can help build trust in the health system—and in the government it represents.7 A convincing economic case for PHC has been made repeatedly. Most of the available evidence comes from high-income countries. In these contexts, it has been shown that by providing key services at the lowest appropriate level of the health system, PHC can decrease the need for unnecessary hospital admissions, prevent avoidable readmissions, and limit inappropriate use of emergency departments.8 In low-income to middle-income countries (LMICs), an expanding body of evidence shows the cost-effectiveness of many interventions that are typically delivered through PHC. Indeed, a 2018 analysis classified 198 (91%) of 218 essential UHC interventions as PHC9 and another report estimated that up to 75% of the projected health gains from the SDGs could be achieved through PHC.10 Expanding a core set of integrated interventions for women's and children's health (narrower than PHC) is calculated to generate economic and health benefits in low-income countries valued at 7·2 times more than the costs; the value increases to 11·3 in lower-middle-income countries.11 A study of 67 LMICs projected that investing in PHC over the period from 2020 to 2030 would avert up to 64 million deaths.10 There is also a strong case for public investments in common goods for health, including public goods12 (which, in the economic sense, are services and functions that are both non-rival and non-exclusive), and in functions that generate strong positive externalities. These goods, which include the essential public health functions in PHC, require public funding as they are otherwise subject to market failure. Yet despite its fundamental importance and incredible promise, PHC is not doing well in many countries, especially LMICs. The global community first proclaimed its commitment to multisectoral and integrated PHC in the 1978 Alma-Ata Declaration. However, this commitment was quickly derailed, with funding and technical support flowing instead into vertical and disease-specific programmes.13 Despite periodic attempts to refocus on PHC, vertical programmes and hospital-based and specialist-based care models have regularly been prioritised over PHC. Funding for PHC is generally insufficient, access to primary health-care services remains inequitable, services are of inadequate quality, and patients often have to make out-of-pocket payments to use them. Health-care worker shortages persist, particularly in rural areas where the need is often greatest, and in many countries supplies of medicines, equipment, and other necessary commodities are grossly inadequate.6 This situation reinforces a cycle of neglect of PHC: when primary health-care services are unreliable, of poor quality, and not accountable to system users, it leads to poor uptake and low levels of trust in community-level health care. Users choose to bypass primary health-care services, which then receive even fewer resources. To successfully provide PHC at community level, national and local health-care systems need to be reimagined and restructured, beginning with placing the needs and preferences of people (including the intended users and providers) at the centre of the system design.6, 14 Health financing arrangements provide the fuel for health systems: they establish the amount of resourcing available and the way in which risks are shared among those who are ill and those who are well, the ways that funds flow through the system to frontline providers, and the payment systems that create incentives for providers. Together, these arrangements shape the equity, effectiveness, and efficiency of PHC. This report focuses on how to get the financing arrangements right to serve and fuel effective, efficient, and equitable PHC service delivery. As will be discussed throughout the report, establishing the right financing arrangements for effective and equitable PHC can both support and drive other necessary transformations. This Commission contends that health financing arrangements for PHC—how to mobilise sufficient resources to support PHC objectives, how to ensure that resources reach frontline providers in ways that align with PHC objectives, and how to design financial incentives that encourage the delivery of, and access to, high-quality, equitable, integrated and efficient PHC—should be centred on people, and focused on equity. Panel 1 presents descriptions of two key terms that are used throughout the report: health financing functions and health financing arrangements.Panel 1 Health financing functions and arrangements Three core health financing functions are mentioned throughout the report: • Mobilisation of funds: the collection of revenue (from taxes, insurance contributions, user fees, donations, or other means) that is used to pay for delivery of health services. Resource mobilisation is addressed in detail in section 3. • Pooling: accumulating prepaid funds (such as social security contributions, taxes, or health insurance premiums) to pay for health services for a group of people. Pooling is addressed in section 3. • Purchasing: the mechanisms by which mobilised and pooled funds are transferred to providers who deliver health services. Purchasing involves three elements: specifying what services will be purchased (often called the benefit package), identifying which providers are eligible to provide these services, and defining the set of arrangements through which providers are contracted to provide the services. How providers are paid to provide primary health care (PHC) is the focus of section 5. We refer in the report to a number of different ways of paying PHC providers: • A line-item budget is when providers are given prospectively a fixed amount of funds to cover specific line items, such as medicines and utilities, for a period (usually a year). • A fee-for-service payment is when providers are reimbursed for each individual service provided. • A capitation payment is when providers are given a fixed per-person payment, determined and paid in advance, to deliver a defined set of services to each enrolled individual for a specified period of time. • A pay-for-performance system is when providers are given bonus payments (or penalties) for achieving service coverage or quality targets. We use the broader term health financing arrangements to refer to both the core health financing functions and the ways in which they are organised and interact. These arrangements include the public financial management processes through which resources flow to frontline providers. Throughout the report we pay particular attention to:15 • Budget formulation: the process of determining, soliciting, and securing sufficient public funding for PHC and the health system overall. • Resource allocation: the process of assigning available resources to specific uses (in this case, to PHC). • Budget execution: how the funds budgeted for services flow through the public system to providers. In this Commission, we aimed to present new evidence on levels and patterns of global expenditure on PHC (throughout the report, the terms expenditure and spending are used interchangeably), including describing how PHC is currently organised and paid for; analyse key technical and political economy challenges faced in financing PHC; identify areas of proven or promising practices that effectively support PHC across the key health financing functions; and identify actionable policies to support LMICs in raising, allocating, and channelling resources in support of the delivery of effective, efficient, and equitable PHC that is people centred. Section 1 provides a general introduction to PHC policy and challenges. It then characterises global and national challenges, as well as opportunities, related to financing PHC. Section 2 describes the current financing landscape for PHC, detailing existing patterns of expenditure, provider payment, and related organisational features. Section 3 elaborates on mobilising sufficient resources for health through progressive means and then pooling resources to enable cross-subsidisation between those who are ill and those who are well. Section 4 focuses on how to ensure that resources mobilised for health are allocated to PHC, and emphasises the importance of engaging with the multiple budget tools available to Ministries of Health to ensure that resources reach frontline providers. Section 5 highlights the importance of structuring incentives for PHC providers so that they are motivated to provide PHC that is people centred, and proposes a strategic pathway of steps that countries can take to establish appropriate incentives. Section 6 describes the importance of, and notes strategies for, addressing the political economy of financing PHC. Finally, section 7 presents a synthesis of the vision for people-centred financing arrangements for PHC, summarises possible pathways for working towards this vision, and provides recommendations and proposes actions for different stakeholders committed to supporting LMICs to spend more—and to spend better—on PHC. It is the Commission's hope that this report will serve as a resource to policy makers around the world who are committed to this crucial endeavour. We prepared this Commission through an extensive process of study and debate on good and promising practices in financing PHC. The 22 expert members, representing 19 nationalities, have amongst them experience working in national governments, technical agencies, bilateral and multilateral donors, universities, and independent think tanks. Assisted by a technical team based at the London School of Hygiene & Tropical Medicine, London, Commissioners drew on the following sources of evidence: case studies prepared by national consultants on innovations in PHC financing in seven LMICs (Brazil, Chile, China, Ethiopia, Ghana, India, and Philippines) and three high-income countries (Estonia, Finland, and New Zealand); a compilation of Organisation for Economic Co-operation and Development (OECD) and WHO health expenditure data to conduct expenditure analysis; a new survey of PHC organisation and provider payment in LMICs; literature reviews, including systematic and scoping reviews, of existing knowledge on financing PHC; and an expert roundtable on digital technologies and PHC financing. Additional publications based on these products from the Commission are available on the Commission's website. Section 1: Financing primary health care in the 21st century—challenges and opportunities Defining PHC In different contexts, PHC has been operationalised in different ways: as an approach to the delivery of health care that reorients the system away from hospitals and specialist care to practitioners working at community-level outpatient facilities; as a coordination mechanism which links primary care, community care, specialised care, wider public health interventions, and long-term care services;16 as a package of health services, often defined using cost-effectiveness as a primary criterion; as a service delivery level or platform, together with the human and other resources needed for it to function effectively; or as a system which combines a platform, a service package, and an approach that emphasises an orientation to meeting the needs of the population. For the purposes of the Commission's health financing analyses, we found it necessary to link service delivery arrangements and orientations of PHC with the way resources are directed through the financing system to reach frontline providers. Resources typically flow to service delivery platforms. For this reason, PHC as a platform is our favoured operational definition of PHC. It typically includes both community-level and first-level health care. While it is true that some PHC services might be provided in hospital outpatient departments, it is the contention of this Commission that, over time, countries should aim to shift most PHC services out of hospitals to the appropriate community-level or first-level platforms where they can be delivered cost-effectively. PHC is being transformed by new technologies that have the potential to overcome persistent challenges and radically change how people engage with health services. For example, digital technologies are streamlining procurement of commodities, improving supply chains, supporting health-care providers' adherence to clinical guidelines, and enabling tracking of patients who would otherwise be lost to follow-up.17, 18 New mobile and telemedicine technologies are helping patients to remotely access health information, medical advice, and their own health data. These technologies might help patients and their families to take greater responsibility for their own health and enable new, more horizontal relationships between patients and providers. Similarly, opportunities to pay insurance premiums digitally, such as via mobile phones, may help to mobilise additional financing for health. Technology-driven transformations bring some risks, including increasing health inequalities and fragmenting financing and delivery of care. However, they also offer avenues for making PHC more convenient, accessible, affordable, and high quality. PHC will continue to evolve. For this evolution to fulfil the potential of PHC, health-care providers must expand their areas of focus and develop new skills, and health systems must develop new ways of delivering services across the life course, including incorporating preventive and supportive services. Innovations such as new digital and telehealth platforms must be deployed to support individuals and their families to manage their own health. Governments and communities must recognise and foster the role of PHC in essential public health functions, and PHC must engage with individuals and the wider community to co-produce forms of delivery that will meet people's needs. Appropriate use of technology will be key to support those delivering and those using services—eg, enabling task shifting between different cadres of health workers and delivering care that is flexible and closer to people's homes. COVID-19: changing the context and highlighting lessons for PHC The COVID-19 pandemic has brought the need for PHC that is well financed into sharp focus in several ways: • It underscores the relationship between health and the economy. In particular, it has highlighted that failing to invest in health, including PHC, can have dramatic economic consequences. • Countries with stronger PHC systems were able to respond faster and more effectively to the pandemic.19, 20, 21 For example, Japan, Vietnam, and South Korea were better prepared to carry out COVID-19 surveillance because they were able to capitalise on existing public health capacity for contact tracing.22 Close partnerships between multidisciplinary PHC providers and local governments allowed rapid responses by reassigning roles while still maintaining other public health services, as seen in France23 and Catalonia, Spain.24 This shows that PHC systems provide a foundation for effective management of health crises. • The PHC system is a good platform for public health measures to control infectious diseases. The pandemic brought renewed attention to the vital importance of common goods for health,25 including the essential public health functions that are a component of PHC. Essential public health functions include surveillance systems, test-and-trace systems, quarantine functions, and vaccination. • Going forward, COVID-19 can only be overcome through action at the PHC level. For example, COVID-19 vaccinations will be provided through PHC platforms as provision shifts from a vertical campaign mode to a routine service. Management of mild-to-moderate illnesses related to COVID-19 will also be through PHC. • The COVID-19 pandemic has accelerated service delivery changes that were already underway. In particular, health care has rapidly adjusted to incorporate remote consultations, ramped-up support for home care, task shifting to lower-level cadres and structures, and expanded digital monitoring of health status, among others. • Above all, by highlighting the structural inequalities that exist within and across countries, the COVID-19 pandemic has emphasised the need to work for equity, solidarity, and social justice for all—these principles are central to the PHC approach. Many of these lessons were highlighted in previous health emergencies, such as the 2014 Ebola outbreak in West Africa. However, the global scale of COVID-19, with its accompanying global and national responses, have conclusively shown that health issues can be made a top priority and that rapid changes to health systems and financing are possible. Governments and international funders alike created new flexibility in health financing arrangements, including rapid budget reallocations, mobilisation of new funds, and use of flexible purchasing arrangements. Some of these new arrangements should be retained and expanded. However, they have also exposed new areas of financial risk and vulnerability, as well as highlighting the need to continually focus on transparency and accountability.26 COVID-19 has shown that the need for well-financed, well-functioning PHC has never been greater. Yet many aspects of the pandemic response have instead led to a greater concentration of resources on hospital care, vaccines, and other so-called silver-bullet approaches27, 28 instead of prioritising basic public health interventions such as test-and-trace, disease surveillance, and population-based preventive measures. This presents real risks to PHC financing and delivery. The financing requirements of the response to COVID-19 (both by the health system and in the economic response) have placed unprecedented pressure on government budgets, while spending capacity has decreased due to declines in revenue and borrowing.29 For example, the immediate financing needs for additional funding for COVID-19 prevention, treatment and surveillance in sub-Saharan African countries were estimated at about 3% of gross domestic product (GDP), or US$53 billion.30 At the same time, the International Monetary Fund estimated that economies around the world contracted in per-capita terms by an average 5·9% in 2020 as a result of COVID-19,31 driving an untold number of households into poverty and reducing their ability to pay for health care. Spending on routine health services has fallen in many countries32 and generating more public resources for PHC will be challenging under conditions of fiscal restraint. In 90% of 105 countries surveyed by WHO, the pandemic badly disrupted many essential services that were not directly related to COVID-19, particularly mental health and reproductive, maternal, neonatal and child health care.33 Among 22 low-income countries, ten (45%) reported disruptions in at least 75% of essential services—this represents far more disruption than was reported in LMICs (30%) and upper-middle and high-income countries (8%).34 In the Democratic Republic of the Congo, for example, by October, 2020, up to 33% of the health budget had been redirected to the COVID-19 emergency response.33 Detailed accounts of the effect of COVID-19 on health financing in two other countries in sub-Saharan Africa, Sierra Leone, and South Africa, can be found in the appendix (p 2). The COVID-19 pandemic has thus shifted the landscape of possibilities for financing people-centred PHC in particular and health more broadly. Although it generated some new opportunities, it also created vast new challenges—and provided a glimpse of the potential havoc that future crises (health and otherwise) can create, and the consequences of not prioritising equity.35 Health financing systems need to be resilient to allow the surge capacity needed to respond to shocks while maintaining access to essential services. Determining PHC packages To examine and operationalise financing arrangements for PHC requires clarity on what services are being financed. The PHC package can be conceptualised at three levels. At the highest level, each government and national health system must articulate its own vision for comprehensive PHC that addresses its population health needs. Fulfilling a stated vision requires drawing on resources from both public and private sources. The Alma-Ata Declaration's vision of PHC also encompasses contributions from other sectors to address social determinants of health. In this report we focus on choices made within the health budget but recognise the need to identify mechanisms for securing contributions from outside the health sector, including education, water, and sanitation. At the benefit package level, each government and health system must identify which services it can afford to provide either for free or with partial coverage. In many low-income settings, external funds will be needed to augment government financing. At the provider payment level, each health system must determine which services it will pay providers for, at what level of payment, and via which provider payment mechanism (see section 5). Vertical programmes that provide some PHC services might be excluded from this payment system. The specific PHC package that is financed and delivered in any particular setting will be determined by a country's (or region's) fiscal capacity, population health needs, and political decisions about priorities. It must include both population-based essential public health functions and personal health services. Cost-effectiveness criteria should inform these choices, but a pragmatic approach is needed when combining services at an operational, or service delivery platform, level. PHC finance and delivery are linked Directing resources to certain levels, structures, and providers makes it possible for them to function—and it also strengthens them so they can continue pulling and absorbing resources for appropriate and effective care. Conversely, inappropriate health financing arrangements can constrain effective care, and drive users to seek services that should be offered as PHC from higher levels of the system or from unregulated providers. Getting financing functions right is important. But numerous countries' experiences have shown that PHC financing reforms work best when the organsation of PHC delivery is improved at the same time. This might be done by, for example, creating new cadres of health worker, or by incentivising multidisciplinary team approaches. Organisational reforms both enable the absorption of additional resources and make PHC more people centred. We therefore argue that countries need to address financing levels of PHC, financing arrangements, and delivery structures at the same time. Financing is political Designing financing arrangements is more than just a technical challenge—it also involves choices that are inherently political, in the broad sense of the term. Political, socioeconomic, and cultural conditions are part of the context in which PHC financing reforms take place and are integral to whether and how reform occurs. Increasing the allocation of resources to health might require taking resources away from other sectors. It might also necessitate changing the roles of hospitals so they are more supportive of PHC and share responsibility for population health. Strengthening financing arrangements for PHC to better reach frontline providers and communities should mean that as resources increase, the relative distribution of resources and power will favour PHC providers compared to hospitals and specialists to ensure improvements in PHC services. Such shifts are complex because they run counter to political pressures that often favour investing in readily visible improvements, such as building facilities and reducing hospital waiting times. Whether changes are instituted as top-down radical changes or via bottom-up incremental modifications, they require shifts in power and influence at all levels. In all cases, leaders pushing to change PHC financing must attend to the political economy of PHC financing reform (including making the political case for change and building the coalitions to enact it). Political economy considerations are woven throughout the report, and section 6 specifically addresses political economy analysis for PHC financing reform. Building on strong health system foundations Financing is only one, albeit an important, element of well-functioning health systems. It strongly influences, and is influenced by, other key health system building blocks. Having these other building blocks in place is crucial. These include governance arrangements that support delivery of people-centred PHC; a health workforce that is trained and supported to provide high-quality care; data, monitoring, evaluation and learning systems to capture and disseminate accurate health and spending information; functioning procurement and distribution supply chains for medicines and other commodities; and public finance management systems supporting every aspect of financing and delivery. Financing community health and community health workers The Commission has not differentiated between PHC and community health—an area that of late attracts substantial donor funding, particularly for the deployment of community health workers.36 Indeed, community health systems can be considered an advanced form of PHC implementation, where care takes place in the communities where people live and work. These platforms also present additional opportunities for deployment of new technologies. Although community health is arguably more focused on accountability and health service delivery in the context of the community (however defined), it still depends on trained staff, supplies, infrastructure, administrative processes, and integration with higher levels of care. Community health cannot exist outside the systems underpinning PHC, and must ultimately be financed on budget through the same mechanisms recommended elsewhere in this Commission. Although some governments or donor-funded programmes might separate community health activities in the context of PHC,36 others will integrate community health workers with their clinical personnel. What is most important is recognising the shared goal of universal access to primary care services and essential public health functions also prioritised by this Commission. A key financing policy choice is whether and how to pay community health workers. Securing sustainable financing for community health worker programmes can be a challenge, particularly as in many countries these rely on external funding.37 Debates about paying community health workers typically focus on the trade-offs between reliance on volunteerism underpinned by intrinsic motivation of volunteers and the need to recognise and remunerate work fairly (and in doing so, addressing gender disparities, as most of the community health worker workforce is female).38, 39 Section 5 includes a brief summary of the evidence on paying community health workers. The private sector and PHC In many LMICS, the private sector is an important source of PHC provision.40, 41 Policy makers frequently express a desire to work with the private sector. However, the term private sector covers a heterogeneous set of providers, ranging from faith-based non-profit facilities that are well integrated into national health systems, to professionally trained clinicians operating private practices, to informally or untrained providers providing unregulated, low quality, even dangerous care. As noted, the Commission contends that universal PHC must rely predominantly on pooled public funds—but this does not preclude engaging with the private sector. Indeed, public financing mixed with private provision is widely adopted in OECD countries, where private providers are the main mode of provision of primary care in half of countries.42 Integrating the private sector into PHC platforms requires mechanisms to channel public funds to the private sector through purchasing arrangements. This pathway requires effective regulation, contracting capacity, and a broader set of purchasing institutions, including accreditation. In low-income countries, where informal private providers predominate and the public sector has insufficient resources for administering effective strategic purchasing and oversight, the best policy option is likely to be to start by providing good quality and affordable services in the public sector, while progressively strengthening regulation and professional bodies and collaborating with the private sector to develop referral pathways, training and qualification, and integration of clinical data.43 As both the public and private sectors develop greater capacities, the private sector can contribute to broader PHC functions, including public health surveillance and other essential public health functions, civil registration processes, the health management information system, and outbreak management. Section 2: The landscape of financing and organisation of PHC Analysing PHC expenditure This section sets the scene for the rest of the report by presenting data on several topics: how PHC is currently financed, how PHC providers are paid, and some of the features of how PHC is organised that are particularly relevant to financing arrangements. Key messages from this section are presented in panel 2 .Panel 2 Primary health care (PHC) financing landscape–key messages • Despite the prominence of PHC in political commitments and policy statements, limited information is available on levels of, or trends in, financial resources for PHC. Different methods of calculating PHC spending are used, making it hard to compare expenditure data from different sources. • Annual government spending on PHC is $3 per capita in low-income countries and $16 per capita in lower-middle-income countries, which falls far short of any commonly used benchmark of the minimum amount needed to provide a basic package of health services. • Much of the variation across countries in estimated spending levels on PHC can be explained by national income level. However, there is also substantial variation in government spending on PHC among countries at similar income levels. • Higher levels of spending on PHC are generally associated with higher levels of service coverage. However, at any given spending level, there is substantial variation in performance, indicating that there is a need to spend better, as well as to spend more, on PHC. • Financing of, and spending on, PHC are fragmented. Governments typically invest in outpatient services, donor funding is used for prevention, and nearly half of private spending (most of which is out of pocket) is on medicines. Although external funds are an important source, particularly in low-income settings, that augment government and out-of-pocket expenditures, they can also cause fragmentation. • Out-of-pocket spending on PHC remains unacceptably high, particularly in low-income countries, continuing to expose households to financial risk. • The most common method for paying public providers for PHC in low-income to middle-income countries is input-based budgets. Capitation-based payment systems for PHC are rare in low-income to middle-income countries. • Public providers have little autonomy over their spending, which limits their efficiency and responsiveness. To prioritise PHC expenditure, it is first necessary to measure what is currently spent. Empirical spending data allow policy makers to track existing expenditure, show how funds are currently being used, and make a case for increased commitments. Without data, it is hard to steer health systems toward stated goals, including equity, and even harder to track progress. The first challenge: defining PHC for financing analysis Analysing the financing arrangements for PHC requires clarity about what is (and what is not) part of PHC. Measuring PHC spending is challenging for numerous reasons, beginning with the breadth of the definition of PHC. As noted in section 1, PHC can include functions outside the health sector. Yet for conceptual and practical reasons, the current standard framework for tracking all financial resources for health in a country restricts the boundaries of health spending to the health sector itself. This framework, the System of Health Accounts, does refer (eg, through memorandum items for health promotion with a multisectoral approach) to expenditures by other sectors that clearly contribute to health, such as water and sanitation, but data on spending in these areas are not collected systematically. Even within the health sector, there is no clear consensus on how to operationalise the definition of PHC in a robust expenditure monitoring exercise. The System of Health Accounts does not include a designated category for reporting spending on PHC.44 Instead, it classifies health spending in various ways, including by type of provider and by type of health-care service. In practice, most countries' definitions of PHC cut across these two classifications. Furthermore, the level of detail possible in System of Health Accounts reporting does not allow for accurate tracking of expenditure through to PHC services; instead System of Health Accounts measures rely on proxies derived from reported aggregates. For example, some countries do not distinguish expenditure on general outpatient services from that on specialised outpatient services, so their PHC expenditure estimates include both. The System of Health Accounts also does not distinguish among different levels of hospital care; however, in many countries district hospitals are part of PHC, while tertiary care is provided at centralised or regional hospitals. When used to estimate PHC expenditure, such approaches can give an impression of relatively high levels and shares of PHC spending. Panel 3 describes three countries' different approaches to organising PHC, and how the differences are reflected in expenditure tracking.Panel 3 Three countries' approaches to defining primary health care (PHC) Despite the existence of a global expenditure classification system, countries vary considerably in their models and conceptualisation of PHC, including how PHC spending is measured. Three examples, Thailand, South Africa, and the UK, are presented to show the breadth of this variety. Thailand * In Thailand, PHC is defined differently in rural and urban areas. PHC in rural areas includes all services provided by the District Health Systems Network, which is comprised of public subdistrict health centres and district hospitals (each of which serves a catchment of approximately 50 000 people). The District Health Systems Network is the first entry point to PHC for all people living in rural areas and provides a comprehensive range of services throughout the life course, including health promotion, disease prevention, and primary care services. It also provides public health functions, such as disease surveillance and response, and home visits, and supports multisectoral action to address social determinants of health and empower citizens and communities. In urban settings, meanwhile, PHC is less well-developed and most of the population uses hospital outpatient care directly, without any gatekeeping. In these settings, PHC is provided by the public and private hospital-based outpatient departments that provide a comprehensive range of primary care services similar to the District Health Systems Network. Based on these definitions, the Thailand National Health Account estimates PHC expenditure as the sum of general and dental outpatient curative and preventive care provided at subdistrict health centres, and at public and private hospitals.45 Between 2015 and 2019, total PHC spending was 38% to 40% of current health expenditure. About 60% of PHC spending was financed by three public-health insurance schemes. Household out-of-pocket payments represented between 4% and 7% of PHC spending, as the publicly-covered benefit package is comprehensive and medicines are fully subsidised. Per capita PHC spending increased from US$85·7 in 2015 to US$114·8 in 2019. South Africa In South Africa, a uniform budget structure for the country's health programming was designed to specifically designate subprogrammes for PHC. The classification system is standardised across provinces and districts. Overall, South Africa spent US$92·9 per capita on public sector PHC in 2019–20. The five main subprogrammes of PHC comprise 18·8% of provincial health expenditure. South Africa also has a large, separately designated, HIV and AIDS subprogramme that adds an additional 10·4%, bringing total PHC expenditure to 32·3% of public health expenditure, or 1·3% of gross domestic product. This budget subprogramme classification differs somewhat from the WHO and System of Health Accounts 2011 classification system: the 2016–17 National Health Accounts for South Africa showed PHC represented 28% of government health expenditure. UK The National Health Service (NHS) in the UK describes PHC as “the first point of contact in the health care system, acting as the ‘front door’ of the NHS”. PHC is most closely linked to doctors in general practice, but also covers dentists, opticians, and pharmacists. Other professionals, such as nurses or physiotherapists, might also be part of a PHC team. Most primary care providers in the UK operate as independent contractors for the NHS. A range of services are contracted and financed through various models such as the General Medical Services or Personal Medical Services, Alternative Provider Medical Services, and Primary Care Trust Medical Services. According to the 2019–20 Annual Report and Accounts of the Department of Health and Social Care, primary care accounted for £12·6 billion (8·6%) of the total gross government expenditure on health of £145·27 billion. Prescribing costs, tracked separately, amounted to an additional £8·5 billion. Many countries' expenditure reporting systems are inadequate for tracking expenditure on PHC. Although both OECD and WHO use the System of Health Accounts 2011 framework to track total health spending, only OECD collects disaggregated health spending (based on both the health-care function and provider classifications) for its 34 high-income-country members. WHO began tracking health spending based on health-care functions in 2016, but does not report health spending classified by provider. In 2016, WHO reported PHC estimates for 93 countries, increasing to 98 countries in 2018 (the most recent data available). Furthermore, breakdowns of public and private spending are only available in the WHO database for 57 countries in 2016, and 61 in 2018. Both the availability and the reliability of expenditure data on PHC in LMICs remain inadequate because of underinvestment in resource tracking systems, as well as the complexity of defining and tracking expenditure on PHC.6 Finally, different international bodies and countries compile PHC expenditure using different definitions (panel 4 ). Not only is there no agreed-on approach to estimate spending on PHC, but there is also no single database that provides globally comparable data for analysing PHC expenditure.Panel 4 Working with Organisation for Economic Co-operation and Development (OECD) and WHO definitions of primary health-care (PHC) expenditure to construct an expenditure database on PHC To prepare our landscape analysis of PHC expenditure across countries, the Commission needed to combine data from various sources, considering different definitions and assumptions. We focused on two key sources: OECD46 and WHO.47 The OECD definition of PHC begins with expenditure on basic health care services derived from the health care function (HC) classification (namely, the sum of spending on general outpatient curative care (HC131), outpatient dental care (HC132), home-based curative care (HC14), and preventive care (HC61-HC64).48 An extended option also includes spending on pharmaceuticals. PHC expenditure is defined as expenditure on these services, limited to those delivered by ambulatory care providers derived from the health provider classification. WHO's definition of PHC uses only the HC classification. It starts with the same basic health care services in OECD's definition, and adds four additional components: • Curative outpatient care not elsewhere classified (HC13). • Outpatient and home-based long-term health care (HC33 and HC34). • 80% of medical goods provided outside health care services (HC5). Because HC5 also includes inpatient-related medicines purchased outside health care facilities, the 80% share is an assumption aimed to capture just the PHC element.44 • 80% of health system administration and governance expenditure (HC7); this is included to represent the share of administrative expenditures related to policy and implementation costs for population-based public health interventions. Because the WHO definition includes spending on hospital-based general outpatient care, pharmaceuticals, and administrative costs, the OECD definition is a subset of the WHO definition. Therefore expenditure levels estimated using the OECD definition will always be lower. The Commission's definition The Commission used a definition of PHC expenditure based on the WHO definition to compile the data presented below. However, we excluded the administration and governance expenditures. This brings our definition closer to that used by the OECD and limits additional arbitrary assumptions. We are not suggesting that administration and governance are not important for PHC; rather, we believe that these inputs are better captured outside the direct measurement of PHC spending. The inclusion of administration costs disproportionately biases estimates of PHC in low-income countries upwards. Even with this restriction on the definition of PHC spending, our estimates have been calculated using a broad definition of PHC that does include some hospital care. Therefore, our estimates should be seen as an upper bound estimate. Details on how we combined data from the WHO and OECD databases to increase the number of countries we covered are presented in the appendix (p 5). Definitions matter. They signal what is prioritised and valued, and they shape norms regarding how services should be organised. They also influence how data are collected and presented. There is a trade-off between using a simple global definition and accounting for country-specific definitions to permit more accurate reporting. In these early days of PHC expenditure reporting, what is most crucial is for each country to choose a way of operationalising PHC expenditure estimates so it can track its progress. Eventually, however, a consistent definition across countries will be needed to allow for cross-country comparisons and global monitoring of expenditure on PHC. In generating our estimates of current levels of PHC expenditure, this Commission was constrained by both the levels of current reporting and the definitions used by the organisations that compile data. The Commission's method of calculating PHC expenditure The Commission's approach to measuring expenditure on PHC is presented in panel 4, along with the thinking behind it. For the purposes of tracking expenditures on PHC in the future, this Commission favours using an operational definition based on service delivery platforms for PHC: population-based public health services, community health services, health centres, and first-level hospitals. This is because financing arrangements typically channel resources to providers and platforms, rather than to interventions or services. We also take the normative position that PHC should not be delivered through higher-level hospitals, because improving financing arrangements should focus on driving resources to and supporting use at the appropriate level of care, which brings services as close as possible to people and delivers them at the lowest cost. To estimate PHC expenditure using a platform-based approach requires data that cross-classify between health care function and provider category. However, because currently available data from WHO are only reported by health-care function, we are limited to estimating PHC expenditure using a service-based (rather than a platform-based) approach. We also made some modifications to the PHC expenditure reported by the WHO for the purposes of our analysis (panel 4). WHO reported data on total PHC spending on 98 countries for 2018 and government spending on PHC on 61 of these. We observed that the WHO database did not provide data on government spending on PHC for all OECD countries. In order to compare spending levels in high income countries with those in LMICs, we constructed PHC expenditure for OECD countries from data on expenditure by financing scheme, using expenditure by government and compulsory schemes to proxy for government spending on PHC. We used exchange rate data from the Global Health Expenditure Database to convert the raw data from local currency. With the addition of reconstructed government spending on PHC per capita from the OECD database following the WHO definition, the total number of countries providing data on government spending on PHC increased to 90 countries. Levels of financing for PHC Despite data limitations, our analyses have identified some notable patterns in the levels and sources of PHC spending across countries. Table 1 presents overall health expenditure data by country income group for 2018. The table shows that total expenditure on PHC in low-income countries is $24 per capita and in lower-middle-income countries it is $52 per capita. Government spending on PHC is even more meagre, at $3 in low-income countries and $16 in lower-middle-income countries, which falls short of the WHO estimate of the per capita recurrent cost for PHC of $65 in low income countries and $59 in lower-middle income countries10 (section 3). Although the share of government health spending allocated to PHC is similar in LMICs and high-income countries, government spending on PHC as a share of total PHC spending is lower in low-income countries than in high-income countries and the same is true of the share of government PHC spending in relation to GDP.Table 1 Summary of health expenditure in 201840, 41 Level of spending per capita (US$) Share of spending (%) (Total) Current health spending Domestic general government expenditure on health Total PHC spending Domestic general government spending on PHC Out-of-pocket spending on PHC External spending on PHC PHC spending as a share of current health spending Domestic general government spending on PHC as a share of total PHC spending Domestic general government spending on PHC as a share of GDP Domestic general government spending on PHC as a share of domestic general government expenditure on health Out-of-pocket spending on PHC as a share of total PHC spending External spending on PHC as a share of total PHC spending Low-income group 40 (n=16) 8 (n=16) 24 (n=16) 3 (n=14) 12 (n=14) 8 (n=14) 59% (n=16) 13% (n=14) <1% (n=14) 33% (n=14) 44% (n=14) 35% (n=14) Lower-middle-income group 104 (n=25) 44 (n=25) 52 (n=25) 16 (n=24) 23 (n=24) 8 (n=24) 52% (n=25) 29% (n=24) <1% (n=24) 36% (n=24) 49% (n=24) 14% (n=24) Upper-middle-income group 416 (n=20) 242 (n=20) 169 (n=20) 73 (n=19) 65 (n=19) 6 (n=19) 42% (n=20) 45% (n=19) 1% (n=19) 34% (n=19) 39% (n=19) 5% (n=19) High-income group 3310 (n=34) 2355 (n=34) 1312 (n=34) 840 (n=33) 318 (n=33) 0 (n=33) 42% (n=34) 59% (n=33) 2% (n=33) 36% (n=33) 28% (n=33) <1% (n=33) Total 1306 (n=95) 907 (n=95) 523 (n=95) 328 (n=90) 139 (n=90) 5 (n=90) 48% (n=95) 41% (n=90) 1% (n=90) 35% (n=90) 39% (n=90) 10% (n=90) GDP=gross domestic product. PHC=primary health care. OECD=Organisation for Economic Co-operation and Development. For total PHC spending, WHO provides data for 98 countries and OECD provides data on high income countries that overlap with WHO data but with an additional five countries which brings a total of combined data for 103 countries.. However, we excluded eight countries due to the inconsistency in how they reported health spending by functions (France, Greece, Ireland, Italy, Mexico, Portugal, UK, and USA). The total number of countries for which we present data is therefore 95 countries. For government spending on PHC, WHO provides data for 61 countries and OECD provides data for an additional 36 high income countries that brings a total of combined data for 97 countries. However, we excluded seven countries due to the inconsistency in how they reported health spending by functions (France, Greece, Ireland, Italy, Portugal, UK, and USA). This gives a total number of 90 countries for this indicator. For five countries, the Global Health Expenditure Database reports total PHC spending data but no disaggregation by financial source (ie government, external, and private [Bosnia and Herzegovina, Democratic Republic of Congo, Ethiopia, Ghana, and Uruguay]). Only five high-income countries report external funding for PHC (Barbados, Mauritius, Seychelles, St Kitts and Nevis, and Trinidad and Tobago). For other high-income countries external spending is negligible or zero. Averages are unweighted means across countries. To calculate the average of ratios, we calculated the ratio for each country and took the average for each income group. The sum of government, external, and out-of-pocket spending will not be equal to the total spending due to the omission of other types of private spending, such as voluntary private insurance. The gap is bigger in high-income countries because private insurance is more common in high-income countries. Although out-of-pocket spending on PHC is available in the OECD database, WHO only reports domestic private spending on PHC. To estimate the out-of-pocket spending for PHC in low-income to middle-income countries, we took the ratio of total out-of-pocket spending on health to total domestic private spending on health and multiplied it by private spending for PHC for each country. A descriptive analysis of PHC expenditure patterns is presented in the appendix (p 9). There is a strong correlation between income level and health expenditure on PHC, yet there is also substantial variation within any given economic level in how much governments spend on PHC. For example, spending in low income countries ranged from $8 to $46 and in lower middle-income countries, spending ranged from $11 to $120 per capita (appendix p 9). The share of PHC in current health expenditure (which includes expenditure from public, private, and external sources) decreases as countries' income levels increase—again, substantial variation exists at every income level. For example, the share in low income countries ranged from 29% to 86% (appendix p 10). Priority given to PHC within government health spending is similar on average across income groups, although there are variations in commitment at any given income level and this is more pronounced in low-income and lower-middle-income countries. However, the government share of total PHC spending was lowest in low-income countries, where both external sources and private spending (in LMICs, predominantly out-of-pocket spending) has a substantial role in financing PHC. Indeed only 13% of PHC financing in low-income countries is public. Financing for PHC in low-income and lower-middle-income countries is dominated by relatively unregulated private expenditure, most of which is out of pocket (figure 1 ). Only in high-income countries does the average share of out-of-pocket spending on PHC fall below 30% (although even in this group of countries, there is a long tail of countries with substantial out-of-pocket spending for PHC, suggesting that the issue is one of policy, not availability of resources). Figure 1 also compares the out-of-pocket share of PHC spending with the out-of-pocket share of non-PHC spending. At all country income levels, households are more exposed to out-of-pocket spending for PHC than for other health spending. This finding suggests that pooling arrangements provide less coverage for PHC and that households are more likely to be exposed to catastrophic financial consequences of paying for PHC—this underlines the importance of including PHC in benefit packages. The high level of out-of-pocket spending for PHC is particularly worrisome in LMICs, where the majority of people die from preventable causes that could be managed at the PHC level, and where poor people might be more likely to forego PHC than advanced specialist care. We contend that the lack of pooling for PHC runs counter to a progressive universalism approach to PHC, and exacerbates inequities. Finally, it is important to note that these figures do not account for people who are not able to access PHC at all; this is a general limitation of any equity analysis based on incurred expenditure.Figure 1 Out-of-pocket household spending as a share of total spending for PHC and non-PHC in 2018, by country income Out-of-pocket share of PHC is calculated as out-of-pocket spending on PHC as a share of total PHC spending. Out of pocket share of non-PHC is calculated as out-of-pocket spending on non-PHC as a share of total non-PHC spending. The box represents IQR; the ends of the whiskers represent the minimum and maximum value; the bold horizontal line represents the median. PHC=primary health care. It is also notable that about half of private spending on PHC (most of which is out of pocket in LMICs) is for medical goods purchased outside health services (figure 2 ). Much of this is likely to be for medicines; for example, The Lancet Commission on Essential Medicines for Universal Health Coverage49 reported that more than 62% of pharmaceutical expenditure in LMICs was from private sources, which is likely to be mostly out-of-pocket spending considering the low levels of prepaid and pooled resources.Figure 2 Components of PHC spending by financial source in low-income to high-income countries in 2018 Figures exclude dental care and home care. Medical goods only include medicines and other medical goods purchased outside of outpatient facilities. Private spending includes individuals paying out of pocket and other domestic private sources, such as private insurance (voluntary and compulsory). External spending includes the use of grants, concessional loans, and aid in kind from outside the country. PHC=primary health care In LMICs, the largest share of government expenditure is for outpatient care. Within this spending category, government health spending is highly skewed towards health-worker salaries. WHO's analysis of the Global Health Expenditure Database indicates that for 136 countries, 57% of public spending on health is allocated to wages.50 For PHC, in which few other inputs are used, salaries are likely to be an even higher share. This balance of spending likely represents a source of inefficiency if, after paying the wage bill, insufficient funds are left to purchase other inputs required for health workers to work effectively. In low-income and lower-middle income countries, where external funds are a significant contributor to PHC expenditure, donor funds are predominantly spent on prevention. Figure 2 suggests significant fragmentation of PHC expenditure across financing sources. Does the level of spending matter? This Commission examined whether there is a relationship between overall government spending on PHC and coverage of key services (figure 3 ). We used the universal health coverage (UHC) service coverage index51 as a proxy for PHC service coverage—of the 14 variables indexed, 11 relate to core PHC services (in addition, the index includes hospital bed density, health-worker density, and capacity to implement the International Health Regulations). Two key findings emerge from the analysis. First, higher government spending on PHC is strongly associated with better service coverage. This relationship remains strong after adjustment for GDP per capita (result not shown), which figure 3 suggests could be a potential confounder. Second, for any given level of government spending, particularly below $50 per capita, there is substantial variation in performance on the service coverage index. For example, countries that spent between $50 and $55 have a range of UHC index from 42 to 68. Whether there is a causal relationship underlying the association between government spending on PHC and service coverage is hard to show. However, the data are consistent with the notion that countries should not only spend more but also spend better to achieve improved coverage of core PHC services.Figure 3 Government spending on PHC versus UHC index, 2018 PHC=primary health care. UHC=universal health coverage. The role of donors in financing PHC Low-income countries depend substantially on external sources to pay for PHC (appendix p 11). External funding typically focuses on prevention and treatment of single diseases, which can contribute to fragmentation in financing arrangements and PHC delivery—this is especially evident when externally funded programmes are not part of government planning and budgeting processes. Community health worker programmes, which often form the backbone of PHC delivery, are also highly dependent on donor funding: an estimated 60% of funding for community health worker programmes in sub-Saharan Africa comes from external sources, much of which is funding for vertical, disease-specific programmes.52 The fragmentation of PHC financing related to reliance on donor funding is due to requirements for tracking and reporting separately on donor funded activities. It might also be a source of inefficiency due to the so-called start-stop nature of donor funding (as compared with government budgeting).53 Limitations of our analyses There are limitations to these analyses. First, use of the WHO's broad PHC expenditure definition has the effect of biasing the estimates of PHC spending upwards (because of the inclusion of outpatient services provided in hospitals). Any definition that uses a narrower scope would produce lower estimates of PHC expenditure. For example, in 2019, the OECD reported that spending on primary care within ambulatory settings represented just 12% of health expenditure; this increases to 17% when PHC services delivered in hospital settings are included, and to 34% by including retail pharmaceuticals.48 Second, data are only available for 95 out of 192 countries and for a single point in time. The spending estimates for some income groups are affected by the small number of countries in each group. Most importantly for our purposes, the absence of consistent and comparable data over time means that it is not possible to easily identify which countries are increasing or sustaining their commitments to PHC. Therefore, although these results are informative, throughout the report we also include more qualitative assessments of countries' progression towards adequate financing for PHC. Even with our limited dataset, we argue that LMICs need to spend more on PHC to provide equitable and universal access PHC that is people centred. Further, the Commission contends that adopting a definition of PHC that is consistent with the vision of providing health care at the lowest possible level, and then supporting countries to collect and report data disaggregated in this way, are essential steps toward improving the quality of national data on PHC expenditure. Data, after all, are an essential part of the strategy for monitoring and actively protecting resources for PHC going forward, and for enabling countries to show their progress in increasing their commitment to PHC.52 Organisation and provider payment for PHC Countries differ widely in terms of how PHC is structured and organised, including whether PHC includes community health workers and how they link to health facilities. Yet, although the availability of data on PHC expenditure in LMICs is improving, there is still little systematically collected cross-country data on the financial arrangements of LMIC health systems and provider payment structures. This contrasts with the data collected from OECD countries, for which the Health System Characteristics Survey provides comparable data on how countries finance, deliver, allocate resources to, and govern their health systems.42 To address this important data gap, in this Commission, we did our own cross-sectional survey in LMICs with the aim of collecting data on the key financing-related features of how PHC is organised, and on how PHC providers are paid. The questionnaire was sent in a personal email to health financing experts identified through the networks of the London School of Hygiene & Tropical Medicine, the World Bank, Results for Development, and WHO. The survey could be completed in two ways: either through self-completion or through a videoconference interview (for methods see appendix p 13). The survey was sent to 107 LMICs (one expert from each country), of which there were 75 responses: 22 from low-income countries, 22 from lower middle-income countries, and 31 from upper middle-income countries. Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 present key findings from this survey.Figure 4 Organisation and governance of public PHC providers Data indicate the proportion of countries with each organisational and governance arrangement in place. Data are disaggregated by country income group. PHC=primary health care. Figure 5 Degree of autonomy for public PHC providers Data indicate the proportion of countries in which public PHC providers have autonomy in how they function. Data are disaggregated by country income group. PHC=primary health care. Figure 6 Formal and informal user fees at public PHC providers Data indicate the user fee policy in place in public PHC providers across surveyed countries (categories are mutually exclusive). Data are disaggregated by country income group. PHC=primary health care. Figure 7 Provider payment mechanisms for public PHC providers Data indicate the type of payment system used in public PHC providers across surveyed countries (categories are mutually exclusive). Data are disaggregated by country income group. Population payment refers to capitation. Input-based payment refers to global budget, line item budget, and direct payment of salaries by government. Service-based payment refers to fee-for-service, case-based payment and pay-for-performance. PHC=primary health care. Figure 8 Paying community health workers Data indicate how community health workers delivering services on behalf of the government are remunerated in surveyed countries (categories are mutually exclusive). Data are disaggregated by country income group. How PHC delivery is organised Figure 4 presents results on indicators of how PHC services are organised that have implications for provider payment. The requirement for people to register with a public PHC provider, necessary for population-based forms of provider payment such as capitation, is uncommon in low-income and lower-middle-income countries, but more common in upper-middle-income countries. People are not restricted in their choice of public PHC providers in two-thirds of low-income and lower-middle-income countries. More restrictions on choice of provider (often in relation to the requirement to register with a provider) in upper-middle-income countries suggest that these countries seek to encourage greater consistency in use of PHC providers. Gatekeeping (in which individuals are required, or have strong financial incentives, to be referred by primary care providers to access higher-level services) is used in less than half of low-income countries. It is present in nearly three-quarters of upper-middle-income countries, although in practice these systems might frequently be circumvented. Dual practice (when a doctor works in both public and private sector practices) is common in all LMIC settings. The extent of provider autonomy links closely with provider payment arrangements, because changing providers' incentives without enabling them to make decisions about how to use their resources constrains the potential for improvements in responsiveness and efficiency. Figure 5 shows that public sector PHC providers generally lack autonomy, with fewer than half of the countries included in the survey granting providers autonomy in any domain. More specifically, public PHC providers in fewer than 40% of LMICs have autonomy to manage and retain income, nor do they typically have the autonomy to top-up the salaries of their staff to reward performance. Provider autonomy to hire and fire staff, for example, tends to increase as country income level increases, with PHC providers in more than 40% of upper-middle-income countries able to make such decisions. There is a similar pattern across country income groups when it comes to autonomy to choose the mix of services to provide. Regarding procurement of medicines, however, autonomy is greatest in lower-middle-income countries—this might reflect chronic shortages of inputs. How PHC providers are paid by different funding sources This Commission's survey confirms that user fees (whether formal or informal) are a common source of payment for PHC in low-income and lower-middle-income countries, although many countries have exemptions in place for specific population groups or types of services. Figure 6 shows that the prevalence of user fees as a funding source in the public sector decreases as countries get richer: about 95% of low-income countries report that there is some form of user fee, whereas fewer than 60% of upper-middle-income countries have them. Informal user fees remain a problem in both low-income and lower-middle-income countries; upper-middle-income countries do considerably better in this regard. The survey also provided insights into how PHC providers are paid from pooled funds (most commonly government and social health insurance payers). Figure 7 shows that these two types of payer rely fully on input-based budgets in more than 60% of low-income countries, and on input-based budgets in combination with fee-for-service in a further 30% of low-income countries; the relevance of the type of budgeting used for PHC is discussed in detail in section 4. Service-based payment, which includes fee-for-service, case-based payment or pay-for-performance, are also commonly used by government and social health insurance payers, most notably in lower-middle-income countries where it is present in almost 60% of PHC systems. Population-based, or capitation, payment systems are rarely used in low-income countries. At higher income levels, there is wider use of blended payment methods that combine different payment mechanisms and thus open up the possibility of a greater role for capitation. Capitation payment, alone or in combination with other payment methods, is used in almost 40% of middle-income countries. Our survey also found that in countries where capitation is used, around two-thirds of PHC systems adjust the payment amounts to reflect variations in health needs (data not shown). These findings from this Commission's survey support other evidence about the role of user payments in low- and lower-middle-income countries, and show that countries at higher levels of income are making greater use of capitation or blended payment as part of the move towards paying for people-centred PHC. These new data also provide a baseline for future efforts to track developments in PHC financing in LMICs. Payment of community health workers The Commission's provider payment survey also collected data on how community health workers are paid. Figure 8 shows that more than 80% of lower-middle-income countries and more than 90% of low-income countries reported having some form of community health workers, compared with just over half of upper-middle-income countries. The most common form of payment for community health workers was by salary, although payment per day or per activity was also common in low-income and lower-middle-income countries. A small minority of countries have their community health workers paid in kind only, such as gifts or training; no payment for community health workers was reported only in a small minority (less than 5%) of low-income countries. Evidence on different ways of paying community health workers is presented in section 5. New challenges for financing PHC PHC needs to constantly respond and adapt to new challenges created by changes in the environment, population health needs, and our expanding understanding of the importance of pursuing equity. Improving financing arrangements is one key way to influence adaptations of PHC. The growing burden of non-communicable diseases, the inevitable emergence of new pandemics, rapidly changing digital technologies for delivery and administration, and the growth of the private sector will all force the delivery of PHC—and its financing arrangements—to change as well. Responding to changing conditions also generates many opportunities for policy makers to influence PHC and improve its capacity to provide people-centred care that better supports social goals, such as UHC and equity, on the national and global scales. Financing should, at a minimum, not impede needed changes in service delivery; at best, financing arrangements can serve as a transformational force. This section describes several challenges for delivering PHC and the promising approaches to organising PHC financing that are elaborated further throughout the rest of the report. Financing to facilitate service integration Coordination and integration of services is at the heart of people-centred care. PHC has been described as the coordination mechanism that links primary care, community care, specialised care, wider public health interventions, and long-term care services.16 However, efforts to improve coordination and integration often face challenges at the interfaces of services and specialities; as noted, in LMICs, many challenges have arisen from the absence of integration between PHC and donor-financed vertical programmes, such as those focused on maternal health or HIV and AIDS. Improving coordination among different PHC services is as important as coordinating PHC with higher levels of the health system. Problematic financing arrangements are frequently at the core of these challenges. Different financing mechanisms for different sources of funds, the means of allocation and management of flows of funding, and complicated payment mechanisms all act as major barriers to the implementation of more integrated approaches to service delivery.54 As noted in the Commission's case study on Chile, for example, the absence of strategic coordination between the capitation payment for PHC and the application of diagnosis-related groups for hospital care made clinical coordination and integration of care difficult.55 Where PHC is provided by an unorganised private sector and paid for by out-of-pocket payments, care can be fragmented and create excessive financial burdens. In this situation, for example, every provider might require the same diagnostic tests to be repeated and there can be excess prescription of branded medicines.41, 56 Donor funding can be a further impediment to integration. How health financing arrangements can support integration is explored in section 5. Digital innovations in health financing One major development in PHC, as indeed in all sectors, has been the rapid development of new digital technologies. In this Commission, we reviewed the published and grey literature regarding two questions: where have digital technologies facilitated financing of health and PHC? And how have financing arrangements been adapted to facilitate the digital delivery of PHC (appendix p 17)? After the literature review, a roundtable discussion took place among 13 digital and health financing experts (academics, practitioners, and donors) and the Commissioners captured recent and unpublished experiences and insights (for methods see appendix p 35). Four main messages emerged from this process. First, very little robust evidence exists on the impact of digital technologies on health financing objectives, including financing for PHC. Rigorous research in this area is urgently needed. Second, digital technologies offer great promise to improve the efficiency of resource mobilisation and purchasing arrangements; however, caution is warranted. There are risks, such as fragmentation if it leads to multiple small funding pools or parallel electronic patient information systems. Another risk is misalignment with UHC equity principles; for example, for the moment, digital access is much higher in cities, although this is changing rapidly. Third, to date, digital technologies related to financing have mostly been applied to health-care purchasing. For example, in the Philippines, digitised claims analysis using artificial intelligence is being used to detect fraud;57 in India, provider payment has been facilitated through an online system for data reporting and online pay-for-performance of Accredited Social Health Activists.58 Some applications to revenue mobilisation and pooling have also been documented: in Ghana, mobile phone apps have simplified payment of health insurance contributions and enrolment reminders.59 Fourth, little has been documented about any actual adjustments to financing arrangements to facilitate online and digital delivery of PHC, beyond measures to include remote consultations in insurance benefit packages. However, this area is developing rapidly (for example, the introduction of Babyl, an artificial intelligence-supported digital health service provider, in Rwanda)60 and often without formal assessment or evaluation. Financing arrangements must be adapted if digital solutions are to be included in pooled funding arrangements, and provider payment levels need to be adjusted to cover additional activities required to integrate new technology enabled activities. Most recently, the public health and social measures put in place to address the COVID-19 pandemic, such as lockdowns and quarantines, accelerated the development and use of various forms of digital health-care delivery, including remote consultations and home-based monitoring. Financing arrangements were quickly adapted to enable these changes, such as extending benefit packages to include remote consultation. PHC needs to continue to embrace new technologies that support coordination, interoperability, and regulation; and financing arrangements should continue to adapt to accommodate these. It is also important to avoid a situation where multiple solutions act independently and foment fragmentation. The key will be to allow flexible financing approaches to support these developments, while ensuring that they are thoughtfully integrated into the wider health system to induce efficiency gains without sacrificing equity. This is discussed throughout the Commission. Paying for medicines for PHC The data presented above show the importance of expenditure on medicines—which is sometimes as high as 40–50% of all expenditure and is often paid for out-of-pocket through private pharmacies—in PHC spending in many LMICs (figure 2). However, medicines are often out of stock in public health facilities or not included in provider payments. Patients then pay out of pocket for medicines purchased in private pharmacies, representing a substantial barrier to the use of medically appropriate and high quality medicines. This is particularly apparent for non-communicable diseases.61 Such out-of-pocket expenditure leads to inefficient use of medicines and financial burdens for patients. Effective financing arrangements for medicines for PHC should have two key outcomes: to provide financial protection against the costs of medicines, and to encourage efficient use of resources. Financing arrangements for paying for medicines should focus on four goals. (1) Including essential medicines in the PHC package that is covered by pooling arrangements, including those needed for management of chronic conditions. (2) Ensuring that sufficient funds are mobilised to supply these medicines and deploying all available policy tools to procure medicines efficiently (including essential medicines lists, bulk procurement, and market shaping to secure favourable prices). (3) Encouraging efficient use of medicines through appropriate incentives to providers, dispensers, and patients. (4) Enabling new ways for people to access medicines that allow them greater control and easier access, for example, through self-service vending machines,62 e-prescription and e-pharmacy. Conclusion In section 2, we provided an overview of how PHC is financed and how PHC providers are paid. Although the findings are instructive, it is clear that much of the global expenditure data on PHC are problematic. A consensus on a consistent operational definition of PHC would enable better expenditure analyses—it would also help to prioritise PHC, promoting the normative message that PHC should not be provided at higher level hospitals, and enable countries to better track their progress. Generating agreement on a definition requires taking a position on the trade-offs between, on the one hand, allowing flexible definitions that fully capture local particulars, and, on the other hand, having a consistent definition that enables global monitoring and comparisons. Even if the data presented in this Commission significantly overestimate PHC spending, they still show that the levels of government spending on PHC in low-income and lower-middle-income countries are far from sufficient, and that a considerable share of current financing comes from out-of-pocket, not pooled, resources. Furthermore, the relationship between PHC spending and service coverage, as measured using the UHC service coverage index, shows that existing money could be spent more effectively. These findings underpin the Commission's overarching message: that countries need to spend more and spend better on PHC. Better spending would mean replacing out-of-pocket spending on PHC with pooled public sources (see sections 3 and 4); ensuring that the resources intended for PHC reach the frontline providers; reducing fragmentation of funding; and expanding provider autonomy by improving how providers are paid to better incentivise people-centred PHC (see section 5). Section 3: Mobilising and pooling resources to finance health care This section addresses how to take a people-centred approach to raising more prepaid and pooled funding for PHC by mobilising more funding for health overall. The issue of how resources are allocated to PHC from the overall health budget is addressed in the next section, although these are closely intertwined in practice. This section includes: an overview of the challenges faced in mobilising and pooling financing for health, a vision for increasing resources for health, and some possible strategies to pursue this vision. It is important to note throughout that stakeholders beyond the health sector have the greatest responsibility for raising additional resources for health. The ministry of finance is key because it both has responsibility for raising government revenue and wields significant power in budget-setting negotiations with spending ministries, such as the ministry of health. Key messages from this section are presented in panel 5 .Panel 5 Mobilising and pooling resources for primary health care (PHC)–key messages • Government expenditure on health falls short of what is needed for UHC, which limits the overall ‘pie’ available for the PHC share and forces patients to continue to pay out-of-pocket, posing a persistent barrier to access. PHC should be free at the point of use because even small payments can deter use. This requires progressive removal of user fees and increased public funding. Increasing funding will be challenging in the near-future because of the economic impact of the COVID-19 pandemic, which will constrain government budgets. • Generating additional pooled resources is a challenge: fiscal capacity remains limited by macroeconomic conditions and inefficient revenue collection; however, additional resources will have to come mainly from taxes (general or earmarked); expansion of social health insurance, a strategy being pursued in many low-income to middle-income countries, is constrained by the small size of the formal labour force in many countries; thus donors continue to have an important role to play in low-income settings. • Increasing tax revenue is both a technical issue (how to increase tax capacity and how to broaden the tax base) and a political issue (due to acceptability and compliance): the latter requires skilful engagement in budget politics. • Better spending of available resources is key, although the potential to generate efficiency savings in the health sector is limited within existing institutional arrangements; it also takes time (and often investment) to achieve these savings. • Better pooling arrangements are also needed to reduce fragmentation, secure equitable cross-subsidies and efficient integration between levels of care. Where actual pooling is not possible, intermediate pooling supported by, for example, harmonisation of financing arrangements across pools, or virtual pools supported by digital technologies, can provide intermediate solutions. The Commission has not taken a position on a target spending level for health. That topic has been considered by various bodies that have each proposed their own benchmarks.63, 64, 65, 66, 67 Various governments and donors have committed to these benchmarks, although few have ever actually met the targeted spending level. Regardless of which benchmark is considered, LMIC governments and donors do not yet spend enough on health.33 These benchmarks might be useful for advocacy purposes, and to give a global indication of how much is needed overall. However, they do not recognise that a given level of spending could be necessary but not sufficient to achieve desired outcomes. The benchmarks also disregard the substantial variation in performance across countries with similar levels of government budget allocation to health.68 Whatever the benchmark, estimates of resources required should be based on per-capita needs rather than simply funding facilities or infrastructure. Challenges in mobilisation and pooling of funding for health The existing mechanisms for mobilising and pooling health funds in LMICs have multiple weaknesses that arise from their fragmented and insufficient revenue sources and the inherent challenges of pooling and managing resources. There are several specific problems. The first is limited tax revenues. Public spending, including on health budgets, is often constrained by the general macroeconomic conditions and underlying weaknesses in national systems of taxation that limit the revenue base for public spending (as shown by the low mean tax–GDP ratio in low-income countries of 12%, as compared with nearly 30% in high-income countries).69 The second is social health insurance contributions. Social health insurance contributions represent an important source of health funding in several high-income countries.33 However, the formal labour force in many LMICs is insufficient to support expanding social health insurance contributions; in low-income countries, for example, only 10·2% of the population works in the formal labour sector.70 The COVID-19 pandemic has increased unemployment in the formal sector, with ever more workers shifting to the informal sector where it is more difficult to collect contributions.70 Furthermore even when social health insurance contributions are collected, translating them into additional revenue for health is made complicated by the issue of fungibility (interchangeability) within government budgets. Third, a larger government budget does not necessarily lead to greater allocation to health. Even when governments do manage to expand the total envelope of available funds, either through tax revenues, social health insurance contributions, or donor financing, this does not necessarily lead to greater allocations to health.33 Failure to prioritise funding for health in total government spending might be due to technical issues (such as insufficient capacity within ministries of health to present strong investment cases for health to ministries of finance).71 It might also arise from public finance management structures that are not aligned with health budgeting needs or as a consequence of various political economy factors such as power imbalances and low levels of accountability between health system users and political leaders, and lobbies promoting other public expenditure priorities (see section 6). The fourth problem is inadequate, declining, and fragmented donor funding. Donor funding for health, although valuable and, for the foreseeable future, essential for low-income countries, has been falling since 2015· 33 The economic effects of the COVID-19 pandemic are likely to also negatively affect donors' budgets. For example, in 2021, the UK Government reduced its international aid spending target from 0·7% of gross national income to 0·5%. Furthermore, donor development funding can be unpredictable even in good economic times as it is based on donor preferences. The funding that is made available is typically largely targeted to specific disease programmes (in low-income countries, for example, an average of 65% of donor assistance is allocated to infectious and parasitic diseases).33, 72 Further, donor contributions tend to operate off-budget, undermining countries' ability to plan and manage these resources alongside domestic funding.33, 72 Nevertheless multilateral and bilateral assistance for selected low-income countries remains essential to at least 2030, and an improved global system of multilateral country support is required. The fifth is excessive reliance on out-of-pocket spending. As alluded to previously, and as a result of the difficulty in raising pooled public funding, out-of-pocket payments by individuals are the dominant source of funding for health in low-income countries: they represent, on average, more than 40% of all health spending across all LMICs, and nearly 60% in the 32 low-income countries.33 Out-of-pocket payments can include formal user fees, informal fees levied by individual providers, travel costs, and payments for medicines and other supplies. Because payments made out of pocket are not pooled, they curtail cross-subsidisation between rich and poor, as well as between healthy and sick populations. The global health and economics communities have stated unequivocally that out-of-pocket payments need to be reduced (and user fees for PHC removed) and instead replaced with prepaid and pooled funding.73 Indeed, formal and informal user fees act as barriers (disincentives) to accessing PHC, particularly for the poorest; for those who do choose to pay, user fees can lead to financial hardship.74, 75 The frequent argument that fees paid at the time of service prevent frivolous use of health services without causing harm is not substantiated by existing evidence.76 Requiring even small payments can dramatically reduce use of highly cost-effective interventions.77 User fee exemptions for the poor and vulnerable can mitigate access barriers, although experience in LMICs suggests they can be challenging to implement, as it is hard to identify the poor effectively and the criteria are often gamed or applied inconsistently.78 The sixth is reduced budgetary allocations to health due to the economic impacts of COVID-19. Budget allocations to health were falling in LICs and lower-middle-income countries even prior to the COVID-19 pandemic.33 As the COVID-19 emergency phase ends, many countries can be expected to experience falling government budgets.29 For example, the South African Health 2021 Medium Term Expenditure Framework budget was reduced by almost 10% over one budget cycle as a result of the post-COVID budgetary restrictions.79 Austerity budgets following previous crises have had very harmful effects on health budgets.80 The seventh is difficulty in pooling resources. The incremental nature of UHC expansion often leads to the development of multiple risk pools as different programmes evolve to cover different population groups.81 Fragmentation of revenue streams is also an obstacle to pooling, with general tax revenues collected and used through the budget system and often disbursed as input-based budgets to maintain the health delivery infrastructure, and other sources of revenue pooled in an off-budget fund (such as a public insurer) and disbursed as payments for services. Once multiple pools have been established, it can be politically difficult to merge or integrate them, as this requires trade-offs with some interest groups losing, or perceiving to lose, their advantages. The convergence of these problems in mobilising and pooling funds for health has resulted in large variations in levels of government health spending, even among countries with similar economic status (appendix p 12). The Commission's vision for resource mobilisation and pooling The Commission's vision for people-centred financing of PHC requires an adequately financed health sector funded by expanded public and pooled sources that protects people from financial hardship when seeking care and promotes equity. Achieving the Commission's vision involves reducing out-of-pocket payments (including removal of user fees for PHC) and replacing them with other types of pooled public funds that are raised through progressive means. In line with progressive universalism, the Commission argues for an explicit focus on addressing inequities.82 Mobilising new resources for government spending on health One of the greatest sources of increased government health expenditure has been economic growth, rooted in conducive macroeconomic conditions.83 However, in the near term this is unlikely to be sufficient to secure the additional financing needed due to the adverse global economic outlook generated by the COVID-19 pandemic. Below, we discuss other strategies for mobilising new public resources for health and investigate how countries can translate increased general revenue into increased funding for health. We also comment on the potential for the health sector to improve the efficiency of spending of its existing resources. Increasing the overall envelope Governments and health systems have several options for increasing resources for health, including expanding taxation, expanding social health insurance contributions, and increasing borrowing. The primary means of expanding resources for health in LMICs is to expand the overall available government revenue collected via taxation. This can entail improving collection of existing taxes, increasing the tax base, and expanding the number and types of taxes levied. LMICs, however, face significant challenges when collecting tax revenues. Constraints including lack of infrastructure (including information technology systems) and administrative challenges, such as incomplete property registers, the size of the informal economy, or the inability to trace transactions within it. These limit the capacity of tax agencies, often already weakened by an insufficient number of skilled staff and limited technical capacities. Expanding LMICs' national taxation capacity therefore requires strengthening various institutions, systems, and skills. Countries also need to decide on the appropriate mix of direct (income), indirect (for example value-added tax) and other taxes (including trade taxes), in which there will be a trade-off between administrative complexity and equity. Income taxes can be more progressive than indirect taxes, but require more elaborate systems to enforce compliance.84 More global comparative data on taxation across countries is now available than ever before and is vastly better than what was available 10 years ago.85 LMICs could also focus on taxes that directly effect health outcomes, which are often perceived as politically more acceptable.86 The health sector is increasingly active in advocating for health taxes or excise taxes on unhealthy products such as tobacco, alcohol, and sugar-sweetened beverages. However, the primary purpose of these taxes is to improve population health status by creating disincentives to the use of harmful products. Therefore, where excise taxes are successful, they should eventually result in reduced revenue. However, in the meantime, these can raise revenue that can be earmarked for health if political economy considerations and budgetary rules are aligned.87 According to the World Bank, an increase in excise taxes on tobacco, alcohol, and sugar-sweetened beverages that would result in a 50% increase in prices could raise the tax–GDP ratio by an average of 0·7 percentage points in low-income countries and LMICs.72 Other taxes could also be explored for the health sector, such as taxes on transport and airline levies88 or on carbon emissions.72 Overall, the potential additional revenue that these taxes or levies could bring seems to be limited (up to a maximum of 2·3% of actual spending on health).89 Earmarking these taxes to cover health expenditure is possible29 although it might be resisted, particularly by finance ministries,86 because of the inflexibility it introduces into the budgeting and planning process. A tax on payroll, or compulsory social health insurance (rather than voluntary as is the case of many community-based health insurance schemes) is also used as an earmarked tax for health. Indeed, many countries are working towards UHC through expansion of social health insurance,90 as it holds the potential to improve equity by increasing pooling and introducing pre-payment.91 However, empirical evidence raises some concerns about the introduction of compulsory health insurance contributions in low-income countries.92 First, the necessary conditions for raising additional social health insurance funding are seldom met because most do not have full (or nearly full) employment and few jobs are located in the formal sector, registered, and therefore taxable.90 Second, social health insurance might not be feasible due to the high costs of setting up and administering the collection of contributions92 and might take a long time to lead to the achievement of UHC, as was the case in Germany for example.93 Whether the revenue raised can outweigh the operational costs has seldom been analysed.94 When it has been examined, the findings are inconclusive.90 Third, the cost to employers of social health insurance premiums might also have a negative effect on the labour market, reducing the probability of employment in the formal sector as well as wages.95 The changing nature of work, including demographic changes and structural changes in employment, pose a challenge to the feasibility and sustainability of mobilising resource for health through employment-based models such as social health insurance.96 Finally, social health insurance does not necessarily support progressive universalism, as it can redistribute resources towards wealthier segments of the population. For example, general revenues can be used to subsidise social health insurance that predominantly serves upper-income groups rather than ensuring that subsidies are used to extend coverage to vulnerable segments of the population.90 Whatever tax is chosen, political economy factors, both internal and external, as well as the structure of LMIC economies, always threaten the feasibility of taxation reforms (panel 6 ).Panel 6 Political economy challenges of expanding taxation in low-income to middle-income countries (LMICs) An increase in taxes is never welcomed by taxpayers. Introducing taxation reforms involves complex political economy concerns, including: Internal factors Taxation reform involves extensive bargaining among different actors, including governments, taxpayers, specific industries and sectors, tax intermediaries such as accountants and tax advisers, and revenue collection organisations.97 As such, the process of bargaining with all stakeholders will be central to decisions about the design of tax regimes. Certain groups might be disproportionately affected by different taxes—and this might constrain the feasibility of adopting them. For instance, although a property tax could be an effective way to increase tax revenue, these tend to affect wealthier people who are closely connected with political decision makers.98 Similarly, sugar-sweetened beverage taxes, although important for chronic diseases, affect particular agricultural and industrial sectors. External factors LMICs' ability to raise taxation revenue is currently substantially limited by existing bilateral tax treaties and the associated curtailment of taxes on multinational companies, tax evasion and avoidance, loss of potential revenue from extractive industries, and tax exemptions and incentives given to investors.98 The G20 finance ministers' recent announcement of a minimum 15% tax on multinationals in countries where they operate is a positive step towards addressing this issue.99 Structure of the economy LMICs typically have a large informal sector which is difficult to tax. Further, the economy might be comprised of many small-scale firms that are more likely to be dependent on a few natural resources or commodities. LMIC economies also often rely on foreign aid, which is not taxable. These factors limit tax collection and the tax base.84 Generating additional resources from taxes, including for health, is certainly possible—but it requires a tailored strategy in each country and it takes time. Developing a tax reform strategy requires understanding the types of taxes and rates that are feasible given the political and economic structure of a country, and a whole-of-government effort to successfully push to increase the overall revenue envelope. For example, Tunisia increased its tax revenue from 18·2% of GDP in 1990, to 32·1% in 2018.100 It is now nearly double the average of 16·5% across Africa in 2018.101 This was achieved through the implementation of multiple policies, including: countering harmful tax practices, preventing tax treaty abuse, making dispute resolution mechanisms more effective, tackling money laundering, and cracking down on tax evasion. Aside from the technical solutions that were adopted, political resolve was essential in driving these reforms, and their success.100, 101 Another way to increase the budget available for health is to use government borrowing, as a short term approach. In times of crisis, government borrowing can have an important temporary role in sustaining health and social spending. The ongoing global COVID-19 pandemic, and the associated global recession represent just such a situation. Some reflections on the challenges of mobilising revenue in times of crisis are presented in the appendix (p 3). Deficit financing is typically considered more suitable for capital, rather than recurrent, expenditures. However, it can have a crucial role in supporting countries to emerge from crises while helping to mitigate their social consequences. In the first quarter of 2021, the International Monetary Fund, World Bank, and OECD all called on countries to take on temporary debt to finance human development. These institutions have emphasised the important role of the health sector in dealing with the pandemic, as well as in turning around the COVID-19-induced economic crisis. They also note the importance of not withdrawing social support, health spending, or fiscal stimulus in the midst of a pandemic.102 The increased borrowing will need to be accommodated by fiscal adjustments in the future. Most countries are projected to go through extended periods of fiscal adjustments in the coming years; in some places, the overall government spending envelope is expected to shrink, as will discretionary spending as a share of government expenditures, as debt service requirements grow.29 Grants, concessionary loans, and favourable debt treatment, including debt service suspension and restructuring, will be vital for many LMICs in the coming months and years. Government debt can help maintain liquidity and solvency and support to the spending needed to recover from COVID-19. Although LMIC governments might be reluctant to acquire more debt, it represents a way to respond to crises while expanding support for the health sector. Ultimately, investing in health will benefit the economy, enabling countries to get out of debt more quickly. Another way to increase the overall envelope is by bringing donor funding on-budget and aligning it with national priorities. The COVID-19 pandemic (just as with the 2016 Ebola epidemic) showed that investing in health in LMICs has global significance.29 Donor funds not only need to increase, but they also need to be used in ways that align with government priorities, after the principles of the Paris Declaration.103 Translating more resources overall into additional funding for health Whichever strategies are selected by governments, and specifically ministries of finance, to secure additional resources, the next step is to ensure that the new resources are invested in health. The widespread economic impact of the COVID-19 pandemic has provided clear evidence of the close connection between health and economic prosperity and should strengthen the case for increased investment in health. However, ensuring that adequate resources are allocated to health requires continuous efforts. The allocation of resources to health is an intensely political issue. Multiple political economy factors have a role in whether health is prioritised. The first factor is competing demands. In the face of infinite needs, different sectors in a given country must compete for limited funds. How much goes to health depends both on how much value is placed on health and on what other challenges the country faces. A country struggling with active conflict might prioritise military spending over health. For example, Mali's budget allocation to health decreased from 2016 in the face of a domestic terrorist insurgency.47, 104 Health advocates must both understand the competing demands and determine how best to present health as an economically sound and ethical investment. The second is who benefits from funding health. In most countries, small elite groups of politicians and financiers wield the most control over budgeting and financing decisions. Unless they see benefits in investing in health, little might change. In Turkey, for example, there were nine unsuccessful attempts to establish a national health insurance system for UHC before 2003. Those attempts were blocked in part by legislative gridlock in the parliament and opposition to new legislation by the Constitutional Court.105 UHC was eventually adopted after the rise of the AK party, whose platform centred, at the time, on opposition to persistent inequalities106 (see section 6 for further details). Third is that in certain contexts, health must become part of a popular political agenda in order to be prioritised. This can occur during political transitions, such as following the fall of dictatorships,55, 107 the rise to power of an insurgency (as in Ethiopia),108 or as a result of crisis such as COVID-19 (see section 6 for further details). Whilst this list is by no means exhaustive, it illustrates the need to take political economy considerations into account in tandem with technical solutions to gain support for health. Better pooling of existing and new resources Whether or not total health spending increases, a shift from out-of-pocket spending towards pooled arrangements can radically improve the equity and efficiency of health financing. A defining characteristic of the health sector is the high degree of uncertainty associated with health needs, which vary across populations, over time, and across geographies. This uncertainty makes it necessary to pool risk across populations to protect individuals from financial hardship if they find themselves in the unlucky group that requires expensive health services. Redistribution of resources from people and places of lower need to those of higher need is more effective in larger, more diverse pools. Pooling can occur either in government budgets (at central or decentralised levels), through compulsory insurance schemes, or, potentially, through virtual health insurance pools supported by digital technologies. The opportunities, and challenges, of virtual pools are discussed in the appendix (p 34). Many countries are moving in the direction of merging or consolidating multiple pools to gain the benefits of more effective risk-pooling, which include better redistributive mechanisms and greater equity, stronger purchasing power, and increased administrative efficiency.109 Better spending Inefficiencies in health spending exist. The 2010 World Health Report stated that there were inefficiencies equivalent to 20–40% of health spending, and in the OECD between 20% and 50% of health expenditures might be wasted due to inefficiencies.73, 110 Inefficiency can be allocative (that is, spending money on the wrong interventions, such as high-cost, low-impact health services). Inefficiency can also be technical (failing to get the maximum output from available inputs, for example when patented drugs are purchased in lieu of generic drugs). In theory, addressing these inefficiencies could be a source of resources available to the health sector by releasing resources while maintaining (or even increasing) output. Efficiency can also improve if greater outcomes are achieved with the same inputs; however, this does not increase fiscal space. Efficiency reforms should not be seen as a way to either balance budgets or identify spending cuts in health. However, they do form part of the wider effort to use available health resources to improve health outcomes, and to persuade ministries of finance to increase the health budget. Improving either technical or allocative efficiency is a complex task, fraught with technical and political hurdles, and requiring time. A review done for this Commission111 showed that the efficiency gains that could be achieved in practice through reforms addressing inefficiencies are likely to be smaller in magnitude than suggested in the 2010 World Health Report.73 Furthermore, the available evidence on the timing and feasibility of efficiency-focused reforms is scarce and not generalisable across countries. The methods used for the literature review, together with an overview of possible efficiency reforms identified, are presented in the appendix (p 37). Although the effect of some reforms focused on enhancing spending efficiency might be immediate (such as requiring the purchase of generic rather than branded medicines, or use of bulk tenders for medicines or medical equipment), many others may take years to reap benefits. Many inefficiencies, such as leakage due to corruption or fraud, are structural; tackling them requires addressing historical precedents and social norms in addition to administrative processes.112 Further, it should be noted that addressing some inefficiencies can require upfront investments. This might be one area where digital technologies can prove (positively) disruptive, as long as they do not create new fragmentation (appendix p 16). Finally, although the focus here has been on whether inefficiencies could lead to financial savings (as has been argued by many finance ministers), ministries of health should be working to improve the value for money of their spending. This outcome might be realised through a decrease in cost, as long as the outcome (including quality of care) remains the same. It might also involve an increase in cost with an accompanying improvement in outcomes or simply an improvement in outcomes, with the cost remaining constant. Conclusion The need to shift away from out-of-pocket spending towards pooled public funding is urgent, yet the reforms to increase resource mobilisation and pooling necessary to achieve this objective will take time to design and implement. Investments in strengthening the tax base, expanding the types of taxes levied, and tax collection capacity—all of which fall outside the purview of the ministry of health—will be essential. Digital technologies might hold some promise (for example through virtual pools), although caution is warranted in ensuring that these align with UHC objectives. Improving the efficiency of spending can help, but is not the primary way to create more fiscal space for health. Section 4: People-centred allocation of resources to PHC The Commission's vision for a people-centred approach to allocating sufficient resources to PHC In the Commission's vision, each country's health sector strategically uses appropriate policy tools to direct sufficient resources to PHC to enable a universally accessible system that provides quality services in line with a benefit package appropriate to the level of care. To do so requires mechanisms for funding, budgeting, and financial management to ensure that those resources reach frontline providers and platforms. It will also organise the service delivery system to pull resources to PHC. By doing so, countries can support the delivery of, and equitable access to, people-centred PHC. Key messages from this section are presented in panel 7 .Panel 7 People-centred allocation of resources for primary health care (PHC)–key messages • Increasing the allocation of health resources to PHC is a political decision; it might involve redistributing of resources—in absolute or relative terms—away from other sectors, or within the health sector away from hospitals. • Increasing budget allocations to PHC does not guarantee that resources reach frontline services; protecting PHC allocations to the point where they reach frontline providers requires clarity, active steering, and accountability mechanisms. • Making allocations to PHC more visible in health budgets is an important way to improve tracking of existing resources, secure additional resources, and highlight the importance of essential public health functions. • A range of policy levers are available to increase and protect allocations to PHC. Public finance management tools, particularly those that strengthen budget formulation and execution, can be used to increase PHC budgets, make them more visible in the public finance management system, and ensure that resources reach frontline services. Service delivery arrangements can be used to pull resources to PHC. These arrangements include explicit service standards and guidelines, new configurations of teams, and effective referral systems. In some instances, new cadres of frontline PHC providers have enabled more resources to be directed to PHC. All these improvements can have the effect of stimulating demand for PHC and, with the right financing arrangements, drawing more resources to this level. • Multiple tools can be applied at the same time. Many of them require a clear and context-specific operational definition of PHC. Applying public finance management levers requires engaging with various health and public finance management system capacities. • Institutional responsibility for PHC is typically fragmented across ministry of health departments; as a result no single unit is in charge of securing funding or held accountable for progress. Although it is not necessary to create a new operational unit for PHC to ensure that PHC is prioritised, it must be clear where responsibility for budgeting and planning for PHC lies in the ministry of health. The objective of resource allocation for PHC is to ensure that sufficient resources are directed to, and reach, frontline providers working in PHC platforms, including at community level, as well as supporting essential public health functions. Whether or not the overall fiscal space for health is increased, deliberate efforts can be made to allocate more resources to PHC from existing health budgets, either through the budget process or by strengthening budget execution to ensure that available resources for PHC effectively reach PHC platforms and providers. Putting people at the centre of these arrangements entails allocating resources for PHC based on population needs, rather than allocating resources to facilities, inputs, or vertical programmes. To achieve equity in this allocation process means prioritising the needs of people with the lowest socioeconomic status and least-served geographic areas. This section of the report explores the forces that can draw resources away from PHC, and describes policies that support allocation and protection of resources for PHC. It also addresses the special case of allocating and securing financing for essential public health functions. We acknowledge that various other policies also indirectly influence allocations to PHC, including some key PHC design elements and how PHC is linked to higher levels of the health system. Provider payment methods, which are addressed in greater detail in section 5, are central to this. How resources for health are allocated Allocation of resources to health in general, and to PHC in particular, occurs in different ways in different systems of government. In centralised budget systems, the collection and allocation of funds takes place centrally with, typically, the Ministry of Finance allocating a set amount, based on multisectoral budget negotiations, to the Ministry of Health. Although some of these resources might be ring-fenced for specific purposes, many decisions about allocation within the health sector are the responsibility of the Ministry of Health. It decides how to allocate the available funds to geographic units (such as provinces) or to levels or platforms of care, such as PHC. In decentralised systems, decisions about resource allocation—across sectors and within the health sector—are made by local authorities. This gives the opportunity for resource allocation decisions to be shaped by local needs, disease patterns, and priorities. However, it can also lead to a focus on financing services that are popular or visible, rather than those that bring the greatest population health benefits. A key challenge for decentralised systems is that allocations to health in general, and PHC in particular, might be less visible and traceable at national level than when they are presented as part of a single centralised budget. Therefore, decentralised systems might require additional policy tools (such as common budget structures and reporting systems) to protect, manage, and monitor the use of resources for PHC. The countries of Kenya, India, and the Philippines are all examples of devolved decentralised systems where allocations to, and power over, health are substantially determined at subnational level. Whether a government is organised as a centralised or decentralised system represents a broader political choice that is made outside of the health sector. However, the health sector must adapt its allocation processes to match the wider political structure of the country. As with allocations to health in general, there is no definitive answer about how much is the right amount to allocate to PHC in any given setting. The key issue is estimating the resources required to finance a PHC package that is universally accessible, places minimum financial burdens on users, and that is aligned to macrofiscal capacity. The analysis of the relationship between PHC spending and coverage of priority interventions presented in section 2 showed that there is substantial variation in the level of coverage achieved for a given spending level, particularly in LMICs; section 3 discussed some evidence about spending targets. Costing exercises can help to inform these decisions. However, costing requires a clear operational understanding of what constitutes the PHC platform and services in that particular setting. South Africa has used numerous approaches to cost its PHC system, including normative personnel and infrastructure costing, modelling visit rates and unit costs, complex methods of costing specific disease groups and treatment interventions, comparisons of facilities using cost, efficiency, and quality indicators, and actuarial approaches in designing capitation payments. The politics of resource allocation within the health sector Budget allocation processes are influenced by a range of political forces operating at all levels. Problems, such as those created by political patronage systems,113, 114 can be exacerbated in decentralised systems where local political incentives, power relations, and special interests can carry more weight than the policy priorities of a central Ministry of Health. Vested interests at central level might also skew resource allocation. PHC faces particular problems in attracting sufficient resources because it typically does not elicit strong political support within the health sector and may be excluded from the benefit package of particular insurance arrangements. Responsibility for elements of PHC might be fragmented across agencies or technical departments, with no clear responsibility or accountability for delivering on policy commitments to improve PHC. Other factors combine to limit the political attractiveness of PHC. In particular, the people who would benefit most from expanding and improving PHC (for example, children, women, and people living with chronic conditions) typically have little political power. Furthermore, when PHC is done well, it keeps people healthy and thus becomes nearly invisible. In contrast, allocating resources to hospitals is more politically appealing. The hospital is a highly visible symbol of the health system. A hospital is the site of more, and more expensive, technology, which is perceived as beneficial and important. Hospitals also employ specialist physicians, who tend to have higher professional status and political connections than PHC workers. Hospitals engage in readily apparent medical education and other training activities, and employ large numbers of support staff. They tend to be located in urban areas, closer to decision makers, and cater to wealthier population groups who wield significant political influence. Therefore, when limited resources are available for health, hospitals and other secondary and tertiary care initiatives are often more successful at securing them. However, this imbalance undermines the efficiency and equity of the health system as a whole. In this section, we simply note that the process of securing budgets for PHC is not merely technical but is also influenced by political forces. Section 6 addresses the political economy of financing PHC in much more detail. Forces that impede allocation to PHC and the policy levers that protect resource flows A number of technical factors can also impede increased allocations to PHC or divert funds away from PHC. Addressing these is essential to protect resources for PHC and ensure they reach the frontline. This can be done through careful system design and, in particular, the use of various categories of policy tool (figure 9 ).Figure 9 Policy tools to increase and protect resource allocations to PHC PHC=primary health care. The first category of tools is related to budget formulation. Programme budgeting, budget rules and statutory appropriations, and conditional grants influence how budgets are made and aligned with policy priorities. A second category of tools relates to budget execution, or the effectiveness with which funds flow through the system. These include the use of population-based resource allocation formulae and direct facility funding. Some purchasing tools, such as benefit specification, provider payment, and contracting arrangements, also fall in this category because they influence the flow of funds. The third category concerns how PHC services are organised and how they relate to and interact with the rest of the health system. This category includes the adoption of an operational definition of PHC, the use of norms and standards to establish resource requirements, service delivery models, and effective referral and gatekeeping functions. The first two categories, budget formulation and budget execution, are part of public finance management processes; for their part, service delivery arrangements will influence both the estimation of budget needs and the routes through which funds move through the health financing system. These policy levers clearly are not mutually exclusive; indeed, they can be undertaken in parallel. We outline a short overview of each of these levers in the following sections. The capacities (of the health system and broader public finance management processes) required to deploy these tools are presented in the section's conclusion. Budget formulation The primary instrument for allocating public funds to health programmes is the government budget. The budget formulation processes used by many governments can sit at odds with the unpredictability and complexity of resource needs for health,115 with budget formulation and stated policy objectives often disconnected.116 In practice, the process of budget formulation is commonly reduced to incremental adjustments to the previous year's budget. Health budgets also suffer from a lack of credibility, as they are often believed to be based on inaccurate data and incomplete budgeting analysis.117 Even when PHC is nominally prioritised in national policies and plans, budgeting structures and processes can inhibit the allocation and flow of funds to PHC providers. In most countries, PHC is not a visible line item in the national budget, making it hard both to pinpoint how much is being allocated and to monitor whether PHC funds are actually received. Instead, PHC budgets are often embedded in, and absorbed by, the budgets of hospitals or local governments. Budgets frequently follow a line-item structure, with resource flows tied to inputs rather than to activities, levels of care, or population health needs. In these systems, resources get to frontline PHC providers mostly through in-kind, easily quantified provision of medicines and supplies, and staff salaries. Establishing appropriate programme budgeting can help to make financial allocations to PHC more visible in the budget. Oversight of resource allocation to PHC is especially complicated when multiple financing agents and purchasers are involved. This occurs, for example, if a social health insurance fund is administered separately from the government health budget. The social health insurance fund is then outside the influence of the ministry of health and different provider payment mechanisms might apply. Likewise, in decentralised systems, central authorities have little oversight or influence over budget allocations by decentralised authorities. In China, for instance, PHC is the responsibility of two different agencies: the local health bureau for essential public health services and the social insurance office for medical services. Each has a separate budget and uses different provider payment methods (see section 5). In the Philippines, fragmented funding makes it difficult to prioritise PHC (panel 8 ). As noted in section 3, donors can contribute to fragmentation of budgets when their substantial contributions remain off-budget and separate from national plans. Ethiopia provides an interesting example of a national initiative to harmonise all sources of funding on one budget, which is called One Health, One Plan, One Budget.118 Panel 8 Multiple funding flows and fragmented budgeting processes for primary health care (PHC) in the Philippines' decentralised system The Commission's case study on the Philippines examined the arrangements for funding PHC in a decentralised political system. Health is a devolved function in the Philippines: local governments receive a share of general tax revenue and are also able to raise revenue through local taxes. Decisions about the allocation of these funds between health and other sectors are taken at the municipal level. In addition to a share of these general funds, health receives some additional supply-side funding from the central government for infrastructure and for employment of essential staff. Local government also receives reimbursement for PHC from PhilHealth, the national social health insurance programme. Resources are allocated among sectors through a process that is led by the elected municipal mayors. Budget rules provide for mandatory shares of the budget to some sectors, such as gender and development, but no such rule is imposed for health. The local health plan is the instrument used to consolidate PHC funds from different sources. But priorities can diverge between political and technical participants. The fragmentation of PHC funding is exacerbated by the different timelines for the national and municipal planning cycles, and different procurement rules applied to different funding sources. In practice, reimbursement from PhilHealth is unpredictable due to administrative complexity; this revenue is also affected by low claims by beneficiaries who are unaware of their entitlements. The outcome of this process is that municipalities spend only 7% of their budget on health, against a benchmark of 15% set by the central health ministry. Two new initiatives under the Universal Health Coverage Act are intended to help consolidate and align resources for PHC: the development of a new primary care provider network, and a special health fund that is pooled at the province level. Local actors remain sceptical about whether these technical solutions can address the political issues of resource allocation. Several policy levers are available to address these problems and strengthen budget formulation. (1) Programme budgets: this approach organises the budget according to programmes (eg, a service or group of services) rather than by inputs.119 Programme or performance-based budgeting serves to clarify programme objectives, and improves monitoring and accountability, as each programme can also have associated performance measures.120, 121 South Africa uses a programme-based budgeting approach alongside a higher-level item classification, and has a specific budget for PHC comprised of seven subprogrammes within the District Health Services programme, with substantial flexibility in shifting funds. This standardised programme structure applies across all provinces and districts. Each PHC facility can spend up to their budget limit in line with an agreed spending schedule. The five basic PHC subprogrammes (clinics, community health centres, community based services, district management, and other community services) comprise 19·2% of total public health expenditure (31·1% when including the HIV or AIDS and PHC facilities subprogrammes). Budgeting for the PHC programme can be a means of creating greater visibility and protecting resources for PHC. When applied generally across a government, programme or performance-based budgeting seems to be effective in improving resource allocation to health and supporting more productive negotiations between the Ministry of Finance and line ministries. Input-based budgets maintained alongside programme budgets facilitate financial control. However, programme-budgeting requires both considerable budget management capacity within the spending institutions and good costing information. It might also lead to new rigidities in budget execution if controls are carried over from input-based budgeting.120, 122 Although there is scope for expanding programme-based budgeting, there is little experience in how to address specific demands of the health sector.123 Programme-based budgeting for PHC also assumes that the package of services, as well as the level at which given services are to be delivered, are defined. As outlined in section 2, this is a key challenge for policy makers. Although experience with programme budgets is growing, there is still much that needs to be tested.120 (2) Budget rules and statutory appropriations: this is another approach to ensure sufficient budgets for PHC. Budget guidelines can mandate minimum budget shares for specific sectors (as in the case of the gender and development sector in the Philippines; panel 7). Statutory appropriations are a legally-mandated standing budget provision (so-called appropriation), which is not dependent on the passing of a legislative appropriation bill. For example, in Nigeria, the National Health Act earmarks 1% of the federal government's consolidated revenue to fund the Basic Health Care Provision Fund.124 Such rules and statutory appropriations have the advantage of protecting part of the health budget from political processes. (3) Conditional grants: in both centralised and decentralised systems, the central government can influence resource allocation towards policy priorities through conditional grants that impose restrictions on the use of funds. Conditional grants can, additionally, be used to create incentives for specific spending (eg, matching rules for National Health Mission grants in India; panel 9 ), impose governance requirements (eg, audits or reporting requirements), or be supplemented with performance targets (eg, Plan Nacer and Sumar in Argentina).126 These can be used to influence devolved or other local units to invest in certain programmes, leveraging centralised funds to expand overall PHC resources.Panel 9 Conditional grants influence resource allocation in India In India's system of government, health is the responsibility of the subnational state governments. The Commission's case study on India examined mechanisms used by the central government to encourage states to invest in primary health care (PHC).125 The National Health Mission, which was previously known as the National Rural Health Mission, is the main central programme for strengthening PHC. Under the National Health Mission, the federal government created a number of mechanisms to encourage states to fund PHC, including: • A matching rule for distribution of central level grants, which initially required states to contribute 15% of the total funds; the required contribution increased to 40% in 2016–17. • A planning process which included allowing the central government to conduct detailed reviews of PHC plans and budgets. • A system of performance assessment and accountability that made the release of a portion of the approved resources contingent on achieving a given level of performance. The performance-based component increased from 10% to 20% of the total National Health Mission funding and conditions were revised in 2018 though not yet implemented due to COVID-19 disruption. Over the period 2008 to 2019, state spending on PHC spending did increase. However, while the central government influenced the pattern of PHC spending, state-level respondents did not feel that these policy levers necessarily increased effective use of funds. Budget execution Budget execution processes in LMICs are often highly bureaucratic and focus on financial accountability rather than achieving outcomes. This focus on financial control might be due to concerns about corruption. The many stages of fund disbursement, as well as frequent delays in approval and release of funds, tend to reduce the amount and timeliness of funding that actually reaches PHC facilities and providers. The more remote the health facility, the fewer resources eventually arrive. This is not just an allocation problem; significant leakage and delay can occur during the many steps involved when disbursing funds from the central ministry of finance to a regional authority, then to local levels of government, and at last to facilities.127, 128, 129 A particular challenge for PHC is that discretionary item allocations suffer the most from leakage and other execution problems; this leads to few resources being made available for the general operating costs of PHC facilities.116 Low levels of provider autonomy (section 2) creates further rigidities in how resources are deployed. Various policy tools are available that can help to avoid or ameliorate these difficulties. (1) Resource allocation formulae: These are used to allocate resources (among geographical units and levels of care) are policy levers that can promote equity in allocation. In its simplest form, a resource allocation formula allocates an equal per-capita amount across the recipient units. These formulae can be refined by adding adjustments, such as for differing health needs or local cost differences.130 For instance, health budgets in the English National Health Service are allocated among geographical units using a needs-adjusted formula. This approach inspired many countries to develop similar arrangements. Resource allocation formulae can also be used as part of budget rules and conditional grants. Resource allocation formulae have been effective in directing and protecting resources for PHC in both Chile and Brazil. In Chile, a per-capita amount for PHC is allocated to municipalities to operate PHC facilities (panel 10 ).55 In Brazil, the budget for the Family Health System is allocated to municipalities using a formula (panel 11 ).107 Panel 10 Capitation-based resource allocation in Chile The Commission's Chile case study examined the development, design, implementation, and effect of the capitation-based system for allocating resources to primary health care (PHC) in Chile.55 After its return to democracy in the 1990s, Chile replaced a system based on fee-for-service for PHC with one based on a per-capita-based allocation using a resource-allocation formula. The payments transferred to local governments are calculated according to the size of their population, adjusted to take account of the age, poverty, and rurality of a population. The local authority then allocates resources to PHC providers to cover the costs of salaries and services in accordance with the government's Family Health Plan. By 2019 this per capita allocation represented nearly 65% of the overall financing transfers for PHC. The payment system both serves to allocate resources to geographical units and is a means to pay providers. This capitation-based allocation has improved equity through the provision of more resources to poorer municipalities. One notable limitation of the system, however, is the low degree of financial integration between PHC and higher levels of care. Because diagnosis-related groups are used to pay for hospital services, clinical coordination and integration of care are more difficult. The risk adjustment mechanism is also considered to be quite crude, limiting possibilities for redistribution. Panel 11 Brazil uses a population-based mechanism to allocate resources to the frontline The Commission's Brazil case study set out to understand the health financing arrangements that shaped the Family Health System.107 The Family Health System scaled up the provision of PHC through multidisciplinary teams who provided community-based services in a geographical area, shifting the way health-care services are delivered in Brazil. Financing of the Family Health System was through a direct transfer from the federal level to municipalities, known as the Piso da Atencao Basica (meaning Floor for Basic Care). The transfer was calculated as a fixed per-capita amount based on municipal population and a variable component linked to federal priorities, including scale-up of the Family Health System model. From 2011, adjustments to the formula were introduced to allow more funds to be allocated to more deprived municipalities. Through this mechanism, regular and predictable resources were provided monthly to all municipalities for delivering primary health care (PHC). Adjustments to the Piso da Atencao Basica arrangements were made over time to encourage municipalities to adopt the Family Health System model, expand the scope of PHC services provided, and to address health inequalities. This approach to financing PHC had clear impacts on reducing inequality in funding; although this effect was mitigated to a degree by the requirement to have a substantial municipal contribution to PHC funding. Supported by these financing arrangements, the number of Family Health System teams grew from 2054 to 43 286 between 1998 and 2020, covering 133·7 million people (63·3% of the population). A number of studies point to the effectiveness of the family health services in improving access to health care, reducing health inequalities and improving health outcomes.131, 132 However, with substantial disparities across municipalities in financial, administrative and technical capacities, inequalities across the country have persisted, and the availability of qualified health professionals in poorer and rural areas has constrained expansion of family health services. Recent developments, including fiscal austerity from 2016 onwards, political pressures to concentrate resources in specialised and hospital care, and the merger of financing blocks for PHC with secondary and tertiary hospitals threaten the achievements in financing an innovative PHC model over the past 20 years. A systematic review of the use of resource allocation formulae found that they enhanced equitable allocation of resources across provinces or smaller administrative units in Chile, Colombia, Zambia, and Zimbabwe.133 The appropriate mix of local and central financing matters for equity, as greater dependence on local resources undermines equity. Various other forces can also constrain the equity achieved through the use of resource allocation formulae, including: failing to account for local differences in costs; failing to account for the absorptive capacity of each geographic area; and overlooking the up-front investments required to expand service provision. Scarcity of consistent and robust data to inform the components of resource allocation formulae (eg, health status levels) at a given point in time can hinder the application and implementation of resource allocation formulae in LMIC.134, 135, 136 Furthermore, simply applying resource allocation formulae does not ensure equitable distribution of resources. Okorafor and Thomas137 found that resource allocation formulae for PHC in South Africa were inequitable due to weak managerial capacity at lower levels of government, poor accounting for PHC expenditure, and lack of protection for PHC funds with regard to other service areas. (2) Direct facility funding: where public finance management systems fail to effectively channel money to PHC platforms and providers, whether due to leakages or other reasons, a second policy tool is to give money directly to the PHC level. Under direct facility funding, a health-care facility receives some core funding from the central level directly into its own bank account, usually to enable the purchasing of medicines and other supplies, or to pay for operating costs such as utilities. When coupled with autonomy to spend according to local priorities and sound facility financial management, direct facility funding can improve efficiency and quality of care.138 Direct facility funding can also serve as a means to integrate multiple sources of financing at the facility level. For example, in Tanzania, the same formula is used to allocate resources from on-budget donor funds and the government budget (excluding salaries and medical supplies). This approach was used in Kenya in the late 2000s,139 Nigeria,140 Tanzania,141 Burkina Faso,138 and Uganda.142 Direct facility funding can ensure that funds reach the periphery; it can have the added benefit of making PHC providers more visible in the public finance management system, raising the profile of spending on PHC. More evidence is required, however, about the effectiveness of direct facility funding in actually channelling funds to facilities, and by association facilitating the removal of user fees.138, 143 (3) Strategic purchasing: This incorporates specification of the benefit package, selection of eligible providers, and choice of provider payment methods. Done well, strategic purchasing promotes effectiveness, efficiency, and equity in a health system. However, a poorly (or too narrowly) specified benefit package can cause patients, and therefore resources, to drift up the health-care system towards the hospital level. A benefit package that is too narrow, for example, might exclude routine management of chronic conditions. This can drive patients to seek care at levels of the system that are higher than necessary and expose them to risks of substantial out-of-pocket spending on medicines. However, various policy levers are available to address purchasing problems. First is benefit specification. Having an explicitly defined and appropriate benefit package (that was developed using realistic costing) is a way to secure and protect allocations to PHC. In Thailand, for example, capitation payments for PHC are based on a defined benefit package. Its Health Interventions and Technology Assessment Programme is a world leader in health technology assessment. Benefit specification can also require co-payments when patients bypass PHC, which helps to direct patients to the appropriate level of care. Second is provider payment mechanisms, which determine how money is paid from pooled resources to service providers. As such, the provider payment system has a substantial influence on how resources are ultimately allocated across providers. As is elaborated in section 5, adopting a payment mechanism that directs money to PHC, such as capitation, makes PHC expenditure more visible, more equitable, and helps to protect allocations. Poor design of provider payment systems can also incentivise bypassing PHC, such as when providers are rewarded for referring patients to higher levels of the health system. Resources can also be diverted away from PHC when mechanisms for paying providers have money following patients (for example, with fee-for-service payments) or are reinforced by an absence of gatekeeping or open-ended budgets at higher levels of the health system. The effectiveness of a provider payment system as an allocation tool depends on the existence of complementary policies and properly aligned incentives. These include the coherence of provider payment systems across levels of care and payers, without which systems can decrease access, generate waiting lists, and overall decrease efficiency in the use of funding to provide care. It also can include health system organisation features such as gatekeeping and user incentives. Payment systems that constrain the budget at higher levels of the system (sometimes called closed-ended payments) can help to protect resources for PHC; this approach is used in Thailand. PHC subpools can also do this function; for example, in Taiwan an umbrella budget for the national health insurance has five subpools, including one dedicated to PHC provided at independent ambulatory care clinics.144 Third is contracting and monitoring. Contracts between purchasers and providers can be designed to include provisions that help channel funds to PHC and constrain resources from being paid to hospitals. For instance, in Estonia, the national health insurance fund guarantees a minimum amount of revenue (equivalent to the per-capita amount for 1200 individuals) for a defined list of PHC providers working in non-urban areas. This channels funds to PHC providers in sparsely populated areas and ensures that they can cover their fixed costs. More detail on Estonia's approach to PHC is provided in sections 5 and 6. Some contracts include volume caps on hospital payments, whereas others specify a facility level for payment for specific services. For example, in both China and Indonesia, purchasers will not pay for a service delivered at a level higher than appropriate. Contracts can also cover referral rules to limit unnecessary referrals to higher levels. At times of crisis such as the COVID-19 pandemic, keeping money flowing through the system to frontline providers requires flexibility and adaptation of budget execution processes. Such flexibility also carries some risk. Keeping the focus on PHC is essential (appendix p 4). Service organisation In many health systems, inappropriate incentives for both patients and providers have been created that encourage the bypassing of PHC, resulting in pulling resources away from frontline providers. The incentives to bypass PHC are reinforced by the vicious cycle cited in section 1: weak political support for PHC leads to chronic underfunding, which affects the capacity of PHC providers to offer quality services; this, in turn, causes users to mistrust PHC and they bypass this level of care, turning instead to hospitals and specialists. These incentives might not have been created with the intention of undermining PHC. However, without strategic and coherent organisational linkages among levels of care, resources can drift up the system. For example, when user fees are set such that the cost of PHC is similar to the cost of higher levels of care, patients may feel encouraged to bypass PHC. In places that lack easily accessible and trusted PHC to serve as a first point of contact, patients are also likely to seek care directly (if often unnecessarily) from specialists. Some innovative approaches intended to reinforce PHC by providing greater technical support to lower levels of care, such as the Health Care Alliances introduced in China, actually had the unintended effect of driving resources back to higher levels of the system.145 Various policy levers are available that, by improving the organisation of services, help to drive users and resources back to PHC: (1) Organisational definition of PHC: a clear operational definition of PHC helps to steer resources towards it by defining what functions must be supported and by providing a category for tracking PHC expenditure. In Indonesia, for example, the national health insurance programme defines a clear set of PHC facilities (public and private) from which the PHC benefit package can be delivered. (2) Norms, standards and guidelines: establishing clear norms standards and guidelines for PHC (including population coverage levels) can be a useful way to support allocation of resources to PHC. These declarations of how PHC is supposed to be delivered lend themselves to measuring resource requirements and tracking progress. The combination of a clear operational definition of PHC and service delivery norms makes it easier to cost PHC and to determine what level of funding is needed to deliver it. In Ethiopia, the expansion of PHC followed a stated objective: each PHC Unit (made up of one health centre and five health posts) would serve 25 000 people, including community-level services provided by two health extension workers per kebele (5000 people) in each health post. This clear standard enabled the government to estimate its resource needs and then translate the estimates into a costed plan that involved the construction of 15 000 health posts and 3200 health centres. The plan proved to be a powerful instrument when negotiating with both donors and government finance officials. The investments required for human resources to staff the new units were also determined based on the plan. Challenges remain, of course; these facilities are still developing the capacity to deliver the full set of PHC services, and there remain shortages of certain cadres of staff, particularly doctors and midwives.146 (3) Service delivery model: changing the service delivery model such that it effectively pulls resources to the PHC level. With appropriate financing mechanisms, service delivery arrangements that strengthen PHC can stimulate demand and help to pull resources to PHC. Different approaches have been implemented. For example, on the one hand, Brazil's Family Health System introduced multidisciplinary teams that operate at municipal level and are financed through the per-capita allocations paid directly to municipalities.107 Ethiopia, on the other hand, has developed a new model of PHC service delivery with the introduction of a new cadre. Health extension workers receive a year of preservice training and are paid a government salary to work from health posts at the village (kebele) level.118 (4) Referral systems and gatekeeping: protecting resource allocations to PHC includes directing patients and resources to the appropriate levels of care. Gatekeeping policies, in which patients must be referred from PHC providers to access specialist care, and measures such as empanelment and registration which link patients to providers, can help to influence care seeking and direct patients to use the lowest level of care at which their condition can be effectively managed. However, effective gatekeeping requires having functional referral systems in place so that patients can be rapidly sent to the appropriate level of care. Many health systems, especially in LMICs, do not have those systems in place. Where strict gatekeeping is not feasible, financial disincentives or incentives—such as higher co-payments for higher levels of care—can deter patients from bypassing PHC. In France, for example, a preferred doctor scheme was introduced in 2005, aiming to reduce the number of visits to outpatient specialists. Those who continued to self-refer to a specialist incurred a higher co-payment than those who were referred via their preferred doctor, which led to fewer specialist consultations.147 These and other policy levers will be more or less appropriate and effective depending on the specific context of each health system. Further, many problems exist within these approaches that have yet to be successfully addressed. Therefore, identifying new ways of improving the allocation of resources to support high-quality PHC service delivery remains a key challenge for policy makers. Financing essential public health functions Essential public health functions and global common goods148 are part of PHC. They include activities whose importance has become particularly evident during the COVID-19 pandemic, such as public communication, disease surveillance, testing, contact tracing, support for affected individuals or communities, basic laboratory services, safe water supplies, sanitation, and hygiene. Mitigation of the impacts of environmental pollution and climate change on population health might also fall under the rubric of essential public health functions, with implications for health financing that is people centred. Financing population-focused interventions require different arrangements, sectoral engagement, systems, and capacity from individual primary care; however, many of the same principles and policy levers apply. Any national strategy for delivering essential public health functions should: create coherent priorities, clarify who is responsible for what, and align budgeting processes to ensure that these are adequately funded.149, 150 Financing for public health services frequently receives inadequate attention in government budgets; however, this financing is crucial for prevention and addressing the determinants of disease. Essential public health functions can also suffer from fragmentation and ineffective organisational arrangements, with multiple payers involved and different government and non-government organisations responsible for delivery. Financing and formalising the regulation of private sector activity with population health impacts might be particularly important in considering public finance allocations to PHC at national and subnational levels. For example, subnational jurisdictions that benefit financially by taxing or partnering with polluting or extractive industries whose waste harms population health could reasonably be expected to fund essential public health functions or the additional services needed to mitigate these impacts. In this Commission, we argue that a systems approach to financing is especially important for essential public health functions. This approach starts with clarifying who has responsibility for budgeting, planning, and ensuring that the essential public health functions are adequately resourced and delivered. As in other aspects of PHC, alignment of donor funding with national budgets and plans is also crucial. Decentralisation might result in insufficient resources for essential public health functions. Uniform national (and even regional) standards are essential; this requires system-wide coordination and standards across subnational units. However, the relative invisibility of these services and their benefits, and the required coordination of multiple sectors, might make them unappealing to local governments. The policy levers detailed above can also be applied to securing financing for essential public health functions. For example, programme budgets can help to connect resource allocations to priorities and targets, whereas fiscal rules such as matching grants or resource allocation formulae can protect resources for essential public health functions. In some cases, however, it will be necessary to create new institutions to finance and deliver essential public health functions. For example, the Thai Health Promotion Foundation, which receives a share of the proceeds from national excise taxes, has responsibility for providing population-based health promotion activities. Private sector activities with an impact on population health should contribute to essential public health functions as appropriate. Institutional responsibilities and supporting functions One challenge in financing PHC is the institutional design of many ministries of health, which are structured around health programmes (eg, department of maternal and child health, department of communicable diseases, and department of non-communicable diseases) rather than functions, levels, or service delivery platforms. Although it might not be operationally necessary or feasible to have a specific department for PHC, it is important that the responsibility for setting priorities for spending on PHC, and monitoring these, sits within an identified department or unit. A key supporting function for resource allocation to PHC is the public financial management system through which budgets are developed and executed. Such systems are usually not unique to the health sector, and their strengthening may benefit from cross-sectoral initiatives to improve broader social sector financing and budget processes. Specific adaptation for the health sector might reduce paralysing bureaucracy. Strengthening public financial management, as well as PHC provider platforms' capacity to deal with finances, requires technical skill and the ability to collect, analyse, and interpret data on the population and its health needs to cost PHC benefits and required services. Another supporting function is the efficient deployment and management of the health workforce, including cadres at community level. Conclusion When based on a clear operational definition of PHC, these budget formulation, budget execution, and service organisation policy tools can be deployed in concert to help ensure that sufficient resources are allocated to meet the needs of PHC that is people centred, and to protect the resources so that they reach the PHC delivery platforms. In Thailand, for instance, the budget for PHC is safeguarded through the combined effects of a defined PHC benefit package, ring-fencing of the PHC budget (and constraints on the budget for higher levels of the system), and capitation payment for PHC. How can countries begin to move in this direction? We suggest that it begins at the budget formulation stage. Working towards developing a programme budget would be helpful to PHC, using other policy levers as necessary. At budget execution stage, a well-developed resource allocation formula can be a useful starting point for improving allocation of financing to PHC. Even a simple per-capita formula, with risk equalisation and performance and quality incentives added as the system develops, can begin to foster equity in universal coverage of a basic package of primary care services. For a formula to be effective, however, other reforms that link budget allocations to PHC are also needed. For any of these policy levers to be feasible, various health system and financial capacities need to be strengthened as well, such as budget management capacity at the Ministry of Health, high-quality data on health status and needs, and effective accounting practices. Figure 10 sets out these capacities.Figure 10 Health and public finance management system capacities needed to exercise public finance management policy levers PFM=public finance management. Direct facility funding also requires individual facility bank accounts. Decentralised systems are closer to populations than centralised ones and have the potential to provide more flexible and responsive PHC services. However, they are at higher risk of local management problems and inequities. It can also be more difficult to influence resource allocation in decentralised systems. Regardless of the type of system in any country, strong monitoring, performance management, and enforcement of appropriate budget and public finance management systems to ensure that resources reach PHC are as important as the initial process of resource allocation. The Commission recognises that it is often difficult to make major or rapid shifts in allocation of resources in existing systems. In most cases, the discussion must be about incrementally influencing spending. However, as the COVID-19 pandemic has shown, sometimes major disruptions can create space for major reforms (and across multiple sectors) if appropriate policies have been identified in advance. Ultimately, these are complex processes that pose significant implementation challenges. Further, they rely on having data on variables such as costs and activities, together with numerous other supporting functions and systems. Ultimately, all of these systems are driven by human behaviour. Thus the following section addresses a key feature of every complex system for financing PHC: how it organizes the incentives for people to deliver and access high quality people-centred PHC. Section 5: Getting incentives right for PHC Provider payment and incentives are another tool to ensure resources reach frontline providers and are used most efficiently. Countries can do more to create incentives that direct the behaviour of organisations that provide PHC and users towards people-centred PHC. In this section, we propose our vision of the PHC payment system based on a concrete set of principles. We describe the pathway countries can follow to make progress towards this vision and lay out the basic functions that need to be strengthened along the way. We also consider what motivates individual health workers, including the need to foster a culture of professionalism. Finally, we examine the role of provider payment policy in reducing financial barriers for those in need of PHC. Key messages from Section 5 are presented in panel 12 .Panel 12 Getting incentives right for primary health care (PHC)–key messages • Incentive policies for providers and users are inextricably intertwined: PHC provider payment policies are integral to the elimination of user fees and informal payments for PHC services. • Incentives alone cannot solve all PHC financing problems, but they should at least not work against PHC service delivery objectives. • The Commission's vision of how PHC provider organisations should be paid is a context-specific blended payment model with capitation at its centre because that is most aligned with the principles and objectives of PHC. • The blended payment model purposively combines capitation with elements of other payment methods (such as fee-for-service or performance-based bonuses for selected high priority services, and budgets to cover unavoidable fixed costs) to maximise beneficial incentives and offsets perverse incentives of each payment method, while ensuring other service delivery objectives, such as access, are met. • Countries should only embark on provider payment reform when they are ready. The transformation of the PHC provider payment system is a complex process with distinct political economy challenges. The aim is to make incremental progress that involves continually strengthening supporting systems as the payment model evolves. The need to get incentives right An incentive is an economic signal that directs individual health workers, health provider organisations, and patients towards self-interested behaviour. We know that incentives influence the performance of PHC providers and the behaviour of users.151 However, getting incentives right is not a panacea. As noted in section 4, the key problem in many LMICs is that insufficient resources actually reach PHC providers. No amount of tweaking incentives will help when newly qualified professionals often look to establish lucrative specialist practices and facilities are poorly equipped, making it difficult to staff primary care clinics. Dual practice, widely observed in LMICs, is a symptom of precisely this problem.152 To deliver high-quality PHC, it is essential that doctors, nurses, and other cadres of staff are valued, with adequate remuneration and conditions of work to attract them into PHC as a long-term career choice.96, 153 In many countries, the way health-care providers are paid often works against the objectives of PHC. In systems that pay providers a fee for a service, they typically set higher payment rates for specialty services, giving providers a financial incentive to prioritise curative, rather than preventive, care. In many LMICs, people lack trust in PHC and choose to seek care at higher levels, even if they have to pay more. Addressing these problems is especially difficult when funding and provider payment systems for PHC are fragmented. Case studies from LMICs have documented that the typical PHC provider receives funding from multiple payers using different payment systems for different population groups.154 In addition to creating administrative hurdles for PHC providers, when payments are poorly coordinated, the incentives generated might not align well (or might even conflict) with the objectives of PHC. These incentives can instead drive providers to prioritise certain patient groups in ways that exacerbate inequities or health services that are of low value to patients but lucrative for providers. Some of these problems are evident in China, as described in panel 13 .Panel 13 China's experience with fragmented payment structures for primary health care (PHC) The Commission's China case study focused on the fragmentation of PHC financing.145 In China, PHC is mainly provided at village clinics and township health centres in rural areas and at community health centres and stations in urban areas. Financing for these institutions comes mainly from two sources: social health insurance and the essential public health fund. Social health insurance pays for a medical care package largely using fee-for-service payment, whereas the essential public health fund pays for a package of public health services using a population-based method (capitation). These two sources of funding are managed by different government authorities at both the national and local levels: social health insurance is managed by the Department of Social Medical Security, whereas the essential public health fund is managed by the Department of Health. This fragmented payment system has been a barrier to integrated PHC for several reasons. First, a lack of coordination between the two funds' administrative authorities has resulted in separate delivery of medical and public health services. Second, fragmented funding can make it difficult for different cadres of PHC providers to coordinate their services, even when they are working within the same institution. For payment purposes, PHC providers try to maintain clear boundaries between the medical and public health services they offer; this is difficult, particularly in the case of services for non-communicable diseases. For example, payment incentives might lead doctors to focus only on providing medical services, even if they should also be playing important roles in disease prevention and public health services. Third, separate information systems have been established for the medical and public health services, making it difficult to manage the health of individuals and communities holistically. Finally, the performance of various PHC providers is evaluated separately by social health insurance and the essential public health fund, impeding the health system's ability to determine whether it is achieving its objectives. The existing incentives are not aligned to encourage medical and public health providers to coordinate, or even share information, with each other, although they are serving the same community members. A number of counties in China have recently begun to experiment with changing this fragmented financing situation by pooling the two sources of funding for PHC to pay family doctor teams. The intention of this innovation is to encourage greater continuity and integration in PHC. Chronic underfunding of PHC, fragmented revenue streams, and misaligned provider incentives all contribute to the fundamental problem mentioned in sections 2 and 3: that many users in LMICs pay out-of-pocket fees for PHC services, which act as a barrier (disincentive) to accessing PHC, particularly for the poorest; for those who do choose to pay, user fees can lead to financial hardship.74, 75 Each option for paying PHC providers generates certain incentives that have been described in the literature.155, 156, 157 The key insight from the empirical evidence on the effectiveness of different payment methods is that no single payment method is perfect. Each payment method carries advantages and disadvantages.158, 159, 160, 161 Many countries have therefore moved towards a blended payment system, which combines elements of multiple payment methods, in part to maximise the beneficial incentives and minimise the perverse incentives of each option. We describe the main categories of payment method used for PHC. • Tying payments to inputs, as with a line-item or global budget, is a passive form of purchasing. It provides a facility and its staff with a stable income, which is especially important in hard-to-serve areas, and contains costs. However, this provider payment method generates no strong incentives for providers to address the health needs of the population in the catchment area. Input-based budgets are often rigid, so providers cannot easily move funds across budget lines to respond to local needs (eg, they cannot choose to cut utility costs to spend more on medicines). • Paying for services, typified by a fee-for-service system (not to be confused with user fees) in which the provider receives a payment from the institutional payer for each service provided, prioritises meeting users' demands. However, it carries many disadvantages, including an incentive to provide more care than is needed (particularly services with higher fees) and rarely prioritises preventive care. Pay-for-performance is a common add-on to other payment methods, whose purpose is to incentivise high-quality care through bonuses for reaching service coverage or quality targets, but can in principle result in gaming and the neglect of aspects of care that are not being measured.162, 163 • Population-based payment, which in this Commission we refer to as capitation, gives providers a fixed per-person payment, determined and paid in advance, to deliver a defined set of services to each enrolled individual for a specified period of time. Under capitation, continuity of care, a key prerequisite for successful PHC, is built into the reimbursement mechanism. Providers have an incentive to attract more patients to their practices and contain costs. However, a provider's revenue might not be enough to cover the costs of serving the population if some groups have higher needs than anticipated by the payment formula or if payment rates are set too low. In this case, there might be an incentive to avoid enrolling higher-need individuals, refer patients unnecessarily to specialists for care that could be provided in primary care, and skimp on quality of care. The payment system itself is not the only force creating incentives for providers. Broader purchasing arrangements, including contracting, monitoring of provider performance, and population enrolment, can also generate both financial and non-financial incentives. For example, conditions of contracting, such as accreditation status or data reporting requirements, might create incentives for providers to improve their quality standards and upgrade information systems. Contracting arrangements might also specify service delivery standards, often tied to national clinical guidelines; this creates additional financial incentives if those standards must be met for providers to be paid. The Commission's vision of a people-centred payment model for PHC Blended payment with capitation at its centre The Commission's vision for PHC provider payment is a context-specific blended payment model built on capitation. This structure embodies principles that the Commission argues should form the core of PHC. Payment systems should allow adequate resources to flow to the PHC level in ways that: are equitable; match resources to population health needs; create the right incentive environment to promote the full PHC spectrum of prevention, health promotion, and management and treatment; foster people-centeredness, continuity and quality of PHC; and are flexible enough to support changes in service delivery models and approaches. Capitation is a prospective population payment system.164 Because it is not tied to specific inputs or the volume of services delivered, capitation payment gives providers flexibility to coordinate and optimally manage care for individuals and populations. It is the only payment method that is based on the principle of equity, as its starting point is an equal fixed payment per person, which can then be adjusted based on health needs or other factors. Capitation payment is the only method that pays PHC providers for managing population health and prioritises preservation of good health, rather than delivering individual services to address health problems. As a prepayment-based system, capitation also provides a predictable and stable revenue stream to PHC providers that can be used to flexibly deliver services in responsive ways.156, 165 As PHC service delivery models become more community-based, patient-driven, and technology-enabled, payment methods need to be flexible enough to adapt to more varied, complex, and dynamic service delivery. Payment should compensate providers for delivering all services specified in a PHC package, some of which might not appear in typical fee-for-service lists or are not delivered in facility-based settings (such as essential public health functions, telemedicine, care management, or patient engagement). Capitation payment is flexible and can be redirected quickly in support of the service delivery model, as under this model providers have a large degree of financial and managerial autonomy. If capitation has so many benefits, why does the Commission recommend a blended payment model? As noted above, capitation payment also has some clear drawbacks, such as encouraging underprovision, selection bias towards low-need patients, and unnecessary referrals to other levels of care.157 Blended payment models bring the benefits of capitation as the starting point and then use elements of other payment mechanisms to deliberately offset capitation's disadvantages and support achieving other specific health system objectives.157, 166 Blended payment models for PHC typically include a budget payment to cover unavoidable fixed costs, particularly in low-population or hard-to-serve areas; some fee-for-service carve-outs for health conditions or services that are high priority or at higher risk of being underprovided in capitation; and, in some cases, performance-based payment to incentivise reaching coverage targets for priority services and improving quality of care. Other complexities may be added to align with evolving and innovative service delivery models (panel 14 ).Panel 14 Paying for integrated care Differences in health financing arrangements, including payment methods, are frequently cited as a major barrier to more integrated approaches of service delivery. Successful integration requires sustained investment in staff and support systems, funding for start-up costs, and flexibility to respond to needs that emerge during implementation.167 A review of the evidence (mostly from high-income countries) found that a range of mechanisms have been used, often in combination, to achieve better service integration.54 This includes the commitment of dedicated resources to support the development of innovative care models, such as through targeted payments to finance infrastructure for provider networks, or the use of start-up grants to promote care coordination and integration activities.168 Countries are also increasingly experimenting with what has been referred to as value-based payments, which seek to link provider payment to a predefined set of evidence-based clinical process or outcome measures.166 Examples of value-based payment include bundled payments, shared savings, and global budgets. Bundled and global payments are disbursed as a single payment in form of a lump sum per period for a specified population (global payment) or per episode or condition per patient (bundled payment) to a collective of providers. By linking payment to clinical, process, and outcome measures, providers are incentivised to increase efforts to improve patient care and process efficiency. As the payment is transferred as a single lump sum, regardless of the number of services provided, value-based payments are expected to promote care coordination and integration across providers and so reduce wasteful duplication of services and unnecessary hospital use. The Netherlands and various states in the USA have introduced disease-based bundled payment schemes for mostly chronic conditions, such as type 2 diabetes or cardiovascular disease.169, 170 These involve reimbursing providers for a package of services on a predefined patient pathway per patient and for periods of up to 1 year. Global payment models include shared savings programmes and comprehensive care payments. Shared savings programmes essentially mean that the payer and providers share the risk of rising expenditure, that is, providers that successfully lower their growth in health-care costs while continuing to meet quality standards will be able to keep the savings and reinvest them. Examples include the Medicare Shared Savings Program in the USA and the Healthy Kinzigtal integrated care programme in Germany.171 Global payment models involve fixed payments for the care of a patient during a specified time period. Several countries have additionally introduced pay-for-improvement, pay-for-coordination, or pay-for-performance schemes in primary care, incentivising chronic and coordinated care in particular, although the evidence of their benefits remains mixed. Finally, a number of countries have experimented with different financing mechanisms, such as shifting responsibility for funding of particular components of service delivery between funding agencies.172 Others have introduced pooled funds to integrate health and social care or structurally integrated budgets, in which responsibilities for health and social care are combined within a single body under single management, such as within the Integrated Health and Social Care Board in Northern Ireland. The evidence of what works remains patchy. However, an important lesson is that “integration costs before it pays”.167 Indeed, evaluations of novel schemes often find an increase in cost, mainly because the new service delivery model uncovers unmet need.173 The creation of new coordinating mechanisms will not compensate for lack of resources. The injection of one-off extra funding to pay for new services will not necessarily ensure long-term sustainability, particularly where new approaches fail to be incorporated into routine care. Pay-for-performance has received considerable attention from policy makers and researchers in the last two decades.174, 175, 176 Explicit performance incentives encourage providers to focus on aspects of PHC that are unlikely to be incentivised by the global base payment and might be prone to quality skimping or underprovision. However, the current evidence suggests that improvements that result from pay-for-performance schemes are often less than anticipated.174 Financial incentives should be relatively low powered to prevent disproportionate focus on rewarded tasks and to ensure sustainability.177, 178, 179, 180 Performance monitoring should happen alongside implementation of the blended payment model. Under capitation, the payment is divorced from activity, meaning a concerted effort needs to be made to monitor how well health-care providers are doing. Indeed, a key advantage of pay-for-performance is that it can contribute to better accountability, such as improved measurement of provider activity and performance, and a more informed dialogue between purchasers and providers.181 Payment levels and flows For a blended provider payment system to create meaningful incentives that affect providers' behaviours, adequate funding needs to flow through funding streams and not create conflicting incentives. Providers must also have the autonomy to manage funds and respond to incentives. Capitation payment rates on which a blended system is based should reflect adequate funding levels to purchase the inputs needed to deliver the package of PHC services according to quality standards laid out in national treatment guidelines. At the same time, payment levels must also align with the resources available from pooled public sources and the political priority placed on PHC. As funding to the health system overall grows, and as the skills of health workers to deliver the package of PHC services increase, capitated rates can be increased to channel a larger share of overall funding to PHC. Capitation payments should be managed at the lowest level where they can be used effectively to provide the full range of services to address population health needs. In some systems, such as in Estonia182 and England, this is a frontline PHC provider organisation. Other countries, including Ghana and Kenya, are experimenting with establishing groups or networks of PHC providers to manage capitation payments. These platforms can share some functions, such as information management and quality assurance, and close capacity gaps. Providing capitation payments to a multidisciplinary provider group has been shown to foster coordination across the continuum of care.166 At a higher level still, in Brazil and Chile, local government authorities manage capitated funds for the delivery of PHC, and individual health workers can receive bonuses when their team or facility achieves performance targets.55, 107 In Thailand, capitation payments go to a contracting unit for primary care. More than likely, funds will need to be directed to multiple levels simultaneously to enable the provision of both individual-focused care and population-level services including essential public health functions (see section 4). Funding flows from multiple sources need to be harmonised to align the incentives for providers. Although it is usually not feasible (or necessarily desirable) to merge all funding flows, a coordinated and deliberate payment system can help to achieve greater coherence at the provider level. A good example is Tanzania's system for providing direct facility financing using health-sector basket (pooled) funding from donors who have agreed to finance the health sector budget through the central treasury.141 Interface with payment at other levels of care The way outpatient specialty and inpatient services are paid for can also influence the overall incentive environment for PHC providers. For example, if PHC providers are paid a fixed capitation payment, but hospitals are paid based on the volume of services provided, the potential adverse incentive of capitation payment to increase referrals is reinforced by the adverse incentive for hospitals to increase the volume of care, including admissions.55, 165 As mentioned in section 4, additional policy measures might therefore be needed to harmonise incentives across the levels of care, including gatekeeping requirements to enforce referral guidelines, ring-fencing the payment pool for PHC, or introducing payment caps at the level of the hospital to reduce incentives for unnecessary admissions.183 Progressing towards a blended capitation-based payment model An effective capitation-based blended payment is a sophisticated provider payment model that relies on a complex set of policies, implementation arrangements, and purchaser and provider management capacities. Reaching this stage requires a clear vision backed by strong political commitment, significant time, and consistent investment. As with the introduction of any new provider payment system, the responses of providers to the rollout of a capitation-based blended payment model cannot be fully anticipated. However, their responses are likely to be different if the payment model is evolving from an input-based budget (in which providers have little autonomy and might welcome the flexibility of capitation) or from fee-for-service payment (in which providers have more control over their revenue and might resist the move to capitation). Moving towards a blended payment model, as with any reform process, requires anticipation and deft management of complex political economies, collection and analysis of data to address emerging issues, and flexibility to address unintended consequences in a timely manner. This process can seem dauntingly complex—however, the alternative is to remain with a status quo that is failing to provide the incentive environment required for delivery of PHC to improve health outcomes and equity. Provider payment reform is incremental and rarely is there a perfect time to start. Evolution of the payment system Many payment systems have evolved to reach blended payment models in similar ways, regardless of their starting points. In most countries, the introduction of equity-orientated and efficiency-oriented payment system reforms starts with a basic population-based capitation model. Typically, these systems are transparent, involve simple per-capita payments, and are easy to administer in places where data automation is limited. Most payment systems then eventually introduce risk adjustments. Complexity continues to increase over time as additional payment methods are added. Figure 11 presents the pathway of how countries can pursue the Commission's vision of a blended payment system, showing the interim steps in the evolution. Figure 11 also indicates the basic functions that should be strengthened over time to support payment reforms. Learning from other countries' experiences can help countries committed to progressive policies to hasten the development of their own context-appropriate payment models.Figure 11 Strategic pathway for moving to a blended, capitation-based payment The trajectory towards a population-based payment model involves several concrete steps. First, establish a baseline capitation payment system. For capitation to promote equity and create clear incentives, the payment amount should be based on a formula that links the payment parameters (base per-capita rate, number of enrollees linked to the provider, and any individual or provider-level adjustments) to a defined package of PHC services. Each payment parameter can range from simple to complex. They function as strategic levers to maximise the potential benefits of the payment system while minimising adverse incentives and unintended consequences. Second, define the PHC package. As mentioned in section 1, defining a package of PHC services linked to capitation payment creates an opportunity for each country to clarify what its definition of PHC includes (as well as setting boundaries between primary care, outpatient specialty services, and secondary care). Specifying what is included in the PHC package and where it is provided can drive shifts in service delivery priorities and promote the integration of vertical programmes into PHC.155 The PHC package linked to capitation (which may be a subset of the broader PHC package) can also be expanded to serve as the platform for financing essential medicines,155 either directly or through an outpatient drug reimbursement scheme, because medicines are a key driver of out-of-pocket expenses in LMICs. Third, manage enrolment. A capitation payment system relies on all individuals being enrolled (registered) with a given provider for a fixed period. The assignment of a fixed and defined population to a single PHC provider is an advantage of the system. PHC services contribute to improving the health of communities by organising around populations rather than only serving individuals who actively seek health care. Individuals can be enrolled with providers through administrative assignment (as defined by a geographical catchment area) or by their own choice (known as open enrolment). Open enrolment during select time periods allows financing to support users' choices; in principle, this creates incentives for providers to be responsive to patients and provide high-quality services.155, 184, 185 Fourth, adjust for risk levels. Risk adjustment is a correction tool that uses a measure of risk variation to compensate health providers appropriately for the expected costs of providing necessary services for their enrolled populations. The calculations account for variation in health need, typically by using data on different baseline characteristics such as levels of health, sex or gender, chronic disease risk, and socioeconomic status. Risk adjustment protects higher risk and sicker patients from the incentive providers have to avoid caring for them when their care is predicted to be especially costly. Other adjustments, such as for geographic area, might also be included if there are significant cost variations for delivering the same package of PHC services in different locations, such as in rural and remote areas where fixed costs for transportation or use might be higher. Fifth, blend payment methods. Countries almost always find that the precise blend of payment changes as the system matures. Various factors, including history, culture, priorities within the PHC system, or shifting disease patterns in a country, can drive these changes.186 In Estonia, for example, the relative contribution or blend of different payment methods (capitation, fee-for-service, fixed basic allowance, and pay-for-performance) has evolved over time, as described in panel 15 and figure 12 . In Aotearoa, New Zealand, similar evolutions are also occurring, as outlined in panel 16 .Panel 15 Development of the primary health-care (PHC) payment system in Estonia The Commission's Estonia case study focused on the process through which the capitation-centred provider payment model for family doctors was developed after independence from the Soviet Union in 1991.182 Estonia is a high-income country with an ageing population of 1·3 million people. The country's health system has been lauded for achieving good health outcomes at low cost. Public funding represents the predominant source of health financing, constituting approximately three-quarters of total health expenditure. The current system has been evolving over nearly three decades. In the late 1990s, Estonia undertook major payment reforms in parallel with organisational changes to the health system. Everyone in the population was registered with a PHC provider, either a family doctor, general internist, or paediatrician. Family doctors worked as private practitioners contracted by the national health insurance fund. The previous fee-for-service system was replaced with a capitation system, initially based on a flat per-person rate that was subsequently age adjusted. The capitation rate was intended to cover the salaries of a practice's family doctor and a nurse, as well as a defined set of equipment and certain laboratory tests. A basic allowance covered the costs of equipment, facilities, and transportation. Some fee-for-service payments were retained for a defined list of diagnostic tests and procedures. An additional lump sum was provided to cover the expenses of family doctors working in rural areas. This new payment system was designed to incentivise family doctors to take more responsibility for diagnostics and treatment and provide continuity of care; the system also compensated doctors for the financial risks of caring for older people and working in more remote areas. Moving to capitation-based funding represented a major shift from the previous fee-for-service payment mechanism, in which doctors and health-care institutions were incentivised to perform a large number of diagnostic procedures. The shift in payment model was introduced along with a new organisational and contractual mechanism that has successfully promoted PHC while increasing the freedom and independence of PHC providers. The payment system has continued to mature as new elements are added, as shown in figure 12. In the 2000s, a voluntary pay-for-performance element was added that was designed to motivate family doctors to widen their scope to include more prevention services (eg, childhood vaccinations) and chronic disease management (eg, hypertension care). This reform was widely accepted by providers: the proportion of family doctors participating in the scheme rose from 50% in 2006, to 97% in 2014. In 2015, participation in this quality-focused bonus scheme finally became obligatory for all family doctors, and individual performance results became public information. The basic allowances have also increased to cover rising costs of management and information systems and to motivate individual providers to form groups and expand the scope of services offered. Figure 12 Estonia's blended PHC payment system in 2003 and 2019 PHC=primary health care. Panel 16 Capitation in the reform of Aotearoa New Zealand's primary health-care (PHC) system The Commission's case study on Aotearoa (New Zealand) focused on capitation payment for general practitioners.187 Capitation was suggested as a way of paying PHC providers as far back as the 1930s, when the government embarked on a major reform to introduce free, integrated, comprehensive health care for the entire population. However, due to resistance from general practitioners, the capitation plan was scrapped and general practitioners remained as independent small businesses with their services funded by a fee-for-service government subsidy and out-of-pocket user fees. By the 1990s, the subsidy was funding only 20–30% of general practitioner costs, with the rest coming from user fees. In the early 2000s, the government developed a Primary Health Care Strategy, the first major PHC policy since the 1930s. The government established a new type of not-for-profit entity, the Primary Health Organisation, which enlisted PHC providers on a voluntary basis. This allowed the health system to shift to universal weighted capitation at the Primary Health Organisation level. The shift ensured that all citizens could receive subsidised care in a way that accounted for need. The move to capitation was also designed to control government expenditure on PHC and expand the range of services that could be delivered by nurses. Over time, the Primary Health Care Strategy enabled the government to increase PHC funding and allocate a greater proportion to Primary Health Organisations working with higher-need populations. Large decreases in unmet need for general practitioner services were observed in the first 5 years, including for Māori people. However, ongoing issues also persist. Although the funding was allocated using a risk-adjusted capitation formula, it insufficiently accounted for variation in needs related to ethnicity and deprivation. Despite a number of reviews, the formula has not been improved, in part because of concerns that further risk adjustment would create so-called winners and losers and undermine support amongst key groups. Instead, numerous ad-hoc changes have been made, resulting in complicated funding arrangements that can be confusing. There have also been concerns about continued user fees. Although they decreased initially, particularly for those on low incomes, the decreases were less than anticipated given the large increases in funding and the continued existence of charges might be blunting the provider incentives that were meant to be created by capitation. Major ongoing reforms to the structure and delivery of health services seek to address some of these challenges.188 Strengthening basic functions LMICs considering shifting to a more strategic payment model for PHC need to consider when and how to embark on the process. The Commission emphasises that the right choices of when and how to transform the PHC payment model depend on the country context. Experience suggests that if a country is not ready (with conducive political will and some basic functions in place), payment reforms can be disruptive and potentially harmful. In many LICs, the priority issue is the amount of coordinated funding reaching PHC provider level. These countries should focus first and foremost on getting more funds to PHC providers that can be used flexibly to meet population health needs. This might involve changes to the funding allocation mechanism, as discussed in section 4. In other countries, strengthening basic functions alongside an assessment of the current provider payment system is an appropriate first step.156 Even a perfectly-designed capitation-based blended payment system cannot work without basic supporting functions in place (figure 11). These will need to develop and evolve as the payment system becomes more sophisticated. It is possible that digital technologies can help to support the evolution of the provider payment system but this remains to be seen (appendix p 35). We describe the main basic supporting functions. • Routine data capture and electronic record systems that are interoperable across datasets, ideally for the whole country, are needed to support complex payment systems. This is an iterative process: requiring data for payment leads to better data, which in turn allows for implementation of a more sophisticated payment system. The capacity to analyse and interpret data should be strengthened if policymakers are to act on data-driven evidence. Data are necessary for payment calculations and monitoring of population enrolment, population characteristics, and service delivery performance. The population enrolment list or database must be accurate because payments to providers under capitation are influenced by the number of individuals enrolled with that provider. The method of creating the list and giving providers access to it should be transparent so providers trust the list and their final payment amounts. Data are also necessary for population characteristics monitoring. To adjust population-based payment rates for need, detailed data on the key characteristics of the registered population are required. Last, data are necessary for service delivery performance monitoring because it is important to capture information on clinical care, health-care use, and patient experience to understand how well PHC providers are addressing population needs and to monitor undesirable behaviours, such as underprovision and excess referral. Performance monitoring should be aligned with the benefit package so that providers can be held to account on the services they deliver. • Giving providers more autonomy to make decisions about how to provide PHC supports their flexibility to respond to incentives.11 Fostering provider autonomy includes the acquisition of new skills, many of which might be non-clinical—for example to do with coordination and communication. It also necessitates building their capacities to understand and manage incoming resources. In systems where providers have little management autonomy or do not have the skills to manage new procedures, the results of new purchasing and payment methods will be either be diminished or perverse. For example, in Indonesia, the purchaser for the national health insurance system pays PHC providers by capitation. However, there are strict rules about how public providers can allocate those funds among staff payments and other operational costs.116 In addition, providers that receive funds from multiple revenue streams must know how to allocate and account for them separately. Introducing complicated rules without training providers to understand them greatly diminishes the potential of capitation payment systems to encourage better and more efficient use of resources for service delivery. • Public financial management systems must be flexible and straightforward. Many governments' public finance management systems are notorious for their rigidity regarding the use of funds and the complexity of their accounting and financial management requirements. Traditional public finance management systems might even prohibit prepayment to PHC providers, which is an essential design feature of capitation. However, when health facilities have strong financial management capacity and authority to make some financial management decisions, they are more likely to adjust service provision and deploy inputs based on the needs of the population.115 Payment system reform also requires simplification, by reducing administrative layers and burdens, and harmonising funding flows. This is key to ensuring that resources reach facilities on time and in full, and are appropriately tracked. Delays can be corrosive. In Ghana, for example, delays in fund transfers eroded trust in the capitation system.189 One potential solution is to establish facility bank accounts, which might require changing the legal status of facilities within the public finance management system, to enable direct payments to facilities,128 as has been tried in Tanzania,190 Uganda,128 and Nigeria.128 • The purchaser—whether it is the government or a social health insurance agency—should have the institutional authority and technical capability to enter into legally binding agreements with health-care providers that specify the characteristics and minimum requirements of contracted providers, services that providers will deliver, the methods and terms of payment, reporting requirements, and processes to resolve disputes.156 The experiences of countries that have attempted reforms of provider payment systems also highlight the need for PHC providers to be involved in the design of policies and adequately sensitised so that they understand the changes and lend their support to implementation.191, 192 Patients also need to be given information on their eligibility and entitlements. A method to monitor when providers incur excessive financial risks is required, and risk mitigation strategies considered. Finally, it is important to have a policy on the portability of benefits to determine how patients can access services when away from their registered facility and how their temporary providers will be paid. Managing the politics of provider payment reform Introducing a new provider payment system such as capitation is a highly technical endeavour. It is also a complex political process because making such major system-wide changes affects many of the stakeholders in the health system, including every PHC provider and patient served by the health system. It requires anticipating the effect of the new system on major stakeholders, including medical professionals (both general practitioners and specialists), social health insurance administrators, private sector providers, and the pharmaceutical industries. Introducing reforms requires significant interagency coordination within the government and possibly with donors. The political economy challenges depend on a country's starting position. If the status quo is input-based budgets, a key stakeholder to engage is the ministry of finance because of its central role in defining the budget approach and public financial management rules. In countries with institutionalised fee-for-service, health providers might try to impede the reform, because population-based payment involves a transfer of risk to providers and is often perceived to be less lucrative for them as income is no longer tied to services. The various possible pitfalls a country can encounter when attempting to change provider payment systems are evident in the experiences of several countries. The American Medical Association has opposed capitation (amongst other reforms) since the 1930s, when primary care physicians created a new insurance company to prevent hospital insurance plans from entering the primary care sector and influencing control over fees.193, 194 Medical associations in South Korea (where all physicians are paid on a fee-for-service basis) have complained about price regulation by the government and successfully pushed back on cost control reforms, such as diagnostic related groups and global budgets.195, 196 Providers' resistance has been influential in thwarting or undermining reforms in Taiwan,195 USA,193 Turkey,197 and Ghana (panel 17 ). Patient pushback (often fuelled by provider activism) against limits to provider choice under capitation has been documented in some of these countries.Panel 17 Lessons learnt from the capitation pilot in Ghana The Commission's Ghana case study examined a pilot of a capitation scheme in one region of the country between 2012 and 2016.189 The national health insurance scheme was introduced in Ghana in 2003, to replace the so-called cash and carry system based on user fees that had prevailed in the country after a free health care for all policy was abolished in 1969. The national health insurance scheme started with a fee-for-service payment for all covered services including primary health care (PHC), but that resulted in cost escalations that threatened the sustainability of the scheme. There was, for example, a large increase in annual spending, in the first 5 years of the scheme, driven by increases in the number of claims per insured member.198, 199 In response, the national health insurance authority introduced diagnostic related groups for outpatient and inpatient services and itemised medicine fees in 2008. However, even though the number of claims per member was reduced by 13%, overall costs kept rising and the complicated claims management process led to delays in processing claims. In 2012, the national health insurance authority piloted a capitation payment system for PHC in the Ashanti region. However, the pilot was suspended after 5 years and various challenges. Agitations and protests by providers had begun at the start of the reform.189 Their objections were especially consequential because 2012 was an election year and the government was highly sensitive to any social unrest. In response to the providers' agitation, the Ministry of Health and the national health insurance authority made major policy compromises, including a reduction in the package of services and a 22% increase in the per capita rate.189 These and other compromises could possibly have been avoided if the implementers had considered various political, social, and economic factors from the outset. For example, they might have faced less opposition if they had chosen a pilot site that was less politically sensitive and not dominated by politically powerful private health care providers. Many lessons were learned from the pilot about stakeholder engagement and building trust among providers and users. It also showed the importance of the supporting systems needed to implement capitation payment, which continues to be among policy options considered by the national health insurance for future reforms. The experience of the capitation pilot has also triggered discussions about service delivery reforms to form PHC networks to close gaps in provider clinical capacity, which also posed a challenge to the successful implementation of capitation payment. In addition to the wider context, specific design characteristics of the proposed changes can also hamper or foster reform efforts. In HICs, capitation has often been implemented as part of a larger health reform, such as a shift towards family medicine.200, 201, 202 This was the case in Estonia, where the early establishment of a strong association of family doctors helped build support for PHC reform. In LMICs, capitation has more often been introduced as part of financing, rather than delivery, reforms. Several countries have combined the introduction of a capitation-based system with the creation of national health insurance programmes; examples include Indonesia,203 Thailand,204 and Kyrgyzstan.205 Changing the provider payment system also requires engaging with health system users. Many will have concerns, such as about whether the new system would limit their choices or introduce new forms of gatekeeping. However, one of the key potential benefits of capitation is fostering longer-term relationships between health system users and PHC providers; in principle this can bring more personalised care management and attention to prevention.155 Maximising this benefit and minimising perceived limits on choice require concerted patient education efforts and facilitating informed choice. As is discussed further in section 6, advocates for changing the provider payment system can define political strategies to strengthen support and neutralise opposition to the proposed reforms.197, 206 Motivating health workers Thus far, our discussion of incentives has been focused on PHC provider organisations. Health care is delivered by individuals or teams of health workers; therefore, getting individual incentives right means health workers will be motivated to provide quality care and refrain from engaging in dual practice. Because community health workers have a vital role in the PHC system of many countries, panel 18 addresses the question of how this cadre of health worker should be paid.Panel 18 Paying community health workers WHO recommends that community health workers should be remunerated for their work “with a financial package commensurate to the job demands, complexity, number of hours, training, and roles that they undertake”.207 This of course leaves open the question of how community health workers should be paid. Before addressing this question, it is important to highlight that community health workers in many low-income to middle-income countries feel they are underpaid and poorly compensated for their time and effort, and their pay is often delayed.208 Most community health worker programmes offer some kind of financial incentive, with the choice between salary, monthly payments, or performance-based payments.209 Monthly payments are often a way of avoiding giving further benefits associated with salaried government employees. It is notable that Ethiopia, Ghana, Malawi, and Nigeria have large programmes in which community health workers have the status of formal civil servants. Whether community health workers should be paid by salary or performance-based financial incentives must consider more than effectiveness. However, even if we focus on the narrow question of effectiveness, we are unaware of any studies that have compared the performance of community health workers under these two alternative ways of paying. By contrast, there is a growing body of evidence that suggests some types of performance-based financial incentives are more effective than others in improving delivery outcomes of community health-worker services.210 Large payments appear to be more effective than small ones, but they tend to shift effort away from non-incentivised activities. Giving performance payments to both community health workers and supervisors, and combining incentives with information to the community have been shown to be effective. Financial incentives have not worked when community health workers had limited control over the incentivised task, there were complex rules around disbursement of the incentives, and the focus was on selling products to poor households. Relatedly, there is good evidence that non-financial incentives (social recognition, trust, respect, and opportunities for growth and career advancement) can improve community health workers' performance and reduce attrition.211, 212, 213, 214 Financial incentives for community health workers need to be approached with caution, particularly the exclusive use of them. Although they can be effective, there is also scope for unintended consequences. Such a conclusion is reflected in WHO's suggestion that community health workers should not be paid “exclusively or predominantly according to performance-based incentives”.215 One frequently overlooked observation is that financial incentives generated by the provider payment mechanism at the organisation level might not be passed on to individuals. It depends on how individual health workers are paid. For example, pay-for-performance can provide a strong incentive at the provider level, but if most of the health workers are paid a flat salary, the effect on individual motivation might be less than expected. In low-income countries, salary levels are clearly important for health worker motivation but as influential perhaps are the delays in the monthly payment of salaries that are commonplace. Because of the constraints of any payment system, the way in which services are provided depends to a large degree on the training and professionalism of the clinical workforce. It is essential that payment systems support and reinforce the professionalism of staff and their commitment to providing high-quality care. If the design of payment systems undermines professionalism by, for example, incentivising overprovision or discouraging continuity of care, staff can become demotivated.216 This is one reason why representative clinicians should be involved in the design of payment systems. Health workers place great stock in their strong culture of professionalism, often supported and cultivated during training.217 This culture includes an orientation to the needs of patients, periodic self-reflection, and peer review. This relates to the importance of preserving provider autonomy. Similarly, there are direct non-financial incentives that can improve quality of care, such as opportunities for providers to share data on their performance with other professionals or with the public,218 and social recognition.212 Health workers appreciate having data that validates their perceptions that they are doing a good job; comparing them with their peers can be a powerful incentive to improve their practices. Addressing out-of-pocket payments for PHC The role of provider payment policy Progressive universalism, in which pooled funds should first be used to cover PHC to reduce out-of-pocket payments and replace the lost financing, requires action across all the health financing functions. In particular, removing financial barriers for PHC involves more than just changing user fees policy. It means ensuring patients do not face informal fees and are not sent to pharmacies to purchase medicines because public health providers are under-resourced. Provider payment rates and health worker salaries must be high enough to eliminate the need for user fees and informal payments. In this sense, provider payment policy is integral to the elimination of user fees and informal payments for PHC. In some countries, where escalating health expenditure from excessive use of services is a key concern, there might be a role for cost sharing but its effect on people with the lowest socioeconomic status should be carefully considered and mitigating measures put in place. Incentivising the users of PHC A key benefit of reducing out-of-pocket payments for PHC is to incentivise greater use of cost-effective health interventions, particularly amongst those with the greatest health needs. However, in some contexts, removing user fees is not sufficient to meaningfully expand financial access to care. Various programmes in LMICs have therefore introduced additional financial incentives for patients to use highly effective PHC services, such as immunisations and antenatal care. There is good evidence that offering cash or other financial incentives increases use of primary care services—although whether use translates into improved health outcomes is less clear.219, 220 Incentivising use of PHC will be more effective when the services are more readily available and of high quality. If the policy objective is to increase use, demand-side financial incentives should be considered before changes to provider incentives. However, there is little point in increasing demand for PHC if the available services are not ready or of sufficient quality to meet the population's needs. Conclusion This section has presented the Commission's view that provider payment for PHC should be based on a blended payment model, adapted to each context but with capitation at its centre. It has presented a strategic pathway for countries to progress towards this model, recognising that each country has its own unique starting point. It has recognised that incentive policies for providers and users are linked, and the crucial importance of simultaneously moving towards an elimination of user fees and informal payments for PHC. Progressing this vision requires mobilisation of additional resources for health, allocating these resources to PHC, and ensuring that they work their way through the public finance management system to reach frontline service providers. The success of this technical strategy depends on a political strategy to set the vision, understand the interests of different stakeholders, and actively manage these interests. Section 6: The political economy of financing PHC Transforming financing to support efficient and equitable PHC is often approached as a technical problem. Political and socioeconomic factors affecting financing reforms are frequently described as bottlenecks, barriers, or contextual factors. However, as has been highlighted throughout the report, the Commission recognises that these political economy elements are in fact central to any effort to understand, improve, or reform PHC financing. In this section, we make the case that political economy analysis must be undertaken in conjunction with technical analysis and strategising for PHC financing reform, as part of a national mapping of the PHC financing ecosystem (section 7). This integration is an important first step in resolving some fundamental questions, including: if PHC is the best approach to achieve UHC, why is it not systematically and adequately prioritised in national budgets? And why are purchasing and other health financing reforms successfully implemented in some countries, and resisted in others? The Commission's vision of political economy At its core, political economy brings together systematic explorations of politics and economics and power dynamics between stakeholder groups. Different schools of thought have focused on the so-called economy of politics, the economic constraints influencing political decisions, or understanding how the political context shapes the implementation of economic policy. The materialist approach, for example, focuses on the material conditions of a society's mode of production and argues that these determine how politics, economics, and social processes evolve.221 The new political economy approach states that policies can be analysed through the prism of neoclassical economics.222 Other scholars take an actor-based approach, seeking to identify the winners and losers from policy processes by studying incentive structures influenced by economic interests. The laws and political conditions that shape the material world change over time, thus political economy is essentially an historical science.223 Drawing on these traditions, political economy analysis in the health policy field has tended to focus on politics,224, 225 particularly on power relations among different interest groups and their relative abilities to influence reforms.225, 226, 227 Political economy analysis in health has also tended to analyse the outcomes of processes at a point in time and with a focus on a particular policy or specific issue, typically examining either the contestation and coalitions among interested parties that drive health system operations and reforms,226 or the nature and strength of political institutions that could stop the legislative process that underpins the enactment of reform in political decision-making.105, 228 Some authors have argued that the focus on political dynamics is too narrow229 and that understanding the roles of individuals within political structures has frequently been overemphasised. Political economy analysis is also useful as a broader frame that examines the context, structures, and relationships that generate systemic features; what Jeremy Shiffman, a political scientist, focusing on the politics and global health governance of health policy-making in low-income countries calls the “enduring political and social arrangements not easily altered by the actions of individuals”.230 In this section, the Commission takes a relational view of political economy analysis. Our approach focuses on identifying how key actors—individuals, social groups, organisations, governments, and other stakeholders—relate to each other over time in determining access to, and distribution of power and resources, and how economic and social factors structurally influence these relationships. Applying political economy analysis to PHC financing helps to understand why resources for health are raised and allocated in particular ways to PHC for example, and what competition and contestation occurs throughout these processes. We conclude that political economy analysis has practical value in seeking to explain why efforts to improve efficiency and equity of PHC financing reforms have faced challenges, and to identify prospectively feasible strategies for particular political and socioeconomic contexts. Key messages from Section 6 are presented in panel 19 .Panel 19 Political economy of primary health care (PHC) financing–key messages • Political, sociocultural, and economic conditions are as important as technical elements in the design and implementation of efficient and equitable financing for PHC. These political economy factors represent both constraints and opportunities. • Advancing financing for PHC that is people centred relies on politically informed technical strategies, meaning that policy making in PHC financing and reform must be underpinned by political economy analysis. • Prioritising PHC is possible at every income level and in any type of political governance model, given the presence of effective political alliances and socioeconomic conditions. • Designing politically informed technical strategies requires navigating the evolving political economy context. Political economy analysis can inform the adjustment of the technical strategies to identify the pathways and challenges to the proposed change, taking the long view, and identifying the structural social or economic so-called red lines to be worked around. • Developing PHC financing policy (whether incremental adaptations or a substantial transformation) and ensuring strategic investment in PHC require coherent policy aligned with the interests of key actors through collaboration and building coalitions among stakeholders (leaders, politicians, clinicians, technocrats, donors, and civil society representatives) and across sectors. This development might require achieving consensus or strategic compromise on how to expand access to PHC. • Having a clearly articulated long-term vision is essential for making progress towards efficient and equitable PHC financing. Consistency, adaptation and staying on course are required when countries pursue long-term reforms, while retaining flexibility to take advantage of opportunities for change created by political and socioeconomic events such as political transitions and shocks, or emerging alliances. • Sequencing is key. Planners must have the technical fundamentals and strategies ready in anticipation of windows of opportunity, which arise as a result of political dynamics and social and economic forces. A pragmatic approach to managing political economy considerations The Commission builds on the application of political economy analysis to health financing by explicitly considering broader social, political, and economic features of a context that can influence the success or failure of PHC financing functions and reform efforts, as well as their evolution. Our political economy analysis framework, shown in figure 13 , takes into consideration three domains that influence financing for PHC. First, politics: including the range of actors (individuals, formal and informal organisations, and institutions) and their respective power, their relationships and contracts, their legitimacy, as well as contestation leading to the enactment of policies. Second, social conditions: encompassing social values, informal networks, class, caste, or other social constructs that can influence, for example, the options for distribution or redistribution of resources, including acceptance of, or resistance to, reforms such as greater pooling of resources. Third, economic conditions: including a country's level of economic development, production structures, levels of taxation and levels of aid, that facilitate or hamper the mobilisation of resources to PHC.Figure 13 A political economy analysis conceptual framework for health financing reform These domains are interdependent. Further, they are characterised by dynamic structures and processes that evolve over time—sometimes gradually and in other cases rapidly. Each is discussed in more detail in relation to particular financing functions. Political conditions shape financing for PHC Politics and political conditions have crucial roles in explaining individuals' and institutions' behaviours. Three key political conditions influence the success of PHC financing initiatives in LMICs: the drivers of change, the mix of political actors, and the historical roots of the existing PHC financing system. Political drivers of change Change can be driven by different actors who represent various political powers, economic interests, or social movements. In some settings, such as Brazil and Costa Rica, strengthening PHC financing has been part of a consistent political drive to guarantee basic human rights and equity put forward by social movements that represent, or advocate for the interests of, grassroots populations, including poorer and marginalised groups.231 In both countries, providing essential public services, including investing in PHC, became part of the social contract between the state and its citizens and was supported by a diverse coalition of actors and enshrined in their constitutions.107, 232 In other cases, the change was driven by political leaders seeking to serve the interests of specific constituencies and expand their legitimacy and influence. For example, in China, political leaders initiated equity-oriented health financing reforms as a means to meet ambitious development goals. Similarly in Ethiopia, the federal government's development policies emphasised poverty reduction and support for community-based initiatives in the health, education, and agriculture sectors. This was enabled through a significant public investment in a comprehensive PHC platform organised around a new cadre of health extension workers233 and the Health Development Army of community volunteers. The health extension workers served as a focal point for expanded investments in health on the part of the government and its development partners, which are supported by the volunteers and local governance structures. The concept was rooted in the ruling party's political strategy of prioritising rural interests and seeking to unite ethnically diverse constituencies. This rural-focused strategy had previously enabled the party to govern territory successfully when operating as an insurgency,108 and was rooted in its Maoist and Marxist political ideologies. Technocratic elites (actors who have both technical expertise and political leverage) and professional organisations, accompanied by effective bureaucracies with the ability and interest to operationalise reforms, can also initiate and drive change. Aligning diverse political actors Early involvement of actors from several sectors, including leaders, technical experts, and social activists from inside and outside the health system, has proven instrumental in transforming PHC financing in many countries. Enhancing health sector governance is an essential first step. Particularly critical is attracting a broad range of actors in support of PHC financing reforms to augment the pool of technical knowledge and skills that can support system functioning and transformation, and to foster unity among diverse interest groups. For example, as shown in the Ghana case study (presented in section 5), the introduction of capitation as the method of provider payment needed concurrence among the Ministry of Finance, medical professionals' associations, and patients.189 Getting their buy-in requires considering the economic interests of different actors. Therefore, managing up and building coalitions with the ministry of finance and other agencies as key partners, as well as addressing opposition, are essential when planning and implementing, for example, the introduction of capitation-based payment systems. There are five key considerations related to bringing together the necessary mix of actors. • Political will and legitimacy of the actors driving change are essential preconditions to generating a political process that enables collaborative policy making in support of effective policy. Ideally, this is combined with strong technical leadership at national and subnational levels of the health system. The leaders, benefiting from broad political support, and networks are key,2 and they do not need to be exclusively from the health sector. In Sierra Leone, for example, the removal of user fees in 2010 was actively led by the President Koroma, and supported by a wide range of actors, including the ministries of health and finance, donors, and civil society.234 • Policy development requires the engagement of an enthusiastic group of pioneers or policy entrepreneurs to provide both technical expertise and the vision for the reform. The central role of policy entrepreneurs or political champions in taking forward major policy initiatives such as UHC was also documented in Nigeria.235 Similarly, health financing reforms started in Estonia in the late 1980s when more opportunities for local decision-making started to arise within the Soviet Union.182 After the political transition, leaders from the University of Tartu, supported by external actors, drove the new vision of how PHC family practices should operate. The Ministry of Health funded early pilots before health financing transformations were initiated; the Ministry of Social Affairs and the Estonian Health Insurance Fund were also important supporters.182 Key individuals provided a strong vision and stewarded diverse reform processes, while enlisting a critical mass of professionals (namely, health-care providers and administrators interested in developing a sustainable financing system that would guarantee stable, earmarked funding for health) to provide support throughout the design and implementation of the reforms.236 • Political leaders and policy pioneers alike need to be supported by effective bureaucracies. Several countries that have achieved good health at low cost through investments in, and transformation of, PHC have bureaucracies characterised by strong managerial and implementation capacity, institutional memory, and openness to change. In both Thailand237 (Thailand's building of multi-actor coalition supporting PHC is outlined in the appendix [p 43]) and Estonia,182 bureaucratic elites and technocrats who could draw on their personal legitimacy provided both technical and political support, working with political actors and external experts to enable complex financing reforms. Conversely, in China, PHC financing reforms have been hindered by bureaucratic fragmentation, with local bureaucracies insufficiently incentivised and invested in developing solutions to the fundamental problems hampering PHC and with preference to invest in hospitals that attract political capital locally.145 • Strong civil society engagement is frequently catalytic in generating and sustaining broad-based political platforms that build support for and sustain momentum towards PHC financing reform. Civil society movements, including workers' unions and employers' associations, have led to the right to health (and in particular PHC) being enshrined in the constitutions of many Latin American countries (as previously highlighted in Brazil and Costa Rica). In Thailand, civil society advocacy was an important factor in creating structures at both national and community levels. In South Africa, civil society organisations such as the Treatment Action Campaign were crucial in bringing about the introduction of antiretroviral therapy.238 • Creating and sustaining structures and institutions that promote and support collaboration and dialogue was important in Thailand, Kyrgyzstan, Turkey, and other countries. Engaging key actors effectively in efforts to improve PHC financing functions entails aligning their interests and clearly defining their roles. In practice, the coordination function is often taken on by a dedicated transformation team. Although this team might have access to key political players, it also needs to be sufficiently independent to provide impartial advice and guidance and to weather shifts in power. For example, Thailand and Kyrgyzstan created research units within their Ministries of Health that were instrumental in guiding policy towards the achievement of UHC and sustaining an orientation towards equity.2 In Turkey, Minister of Health Akdag created a reform team that oversaw the implementation of the Health System Transformation Plan. Similarly, in Rwanda dedicated institutions directly supported decision making and coordinated national programme implementation related to PHC financing, including at the National University of Rwanda, School of Public Health and the Rwanda Biomedical Centre, Kigali, Rwanda.239 Such structures have been instrumental in generating evidence for financing reforms while maintaining independence from political processes. Finally, collaboration is at the heart of the concept of a whole-of-government approach that transcends line ministries and other agencies' typical portfolio boundaries to achieve shared multisectoral goals.240 This is particularly crucial to PHC and its financing, in which engaging the whole government involves, in part, recognising the relative power of different ministries involved (particularly health, finance, and other social sectors such as education or water and sanitation), and ensuring that their interests align. In LMICs, the range of key stakeholders involved in designing and implementing PHC policies and programmes can be even broader, including key civil service agencies, donors, professionals, user associations, and civil society organisations—often disrupting the opportunities for joint action. This is particularly evident in areas such as the essential public health functions, a key component of PHC, under the remit of multiple sectors. Importantly, these intersectoral approaches and linkages need to underpin not only national strategic policies but be translated into structures and operational models at subnational levels such as district and local. In Brazil, the municipalities are a key focal point in planning and implementing community-led multisectoral actions in response to local health needs and are governed by stakeholder committees.241 At the local level, the increasing importance of community health workers and multipurpose volunteers, whose role often incorporates tasks to promote social development and address social determinants of health which are core to PHC, and their linking in or integration within the health systems has meant that funding can be channelled more effectively to delivery of essential services (see sections 1 and 5). For example, in Ethiopia, this is seen in the collaborative community-based multisectoral models involving health extension workers, facility-based PHC providers, community governance structures, and volunteers working in close coordination to improve health and wellbeing of the population. This is underpinned by a national strategy for PHC and the Health Sector Development Program acting as a blueprint for how different actors should link to each other. However, coordination among ministries alone is insufficient for effective PHC financing policy and strategic investment—it also requires coherence across policies. This can be achieved by having an overarching vision for PHC, including strategies for improving routine operation and enabling reform. Despite general recognition that comprehensive PHC requires multisectoral action, in practice coordinating financing across sectors and at different levels of government is relatively uncommon. One exception is China, where a cross-ministerial health system reform administrative mechanism has shown promise, with local actors (such as local vice governors and mayors responsible for PHC at the local level of government and the local health commissions) contributing to the ongoing development of integrated policies.145 In Brazil, participatory management councils at the municipal level were even enshrined in the country's constitution, enabling them to manage powerful local interests.241 Sustained vision, flexible strategies PHC financing is subject to contestation that is dynamically changing over time. Policies promoted by one political coalition can be reversed when a new party comes to power. However, PHC financing is also path dependent, so initial decisions can determine the range of options that are available later. The importance of historical roots in fostering PHC that is people centred is seen in the UK's experience, where small general PHC practices funded by voluntary insurance were integrated into a national health system but preserved their autonomy.242 In many LMICs, colonial histories have had a major role in defining how health financing systems are organised. For example, Algeria and Morocco both have social insurance schemes that are similar to the Bismarck model, and the French system. Egypt, meanwhile, has a social assistance scheme inspired by the English Poor Law. Gaza uses an Egyptian insurance scheme, stemming from Egypt's occupation of Gaza through 1967.243 The structure of the health system and payment for this system remains based on the legacy of the colonisers.243 Although path dependency can mean that changes in PHC financing occur over a long period of time, with a series of reform steps building on each other, in some situations the direction of reform has changed at important junctures through radical socio-political or other crises, as in China (how path dependency has influenced China's PHC financing is outlined in the appendix [p 43]). Similarly, countries in Eastern Europe and the former Soviet Union had opportunities to rapidly overhaul resource mobilisation and purchasing arrangements for PHC after the political transition of 1989 and the collapse of the Union of Soviet Socialist Republics (political transition as an important juncture for PHC financing transformation in the Eastern European region is outlined in the appendix [p 44]). Although there is no doubt that compromise and aligning of actor interests and focusing on what is feasible are often required for changes to be politically viable, some of those compromises, although expedient, can greatly limit options later. Such limiting options for PHC financing include a focus on expanding population coverage by starting with formal sector employees, prioritising hospital services in the benefits package, and allowing for fee-for-service payments in the public sector. Because of the dynamic nature of political processes, having a clear long-term vision, upheld and publicly stated over time, has been important in supporting the transformation of financing to enable PHC delivery models. Fostering effective financing functions and reform in line with this vision entails consistently engaging with politics and continually designing new technical solutions to emerging problems. In some cases, technical solutions can be developed while waiting for a window of opportunity to consider them to appear; in other cases such as Turkey's (Turkey's political strategy to strengthen PHC and its financing is outlined in the appendix [p 45]), rapidly changing political conditions create demand for novel technical solutions. Once a clear vision has been agreed, moving towards PHC financing requires strategies for staying on course while retaining flexibility. Therefore, it involves maintaining direction of travel through periods of political stability and economic growth, taking advantage of windows of opportunity created by adverse political events and crises (such as COVID-19), and persisting through periods of stagnation. The balance of power among different groups also evolves over time, leading to the emergence of new agendas, new actors, and new coalitions. Reformers who have candidate technical solutions can slowly gather support for reforms that are initially unpopular or require strategic compromise. The case of the Seguro Popular in Mexico, for example, highlights the importance of strategic compromise and targeted negotiation to move reform processes forward, from abandoning the idea of merging the Mexican Social Security Institute and other social security programmes into a single organisation, to allowing enrolment without premium payment for almost the whole population.225 Brazil, Ethiopia, and China offer examples of countries where comprehensive reforms have been implemented and refined over decades, maintaining a consistent direction while adapting to political and socioeconomic transitions and considerable uncertainty.107, 118, 145 Examining the early efforts of countries that have successfully implemented long-term PHC financing and delivery transformation reveals the importance of building strong foundations to support ongoing changes. These foundations can be technical (ie, ensuring that the technical features of the reform were ready to be used at the earliest opportunity, as in Turkey), or can involve investing in PHC delivery capacity ahead of the reform, which helped Thailand, Ethiopia, and Estonia to absorb additional financial flows and enable large-scale shifts to PHC-focused health systems. In both cases, this involved creating new PHC cadres (eg, health extension workers in Ethiopia and family practitioners in Estonia) that were deployed nationally, instituting training and supervision structures, and embedding best practice norms and clinical standards. In addition to training human resources and extending PHC infrastructure, these activities generated visibility for the early reforms that improved buy-in among different constituencies. Thailand developed standard designs for PHC infrastructure and built many facilities, making services widely accessible with over 10 000 PHC facilities—one per 6000 population—linked to strong community-based volunteer programmes and widespread public health interventions such as dengue control.244 These reforms created public support for PHC and willingness to allocate new resources. Finally, identifying strategic compromises has also characterised many countries' efforts in financing PHC. For example, in Estonia, policy makers passed iterative step-wise health-care reforms from the early 1990s to the end of the 2000s, strengthening the PHC system. In the early stages they adopted an explicit tactic of lying low and not inviting publicity and identifying paths of least resistance for implementation until a critical mass of supportive PHC providers (general practitioners) emerged.245 Compromise can be achieved through iterative processes of piloting, evaluating, and adapting PHC financing innovations before their scaling up—Rwanda, Burundi, and Burkina Faso all did this when introducing results-based financing.246 In China, provincial-level pilot programmes were evaluated, results disseminated, and experiences shared.145 In Thailand, high-level bureaucrats started implementing reforms (including payment reform) as part of a national pilot, and were able to refine the policies before they had to be approved by Parliament. In Brazil, reformers started implementing pilots of the proposed reforms in selected municipalities. These pilots were then enabled by the 1988 Constitution, which established a national health system for the entire country.245 Social and economic conditions influence PHC financing In addition to political conditions, inter-related social and economic factors can also support or hamper routine operation and reform of PHC financing (figure 13). Social conditions A range of social conditions influence financing arrangements. These include: the degree of inequality in a society, the availability of health workers with the capacity to implement reforms, prominent social grievances that propel certain issues to centre stage, and the strength of the social contract between the state and the population, among others. Inequalities within a nation or society can provoke dissent against the status quo and foster support for reforms aimed at redressing the problems. China's 2009 health reforms stemmed from widespread complaints from the population about severe inequality in access to health care, as out-of-pocket payments continued to cripple households financially, accounting for about 60% of total health expenditure.247 Similarly, in Brazil, the thirst for greater equity among segments of the population and better representation after decades of dictatorship propelled successful adoption of broad health system reforms based on principles of health care as a citizen's right and a government responsibility. This formed the basis for a universal, comprehensive, and decentralised health system open to both community participation and, to some extent, private sector initiatives.107 Social grievances against the state or government bodies could also influence the success of a financing reform. Turkey tried to roll out a Family Doctors Programme, which included a swathe of financing and other health reforms, after unpopular decentralisation and privatisation policies had been implemented in the health sector. Those processes had created significant dissatisfaction amongst the population, as well as medical associations and physicians. The Family Doctors Programme was perceived as a similar strategy and met resistance.248 In Nigeria, a distrust of the national government led to the resistance of civil servants at the subnational level towards making contributions to the national health insurance scheme that would guarantee access to PHC, prevented the adoption of the scheme by subnational governments, delayed coverage expansion, and left the scheme as a voluntary programme.249 The strength of the social contract between the population and the state can affect how reforms are received. In the state of Kerala, India, for example, the strong social contract between the state leadership and the population has been partly credited for its early successes in tackling the COVID-19 pandemic.250 In the Middle East and North Africa, political regimes that came to power after independence also established social contracts that included providing material benefits, such as expanded access to primary and secondary health care.243 The weakening of these social contracts over time has driven widespread resistance to more recent reforms, and contributed to the 2011 uprisings across the region. The existence of certain capacities (in the society at large and the health sector in particular) is also important when promoting change. Any health financing reform, such as the introduction of a new budgeting approach or a new provider payment structure, requires skilled staff to communicate and manage the transition and to implement the new approaches. As mentioned in section 3, having insufficient technical budgeting capacity at the Ministry of Health, for example, weakens its bargaining position during national budgeting cycles. This was the case in India, where the scarcity of capacity of the states to prepare health budgets and plans aligned with the central government's expectations has weakened their ability to obtain resources for PHC.126 Importantly, crises of any type can be transformed into opportunities for PHC reform if reformers are poised to act. In the UK for example, the National Health Service was created following the hardship of World War 2.251 In Costa Rica, a 1991 measles outbreak led to employers across the country being forced to pay for private care for their workers due to weak public PHC.252 Employers then threatened to stop making their social security contributions, contributing to government investment in a comprehensive PHC system. The Chinese health reform of the early 2000s, which involved substantial investments in PHC, was triggered in part by the 2003 SARS epidemic.253 The COVID-19 pandemic has been a particularly severe global shock that has affected societies and economic outlook, as discussed throughout the report. The political economy of the pandemic itself, and the response to its devastating consequences, has also started to be scrutinised, and will require further exploration.254 For now, it has triggered initiatives and debates on how PHC needs to be transformed to cater for changing needs.255 Economic conditions National and global economic conditions have significant influence on health financing. These conditions include the structure of the economy, economic cycles of stagnation, recession, or growth, the structure of the health care provider market, the size and dynamics of the private sector, and the importance of aid as a source of financing for health. As discussed in section 3, the structure of a country's economy has a substantial impact on financing for PHC. In low-income countries, only a small portion of the population and private sector organisations are subject to taxes. In Tanzania, for example, just 286 organisations contribute 70% of domestic tax revenue.98 Those who do pay taxes in countries with small tax bases have substantial power in driving what reforms can be implemented. The level of informality in the labour market dictates whether any type of labour employment tax or health insurance contributions will generate sufficient revenue to support social health insurance. In Ethiopia, for example, attempts to implement social health insurance have stalled for many years in part due to the high level of labour informality. In addition to the structure of a country's economy, where the country is in the economic cycles of growth and contraction can influence the success of a reform aimed at increasing financing for PHC. In Brazil, the fiscal space generated from sustained economic growth during the 2000s enabled the country to increase its public health expenditures.107 In Chile, similarly, important PHC reforms were made possible by high economic growth during the first decades of the country's return to democracy.55 However, in Finland, the collapse of trade with the former Union of Soviet Socialist Republics in 1992 led to a steep economic decline and, as a result, national expenditure on health care was slashed by 12% in 1991–94.256 The structure of the health-care provider market also influences reforms, particularly those to do with purchasing. On the one hand, in Estonia, the Association of Family Doctors had a major role as a partner to the national insurance agency in designing and implementing laws to support PHC financing.182 In Ghana, on the other hand, the design and implementation of the national health insurance scheme was influenced by the strong bargaining power of providers, who preferred being paid on a fee-for-service basis and resisted a system change that would affect payment mechanisms for PHC.257 Finally, the size and dynamic nature of the private sector is another critical economic factor that can influence health financing reform efforts. In Thailand, for example, the capitation system was designed to improve quality of care through fostering competition among PHC providers. This was possible in Bangkok, where a large number of private providers were willing to accept patients with insurance (as the capitation system was part of the national health insurance scheme).184 In Ghana, on the other hand, the shortage of public and private providers in rural areas constrained competition during the capitation pilot in Ashanti region.189 Applying political economy analysis to advance financing for PHC As noted, political economy analysis seeks not only to explain but also to derive practical implications for strategic policy making. It explains how the political and socioeconomic contexts shape what is possible or not in developing and implementing key policies. Although some cross-country lessons can be drawn (with caution) from the country examples discussed throughout this report, the critical influence on PHC financing functions of the political, social and economic conditions that make up the political economy are best understood within each national, and often subnational, context. Analysis of these factors should be an integral ongoing part of implementing feasible interventions to improve efficiency and equity of PHC financing. A common thread is that there is no blueprint approach to changing PHC financing, and even within a country, the political economy context also changes over time, often rapidly. Political economy analysis can be used to inform proactive or responsive strategies for adaptive management of the interests of different actors and formulating strategies that fit the social and economic conditions in support of health financing reforms. A political economy lens focuses on understanding the structural and socioeconomic conditions underpinning decision making and conflicting interests. This understanding might lead to, for example, the use of strategies to strengthen actors with limited power but who would benefit most from more effective PHC financing. In other cases, political economy analysis may indicate the need to anticipate and manage resistance from those who benefit from the status quo, and to identify opportunities to form coalitions. Practical approaches to using political economy analysis to underpin routine operation or transformation of PHC financing are outlined in the appendix (p 45). Designing politically informed technical strategies starts by asking the right questions to navigate the complex political economy context (Sparkes S, WHO, personal communication).258 In this Commission, we have formulated a series of key questions that should be asked throughout the policy cycle, and can be used for setting the agenda as well as designing and implementing people centred PHC financing policies: • What is the problem to be addressed? What ideas exist for improving PHC financing? What technical strategies would achieve this over time? • Who are the stakeholders with an influence over the problem? What are their positions on the topic, and what is their relative power? • What are the political dynamics at play? • What could help to shift incentives to promote the changes pursued? • What social and economic conditions that underpin the political process could present opportunities or constraints for the proposed change? • What are the most likely pathways for change? What are possible entry points to move the reform forwards? If there is a potential window of opportunity, how can it be used to generate and sustain political momentum? • How do the proposed strategies take into account path dependency? • How should the strategies be sequenced? Although awareness of and the need to conduct political economy analysis is a part of developing policies for sustainable and equitable PHC financing, country-level capacity to do so can be scarce. Investing in basic supportive functions, which include technical capacity to do political economy analysis, ability to engage with actors and policy processes, and translate knowledge to policy is key. Importantly, there is often knowledge of political economy analysis developed in other sectors that can be used in policy development in the health sector. Support for, and retention of professionals who can develop and integrate specific aspects of political economy analysis in their work should be ongoing. Subnational entities such as districts can be incentivised to co-fund training and recruitment of managerial or research cadre able to use particular political economy analysis skills, relevant to context, as a part of leadership programmes. We argue that sustaining political economy analysis is key as a part of investment in both system transformation and strengthening routine financing systems; it is particularly important in effectively responding to large-scale shocks such as pandemics or political transitions. Conclusion Technical strategies for efficient and equitable financing for PHC are neither designed nor implemented in a vacuum. They are critically shaped by political, economic, and social conditions—and the dynamic nature of these forces can create opportunities to maximise impact, or impose barriers that constrain success. Applying a political economy lens to technical solutions that explicitly recognises the evolving roles of actors beyond the health system, their relative resources and power, as well as the economic constraints and social relations, is therefore necessary to strengthen the PHC financing architecture. Section 7: Recommendations The Commission's deliberations have analysed how health financing arrangements can be used to drive national health systems to provide equitable, comprehensive, integrated, and high-quality PHC, delivered through platforms that are responsive to the needs of the populations they serve, and fully aligned with the objectives of UHC. We argue that countries should invest more and invest better in PHC, and that the financing arrangements that support PHC—from mobilisation and pooling of resources, to budgeting, allocation, and purchasing—must place people at the centre. They must also be driven by a focus on equity and social justice, in line with the original Alma Ata vision. We articulate the features of people-centred financing for PHC below. We recognise that the opportunities to reorient health financing policies towards PHC depend on the economic, social and political features of a particular regional, national, or sub-national context, and that there is no single pathway to achieving optimal PHC financing (figure 14 ).Figure 14 Framework for people-centred financing of PHC An overarching position of the Commission has been the principle of progressive universalism: This means that governments should prioritise equity by providing universal access to affordable, quality PHC services, and particularly on ensuring that disadvantaged groups are reached first. Guaranteed entitlements can expand beyond PHC as fiscal capacity increases. Beyond this overarching principle, the Commission has identified four key attributes of people centred financing for PHC. First, public resources should provide the core of PHC funding with minimal reliance on direct payments when services are accessed. In most LMICs, this level of public funding can only be generated through increasing allocations to PHC from general tax revenue. Revenue raising mechanisms should be defined based on ability to pay and be progressive. While each country is at a different starting point for a shift to predominantly public funding, strategic and purposeful change to national health financing systems over time can enable gradual progress. In the meantime, low-income countries will require continued development assistance to secure a sufficient resource envelope to enable population coverage of essential PHC services. Second, pooling arrangements should cover PHC. By supporting PHC with pooled public funds, out-of-pocket payments can be reduced to levels at which they do not pose financial barriers to accessing needed care, impoverish households, or push households deeper into poverty. Pooling enables cross-subsidy among those who are well and those who are ill, and among the poor and the wealthy. Third, resources should be allocated equitably across levels of service delivery and geographic areas, and be protected to reach frontline PHC service providers and patients. Resource requirements for PHC should be estimated on the basis of accurate assessments of population health needs. Countries should deploy strategic resource allocation tools (including needs-based per-capita resource allocation formulae) in budget formulation, budget execution, public financial management policies, and service delivery arrangements to channel and protect the flow of resources for PHC. Fourth, provider payment mechanisms should: assign resources based on people's health needs; create the right incentive environment to promote PHC that is people-centred along the spectrum of prevention, health promotion, and treatment; foster continuity and quality of care; and be flexible enough to respond to changing needs of patients, families and communities. This is best achieved through a blended provider payment system with capitation at its core. Table 2 presents a matrix which maps how financing arrangements can be deployed to achieve the dual goals of people-centredness and equity.Table 2 The Commission's vision of financing functions and arrangements for PHC that is people centred and equity driven Mobilisation Pooling Allocation Purchasing People-centred characteristics Resource requirements for PHC should be estimated based on what is needed for each person to access; PHC that is people centred as defined in the country's context Everyone is included in the pool Resources are allocated based on population needs for PHC that is people centred (rather than on facilities, inputs, or vertical programmes) Purchasing arrangements and provider payment mechanisms are linked to making PHC that is people centred available to people and flexible enough to accommodate different modes of delivery; funds flow to and are managed by frontline providers as defined in the country context Equity and progressive universalism characteristics Revenue-raising mechanism is defined based on ability to pay and is progressive Cross-subsidisation occurs between poor and wealthy populations and healthy and sick populations; use of pooled funds prioritises making PHC accessible, with financial protection and subsidies directed to the poor The mechanism used to allocate public funds prioritises the needs of the poorest segments of population, and areas (geographical or health) of greatest need Per capita payment (capitation) is the starting point, which makes the same amount of funds available to providers to deliver the PHC package for each person (adjusted upward or downward according to health needs) Practical implications and anticipated outcomes Reduces out-of-pocket expenditure; progressive taxation policies Merge or consolidate existing pools into larger pools (including formal and informal sectors, poor and rich; coverage dominated by public financing); who and what are covered by the pool expands in the most equitable and PHC-centric way; progressively move to universal health coverage according to the macro-fiscal capacity of the country, starting with access to PHC for all and financial subsidies directed to the poorest and most vulnerable; access to more services beyond PHC and subsidies for more population groups can expand as macro-fiscal capacity expands Budgeting is based on needs-based per capita allocations to enable access to PHC that is people centred (rather than to facilities, inputs, or vertical programmes); protect resources going to PHC through existing policy tools, such as programme budgets, resource allocation formulae, conditional grants or statutory rules; define a benefit package that prioritises coverage of the needs of poorest segments of population; ensure resources reach frontline providers (through direct facility financing, for example) and improve public finance management systems more broadly; organise service delivery to pull resources to PHC, for example by creating new cadres of frontline PHC providers, defining explicit service standards, or instituting effective referral systems Establish a blended payment model with capitation at its core: start with a baseline capitation payment . The payment amount should be determined using a formula that links the payment parameters (base per capita rate, number of enrolees linked to the provider, and any individual or provider-level adjustments) to a defined package of PHC services; define a PHC package; adjust the risk level to prioritise those in greatest need PHC=primary health care. Together, these attributes form the foundations of a resilient and responsive health financing system. As has become evident during the COVID-19 pandemic, effectively financing PHC during a crisis period relies on the existence of a health system that is capable of surging to tackle new priorities while also continuing to deliver existing services. Resource mobilisation, pooling, allocation, and purchasing systems must be able to respond quickly to deliver additional resources while protecting allocations to PHC. We acknowledge that there is no single pathway to achieve optimal PHC financing, and that every country is at a different point in orienting its PHC policies and financing. Moreover, it will take time to adapt financing arrangements, and these will need to continue to adapt to changing conditions and needs over time, and to respond to shocks. However, by moving deliberately towards publicly-financed, progressively universal, and population-based health financing, countries can support the expansion and improvement of PHC. In support of this vision, the Commission makes five recommendations to local, national and global policy makers and other relevant stakeholders (panel 20 ). We recognise that many of the recommendations could potentially be applicable to health system reform in general. However, we continue to focus specifically on PHC, while endeavouring to avoid separating it out from the rest of the health system.Panel 20 Our vision for people-centred financing of primary health care (PHC) PHC needs both more and better resources. The Commission's vision is of a people-centred system for financing PHC. This system should be capable of collecting, pooling, and allocating resources to purchase services that ensure that all people (community members, patients, and providers) are able to benefit. Progressive universalism—ensuring that, at every step, people who are poor or vulnerable gain at least as much as those who are wealthy or privileged—is at the core of this vision. Achieving this vision requires: • An adequately-financed health sector, funded by expanded public and pooled sources, that protects everyone from financial hardship when seeking care. The Commission argues for an explicit focus on addressing inequities first. This entails that revenue will be raised based on ability to pay and through progressive means. • Pooled funding will cover PHC, to enable everyone to receive PHC that is free at the point of use. Pooling of resources will support cross-subsidisation among those are well and those who are ill, and among the poor and the wealthy. • A strategic use of all available policy tools to direct sufficient resources to PHC to enable a universally-accessible system that provides high-quality services according to a defined benefit package appropriate to the level of care and aligned with macro-fiscal capacity. In line with our core focus on people-centeredness and equity, the Commission proposes that resources are allocated based on population needs, prioritising the needs of the poorest segments of the population. To do so will require mechanisms for funding, budgeting, and financial management that ensure that resources reach frontline providers and platforms. • A context-specific blended payment model built on capitation. Payment systems should allow adequate resources to flow to the PHC level in ways that: are equitable; match resources to population health needs; create the right incentive environment to promote the full PHC spectrum of prevention, health promotion, and management and treatment; foster people-centeredness, continuity, and quality of PHC; and, are flexible enough to support changes in service delivery models and approaches. • A nuanced understanding of the political economy of each country throughout the development and implementation of all policy to accompany the technical approaches to ensuring people-centred financing for PHC. (1) Establish people-centred financing arrangements for PHC that have four key attributes Establish financing arrangements for PHC following a principle of progressive universalism and incorporating the people-centred attributes outlined above. These are: public resources provide the core of PHC funding; pooled funds should cover PHC; resources should be allocated equitably and protected so they reach front-line providers; and provider payment is through a blended mechanism with capitation at its core. (2) Take a whole-of-government approach to spending more and spending better on PHC Key actors and stakeholders should be involved in designing and implementing people-centred PHC financing reforms. Although the specifics will vary depending on the national context, some general roles and responsibilities can be identified. The ministry of health should lead efforts to prioritise PHC. Leadership involves promoting technical strategies embodying the above principles, ensuring that sufficient resources are made available, and elaborating and pursuing political strategies in support of expanding and improving PHC financing. The ministry of health should ensure it has the technical expertise to make the case for more funding for PHC. To ensure accountability, which sections within the ministry of health are responsible for the financing and delivery of PHC should be clarified. The ministry should take responsibility for engaging the commitment of the other sectors (such as education and water and sanitation) whose activities relate to PHC. The ministry of finance should enable the mobilisation of sufficient revenue to adequately finance people-centred PHC, as defined nationally. The ministry of finance should work with the ministry of health and other agencies to develop flexible and responsive public finance management systems that make allocations to PHC visible, protect resource flows to reach the frontlines, and allow strategic provider payment systems that evolve as capacity grows and service delivery models mature. Local government agencies should serve as bridges between local populations and central government ministries. Local authorities are well-positioned to identify and communicate populations' needs so they can be accurately captured in allocation formulas. Local agencies are also responsible for integrating multiple funding flows and ensuring they are applied to addressing local priorities. Communities and civil society groups should demand changes. Communities should engage in efforts to hold PHC providers accountable, and be included as partners in monitoring progress. To that end, a key priority for civil society groups should be building the capacity of communities to undertake these functions. Governments should facilitate this accountability by producing data to enable monitoring. Health-care providers and their representative organisations should actively participate in efforts to reform PHC financing arrangements. Providers must engage in the design of payment reforms, understand the implications of proposed changes to provider payment systems, and take any and all opportunities to provide people-centred health care fostered by reformed financing arrangements. Donor and technical agencies should provide financial resources and expert technical support to countries that need assistance to jumpstart changes in PHC financing. At country level, agencies should ensure that, at the very least, their actions do no harm. Agencies should immediately change their approaches to providing financial and technical assistance to reduce fragmentation. At best, donor and technical support agencies can be strategic partners to the national governments working to improve their financing systems to support PHC that is people centred. (3) Strategically plot out a pathway towards people-centred financing for PHC, including supporting basic health system functions Each country should articulate a vision for financing PHC. Having a clear vision allows decision makers to plot a strategic technical path and identify what political engagement is needed from stakeholders throughout the system to support progress. National funding champions of PHC should proactively explore and recognise the political, economic, and social conditions at subnational, national, and global levels to effectively navigate towards the vision through the evolving political economy context. This political economy analysis should begin at the outset of any reform process. Implementing politically-informed technical strategies (ie, strategies rooted in a thorough understanding of the political economy context within which the technical approach sits) should involve regular mapping of the political landscape, assessing opportunities to align interests and build coalitions among actors from different sectors and administrative domains, and communicating and working towards collaboration and strategic compromises in support of key technical policies. This might require strengthening the skills of people working in government, and in academic and donor partners, to undertake political economy analysis. The country's vision should be operationalised by mapping out a clear set of steps to pursue its chosen course, while also preparing to capitalise on unexpected opportunities and creating room to manoeuvre as needed to adapt to political and socioeconomic changes, crises, and other shocks. All stakeholders should undertake ongoing efforts to strengthen the health system's basic functions, including data collection and analysis of resource flows and health impacts; monitoring, evaluation, and learning systems; public finance management systems that enable resources to reach frontline providers; and capacity of providers to manage funds effectively. Investment in such basic functions is needed in conjunction with investment in PHC. (4) Global agencies should reform the way PHC expenditure data are collected, classified, and reported A new reporting item (ie, memorandum item) on PHC should be defined and included in countries' annual reporting of health expenditures to WHO for the Global Health Expenditure Database. In the meantime, current reporting should be adapted to be based on a cross-classification of functions and providers (such as the one used by the OECD). This will provide more specific and useful data by, for example, allowing for differentiation between hospital and ambulatory providers, and enable an operational definition of PHC based on service delivery platforms. The categories currently included in the calculation of PHC expenditure should be revised. In particular, how administrative costs are included and how outpatient services in hospitals are classified should be reconsidered, as current practices skew estimates of PHC spending upwards. Most importantly, each country should establish a clear definition of PHC expenditure that is compatible with how its health system organises services; it can then use this definition to track spending over time to monitor progress. (5) Conclusion The Commission recognises that its work represents the beginning, not the end, of a research agenda on financing people-centred PHC. The Commission's explorations raise many additional questions, starting with those presented in panel 21 . The Commission proposes the following next steps and starting points for additional exploration by key stakeholders, including academic researchers, technical experts, policy makers, donors, and others: • Creating a tool for national mapping of PHC financing ecosystems to create a firm data foundation for developing appropriate technical and political strategies to advance people-centred financing in support of PHC. Collaboration among researchers, technical experts, and policy makers is needed to develop a robust tool and method. • Exploring and devising innovations to support better PHC financing, including adopting implementation science and other operational research methods. • Securing funding from governments and other donors for rigorous research on the research questions suggested in panel 21, as well as others that will arise. Panel 21 Research questions on financing for primary health care (PHC) that is people centred The Commission's work has answered some questions, particularly on technical aspects of financing arrangements. But it has raised many others, most especially on how to operationalise our recommendations. These are the ‘how do we do this?’ questions. Going forward, the research agenda on financing PHC should study the outcomes of proposed reforms and examine implementation at country and local level. Key topics and questions include:Innovative approaches to support the delivery of PHC: • What are the best ways to channel funds directly to facilities? • What are the effects on health outcomes of getting money to frontline providers? • How can digital innovations be used in health financing systems, while promoting universal health coverage and minimising fragmentation?Spending more and spending better on health in general and PHC in particular: • What bottlenecks in health financing reforms arise in particular settings? What are the best approaches to counteracting them? • What strategies that have been effective in ensuring financing for essential public health functions and linking them to PHC can be replicated, and how? • What methods and data work well in measuring implementation of health financing reforms and tracking how funding flows change (including volume, recipients, timeliness, and equity)? How can these methods be used at the local, national, and global levels? • Why are reforms addressing even well-known financing inefficiencies difficult to implement, and how can the reforms addressing these inefficiencies translate into additional financial resources for PHC? • How can researchers collaborate effectively with policymakers to evaluate potential solutions and respond to their priorities and concerns?The political economy factors of financing PHC: • What strategies have been used to effectively manage political economy considerations of providers and patients in health reforms? What are patients' and providers' understandings and attitudes towards proposed financing reforms? • How can local actors be supported to foster investment and allocation in PHC? What power shifts are needed and which technical capabilities are most relevant? • How do local and central government bodies interact in designing and implementing health financing reforms? What political economy considerations are important when seeking to influence resource allocation in decentralised settings? • What are the political economy considerations for efforts in LMICs that seek to address fragmentation in pooling and shifting provider payment mechanisms towards capitation? What approaches can be effectively used to manage these factors? In this report, we have set out a vision for placing people at the centre of the arrangements for financing PHC. This financing vision serves a greater ambition: health systems that provide equitable, comprehensive, integrated, and high-quality PHC delivered through platforms that are responsive to the needs of the populations they serve and fully aligned with the objectives of UHC. Declaration of interests KH, DB, NB, DE and TP-J were funded by a grant from the Bill & Melinda Gates Foundation. DH has received funding from the Bill & Melinda Gates Foundation for various activities at UNICEF, including on health system strengthening and community health, both of which are mentioned in this report. In the period during which the report was developed, AE and HWa were employed by the Bill & Melinda Gates Foundation and were involved in data analysis, interpretation and writing of the report. All other authors declared no conflict of interest. Supplementary Material Supplementary appendix Acknowledgments This report is based on research funded by the Bill & Melinda Gates Foundation. The findings and conclusions contained within are those of the authors and do not necessarily reflect positions or policies of the Bill & Melinda Gates Foundation. Brazil case study was undertaken by Adriano Massuda, Ana Maria Malik, Gabriela Lotta, Marina Siqueira, Renato Tasca, Rudi Rocha (Getulio Vargas Foundation, Rio de Janeiro, Brazil). China case study was undertaken by Jin Xu, Beibei Yuan (Peking University School of Public Health, Beijing, China). Chile case study was undertaken by Cristóbal Cuadrado (University of Chile, Santiago, Chile; University of York, York, UK), Alejandra Fuentes-Garcia, Ximena Barros, Maria Soledad Martinez, Jorge Pacheco (University of Chile, Santiago, Chile). Ethiopia case study was undertaken by Mirkuzie Woldie Kerrie (Addis Ababa University, Addis Ababa, Ethiopia), Girmaye Dinsa Fenot (Harvard T.H. Chan School of Public Health, Boston, MA, USA), and Kiddus Yitbarek (Ministry of Public Health, Addis Ababa, Ethiopia). Ghana case study was undertaken by Eugenia Amporfu, Eric Arthur (Kwame Nkrumah University of Science and Technology, Kumasi, Ghana). India case study was undertaken by V R Muraleedharan (Indian Institute of Technology Madras, Chennai, India) and Shankar Prinja (Post Graduate Institute of Medical Education and Research, Chandigarh, India). Philippines case study was undertaken by Leizel Lagrada-Rombaua, Joyce Encluna, Emanuel Gloria (GECC Development Services, Pasig City, Philippines). Estonia case study was undertaken by Kaija Kasekamp, (PhD candidate, University of Tartu, Tartu, Estonia); Triin Habicht, (WHO Barcelona Office for Health System Financing, Barcelona, Spain); Ruth Kalda (University of Tartu, Tartu, Estonia). Finland case study was undertaken by Ilmo Keskimäki (Research Professor, Finnish Institute for Health and Welfare, Helsinki, Finland; and Professor of Health Services Research, Tampere University, Tampere, Finland). New Zealand case study was undertaken by Jacqueline Cummings (Independent Health Services and Policy Consultant, Wellington, New Zealand). Shaheda Viriyathorn, Anond Kulthanmanusorn, Walaiporn Patcharanarumol, and Viroj Tangcharoensathien (International Health Policy Program, Ministry of Public Health, Bangkok, Thailand) provided additional data for sections 2 and 4. Research assistants that contributed to this Commission were Mazvita Briony Pasipanodya, The London School of Hygiene & Tropical Medicine, London, UK; Manon Haemmerli, The London School of Hygiene & Tropical Medicine, London, UK; Lauren Hashiguchi, The London School of Hygiene & Tropical Medicine, London, UK; F Felipe Vera, Pontificia Universidad Católica de Chile, Santiago, Chile; Henry Cust, The London School of Hygiene & Tropical Medicine, London, UK; Rotimi Alao, The London School of Hygiene & Tropical Medicine alumni; Kevin Pene, The London School of Hygiene & Tropical Medicine alumni; Swati Srivastava, Heidelberg Institute of Global Health, Medical Faculty and University Hospital, Heidelberg, Germany. Experts consulted on political economy analysis were Thomas Lavers (Senior Lecturer in Politics and Development, The Global Development Institute, The University of Manchester, Manchester, UK), Carlos Oya (Professor of political economy of development, School of Oriental and African Studies, University of London, London), Susan Sparkes (WHO, Geneva, Switzerland) and Julia Ngozi Chukwuma (Open University, Milton Keynes, UK). For the provider payment survey of low-income to middle-income countries, Alexandra Michele Beith (World Bank), and Agnes Munyua (R4D) contributed to survey design, piloting, and implementation. Advice and assistance in accessing expert respondents were provided by Matthew Jowett (WHO) and Julia Sallaku (WHO). Becky Wolfe (Independent Consultant) copy edited all case study reports. Figures and diagrams were developed by Joanne Duffy (London School of Hygiene & Tropical Medicine) and Becky Wolfe. Anya Levy Guyer, Boston, MA, USA, was freelance technical writer and copy editor for this Commission. The Commission Manager was Brigid Strachan. Contributors The initial conceptual framework for the Commission was developed by KH together with DB and TP-J, and all authors collaborated in refining it. The first draft of the report was written by a core writing team led by KH and included NB, DE, TP-J, and DB. All authors contributed intellectual content to the report concepts, conclusions, and recommendations, and to the writing and editing of subsequent drafts. Support to case studies was provided by QM (China), RS and AE (India), JVM (Chile) and EN (New Zealand, Estonia, and Finland). The provider payment survey was designed by CK, TP-J and DE, with additional input from CC, MR and MdA, and data collection from their networks was facilitated by CK and CC. KH, TP-J, DB, DE and NB accessed and verified the underlying data. The opinions expressed in the publication are those of the authors and do not necessarily represent the views, decisions or policies of the funder or the institutions with which the Commissioners are affiliated. All authors had full access to the data in the study and accept responsibility for the decision to submit for publication. The Thailand case study was provided by S Viriyathorn, A Kulthanmanusorn, W Patcharanarumol, and V Tangcharoensathien, International Health Policy Programme, Ministry of Public Health, Thailand. ==== Refs References 1 Organisation for Economic Co-operation and Development Realising the Potential of PHC 2020 Organisation for Economic Co-operation and Development Publishing Paris 2 Balabanova D Mills A Conteh L Good Health at low cost 25 years on: lessons for the future of health systems strengthening Lancet 381 2013 2118 2133 23574803 3 Starfield B Shi L Macinko J Contribution of primary care to health systems and health The Milbank Quarterly 83 2005 457 502 16202000 4 Sacks E Schleiff M Were M Chowdhury AM Perry HB Communities, universal health coverage and primary health care Bull World Health Organ 98 2020 773 780 33177774 5 WHO Closing the gap in a generation: health equity through action on the social determinants of health—final report of the Commission on Social Determinants of Heath 2008 WHO Geneva 6 WHOUNICEF A vision for PHC in the 21st century: towards universal health coverage and the Sustainable Development Goals 2018 WHO Geneva 7 Organisation for Economic Co-operation and Development Service delivery in fragile situations OECD J Dev 9 2009 13 8 WHO Building the economic case for PHC: a scoping review 2018 WHO Geneva 9 Watkins DA Yamey G Schäferhoff M Alma-Ata at 40 years: reflections from the Lancet Commission on Investing in Health Lancet 392 2018 143 160 10 Stenberg K Hanssen O Bertram M Guide posts for investment in primary health care and projected resource needs in 67 low-income and middle-income countries: a modelling study Lancet Glob Health 7 2019 e1500 e1510 31564629 11 Stenberg K Axelson H Sheehan P Advancing social and economic development by investing in women's and children's health: a new global investment Framework Lancet 383 2014 133 154 12 Gaudin S Smith PC Soucat A Yazbeck AS Common goods for health: economic rationale and tools for prioritization Health Syst Reform 5 2019 280 292 31661367 13 Walraven G The 2018 Astana Declaration on Primary Health Care, is it useful? J Glob Health 9 2019 010313 14 WHO Integrating health services 2018 World Health Organization Geneva 15 Andrews M Cangiano M Cole N De Renzio P Krause P Seligmann R This is PFM. CID Working Paper no 2852014 https://www.hks.harvard.edu/centers/cid/publications/faculty-working-papers/pfm 16 WHO The World Health Report 2008: primary health care now more than ever 2008 World Health Organization Geneva 17 WHO WHO guideline: recommendations on digital interventions for health system strengthening 2019 World Health Organization Geneva 18 Health Data Collaborative Knowledge Hub https://www.healthdatacollaborative.org/knowledge-hub/ 2021 19 Kraef C Juma P Kallestrup P Mucumbitsi J Ramaiya K Yonga G The COVID-19 pandemic and non-communicable diseases-a wake-up call for primary health care system strengthening in Sub-Saharan Africa J Prim Care Community Health 11 2020 1 3 2150132720946948 20 Haldane V De Foo C Abdalla SM Health systems resilience in managing the COVID-19 pandemic: lessons from 28 countries Nat Med 27 2021 96 180 21 Sagan A Webb E Assopardi-Muscat N Mata Idl McKee M Figueras J Health systems resilience during COVID-19: lessons for building back better 2021 WHO Regional Office for Europe Copenhagen 22 Roy H Asian-inspired contact tracing strategy backed by Canadians: poll https://www.asiapacific.ca/publication/asian-inspired-contact-tracing-strategy-backed-canadians 2020 23 André F Kergadallan M-L Zheleznyakov E France: community partnership in multidisciplinary PHC 2021 WHO Regional Office for Europe Copenhagen 24 Martí T Peris A Cerezo J Spain: accelerating multidisciplinary teamwork to address emerging primary care needs in three Spanish regions 2021 WHO Regional Office for Europe Copenhagen 25 Yazbeck AS Soucat A When both markets and governments fail health Health Syst Reform 5 2019 268 279 31684822 26 Rahim F Allen R Barroy H Gores L Kutzin J COVID-19 funds in response to the pandemic 2020 International Monetary Fund Washington, DC, WA 27 The Lancet The ACT accelerator: heading in the right direction? Lancet 397 2021 1419 28 Hipgrave DB Kampo A Pearson L Health systems in the ACT-A Lancet 397 2021 1181 1182 29 Kurowski C Evans DB Tandon A From Double Shock to Double Recovery 2021 World Bank Washington, DC, WA 30 Schneider P Pivodic F Yoo KJ How much health financing does Sub-Saharan Africa need to fight COVID-19 (coronavirus)? https://blogs.worldbank.org/health/how-much-health-financing-does-sub-saharan-africa-need-fight-covid-19-coronavirus 2020 31 International Monetary Fund World economic outlook: managing divergent recoveries (April 2021) 2021 International Monetary Fund Washington, DC 32 WHO Role of primary care in the COVID-19 response 2020 WHO Regional Office for the Western Pacific Manila 33 WHO Global spending on health 2020: weathering the storm 2020 WHO Geneva 34 Mathauer I Dkhimi F Townsend M Adjustments in health purchasing as part of the Covid-19 health response: results of a short survey and lessons for the future https://p4h.world/en/blog-covid-19-and-health-purchasing-response 2020 35 Permanand G Kirkby V McKee M Take action at all levels of societies to fix the fractures that left so many people vulnerable to the pandemic McKee M A review of the evidence: drawing the light from the pandemic 2021 European Obervatory on Health Systems and Policies Brussels 36 Leydon N Kureshy N Dini H-S Nefdt R Country-led institutionalisation of community health within primary health care: reflections from a global partnership J Glob Health 11 2021 03037 37 Zulu JM Perry HB Community health workers at the dawn of a new era Health Res Policy Syst 19 suppl 3 2021 130 34641904 38 Hermann K Van Damme W Pariyo GW Community health workers for ART in sub-Saharan Africa: learning from experience--capitalizing on new opportunities Hum Resour Health 7 2009 31 19358701 39 Ballard M Westgate C Alban R Compensation models for community health workers: comparison of legal frameworks across five countries J Glob Health 11 2021 04010 40 Mackintosh M Channon A Karan A Selvaraj S Cavagnero E Zhao H What is the private sector? Understanding private provision in the health systems of low-income and middle-income countries Lancet 388 2016 596 605 27358253 41 Basu S Andrews J Kishore S Panjabi R Stuckler D Comparative performance of private and public healthcare systems in low- and middle-income countries: a systematic review PLoS Med 9 2012 e1001244 42 OECD Health Systems Characteristics Survey https://www.oecd.org/els/health-systems/characteristics.htm 2018 43 McPake B Hanson K Managing the public–private mix to achieve universal health coverage Lancet 388 2016 62 130 27155903 44 WHO Estimation of PHC expenditure: technical note for discussion 2019 WHO Geneva 45 National Health Accounts Working Group Thai national health accounts 2017–19 2021 International Health Policy Program Foundation Bangkok 46 Organisation for Economic Co-operation and Development Organisation for Economic Co-operation and Development. Stat. Health expenditure and financing https://stats.oecd.org/index.aspx?DataSetCode=SHA 2021 47 WHO Global health expenditure database https://apps.who.int/nha/database 2021 48 Mueller M Morgan D Deriving preliminary estimates of primary care spending under the SHA system of health accounts 2011 framework 2019 OECD Publishing Paris 49 Wirtz VJ Hogerzeil HV Gray AL Essential medicines for universal health coverage Lancet 389 2017 403 476 27832874 50 Lauer JA Soucat A Araujo E Paying for needed health workers for the SDGs: an analysis of fiscal and financial space Buchan J Dhillon IS Campbell J Health employment and economic growth: an evidence base 2017 World Health Organization Geneva 51 WHO Primary health care on the road to universal health coverage: 2019 global monitoring report 2019 WHO Geneva 52 Gichaga A Masis L Chandra A Palazuelos D Wakaba N Mind the global community health funding gap Glob Health Sci Pract 9 suppl 1 2021 S9 17 33727316 53 Barr A Garrett L Marten R Kadandale S Health sector fragmentation: three examples from Sierra Leone Global Health 15 2019 8 30670026 54 Nolte E Woldmann L Financing and reimbursement Amelung V Stein V Goodwin N Balicer R Nolte E Suter E Handbook integrated care 2021 Springer International Publishing Switzerland 341 364 55 Cuadrado C Fuentes-Garcia A Barros X Martinez MS Pacheco J Financing primary health care in Chile 2021 Lancet Global Health Commission on Financing People-Centred Primary Health Care London https://www.lshtm.ac.uk/financing-phc 56 Garg S Tripathi N Ranjan A Bebarta KK Comparing the average cost of outpatient care of public and for-profit private providers in India BMC Health Serv Res 21 2021 838 34407808 57 Pague B Alcantara AC Claims management using artificial intelligence: experiences from PhilHealth in the Philippines. iHEA 2021 congress: health economics in a time of global change; 2021; Virtual congress: iHEA https://cdn.ymaws.com/www.healtheconomics.org/resource/resmgr/2021_virtual_conference/program_pdf_s/ihea_2021_-_full_program_boo.pdf 2021 58 Rose AF Kumar A Perception and performance: impact of ASHA-soft on ASHA workers in rural field practice area of a medical college in central Karnataka Ann Comm Health 8 2020 14184 14188 59 Boaheng JM Amporfu E Ansong D Osei-Fosu AK Determinants of paying national health insurance premium with mobile phone in Ghana: a cross-sectional prospective study Int J Equity Health 18 2019 50 30909933 60 Burki T GP at hand: a digital revolution for health care provision? Lancet 394 2019 457 460 31402016 61 Beran D Pedersen HB Robertson J Noncommunicable diseases, access to essential medicines and universal health coverage Glob Health Action 12 2019 1670014 62 McVeigh T South Africa's latest weapon against HIV: street dispensers for antiretrovirals The Guardian July 17, 2016 https://www.theguardian.com/world/2016/jul/17/atms-dispense-antiretrovirals-south-africa-hiv 63 WHO The Abuja declaration: ten years on 2010 WHO Geneva 64 WHO Commission on Macroeconomics and Health Macroeconomics and health: investing in health for economic development/report of the commission on macroeconomics and health 2001 WHO Geneva 65 Jamison DT Summers LH Alleyne G Global health 2035: a world converging within a generation Lancet 382 2013 1898 1955 24309475 66 WHO Sustainable development goals: Health price tag https://www.who.int/news-room/q-a-detail/sustainable-development-goals-health-price-tag July 17, 2017 67 McIntyre D Meheus F Fiscal space for domestic funding of health and other social services 2014 Chatham House London 68 Jowett M Brunal MP Flores G Cylus J Spending targets for health: no magic number 2016 WHO Geneva 69 International Monetary Fund World Revenue Longitudinal Data (WoRLD) July 2021 ed. 2021 International Monetary Fund Washington, DC, WA 70 International Labour Organization World employment and social outlook 2020 International Labour Organization Geneva 71 Doherty J Kirigia D Okoli C Does expanding fiscal space lead to improved funding of the health sector in developing countries? Lessons from Kenya, Lagos State (Nigeria) and South Africa Glob Health Action 11 2018 1461338 72 World Bank High-performance health financing for universal health coverage: Driving sustainable, inclusive growth in the 21st century 2019 World Bank Washington, DC, WA 73 WHO The world health report: health systems financing: the path to universal coverage 2010 WHO Geneva 74 James CD Hanson K McPake B To retain or remove user fees? Reflections on the current debate in low- and middle-income countries Appl Health Econ Health Policy 5 2006 137 153 17132029 75 Gilson L McIntyre D Removing user fees for primary care in Africa: the need for careful action BMJ 331 2005 762 765 16195296 76 Yates R Universal health care and the removal of user fees Lancet 373 2009 2078 2081 19362359 77 The Abdul Latif Jameel Poverty Action Lab The impact of price on take-up and use of preventive health products https://www.povertyactionlab.org/policy-insight/impact-price-take-and-use-preventive-health-products 2018 78 Hanson K Worrall E Wiseman V Targeting services towards the poor: a review of targeting mechanisms and their effectiveness Bennett S Gilson L Mills A Health, Economic Development and Household Poverty: From Understanding to Action 2007 Routledge London 79 National Treasury of South Africa Budget review 2021 2021 South African National Treasury Pretoria 80 WHO Spending on health in Europe: entering a new era 2021 WHO Regional Office for Europe Copenhagen 81 Reich MR Harris J Ikegami N Moving towards universal health coverage: lessons from 11 country studies Lancet 387 2016 811 816 26299185 82 WHO Making fair choices on the path to universal health coverage: final report of the WHO consultative group on equity and universal health coverage 2014 WHO Geneva 83 Tandon A Cain J Kurowski C Dozol A Postolovska I From slippery slopes to steep hills: contrasting landscapes of economic growth and public spending for health Soc Sci Med 259 2020 113171 84 Besley T Persson T Why do developing countries tax so little? J Econ Perspect 28 2014 99 120 85 Organisation for Economic Co-operation and Development Government at a Glance 2021 2021 Organisation for Economic Co-operation and Development Paris 86 Cashin C Sparkes S Bloom D Earmarking for health: from theory to practice 2017 WHO Geneva 87 Elliott LM Dalglish SL Topp SM Health taxes on tobacco, alcohol, food and drinks in low- and middle-income countries: a scoping review of policy content, actors, process and context Int J Health Policy Manag 11 2020 1 15 88 Mathauer I Koch K Zita S Revenue-raising potential for universal health coverage in Benin, Mali, Mozambique, and Togo Bull World Health Organ 97 2019 620 630 31474775 89 Barroy H Sparkes S Dale E Mathonnat J Can low-and middle-income countries increase domestic fiscal space for health: a mixed-methods approach to assess possible sources of expansion Health Syst Ref 4 2018 214 226 90 Yazbeck AS Savedoff WD Hsiao WC The case against labor-tax-financed social health insurance for low- and low-middle-income countries Health Aff 39 2020 892 897 91 van Hees SGM O'Fallon T Hofker M Leaving no one behind? Social inclusion of health insurance in low- and middle-income countries: a systematic review Int J Equity Health 18 2019 134 31462303 92 Wagstaff A Social health insurance reexamined Health Economics 19 2010 503 517 19399789 93 Busse R Blümel M Knieps F Bärnighausen T Statutory health insurance in Germany: a health system shaped by 135 years of solidarity, self-governance, and competition Lancet 390 2017 8828982-9 8828982-17 94 Friebel R Josephson E Forman R Calza S Challenges of social health insurance in low- and lower-middle income countries: balancing limited budgets and pressure to provide universal health coverage https://www.cgdev.org/blog/challenges-social-health-insurance-low-and-lower-middle-income-countries-balancing-limited 2020 95 Baicker K Chandra A The labor market effects of rising health insurance premiums Journal of Labor Economics 24 2006 609 634 96 World Bank Walking the talk: reimagining PHC after COVID-19 2021 World Bank Washington, DC, WA 97 Moore M Obstacles to increasing tax revenues in low income countries 2015 Institute of Development Studies Brighton 98 Mills L Barriers to increasing tax revenue in developing countries 2017 Institute of Development Studies Brighton 99 Organisation for Economic Co-operation and Development OECD secretary-general tax report to G20 finance ministers and central bank governors–July, 2021 2021 Organisation for Economic Co-operation and Development Paris 100 Organisation for Economic Co-operation and Development Revenue Statistics in Africa 2020, Tunisia https://www.oecd.org/countries/tunisia/revenue-statistics-africa-tunisia.pdf 2020 101 Organisation for Economic Co-operation and Development Strengthening tax capacity to increase domestic resources in Tunisia 2021 Organisation for Economic Co-operation and Development Paris 102 International Monetary Fund Fiscal monitor: A fair shot 2021 International Monetary Fund Washington, DC, WA 103 Organisation for Economic Co-operation and Development The Paris declaration on aid effectiveness: five principles for smart aid 2005 Organisation for Economic Co-operation and Development Paris 104 Global financing facility Dossier d'investissement de la sante de la reproduction, de la mere, du nouveau-ne, de l'adolescent et de la nutrition (SRMNEA + N2019-2023) 2021 Global Financing Facility https://www.globalfinancingfacility.org/mali-investment-case-reproductive-maternal-child-adolescent-health-and-nutrition-2019-2023-french 105 Sparkes SP Bump JB Reich MR Political strategies for health reform in Turkey: extending veto point theory Health Systems Reform 1 2015 263 275 31519093 106 Baris E Mollahaliloglu S Aydin S Healthcare in Turkey: from laggard to leader BMJ 342 2011 c7456 107 Massuda A Malik AM Lotta G Siqiueira Tasca R Rocha R Brazil's PHC financing: case study 2021 Lancet Global Health Commission on financing people-centred PHC London (available at https://www.lshtm.ac.uk/financing-phc 108 Croke K The origins of Ethiopia's PHC expansion: the politics of state building and health system strengthening Health Pol Plan 35 2020 131 227 109 Mathauer I Digital technologies for health financing: what are the benefits and risks for UHC? Some initial reflections 2021 WHO Geneva 110 Organisation for Economic Co-operation and Development Tackling Wasteful Spending on Health 2017 Organisation for Economic Co-operation and Development Paris 111 London School of Hygiene and Tropical Medicine Technical team. Systematic review of inefficiencies in financing PHC https://www.lshtm.ac.uk/financing-phc 2021 112 World Bank Background paper for second annual UHC financing forum: Greater efficiency for better health and financial protection https://thedocs.worldbank.org/en/doc/5d7befa83cbafe469a1f9a5d591eb443-0140062021/related/Background-Paper-Second-Annual-UHC-Financing-Forum-FORUM.pdf 2017 113 Manthalu G Nkhoma D Kuyeli S Simple versus composite indicators of socioeconomic status in resource allocation formulae: the case of the district resource allocation formula in Malawi BMC Health Serv Res 10 2010 6 20053274 114 Chee HL Ownership, control, and contention: challenges for the future of healthcare in Malaysia Soc Sci Med 66 2008 214 256 115 WHO Building strong public financial management systems towards universal health coverage: key bottlenecks and lessons learnt Regional workshop on PFM for sustainable financing for health in Africa 2018 WHO Nairobi 116 Cashin C Bloom D Sparkes S Barroy H Kutzin J O'Dougherty S Aligning public financial management and health financing: sustaining progress toward universal health coverage 2017 WHO Geneva 117 Simson R Welham B Incredible budgets: budget credibility in theory and practice 2014 Overseas Development Institute London 118 Woldie M Resource mobilisation and allocation for PHCPHC: lessons from the Ethiopian health system. London: Lancet Global Health commission on financing people-centred PHC https://www.lshtm.ac.uk/financing-phc 119 Robinson M Last DP A basic model of performance-based budgeting 2009 International Monetary Fund Washington, DC, WA 120 Barroy H Blecher M Lakin J How to make budgets work for health: a practical guide to designing, managing and monitoring programme budgets in the health sector 2021 WHO Geneva 121 Abewe C Margini F Mwami E Mwoga J Kwesiga B Transition to programme budgeting in Uganda: status of the reform and prelimnary lessons for health 2021 WHO Geneva 122 Rajan D Barroy H Stenberg K Strategizing national health in the 21st century: a handbook. Chapter 8: budgeting for health 2016 WHO Geneva 123 Barroy H Dale EM Sparkes S Kutzin J Budget matters for health: key formulation and classification issues 2018 World Health Organization Geneva 124 Uzochukwu B Onwujekwe O Mbachu C Implementing the basic health care provision fund in Nigeria: a framework for accountability and good governance https://resyst.lshtm.ac.uk/resources/implementing-the-basic-health-care-provision-fund-in-nigeria-a-framework-for 2015 125 Prinja S Muraleedharan VR How effective has the central government been in nudging the states for financing PHC? An analysis of fiscal federal relations in India. London: Lancet Global Health Commission on financing people-centred PHC https://www.lshtm.ac.uk/financing-phcPHC 126 Gertler PJ Giovagnoli PI Martinez S Rewarding provider performance to enable a healthy start to life: evidence from Argentina's Plan Nacer. World Bank Policy Research Working Paper 2014 The World Bank Washington DC, WA 127 Gauthier B. PETS as a tool to improve accountability and transparency in public services. ODI 2020 Public Finance Conference; London; Feb 26, 2020. 128 Primary Health Care Performance Initiative Improvement strategies model: funds https://improvingphcPHC.org/sites/default/files/Funds%20last%20updated%2012.20.2019.pdf 2019 129 Goryakin Y Revill P Mirelman A Sweeney R Ochalek J Suhrcke M Public financial management and health service delivery, a literature review 2017 Overseas Development Institute London 130 NHS England Fair shares: a guide to NHS allocations 2020 NHS England and NHS Improvement London 131 Love-Koh J Mirelman A Suhrcke M Equity and economic evaluation of system-level health interventions: a case study of Brazil's Family Health Program Health Policy Plan 36 2021 229 238 33386400 132 Hone T Saraceni V Medina Coeli C Primary healthcare expansion and mortality in Brazil's urban poor: a cohort analysis of 1.2 million adults PLoS Med 17 2020 e1003357 133 Anselmi L Lagarde M Hanson K Equity in the allocation of public sector financial resources in low-and middle-income countries: a systematic literature review Health Pol Plan 30 2015 528 545 134 WHO United Nations Children's Fund. Lesotho: Public health sector expenditure review 2017 The World Bank Washington, DC, WA 135 Piatti-Fünfkirchen M Ally M Tanzania: health sector public expenditure review 2020 The World Bank Washington DC, USA 136 Chansa C Workie NW Piatti M Matsebula T Yoo KJ Zambia: health sector public expenditure review 2018 The World Bank Washington, DC, USA 137 Okorafor OA Thomas S Protecting resources for PHC under fiscal federalism: options for resource allocation Health Policy Plan 22 2007 415 426 18006526 138 O'Dougherty S Kutzin J Barroy H Piatti-Fünfkirchen M Margini F Direct facility financing: concept and role for UHC. 5th meeting of the Montreux collaborative on fiscal space, public financial management and health financing 2021 WHO Montreux 139 Opwora A Kabare M Molyneux S Goodman C Direct facility funding as a response to user fee reduction: implementation and perceived impact among Kenyan health centres and dispensaries Health Policy Plan 25 2010 406 418 20211967 140 Nigeria Federal Ministry of Health Nigeria state health investment project (NSHIP): project implementation manual National Primary Health Care Development Agency Nigeria: Adamawa State Primary Health CarePHC Development Agency; 2012 2012 Adamawa State Primary Health Care Development Agency Nigeria p3 11 141 Kapologwe NA Kalolo A Kibusi SM Understanding the implementation of Direct Health Facility Financing and its effect on health system performance in Tanzania: a non-controlled before and after mixed method study protocol Health Res Policy Syst 17 2019 11 30700308 142 Chee G Picillo B Strengths, challenges, and opportunities for RMNCH financing in Uganda 2019 United States Agency for International Development https://www.mcsprogram.org/resource/strengths-challenges-and-opportunities-for-rmnch-financing-in-uganda/ 143 Witter S Bertone M Diaconu K Review of Global Fund experience with facility-level financing. 5th Meeting of the Montreux Collaborative on Fiscal Space, Public Financial Management and Health Financing 2021 WHO Montreaux 144 Cheng T-M Reflections on the 20th anniversary of Taiwan's single-payer national health insurance system Health Affairs 34 2015 502 510 25732502 145 Xu J Yuan B The progress and challenges in China's health financing development towards integrated people-centred health services—a systems perspective 2021 Lancet Global Health commission on financing people-centred PHC London https://www.lshtm.ac.uk/financing-phcPHC 146 Alebachew A Hatt L Kukla M Monitoring and evaluating progress towards universal health coverage in Ethiopia PLoS Med 11 2014 e1001696 147 Dumontet M Buchmueller T Dourgnon P Jusot F Wittwer J Gatekeeping and the utilization of physician services in France: evidence on the Médecin traitant reform Health Policy 121 2017 675 682 28495205 148 Soucat A Financing common goods for health: fundamental for health, the foundation for UHC Health Syst Reform 5 2019 2632763 2632767 149 Sparkes SP Kutzin J Earle AJ Financing common goods for health: a country agenda Health Syst Reform 5 2019 322 333 31684816 150 Rechel B Maresso A Sagan A Organization and financing of public health services in Europe: country reports 2018 WHO Regional office for Europe Copenhagen 151 Flodgren G Eccles MP Shepperd S Scott A Parmelli E Beyer FR An overview of reviews evaluating the effectiveness of financial incentives in changing healthcare professional behaviours and patient outcomes Cochrane Database Syst Rev 7 2011 CD009255 152 Hipgrave DB Hort K Dual practice by doctors working in South and East Asia: a review of its origins, scope and impact, and the options for regulation Health Policy Plan 29 2014 703 716 24150504 153 WHO Global strategy on human resources for health: workforce 2030 2016 World Health Organization Geneva 154 Barasa E Mathauer I Kabia E How do healthcare providers respond to multiple funding flows? A conceptual framework and options to align them Health Policy Plan 36 2021 861 868 33948635 155 Cashin C Gubanova O Kadyrova N Primary health carePHC per capita payment systems Langenbrunner J Cashin C O'Dougherty S Designing and implementing health care provider payment systems: how-to manuals 2009 The World Bank Washington, DC, WA 156 Cashin C Ankhbayar B Phuong HT Assessing health provider payment systems: a practical guide for countries working toward universal health coverage https://www.jointlearningnetwork.org/wp-content/uploads/2019/11/JLN_ProviderPayment_MainGuide_InteractivePDF.pdf 157 Robinson JC Theory and practice in the design of physician payment incentives Milbank Q 79 2001 149 177 11439463 158 Lagarde M, Powell-Jackson T, Blaauw D. Managing incentives for health providers and patients in the move towards universal coverage: technical report. Global Symposium on Health Systems Research; Nov 9, 2010. Montreaux, Switzerland; 2010. Montreaux, Switzerland: First Global Symposium on Health Systems Research, 2010: p15–35. 159 Tan SY Melendez-Torres GJ Do prospective payment systems (PPSs) lead to desirable providers' incentives and patients' outcomes? A systematic review of evidence from developing countries Health Policy Plan 33 2018 137 153 29126109 160 Jia L Meng Q Scott A Yuan B Zhang L Payment methods for healthcare providers working in outpatient healthcare settings Cochrane Database Syst Rev 1 2021 CD011865 161 Roland M Olesen F Can pay for performance improve the quality of primary care? BMJ 354 2016 i4058 162 Roland M Dudley RA How financial and reputational incentives can be used to improve medical care Health Serv Res 50 2015 2090 2115 26573887 163 Rosenthal MB Dudley RA Pay-for-performance: will the latest payment trend improve care? JAMA 297 2007 740 744 17312294 164 Health Care Payment Learning and Action Network Alternative payment model: APM Framework: refreshed 2017. Health care payment learning & action network https://hcp-lan.org/apm-refresh-white-paper/ 2017 165 Langenbrunner J Cashin C O'Dougherty S Langenbrunner J Cashin C O'Dougherty S Designing and implementing health care provider payment systems: how-to manuals 2009 The World Bank Washington, DC, USA 166 Cattel D Eijkenaar F Value-based provider payment initiatives combining global payments with explicit quality incentives: a systematic review Med Care Res Rev 77 2020 511 537 31216945 167 Leutz WN Five laws for integrating medical and social services: Lessons from the United States and the United Kingdom Milbank Q 77 1999 77 110 10197028 168 North J Achieving Person-Centred Health Systems. Evidence, strategies, and challenges 2020 Cambridge University Press Cambridge 169 Conrad DA Vaughn M Grembowski D Marcus-Smith M Implementing value-based payment reform: a conceptual framework and case examples Med Care Res Rev 73 2016 437 457 26545852 170 de Bakker DH Struijs JN Baan CB Early results from adoption of bundled payment for diabetes care in the Netherlands show improvement in care coordination Health Aff 31 2012 426 433 171 Pimperl A Schulte T Mühlbacher A Evaluating the impact of an accountable care organization on population health: the quasi-experimental design of the german gesundes kinzigtal Popul Health Manag 20 2017 239 248 27565005 172 Maarse JAM Jeurissen PP The policy and politics of the 2015 long-term care reform in the Netherlands Health Policy 120 2016 241 245 26872702 173 Mason A Goddard M Weatherly H Chalkley M Integrating funds for health and social care: an evidence review J Health Serv Res Policy 20 2015 177 188 25595287 174 Diaconu K Falconer J Verbel A Fretheim A Witter S Paying for performance to improve the delivery of health interventions in low- and middle-income countries Cochrane Database Syst Rev 5 2021 CD007899 175 Mendelson A Kondo K Damberg C The effects of pay-for-performance programs on health, health care use, and processes of care: a systematic review Ann Intern Med 166 2017 341 353 28114600 176 Eijkenaar F Emmert M Scheppach M Schöffski O Effects of pay for performance in health care: a systematic review of systematic reviews Health Policy 110 2013 115 130 23380190 177 Campbell SM Reeves D Kontopantelis E Sibbald B Roland M Effects of pay for performance on the quality of primary care in England NEJM 361 2009 368 378 19625717 178 Mullen KJ Frank RG Rosenthal MB Can you get what you pay for? Pay-for-performance and the quality of healthcare providers RAND J Econ 41 2010 64 91 21667575 179 Steel N Maisey S Clark A Fleetcroft R Howe A Quality of clinical primary care and targeted incentive payments: an observational study Br J Gen Pract 57 2007 4494549-5 4494549-14 180 Holmstrom B Milgrom P Multitask principal-agent analyses: incentive contracts, asset ownership, and job design J Law Econ Organ 7 1991 242 252 181 Cashin C Chi Y-L Smith P Borowitz M Thomson S Paying for performance in health care. Implications for health system performance and accountability 2014 Open University Press England 182 Kasekamp K Habicht T The milestones of reforming PHC in Estonia 2021 Lancet Global Health Commission on Financing People-Centred PHC London https://www.lshtm.ac.uk/financing-phcPHC 183 Tangcharoensathien V Limwattananon S Patcharanarumol W Thammatacharee J Jongudomsuk P Sirilak S Achieving universal health coverage goals in Thailand: the vital role of strategic purchasing Health Policy Plan 30 2015 1152 1161 25378527 184 Mills A Bennett S Siriwanarangsun P Tangcharoensathien V The response of providers to capitation payment: a case-study from Thailand Health Policy 51 2000 163 180 10720686 185 Santos R Gravelle H Propper C Does quality affect patients' choice of doctor? Evidence from England Econ J 127 2017 445 494 186 Charlesworth A Davies A Dixon J Reforming payment for health care in Europe to achieve better value 2012 Nuffield Trust London 187 Cumming J Aotearoa New Zealand's PHC strategy: equity-enhancing in policy and in practice? 2021 Lancet Global Health commission on financing people-centred PHC London https://www.lshtm.ac.uk/financing-phcPHC 188 New Zealand Department of the Prime Minister and Cabinet. Our health and disability system: building a stronger health and disability system that delivers for all New Zealanders. Wellington: New Zealand, 2021. 189 Amporfu E Ghana's experience changing provider payment to capitation in PHC 2021 Lancet Global Health Commission on Financing People-Centred PHC London https://www.lshtm.ac.uk/financing-phcPHC 190 Barroy H Kabaniha G Boudreaux C Cammack T Bain N Leveraging public financial management for better health in Africa: key bottlenecks and opportunities for reform. 2019 2019 World Health Organization Geneva 191 Robyn PJ Bärnighausen T Souares A Provider payment methods and health worker motivation in community-based health insurance: a mixed-methods study Soc Sci Med 108 2014 223 236 24681326 192 Agyei-Baffour P Oppong R Boateng D Knowledge, perceptions and expectations of capitation payment system in a health insurance setting: a repeated survey of clients and health providers in Kumasi, Ghana BMC Public Health 13 2013 1220 193 Moseley GB 3rd The US health care non-system, 1908–2008 AMA J Ethics 10 2008 324 331 194 Hoffman C National health insurance. A brief history of reform efforts in the US 2009 The Henry J Kaiser Family Foundation Menlo Park, CA 195 Yi I Sohn H-S Kim T Linking state intervention and health equity differently: the universalization of health care in South Korea and Taiwan Korea Observer 46 2015 517 549 196 Barber S Luca L Paul O Price setting and price regulation in health care lessons for advancing universal health coverage: lessons for advancing universal health coverage 2019 WHO and OECD Geneva (accessed Dec 6, 2021). 197 Bump JB Sparkes SP A political economy analysis of Turkey's health transformation program 2014 World Bank Washington DC, WA 198 Wang H Otoo N Dsane-Selby L Ghana national health insurance scheme : improving financial sustainability based on expenditure review 2017 World Bank Washington, DC 199 National Health Insurance Authority NHIS Annual report 2009 http://www.nhis.gov.gh/files/1(1).pdf 2009 200 Gérvas J Fernández MP Cobos LP Sánchez RP Viente años de reforma de la Atención Primaria en España. Valoración para un aprendizaje por acierto/error: Ministerio Sanidad y Consumo http://equipocesca.org/new/wp-content/uploads/2010/08/20-anos-de-reforma-de-la-ap-en-espana.pdf 2005 201 Hsiao W Done N Implementation of social health insurance in Estonia: a case study 2009 World Bank Washington, DC 202 Dan S, Savi R. Payment systems and incentives in primary care in transition healthcare systems: Implications of recent reforms in Estonia and Romania. European Consortium for Political Research Graduate Conference; Bremen; July 12, 2012. 203 Pisani E Olivier Kok M Nugroho K Indonesia's road to universal health coverage: a political journey Health Pol Plan 32 2017 267 276 204 Hort K Gilbert K Basnayaka P Annear PL Strategies to strengthen referral from primary care to secondary care in low-and middle-income countries 2019 Regional Office for South-East Asia, World Health Organization Manila 205 Giuffrida A Jakab M Dale EM Toward universal coverage in health: the case of the state guaranteed benefit package of the Kyrgyz Republic 2013 World Bank Washington, DC 206 Reich MR Campos PA A guide to applied political analysis for health reform https://cdn1.sph.harvard.edu/wp-content/uploads/sites/2216/2020/08/Guide-Applied-Political-Analysis-final-2020.08.29-FINAL.pdf 2020 207 Cometto G Ford N Pfaffman-Zambruni J Health policy and system support to optimise community health worker programmes: an abridged WHO guideline Lancet Glob Health 6 2018 e1397 e1404 30430994 208 Perry HB Perry HB Health for the People: national community health programs from Afghanistan to Zimbabwe 2021 United States Agency for International Development Washington 209 Colvin CJ Hodgins S Perry HB Community health workers at the dawn of a new era: 8. Incentives and remuneration Health Res Policy Syst 19 2021 112 125 34380518 210 Gadsden T Mabunda SA Palagyi A Performance-based incentives and community health workers' outputs, a systematic review Bull World Health Organ 99 2021 805 818 34737473 211 Kok MC Dieleman M Taegtmeyer M Which intervention design factors influence performance of community health workers in low-and middle-income countries? A systematic review Health Policy Plan 30 2015 1207 1227 25500559 212 Ashraf N Bandiera O Jack BK No margin, no mission? A field experiment on incentives for public service delivery J Public Econ 120 2014 111 117 213 Bernal P Martinez S In-kind incentives and health worker performance: Experimental evidence from El Salvador J Health Econ 70 2020 102267 214 Pallas SW Minhas D Pérez-Escamilla R Taylor L Curry L Bradley EH Community health workers in low- and middle-income countries: what do we know about scaling up and sustainability? Am J Public Health 103 2013 e74 e82 215 WHO WHO guideline on health policy and system support to optimize community health worker programmes 2018 WHO Geneva 216 Gravelle H Sutton M Ma A Doctor Behaviour under a pay for performance contract: treating, cheating and case finding? Econ J 120 2010 129 156 217 Roland M Rao SR Sibbald B Professional values and reported behaviours of doctors in the USA and UK: quantitative survey BMJ Qual Saf 20 2011 515 521 218 Prang KH Maritz R Sabanovic H Dunt D Kelaher M Mechanisms and impact of public reporting on physicians and hospitals' performance: a systematic review (2000–2020) PLoS One 16 2021 e0247297 219 Gopalan SS Mutasa R Friedman J Das A Health sector demand-side financial incentives in low- and middle-income countries: a systematic review on demand- and supply-side effects Soc Sci Med 100 2014 72 83 24444841 220 Lagarde M Haines A Palmer N Conditional cash transfers for improving uptake of health interventions in low- and middle-income countries: a systematic review JAMA 298 2007 1900 1910 17954541 221 Coburn D Health and health care: a political economy perspective Bryant T Raphael D Rioux M Staying alive: Critical perspectives on health, illness, and health care Second edition ed. 2010 Canadian Scholars' Press Inc Toronto 65 91 222 Whiteslide H Capitalist political economy: thinkers and theories First edition 2020 Routledge London 223 Engels F Herrn Eugen Dühring's Umwälzung der Wissenschaft 1878 Genossenschafts-Buchdruckerei Leipzig 224 Campos P Reich M Political analysis for health policy implementation Health Syst Reform 5 2019 224 235 31390295 225 Sparkes SP Bump JB Özçelik EA Political economy analysis for health financing reform Health Syst Reform 5 2019 183 194 31369319 226 Roberts M Hsiao W Berman P Reich M Getting health reform right: a guide to improving performance and equity 2008 Oxford University Press New York, NY 227 Pierson P The new politics of the welfare state World Politics 48 1996 143 179 228 Croke K Mohd Yusoff MB Abdullah Z The political economy of health financing reform in Malaysia Health Policy Plan 34 2019 732 739 31563946 229 Hudson D Leftwich A From political economy to political analysis 2014 developmental leadership program Birmingham https://res.cloudinary.com/dlprog/image/upload/research-paper-25-from-political-economy-to-political-analysis 230 Shiffman J Political Context and Health Financing Reform Health Syst Reform 5 2019 257 259 31414931 231 VanderZanden A Pesec M Abrams M What does community-oriented PHC look like? Lessons from Costa Rica https://www.commonwealthfund.org/publications/case-study/2021/mar/community-oriented-primary-care-lessons-costa-rica 2021 232 Pesec M Ratcliffe H BItton A Building a thriving PHC system: the story of Costa Rica 2017 Ariadne Labs Boston 233 WHO PHC systems (PRIMASYS): case study from Ethiopia 2017 WHO Geneva 234 Witter S Brikci N Harris T The free healthcare initiative in Sierra Leone: evaluating a health system reform, 2010-2015 Int J Health Plann Manage 33 2018 434 448 29327367 235 Onoka CA Hanson K Hanefeld J Towards universal coverage: a policy analysis of the development of the National Health Insurance Scheme in Nigeria Health Policy and Planning 30 2014 1105 1117 25339634 236 Atun RA Menabde N Saluvere K Jesse M Habicht J Introducing a complex health innovation, primary health care reforms in Estonia (multimethods evaluation) Health Policy 79 2006 79 91 16406131 237 Patcharanarumol W Tangcharoensathien V Limwattananon S Why and how did Thailand achieve Good health at low cost? Balabanova D Ma MA Good health at low cost' 25 years on What makes a successful health system? 2011 London School of Hygiene & Tropical Medicines London 238 Heywood M South Africa's treatment action campaign: combining law and social mobilization to realize the right to health J Hum Rights Prac 1 2009 14 36 239 WHO Primary health carePHC systems (PRIMASYS): case study from Rwanda, abridged version 2017 WHO Geneva 240 WHO Resolution WHA A68/17: Contributing to social and economic development: sustainable action across sectors to improve health and health equity 2015 WHO Geneva 241 Guerra E Citizenship knows no age: children's participation in the governance and municipal budget of Barra Mansa, Brazil Environ Urban 14 2002 2 242 Light DW Universal health care: lessons from the British experience Am J Public Health 93 2003 25 30 12511379 243 Loewe M Jawad R Introducing social protection in the Middle East and North Africa: prospects for a new social contract? Int Soc Secur Rev 71 2018 3 18 244 Blecher M Pillay A Patcharanarumol W Health financing lessons from Thailand for South Africa on the path towards universal health coverage S Afr Med J 106 2016 4 5 245 WHO Country case studies on primary health carePHC: Estonia: the development of family practice to support universal health coverage 2018 World Health Organization Geneva 246 Witter S Anderson I Annear P What, why and how do health systems learn from one another? Insights from eight low- and middle-income country case studies Health Res Policy Syst 17 2019 9 30665412 247 Meng Q Mills A Wang L Han Q What can we learn from China's health system reform? BMJ 365 2019 l2349 248 Espinosa-González AB Normand C Challenges in the implementation of PHC reforms: a qualitative analysis of stakeholders' views in Turkey BMJ Open 9 2019 e027492 249 Onoka CA Onwujekwe OE Uzochukwu BS Ezumah NN Promoting universal financial protection: constraints and enabling factors in scaling-up coverage with social health insurance in Nigeria Health Res Policy Syst 11 2013 20 23764306 250 Chathukulam J Tharamangalam J The Kerala model in the time of COVID19: rethinking state, society, and democracy World Dev 137 2021 105207 251 The National Archives Originis of the NHS https://www.nationalarchives.gov.uk/cabinetpapers/alevelstudies/origins-nhs.htm 2021 252 Pesec M VanderZanden A Ratcliffe H Integrated people-centred health services case study: comprehensive PHC reform in Costa Rica https://www.integratedcare4people.org/media/files/Comprehensive_Primary_Health_Care_Reform_in_Costa_Rica_January2020_.pdf 2020 253 Yu H Universal health insurance coverage for 1·3 billion people: What accounts for China's success? Health Policy 119 2015 1145 1152 26251322 254 Bump JB Baum F Sakornsin M Yates R Hofman K Political economy of covid-19: Extractive, regressive, competitive BMJ 372 2021 372 373 255 Rifkin SB Fort M Patcharanarumol W Tangcharoensathien V Primary healthcare in the time of COVID-19: breaking the silos of healthcare provision BMJ Glob Health 6 2021 e007721 256 Keskimäki I Development of primary health care PHC in Finland 2021 Lancet Global Health Commission on financing people-centred PHC London https://www.lshtm.ac.uk/financing-phcPHC 257 Otoo N Awittor E Marquez P Saleh K Universal Health Coverage for Inclusive and Sustainable Development : country summary report for Ghana 2014 World Bank Group Washington, DC 258 Barnett A Stockbridge M KIngsmill W Political economy of Africa's power sector. Policy Practice Brief 10, 2016 https://www.thepolicypractice.com/sites/default/files/202020020-0-5/pb10pe_africa_power_sector.pdf
PMC009xxxxxx/PMC9005909.txt	==== Front J. Comput. Educ. Journal of Computers in Education 2197-9987 2197-9995 Springer Berlin Heidelberg Berlin/Heidelberg 231 10.1007/s40692-022-00231-1 Article Evaluation of a mobile-based scaffolding board game developed by scaffolding-based game editor: analysis of learners’ performance, anxiety and behavior patterns http://orcid.org/0000-0003-1783-8830 Hou Huei-Tse hthou@mail.ntust.edu.tw Huei-Tse Hou is a Distinguished Professor of Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Taiwan. He is also the Director of Mini-Educational Game development Group in E-learning Research Center in National Taiwan University of Science and Technology (NTUST MEG). His research interest focuses on e-Learning systems, behavioral pattern analysis and game-based learning systems. Wu Chung-Sheng Chung-Sheng Wu is a master student of National Taiwan University of Science and Technology. His research interests include game-based learning and educational board game design. Wu Chang-Hsin Chang-Hsin Wu is a PHD student of National Taiwan University of Science and Technology. His research interests include game-based learning and educational board game design. grid.45907.3f 0000 0000 9744 5137 Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, #43 Keelung Road, Section 4, Taipei, Taiwan 13 4 2022 2023 10 2 273291 12 8 2021 23 2 2022 25 3 2022 © Beijing Normal University 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Game-based learning with scaffolding is expected to provide learners with an effective and positive learning environment. This study developed a scaffold-oriented educational game editor that allows teachers to design educational game activities that combine physical board game cards. Players could obtain mobile-based scaffolding by manipulating and scanning the cards. We used the editor to develop a game activity, "Return," for a high school chemistry course, and conducted a preliminary empirical evaluation of the mobile-based scaffolding game. Participants were students in a high school in northern Taiwan. The study analyzed the learning effectiveness, anxiety level, and learning behavior patterns of the learners. Results showed that learners' learning effectiveness improved significantly, and their anxiety level decreased after using the game. Analysis of learning behavior patterns revealed that learners were able to fully utilize the mobile-based cognitive scaffolding and real-time feedback provided in the game to try to combine various hidden clues to solve problems. Keywords Game-based learning Educational game editor Scaffolding Learning anxiety Board games http://dx.doi.org/10.13039/501100004663 Ministry of Science and Technology, Taiwan MOST-108-2511-H-011 -003 -MY3 MOST-107-2511-H-011 -003 -MY3 MOST-110-2511-H-011 -004 -MY3 Hou Huei-Tse issue-copyright-statement© Beijing Normal University 2023 ==== Body pmcIntroduction Studies have pointed out that a one-way lecture-oriented teaching method may curb students’ interest in learning and lead to low learning motivation (Lee & Brophy, 1996). When students continue to lack motivation for learning or have excessive anxiety about learning, it may have a negative impact on their cognition and emotion (McKeachie, 1984; Vitasari et al., 2010). Previous research has found that the use of games facilitates motivation (Bawa et al., 2018; McLaren et al., 2017) and scaffolding facilitates cognitive thinking (Hou & Keng, 2021; Maryam et al., 2020). Therefore, designing a teaching activity that integrates game mechanics and scaffolding is expected to simultaneously promote learner motivation, reduce anxiety, and enhance learning effectiveness. In this section, this study discuss the design of scaffolding-oriented game-based instructional activities from the literature review and present the research objectives and questions of this study. Learning anxiety and game-based learning Learning anxiety can make learners feel nervous, challenged, or scared, and it can therefore affect their learning performance (Ashcraft, 2002). Learners' anxiety about the subject matter affects the learning process. It is one of the important factors that cause students’ low self-efficacy (Britner & Pajares, 2006; Lopez & Lent, 1992). For example, Gil-Doménech and Berbegal-Mirabent (2019) found that when teaching in a traditional lecture-oriented way, students showed negative attitudes towards the course and had a low degree of participation. Results of their research indicated that the combination of game-based learning allowed students to generate learning interest and motivation. Combining game-based learning can not only improve learners' learning motivation and learning effectiveness (Bawa et al., 2018), but can also reduce their cognitive load (Chang et al., 2018). Some studies have found that game-based learning is one of the effective ways to improve students' self-efficacy and reduce their learning anxiety (Chow & Yong, 2013; Meluso et al., 2012). Many research results have shown that teaching with game-based learning strategies can improve learners’ learning motivation and learning effects (Bawa et al., 2018; McLaren et al., 2017). Games allow learners to revise learning strategies through independent learning and repeated practice which can cultivate their creativity, problem-solving skills and higher level thinking skills (Hsieh et al., 2015; Hwang et al., 2012; Kim & Chang, 2010). Many research results have indicated that digital game-based learning is more able to improve learning outcomes than traditional teaching (Byun & Joung, 2018; Chang et al., 2018). In addition, the design quality of the scaffolding is a key research issue to promote the cognitive thinking of the learners in game-based learning. Scaffolding design for game-based learning The purpose of scaffolding is to provide temporary support while learners solve problems, but also to help gain skills to be able to solve problems independently in the future (Collins et al., 1989; Wood et al., 1976). Appropriate scaffolding design can improve learners' learning motivation and learning effectiveness (Maryam et al., 2020). Puntambekar and Hubscher (2005) believed that scaffolding has two goals: (1) To provide temporary support when learners encounter problems, and (2) To improve learners' own abilities. Therefore, the scaffold design can guide learners of different levels to break through difficulties when the learning process is blocked, so that they will be able to continue learning. Digital games can provide scaffolding and real-time evaluation functions. The game can provide real-time diagnosis based on learners' various operation records. The game can give feedback based on the diagnosis result as a form of scaffolding to help them acquire the correct knowledge to improve their abilities and produce better learning effects (Shute, 2008). Studies have also proposed a dual scaffolding educational game architecture that combines cognitive scaffolding and peer scaffolding (Hou & Keng, 2021). In the era of mobile learning, digital games can provide timely scaffolding through mobile vehicles. Learners with different prior knowledge can be provided with adaptive assistance when facing different learning tasks in the game. These scaffolds may reduce learning anxiety when learners are faced with challenging tasks, and can promote the balance between challenges and skills. The balance of challenges and skills can help learners reach a state of flow (Kiili, 2006). Previous research has also found that scaffolding can help students successfully reflect on and correct their mistakes, which in turn effectively reduces student anxiety (Mitchell et al., 2017; Kusmaryono et al., 2020), and that scaffolding can increase student motivation and learning outcomes in collaborative game-based learning environments (Chen & Law, 2016). However, many studies have found that digital game-based learning still has its limitations, including the lack of face-to-face peer-to-peer interaction and collaboration (Hwang et al., 2012; Lopez & Caceres, 2010). Board games emphasize face-to-face peer-to-peer interaction. Many studies have also found the positive effects of board game-assisted teaching (Hou & Keng, 2021; Kuo et al., 2020; Li et al., 2018). Board games can adapt to the learning content of various subjects, using cards, boards, and simple accessories, making it easy for teachers to design and use them in their teaching practice. Even teachers who do not have a background in information technology can create their own card game activities to match the course teaching. Therefore, many scholars have adopted board games as a learning aid tool, and have actively invested in the research of board game-oriented teaching (e.g., Kuo et al., 2020; Martín-Lara & Calero, 2020; Wang et al., 2019). However, the information content that can be presented by the graphics cards of board games is limited, and it is difficult for board games to present learners with more multimedia information in an adaptive manner as can be done with digital games. Mobile-based scaffolding for educational game Therefore, if we can combine the features of face-to-face peer-to-peer interaction in tabletop games with the real-time scaffolding of mobile technology (e.g., augmented reality or image scanning and recognition of cards; e.g., Hou & Keng, 2021; Wu et al., 2018), we may be able to promote learners’ collaborative learning motivation through face-to-face game mechanisms, and reduce the anxiety caused by learning challenges through the scaffolding provided by mobile technology. Therefore, the purpose of this study is to investigate the learning effectiveness and learning process of an educational board game using mobile technology to provide scaffolding. However, without the expertise of information software or the use of specific editing tools, it is difficult for teachers to design the above-mentioned board games with mobile-based scaffolding. In addition, the general way of scanning card images to present information on mobile devices (such as common QR-codes) can only be used to scan a single card image to display the corresponding information. The existing software cannot perform operations such as combining, sorting, and matching the scaffold information obtained after scanning the images on multiple different cards. These operations can provide more dimensional scaffolding, can be used to evaluate the learning process and status of learners at the same time, and can provide learners with more appropriate and real-time guidance. Therefore, in this research we also developed an editing tool, the Mobile-Scaffolding-based Educational Card Game Editor (MSECGE), which allows teachers to easily design mobile-based scaffolding card educational games without programming skills. Research purpose and research questions To evaluate the mobile scaffolding educational game framework proposed by this research, this research evaluates the educational game by implementing an empirical analysis. The researcher and a high school chemistry teacher used the editor to design a chemistry education board game, and actually used this game to implement game-based learning activities in a high school chemistry course. This research developed an educational game editing tool for teachers to design an educational game with real-time mobile scaffolding, and use it to design a chemistry educational game that combines cognitive scaffolding, a real-time feedback mechanism, and mobile technology. The game is called "Return," and we analyzed learners' learning effectiveness, learning anxiety (Anxiety about studying chemistry) and learning behavior patterns while playing the game. As for the learning behaviors, sequential analysis (e.g., Hou, 2012, 2013, 2015) was used to analyze a large number of learners' operation records to understand their problem-solving behavior patterns in the game. To explore and evaluate learners' learning effectiveness, anxiety, and behavior patterns in depth, the research questions of this study are as follows:After the mobile scaffolding game-based learning activity, did the learners’ learning effectiveness improve significantly? After the mobile scaffolding game-based learning activity, was there a significant reduction in learners' learning anxiety? What is the overall behavioral pattern of learners in the mobile scaffolding game-based learning activity? Educational board game with mobile-based scaffolding and educational game editor The proposed mobile scaffolding-oriented board game and its editing tool framework are shown in Fig. 1, the educational game is based on scaffolding design principles and cognitive scaffolding theory (Brush & Saye, 2002; Puntambekar & Hubscher, 2005; Shin et al., 2020; Wood et al., 1976). Teachers can quickly set up the appropriate cognitive scaffolding mechanism, multimedia content and instant feedback mechanism, which can be promoted and widely used in the teaching curriculum of various subjects. The Augmented reality (AR) module in this system can scan and recognize some specific image stickers to allow users to obtain specific scaffolding content. Teachers can use this module to set the multimedia content that can be obtained after scanning each sticker scaffold. Teachers can paste image stickers on the board game cards designed by the teachers to allow learners to manipulate, scan, check, assemble, and manage scaffold information to expand the content of pure board game card information. The development of this editing tool aims to allow teachers to design more educational board games that can promote learning effectiveness and reduce learning anxiety. In addition, when using the board game developed by this tool for assisted teaching, the system has an automatic learning action recording function. This function can help teachers conduct formative evaluation (Rowe et al., 2017) to explore the learning process in depth, and it can assist in adjusting teaching strategies and revising the content of teaching materials in a timely manner. Initial tests in the pilot studies (Li, Huang, et al., 2021; Li, Lee, et al., 2021) revealed that there was initial engagement and acceptance of the game activities completed by this editor, but the pilot studies did not describe the theoretical framework and did not conduct a multidimensional analysis of learners' anxiety, learning effectiveness, and learning behavior patterns. Therefore, this study not only completely revised and presented the theoretical framework of mobile scaffolding, but also evaluated the learning outcomes in multiple dimensions.Fig. 1 Framework of educational board game with mobile-based scaffolding and educational game editor As shown in Fig. 1, the tool is divided into two applications, the editor and the player, in which users can use the game editor on a personal computer (PC) to set the scaffolding structure in the game. The designer can set those images of stickers to be scanned, the player can scan these stickers and decide whether to collect them in the application (app), and by clicking, combining, or sorting these stickers, they can get various kinds of information content as scaffolding to complete the game tasks in the board game. Designers can also design the content of these cognitive scaffolds (including text, pictures, audio and video, multimedia links.). The designer will use the cards with these images as part of the classroom gamification activities or tabletop games. The designed scaffolding structure will be stored in the cloud, and a link and QR code will be generated for players to scan the code in the player app to get the game-specific scaffolding data. When playing the game, players can use the player app on a mobile device. In the game, players can use the app to scan the cards and get the corresponding scaffolding stickers. After scanning the cards once, players can click on the stickers, or combine and sort the stickers to obtain and read the scaffolding. For example, in Fig. 2, players can combine three cards (two suspect stickers and one police car sticker) to check and obtain intelligence or check if the case was correctly solved. These scaffolds allow players to access additional learning content or guide students to think in more diverse or deeper ways.Fig. 2 Players manage and use the collected scaffolding stickers In order to evaluate the game and explore the learning process of the players, we designed an educational game named "Return." The game was created using the abovementioned educational game editing tool, MSECGE. This game can be used as a card game teaching activity for chemistry teaching. The context, objectives, procedures and scaffolding of the game are as follows:Game scenario: Learners take on the role of participants in a hot air balloon tournament and go through pre-training. In order to help the participants reach the finish line, return safely to the starting point, and lift the balloon correctly to win the competition, the activity officials provided a board game package called "Return" to help the learners understand the concepts of gas and pressure gauges beforehand. Game procedure and objective: Players must use the mobile devices, cards, tokens, documents, and other accessories in the provided table game package (as shown in Fig. 3) to solve the various tasks in the game in a group collaboration. Players in each group were first given a certain amount of game points, as shown in Fig. 3, and had to read the game story in the Game App and solve the game tasks within a limited time. They can discuss and analyze the puzzles in the game tasks by reading the information in the given game book and using the tables for analyzing cues in the game book. The game points can be exchanged for hint cards. Players can collect scaffolding hints and think about the answers, and scan and combine the correct answer cards to solve the puzzles.Fig. 3 The procedure of mobile scaffold-oriented educational game—"Return" Scaffolding structure: This game is based on Hou and Keng's (2021) dual scaffolding framework for AR educational board games, and provides both peer scaffolding and cognitive scaffolding. Of the two types of scaffolding, peer scaffolding is based on Nelson's (1999) collaborative problem-solving teaching method, which provides tasks that groups need to discuss collaboratively with limited time and resources. To solve problems, groups must discuss clues and make decisions. Peers need to provide each other with more information and perspectives to serve as peer scaffolds. The cognitive scaffolding is set by the designer in the editor; players can scan hint cards to gain knowledge and clues to help solve the problem, and can also scan and correctly combine answer cards to instantly check whether the answer is correct, helping learners to get immediate feedback to assist in learning the concepts and knowledge related to gas and pressure gauges. The above two types of card scanning and combination can be used as cognitive scaffolding. Figure 4 shows a player playing a game and scanning a hint card to obtain a cognitive scaffold.Fig. 4 A player is playing the game and scanning a hint card to get cognitive scaffolding Research methodology Because this study was affected by the Covid-19 pandemic, the limited time and classroom arrangements were sufficient to conduct only the experimental group analysis without a control group. In order to ensure the overall validity of the study, a behavioral pattern analysis of the behavioral history was conducted for cross-checking to improve the overall validity of the study. Therefore, the research design of this case study is a one group pretest and posttest design. The study was conducted by conducting a preliminary evaluation of the learning effectiveness, anxiety and behavioral patterns of players in a game developed using the proposed editor. Participants In this study, a high school sophomore class of 31 students (20 boys and 11 girls; average age = 16.5 years old, youngest = 16, oldest = 17) in northern Taiwan were the participants. These students had not previously learned the concept of gas and pressure gauges, and had never played similar educational games. The game was a collaborative problem-solving educational game, so we divided the students into pairs and conducted a collaborative problem-solving game. Participants were grouped into 16 pairs. (Originally, there were 32 participants, but one of them was classified as an invalid sample because of incomplete participation records and completed scale information). Research instruments The instruments used in the study included a pre- and post-test evaluation of learning effectiveness, anxiety questionnaires, and a coding scheme of behavior patterns. Learning effectiveness test In order to ensure the validity of the questions, the content was designed based on the high school chemistry curriculum standards in Taiwan, and the questions were reviewed by high school chemistry teachers and experts in the field of learning science to ensure expert validity. The scope of the questions was based on the second grade chemistry unit "Gases," and the total number of questions was 20, with 1 point per question. The content of the pretest and posttest was the same. (2) Anxiety Questionnaire The anxiety scale used in this study was referenced and adapted from the mAMAS Anxiety Scale (Modified Abbreviated Math Anxiety Scale), which is an anxiety scale developed by Carey et al. (2017) based on the AMAS developed by previous scholars (Hopko et al., 2003). It was modified to examine the anxiety level of the subject (Chemistry in this study). In order to allow learners to compare their pre-game and post-game anxiety about academic subjects, the anxiety questionnaire in this study was administered after the game with nine questions, each of which asked students about their pre-game anxiety and post-game anxiety (both on a five-point scale). In terms of reliability, the overall reliability was .868 (Cronbach’s α=.868) after analyzing the questionnaires completed by the learners in this study, which indicates a high degree of internal consistency. (3) Behavior pattern analysis coding scheme In order to understand learners' behavior in the game, each mobile device will automatically record each of the learner’s actions. The coding schema in this study (Table 1) was modified based on the coding schema of existing studies on game-based learning or learning behaviors (Hou, 2015; Cheng & Tsai, 2014; Hou & Keng, 2021; Chou et al., in press). Table 1 Behavior pattern analysis coding scheme Code Behavior Details GC Receives scaffold stickers Scans stickers on cue cards and adds them to the player's sticker library GA Gets answer sticker Scans the sticker on the answer card and adds it to the player's sticker library SC Selects scaffold stickers Selects the scaffold stickers the player wants to use SA Selects answer stickers Selects the answer stickers that the player wants to use CS Cancels the selected stickers Cancels the selected stickers CI Obtains scaffolding information successfully Triggers and obtains scaffolding information successfully AI Success in using answer stickers Success in triggering the correct answer FI Uses sticker—no response No trigger information Procedure In this study, a set of board games and mobile devices was used for each pair of students. The study procedure is shown in Table 2, and the process took approximately 100 min. Before the experimental activity started, the purpose and schedule of the experiment were explained to the participants, and they were asked to fill out an informed consent form and then take a pretest of learning effectiveness. The researcher then proceeded to the activity description and game configuration, during which students could ask questions immediately if they had any. Afterwards, participants started the game activities. After the game, they filled out the anxiety questionnaire and finally took the learning effectiveness posttest. There was a certain length of instructional activity time (i.e., 65 min) between the pre and post-tests (Table 3).Table 2 Research procedure Steps Item Time (min) Step1 Fill out the consent form for the test 5 Step2 Conduct the learning effectiveness assessment (pre-test) 15 Step3 Set up the game and explain the game 10 Step4 Conducting the game activity 40 Step5 Fill in the anxiety questionnaire (pre- and post-game anxiety comparison) 15 Step6 Conduct the learning effectiveness assessment (post-test) 15 Tota 100 Table 3 Results of learning effectiveness Variable Pretest (N = 31) Posttest (N = 31) t p Cohen’s d M SD M SD Pre-post- Posttest 3.29 2.81 4.52 3.03 − 3.26 0.003 0.586 p < 0.01 Research findings Learning effectiveness In terms of learning effectiveness, the results of the dependent sample t test are shown in Table 4, with a mean score of 3.29 and a standard deviation of 2.81 for the pretest of learning effectiveness, and a mean score of 4.52 and a standard deviation of 3.03 for the posttest of learning effectiveness knowledge. It was found that the posttest scores of learning effectiveness were significantly higher than the pretest (t = − 3.26, p < 0.01), indicating that the learners had achieved significant improvement in gas and manometer-related knowledge after learning with the educational game, "Return," designed for this study.Table 4 Learners' anxiety about chemistry learning before and after using the game Variable Pre-game anxiety (N = 31) Post-game anxiety (N = 31) M SD M SD t p Cohen’s d Pre-anxiety-post-anxiety 26.32 7.73 23.3 8.18 3.02 0.005** 0.544 p < 0.05, p < 0.01 Learning anxiety In terms of learners' anxiety about chemistry learning before and after using the game, the study used the dependent sample t test to compare learners' anxiety before and after the game. The results are shown in Table 4. The mean score of anxiety before the activity was 26.32 with a standard deviation of 7.73, and the mean score of anxiety after the activity was 23.38 with a standard deviation of 8.18. The results indicated that the learners' anxiety level in chemistry learning decreased significantly after learning with the educational game, "Result," designed for this study (t = 3.02, p < 0.01). Analysis of overall learner behavior patterns In this section, we report the analysis of the behaviors of each group of learners during the use of the game, "Return." We used lag sequential analysis, which is commonly used to analyze learning behaviors, to analyze the patterns of learning behaviors. The data analysis step of this analysis method consists of calculating (1) behavior transfer frequency matrix (2) behavior transfer conditional probability matrix (3) behavior transfer expectation matrix (4) behavior transfer residual (z-score) matrix. In the z-score matrix, a z value greater than 1.96 indicates a statistically significant (p < 0.05) continuity from the behavior of the "row" to the behavior of the "column" in the table (Bakeman & Gottman, 1997; Hou, 2012). Based on the coding results, the significant sequences with z-values greater than 1.96 (p < 0.05) (Table 5) were used to map the overall learner behavior patterns (Fig. 5). This study also illustrates and discusses the key behavioral sequences in the overall learners' main learning and problem-solving behavioral patterns. In particular, these behavioral patterns are used to understand the behavior, manner, and timing of the learner's use of scaffolding.Table 5 Sequential analysis of all learners (Z-value table) Z GC GA SC SA CS CI AI FI GC 12.36 − 3.51 7.26* − 3.56 − 2.05 − 2.01 − 0.87 − 2.77 GA − 1.95 24.68* − 11.56 3.82* − 5.91 − 5.79 − 2.52 − 7.98 SC − 3.65 − 11.67 4.28* − 12.85 11.36* 19.60* − 2.91 2.78* SA − 3.47 − 11.09 − 12.61 12.23* 0.89 − 6.36 9.17* 16.39* CS − 0.59 − 4.62 10.46* − 0.31 − 0.96 − 2.61 − 1.44 − 4.03 CI 7.67* − 1.79 11.12* − 4.45 − 3.29 − 3.22 − 1.40 − 4.43 AI 2.20* 6.02* − 2.67 − 0.45 − 1.33 − 1.30 − 0.57 − 1.80 FI − 0.97 4.39* 4.56* 2.28* − 3.99 − 4.44 − 1.93 − 6.13 p < 0.05 Based on the sequential analysis (see Table 5), the significant sequences with z-values greater than 1.96 (p < 0.05) (numbers in bold in Table 5) Fig. 5 Sequential behavior pattern of all learners In the overall learner behavior pattern diagram, the sequences of GC → GC, GC → SC, GA → GA, GA → SA, SC → SC, SC → CS, SC → CI, SC → FI, SA → SA, SA → AI, SA → FI, CS → SC, CI → GC, CI → SC, AI → GC, AI → GA, FI → GA, FI → SC, FI → SA are all related to the learning and problem-solving processes of learners during the game. In the behavior mode, GC → GC, GC → SC, SC → CI, and CI → GC is a series of sequences in which players get hints for scaffolding. Learners obtain scaffolding stickers by continuously exchanging hint cards for scanning, and use these stickers to assemble corresponding hints to learn related knowledge. These behavioral patterns also illustrate the process by which learners select and use scaffolds to think about these cues to learn the correct knowledge to facilitate problem solving. The SC → SC, SC → CS, CS → SC, SC → CI, CI → SC, SC → FI, and FI → SC in the overall behavior mode are all related to the behavior code SC (Select Scaffold Sticker). Learners continuously select and deselect scaffolding stickers (SC → CS, CS → SC) to try to obtain more clues, and they continue to try multiple times even though the attempts fail to yield valid information. The above behavioral pattern also illustrates that the SC (scaffolding selection) behavior intersects with other problem-solving behaviors, indicating that learners may repeatedly use scaffolding to learn knowledge. The GA → GA, GA → SA, SA → AI, and AI → GA behaviors show that learners obtain answer stickers by continuously scanning the answer cards, and use them to try to assemble the answers to the levels, while the SA → SA, SA → AI, SA → FI, and FI → SA behaviors show that learners continuously select answer stickers to try to solve the levels. The AI → GC behavior mode shows that after solving the puzzle, the learner continues to redeem other hint stickers to obtain hints to continue solving the next level. Discussion In terms of learning effectiveness, learners' posttest of learning effectiveness was significantly higher than their pretest of learning effectiveness after using the "Return" game, indicating that the game led to a significant improvement in learning effectiveness, and had a positive impact on learners' learning of chemistry knowledge. Appropriate scaffolding design can enhance learners' motivation and learning effectiveness (Maryam et al., 2020). This game was designed using the mobile scaffolding architecture and editor proposed in our research, using mobile online scaffolding with a physical board game. This study makes reference to the dual scaffolding structure (Hou & Keng, 2021), and considers that the peer scaffolding provided by the board game and the cognitive scaffolding provided by the mobile scaffolding should have a positive effect on the learning effectiveness, at least to a certain extent. Mobile scaffolding is also more dynamic than cognitive scaffolding of paper cards in general board games, and can be combined with multimedia. In addition, mobile devices as scaffolding providers can be more portable than desktop computers and can be easily implemented in general classrooms. In terms of anxiety, learners' anxiety about chemistry after the "Return" game experience was significantly lower than their anxiety before the activity, representing a significant reduction in anxiety about the subject. Studies have found that the use of game-based learning can increase student self-efficacy and reduce their anxiety (Chow & Yong, 2013; Meluso et al., 2012; Young & Wang, 2014). However, the mechanisms of anxiety reduction have been less explored. In this study, we further propose the use of an immediate online cognitive scaffolding design to try to reduce learners' anxiety when they encounter excessive learning challenges. Scaffolds are helpful in alleviating learning anxiety (Kusmaryono et al., 2020; Mitchell et al., 2017), and the scaffolds in this study provide information about problem solving, which are key cues for learners when they experience difficulty solving problems These cues are key hints for learners when they encounter problem-solving difficulties. When learners actively explore and collect the scaffolds, they may be able to alleviate their learning anxiety when they experience frustration in problem solving. In terms of behavioral analysis, this study conducted behavioral pattern analysis through learners' operation process in the game. The behavioral pattern analysis revealed that learners were able to fully utilize the cognitive scaffold and real-time feedback function provided in the game to try to combine various hidden clues to solve the problem. When the answer was wrong, the behavioral pattern also showed that the players would keep actively searching for the unobtained hints to reflect and correct the solution strategy, and further explore and learn in the game. Using this editor, teachers can design dynamic cognitive scaffolding cues to facilitate students' learning. From the above behavior patterns in this case study (e.g., GC → GC, GC → SC, SC → CI, and CI → GC), we can initially understand that learners, with the help of the game's scaffolding, may have a certain degree of repeatedly choosing, correcting, and combining behaviors to solve problems and crack levels. A previous study also found positive effects of scaffolding in educational games on learners' sequential behavior patterns (Hou & Keng, 2021). The above behavioral patterns in this study also illustrate the potential effectiveness of the proposed scaffolding framework in terms of promoting learning effectiveness and reducing anxiety during the learning process. In addition, previous research found that learners' problem-solving sequences in games with high levels of flow showed more reflective processes (Hou, 2015), and the design of scaffolding based on flow elements is a future topic worth exploring to promote a deeper learning process. Conclusion and Recommendations This study investigated the learning effectiveness and learning process of an educational board game using the proposed mobile-based scaffolding framework. This study also developed an educational game editing tool for teachers to design educational games with mobile scaffolding, and used it to design an educational game for the subject of chemistry with cognitive scaffolding and a real-time checking mechanism. The study analyzed learners' learning effectiveness, learning anxiety, and learning behavior patterns. It was found that the game significantly improved students' learning effectiveness and reduced their anxiety about chemistry. The behavioral pattern was also found to have a scaffolding, reflection, and problem-solving learning process during the game. The following recommendations are made for teachers, educational game developers, and future research:In teaching practice: The scaffolding framework and editor proposed in this study can be a reference for teachers to design relevant educational games. At present, although there are educational games that integrate scaffolding (e.g., Hou & Keng, 2021; Wu et al., 2018), mobile educational game editors that integrate scaffolding editing functions are still limited. It is suggested that design tools can provide not only learning content, but also scaffolding based on cognitive theory. Formative assessment in games (Rowe et al., 2017) is gaining importance, and scaffolding or diagnostics can be useful for formative assessment. The tools in this study can provide scaffolding and diagnostics, as well as learning history recording, which can help teachers to assess the learning process. In addition, since the learning units studied in the game are smaller in size, the learning content is smaller and therefore significant improvement in learning can be achieved in a shorter period of time. For more learning content, it may be recommended that more learning time is needed to ensure the effectiveness of learning. In terms of future research: This study is an initial evaluation of the mobile scaffolding framework, and after confirming the positive impact on learning, a cross-group comparison with various other teaching methods will be conducted in the future. In addition, the system recorded the process when the learners were playing, but it could not record the discussion among learners. Therefore, in the future, we can use audio-recording equipment to record the activity for video-based behavior pattern analysis (e.g., Hou, 2015; Hou & Keng, 2021) to explore deeper learning behaviors. In addition, regarding the limitations of the study, since this study is only a preliminary case study with a single group pre-post test, there may be an effect on the intrinsic validity due to the same test. Future studies may include a control group for comparison to explore the differences with other teaching methods in more depth or to analyze the usefulness of each scaffold in more depth (e.g., Hou, Fang & Tang, in press). In addition, new digital tools may increase students' motivation for the first time usage, and in the future, it is recommended that longer or continuous weeks of instructional activities be taken to assess whether students have a long-term positive effect (e.g., learning effectiveness promotion) on the games and scaffolding. Future studies can explore whether there is a reduction in anxiety in the discipline after a longer intervention time, or even whether a reduction in anxiety can be achieved consistently. In the future, this study will also evaluate teachers' acceptance of the editor for teaching and learning. The effectiveness of using this approach in other regions' educational approaches and policies is also worth studying. Funding This study funded by Ministry of Science and Technology, Taiwan (MOST-108-2511-H-011 -003 -MY3, Huei-Tse Hou, MOST-107-2511-H-011 -003 -MY3, Huei-Tse Hou, MOST-110-2511-H-011 -004 -MY3, Huei-Tse Hou). Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Ashcraft MH Math anxiety: Personal, educational, and cognitive consequences Current Directions in Psychological Science 2002 11 5 181 185 10.1111/1467-8721.00196 Bakeman R Gottman JM Observing Interaction: An Introduction to Sequential Analysis 1997 2 Cambridge University Press Bawa P Watson SL Watson W Motivation is a game: Massively multiplayer online games as agents of motivation in higher education Computers & Education 2018 123 174 194 10.1016/j.compedu.2018.05.004 Britner SL Pajares F Sources of science self efficacy beliefs of middle school students Journal of Research in Science Teaching 2006 43 5 485 499 10.1002/tea.20131 Brush TA Saye JW A summary of research exploring hard and soft scaffolding for teachers and students using a multimedia supported learning environment The Journal of Interactive Online Learning 2002 1 2 1 12 Byun J Joung E Digital game-based learning for K–12 mathematics education: A meta-analysis School Science and Mathematics 2018 118 3–4 113 126 10.1111/ssm.12271 Carey E Hill F Devine A Szűcs D The modified abbreviated math anxiety scale: A valid and reliable instrument for use with children Frontiers in Psychology 2017 8 11 10.3389/fpsyg.2017.00011 28154542 Chang CC Warden CA Liang C Lin GY Effects of digital game-based learning on achievement, flow and overall cognitive load Australasian Journal of Educational Technology 2018 34 4 2961 10.14742/ajet.2961 Chen CH Law V Scaffolding individual and collaborative game-based learning in learning performance and intrinsic motivation Computers in Human Behavior 2016 55 1201 1212 10.1016/j.chb.2015.03.010 Cheng KH Tsai CC Children and parents' reading of an augmented reality picture book: Analyses of behavioral patterns and cognitive attainment Computers & Education 2014 72 302 312 10.1016/j.compedu.2013.12.003 Chou, Y. S., Hou, H. T., Chang. K. E. & Su, C. L. (in press) Designing a cognitive-based game mechanism for mobile educational games to promote cognitive thinking: an analysis of flow state and game-based learning behavioural patterns, Interactive Learning Environment. Chow SJ Yong BCS Secondary school students' motivation and achievement in combined science US China Education Review B 2013 3 4 213 228 Gil-Doménech D Berbegal-Mirabent J Stimulating students’ engagement in mathematics courses in non-STEM academic programmes: A game-based learning Innovations in Education and Teaching International 2019 56 1 57 65 10.1080/14703297.2017.1330159 Collins A Brown JS Newman SE Resnick LB Cognitive apprenticeship: Teaching the crafts of reading, writing, and mathematics Knowing, Learning, and Instruction: Essays in Honor of Robert Glaser 1989 Lawrence Erlbaum Associates 453 494 Hopko DR Mahadevan R Bare RL Hunt MK The abbreviated math anxiety scale (AMAS) construction, validity, and reliability Assessment 2003 10 2 178 182 10.1177/1073191103010002008 12801189 Hou HT Exploring the behavioral patterns of learners in an educational massively multiple online role-playing game (MMORPG) Computers & Education 2012 58 4 1225 1233 10.1016/j.compedu.2011.11.015 Hou HT Analyzing the behavioral differences between students of different genders, prior knowledge, and learning performance with an educational MMORPG: A longitudinal case study in an elementary school British Journal of Educational Technology 2013 44 3 E85 E89 10.1111/j.1467-8535.2012.01367.x Hou HT Integrating cluster and sequential analysis to explore learners’ flow and behavioral patterns in a simulation game with situated-learning context for science courses: A video-based process exploration Computers in Human Behavior 2015 48 424 435 10.1016/j.chb.2015.02.010 Hou HT Keng SH A dual-scaffolding framework integrating peer-scaffolding and cognitive-scaffolding for an augmented reality-based educational board game: An analysis of learners’ collective flow state and collaborative learning behavioral patterns Journal of Educational Computing Research 2021 59 3 547 573 10.1177/0735633120969409 Hou, H. T., Fang. Y. S. & Tang, J. T. (in press) Designing an alternate reality board game with augmented reality and multi-dimensional scaffolding for promoting spatial and logical ability, Interactive Learning Environment. Hsieh YH Yi-Chun L Hou HT Exploring elementary-school students' engagement patterns in a game-based learning environment Journal of Educational Technology & Society 2015 18 2 336 Hwang GJ Wu PH Chen CC An online game approach for improving students’ learning performance in web-based problem-solving activities Computers & Education 2012 59 4 1246 1256 10.1016/j.compedu.2012.05.009 Kiili K Evaluations of an experiential gaming model Human Technology: An Interdisciplinary Journal on Humans in ICT Environments 2006 10.17011/ht/urn.2006518 Kim S Chang M Computer games for the math achievement of diverse students Journal of Educational Technology & Society 2010 13 3 224 Kuo C. C., Fang, Y. S., Wang, S. M., Li, Y. Y., & Hou, H. T. (2020). A preliminary study of a business-management/strategic-planning board game with situated learning mechanisms. In 14th European Conference on Games Based Learning (ECGBL20), 24th – 25th September 2020, Brighton, UK. Kusmaryono I Gufron AM Rusdiantoro A Effectiveness of scaffolding strategies in learning against decrease in mathematics anxiety level Numerical Jurnal Matematika Dan Pendidikan Matematika 2020 10.25217/numerical.v4i1.770 Lee O Brophy J Motivational patterns observed in sixth-grade science classrooms Journal of Research in Science Teaching: The Official Journal of the National Association for Research in Science Teaching 1996 33 3 303 318 10.1002/(SICI)1098-2736(199603)33:3<303::AID-TEA4>3.0.CO;2-X Li, C.T., Keng, S. H., Li, Y. Y., Fang, Y. S., Hou, H. T. (2018). The development and evaluation of an educational board game integrated with augmented reality, role-playing, and situated cases for anti-drug education. Paper presented at ICCE International Conference on Computers in Education (ICCE2018), Nov 26–30, 2018, Manila, the Philippines. Li, C. T., Lee, L. H., & Hou, H. T. (2021a). Designing an augmented reality educational board game learning activity with dual-scaffolding teaching strategy to enhance EFL reading comprehension. Paper presented at IIAI International Congress on Advanced Applied Informatics (IIAI LTLE 2021a), July 11th–16th, Online, Japan. Li, C. T., Huang, Y. J., Lai, Y. R., Zheng, S. T., Wang, C. J., & Hou, H. T. (2021b). Designing an Augmented Reality-based Educational Board Game Integrated with Dual-Scaffolding Framework for High school History Course: The Evaluation of Learning Performance and Flow State. Paper presented at IIAI International Congress on Advanced Applied Informatics (IIAI LTLE 2021b), July 11th–16th, Online, Japan. Lopez FG Lent RW Sources of mathematics s elf efficacy in high school students The Career Development Quarterly 1992 41 3 3 12 10.1002/j.2161-0045.1992.tb00350.x López JMC Cáceres MJM Virtual games in social science education Computers & Education 2010 55 3 1336 1345 10.1016/j.compedu.2010.05.028 Martín-Lara MA Calero M Playing a board game to learn bioenergy and biofuels topics in an interactive, engaging context Journal of Chemical Education 2020 97 5 1375 1380 10.1021/acs.jchemed.9b00798 Maryam B Sören H Gunilla L Putting scaffolding into action: Preschool teachers’ actions using interactive whiteboard Early Childhood Education Journal 2020 48 1 79 92 10.1007/s10643-019-00971-3 McKeachie WJ Does anxiety disrupt information processing or does poor information processing lead to anxiety? International Review of Applied Psychology 1984 33 187 203 10.1111/j.1464-0597.1984.tb01428.x McLaren BM Adams DM Mayer RE Forlizzi J A computer-based game that promotes mathematics learning more than a conventional approach International Journal of Game-Based Learning (IJGBL) 2017 7 1 36 56 10.4018/IJGBL.2017010103 Meluso A Zheng M Spires HA Lester J Enhancing 5th graders’ science content knowledge and self efficacy through game based learning Computers & Education 2012 59 2 497 504 10.1016/j.compedu.2011.12.019 Mitchell KM Harrigan T Stefansson T Setlack H Exploring self-efficacy and anxiety in first-year nursing students enrolled in a discipline-specific scholarly writing course Quality Advancement in Nursing Education-Avancées En Formation Infirmière 2017 3 1 4 10.17483/2368-6669.1084 Nelson LM Collaborative problem solving Instructional Design Theories and Models: A New Paradigm of Instructional Theory 1999 2 241 267 Puntambekar S Hubscher R Tools for scaffolding students in a complex learning environment: What have we gained and what have we missed? Educational Psychologist 2005 40 1 1 12 10.1207/s15326985ep4001_1 Rowe E Asbell-Clarke J Baker RS Eagle M Hicks AG Barnes TM Edwards T Assessing ***implicit science learning in digital games Computers in Human Behavior 2017 76 617 630 10.1016/j.chb.2017.03.043 Shin S Brush TA Glazewski KD Examining the hard, peer, and teacher scaffolding framework in inquiry-based technology-enhanced learning environments: impact on academic achievement and group performance Educational Technology Research and Development 2020 68 2423 2447 10.1007/s11423-020-09763-8 Shute VJ Focus on formative feedback Review of Educational Research 2008 78 1 153 189 10.3102/0034654307313795 Vitasari P Wahab MNA Othman A Herawan T Sinnadurai SK The relationship between study anxiety and academic performance among engineering students Procedia-Social and Behavioral Sciences 2010 8 490 497 10.1016/j.sbspro.2010.12.067 Wang SM Wu CH Hou HT Integrating broad game elements, collaborative discussion, and mobile technology to a gamification instructional activity—a case of high school chemical course International Journal of Learning Technology and Learning Environment 2019 2 2 11 20 10.52731/ijltle.v2.i2.478 Wood D Bruner JS Ross G The role of tutoring in problem solving Journal of Child Psychology and Psychiatry 1976 17 2 89 100 10.1111/j.1469-7610.1976.tb00381.x 932126 Wu, C. H., Chen, C. C., Wang, S. M., & Hou, H. T. (2018, July). The design and evaluation of a gamification teaching activity using board game and QR code for organic chemical structure and functional groups learning. In 2018 7th International Congress on Advanced Applied Informatics (IIAI-AAI) (pp. 938–939). IEEE. Young SSC Wang YH The game embedded CALL system to facilitate English vocabulary acquisition and pronunciation Journal of Educational Technology & Society 2014 17 3 239 251
PMC009xxxxxx/PMC9005910.txt	==== Front Public Organiz Rev Public Organization Review 1566-7170 1573-7098 Springer US New York 622 10.1007/s11115-022-00622-z Article Unpacking the blackbox of responsible pandemic governance: of COVID-19, multilevel governance and state capacity in Ghana – A Review http://orcid.org/0000-0002-4230-3654 Arkorful Vincent Ekow saintvincentino@gmail.com Department of Government and International Studies, Hong Baptist University, Kowloon Tong, Kowloon, Hong Kong China 13 4 2022 117 5 3 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Attempts at mitigating COVID-19 pandemic’s impact has pushed stakeholders’ resolve to incept variegated measures using socially embedded multilevel government structures. Given Ghana’s pandemic governance success, this paper reviews government’s nuanced and disaggregated roles in galvanizing social support towards developing, implementing and coordinating pandemic measures. By highlighting the diversity of state-society inter-agency relations, the current study unearths varying stakeholder engagements and their imperativeness to pandemic governance, and acknowledges multilevel governance as critical to fighting the pandemic. Keywords COVID-19 Pandemic governance State capacity State-society relations Ghana ==== Body pmcIntroduction Perhaps unprecedented, the COVID-19 pandemic outbreak has dealt negative impacts on not only social relations and life, but also global health and economic systems (Arkorful, Lugu, Shuliang, 2021). Without doubt, COVID-19 has subjected the collective global corporate governance systems’ response capacity to a litmus test. Given the urgent situation, various global governments have deployed a range of mechanisms principally targeted at designing and implementing effective policies, exploring opportunities and drawing cross-stakeholder potentials, whilst strengthening existing social support processes and procedures. Apparently, the quest to “unravel the myth” surrounding the novel COVID-19 has erupted a diversity of research interests spanning scholarly fields composed of, but not limited to governance and administration, with a significant number of these studies illuminating government’s indispensability to pandemic control and containment (Arkorful, 2021; Arkorful, Abdul-Rahaman et al., 2021; Assan et al., 2022). Peci et al., (2020), Farazmand & Danaeefard (2020) and Hale et al., (2020) have in this breadth strongly recommended interrogating the efficacy or otherwise of pandemic governance mechanisms, actors, and institutions. The motivation thereof has consequely spurred interesting discourses among scholars, particularly social scientists, pertaining to corporate governance institutions’ salience to crises management (i.e. pandemic governance). Excited partly by observations of centralised states (for instance, the People’s Republic of China) pandemic management and containment success, vis-a-vis the relatively malnoursihed inroads of democratic polities in Europe and North America, the arguments in this context have to some extent converged around the notion of pandemic governance efficiency as contingent on states’ capacity to utilise a melange of strategies and social structures (Diamond, 2020). The juxtaposition of these settings notwithstanding, much as democratic and non-democratic governance systems have inherent strengths and weaknesses in responding to and managing crises, there ostensibly appears a dearth of consensus regarding appropriate and responsive governance system for countervailing social challenges, notably, pandemics. To this end, Ghana’s enviable pandemic control, and the World Health Organisation’s (WHO) touting of its success (Taylor & Berger, 2020) has provided veritable study grounds to broach its approaches to ascertain corporate governance imperativeness to pandemic governance and state capacity definition. By engaging the state capacity theory as a hereustic framework, opposed to erstwhile studies’ (Mao, 2021) limited, yet purposeful conceptual generalisation benchmarked by centralised states’ pandemic control efficacy, the current study makes a case for democracies’ use of interagency relations and multilevel governance structures for crises management (i.e., COVID-19). State capacity as an evolving concept has been used to study government and non-government entities’ social imbroglios mitigation potentials (Dincecco & Katz, 2016). In the current study, capacity is operationalised to reinforce states’ policy formulation and implementation capability – in response to emergency situations. Predicated on Hanson (2018), the study segments four capacity dimensions; administrative, policy design and implementation, coercive and social support solicitation, and state extractive capacity. These critical dimensions underpin government-stakeholder relationship; a typical example of which is central-sub-national state synergy, which dovetails into determining states’ variegated capacities, and overall policy response outcomes. As earlier echoed, with the pandemic inter alia exposing centralised states’ crises management proclivities on one breadth, and democracies’ acclaimed capacity paucity on another, Ghana’s success presents an interesting case for study. In this respect, whiles operationalising capacity theory in reference to states’ pandemic management, the current study transcends its mere replication to contribute to literature by unearthing interagency manifestations of capacity relative to pandemic control and management. Therefore, cognisant of the dearth of studies on state pandemic governance capacity – more especially in the Ghanaian context, the theory is applied to comprehensively illuminate actors, institutions and relations, and how their roles and powers were leveraged for COVID-19 management. To the best of the author’s knowledge, this is the premier study to broach state pandemic governance capacity in Ghana. For purposes deepening understanding, the study discourses on the theoretical underpinning (i.e., state capacity theory) in the ensuing section. 2. Theoretical underpinning: The state capacity theory under review. One of the longstanding lenses for ascertaining state multidimensional capacity is the state capacity theory. Its centrality to governance has spurred a panoply of scholarly conceptualisations, with classical elucidations emerging from Wu et al., (2015) who defines it as comprising political competencies and capabilities which are central to policy success. Whiles Kraay et al. (2010) accentuate state capacity in reference to state’s goal execution, Williams (2021) avers to it as states’ policy implementation capacity. Further, Besley & Persson (2010) conceptualise it as entailing states’ revenue extraction and, development maximisation and sustenance capacity. These streams of scholarship providing evidence of state capacity imperativeness explicate its adverse potential in erupting devastating consequences on states’ engagements (Dincecco & Katz, 2016). Importantly, emerging state capacity studies have atttempted creating typologies. Whiles acknowledging state capacity potency to disaster management, Christensen et al., (2016) delimit capacity advantages comprising, facilitating government’s multistakeholder coordination for public service delivery, and subsequently delineates capacity dimensions including; information, coercive, decision design and execution, and mobilisation capacity. Hanson (2018) delimits state capacity into administrative, coercive and extractive strands. Precisely, administrative capacity, including Christensen et al.,’s (2016) information capacity, refers to the state’s policy formulation and implementation capacity. Common to Christensen et al., (2016) and Hanson (2018), coercive capacity refers to state control over society - by exerting dominance to counterpose dissent. Extractive capacity (referred to as mobilisation capacity according to Christensen et al., [2016]) refers to the state’s revenue raising capacity for socio-economic development. Lastly, decision making and implementation capacity refers to the state capacity to design and execute responses to socio-economic challenges. These differing conceptulisations notwithstanding, the theory’s replication in the current pandemic governance study context is predicated on Christensen et al., (2016) and Hanson (2018). Given its relevance, seminal studies (Lin, 2015; Hanson, 2015) have utilised the theory to appreciate pandemic governance discourses. For Christensen et al., (2016) who contend with state capacity indispensability to government coordination, public service delivery, information dissemination, social control, and policy making and implementation, capacity constitutes a pivot for crises management. Regarding pandemic management engagements, and with recourse to Christensen et al., (2016) and Hanson (2018), the study anchors capacity on a conceptual prism comprising; administrative, policy design and implementation, coercive, and social support solicitation and extractive capacity. In this context, administrative capacity refers to the state’s pandemic governance coordination using appropriate structures and arrangements. Coercive capacity refers to the state’s capacity to enforce and ensure citizens’ adherence to pandemic measures. Policy design and implementation refers to the identification and deployment of policy instruments. Lastly, social support solicitation and extractive capacity refers to state’s pandemic control capacity, mobilising social capital from across government and non-government spheres (Bynander & Nohrstedt, 2019). Taking a state-society perspective, this study ascertains nuanced and disaggregated state-bureaucracy-civil society roles. With autonomy shaped and conditioned by central-subnational relations, Ghana’s constitution grants a range of powers to legislative (law making), executive (policy implementation) and judicial (law interpretation) bodies whose collective functions converge around the aforementioned capacity dimensions. Ensuring responsiveness has informed further decentralisation of functions across Ghana’s six Metropolitan, one hundred and six Municipalities, and one hundred and forty-five (145) districts totaling, two hundred and sixty (260) adminstrative domains, all of which are required to facilitate efficiency and effectiveness in governance, including pandemic management. In this study, capacity is best understood in terms of terms of liberating state-society interagency synergies, manifested in multilevel governance for pandemic governance. State capacity is defined with respect to viable institutional structures, embedded interagency relations in policy delivery, a well as development, and harnessing state-society potentials. Thus conceptualised, capacity is broadly extracted from civil society and governance literature (Arkorful, Abdul-Rahaman et al., 2021). Noteworthy, these capacity forms do not exist in isolation; they are steeped in corporate networks with leverageable potentials for pandemic control – hence the germaneness of cross-sectoral cooperation to effective pandemic policy development and implementation, and overall governance. Actually state capacity in democrcies and non democracies vary. Nonetheless, state-society emblemmatises capacity which is critical to crises managment decision-making (Christensen et al., 2016). With enduring effect on state administrative, extractive, coercive, and policy making powers and functions, states’ crises management capacity is overarchingly contingent on a healthy stakeholder relationship. Whereas state capacity in non-democracies may assume a top-down approach granting government unrestricted powers, decision making in democracies may assume an all-inclusive flexible approach fashioned along flexible top-down lines incorporaing a smogasbord of civil society stakeholders. Invariably, much as coordination is significant to eliciting capacity to confront crises, the urgency to distribute public goods in democracies may engender a purposeful creation of independent subnational structures, which could be a recipe for institutional wranglings – more especially in instances when and where there exist unclear distinction and limit to the exercise of vested powers and functions. Aware of the possible challenges of power diversifications to states and capacity exercise, it is important to recognise that offsetting inherent deficits for pandemic governance may require deploying innovative governance approaches sufficiently incorporating multilevel governance. Observing that governments have designed and implemented variegated anti-COVID-19 responses including quarantine, lockdowns, mask wearing and social distancing, as well as testing and contact tracing, reviewing measures of successful country experiences becomes relevant. Significantly, the success of such measures cannot be discussed disparately from multilevel governance deployment – hence the capacity theory replication in this study. Research Methodology For the current study, the research limits its scope to reviewing and analysing stakeholder initiatives and/or responses towards COVID-19 management in Ghana. Therefore, policy making and implementation, as well as stakeholder engagements are reviewed. The particular focus on Ghana is in part due to its pandemic management success. In this study, the researcher undertook a content analysis of COVID-19 policies using secondary data like situation and news reports, as well as government data. The reliance on these options was based on Bowen (2009) and Wach et al.’s, (2013) recommendation for qualitative policy analysis. Despite saddled with health sector challenges, Ghana’s pandemic governance is touted a success. The pandemic control index for instance ranked Ghana third after China and Sri Lanka respectively (Yicai, 2020). Attributed to stakeholder incorporation, the replication of a flexible pandemic governance model guided and streamlined by state-society relations, underpins the success story. Profound to the pandemic governance lies the uilisation of national and subnational government and non-government structures. With disparate, yet convergent capacities, their integration has inured overhwelming pandemic governance benefits. In view of deaths and confirmed cases, vis-a-vis rising and declining trends (Figs. 1 and 2), with intermittent ups and downs, Ghana’s case is considered a success. Given the peculiarities, notably, of variegated multilevel power structures, an interrogation of these are considered veritable to illuminating state capacity and corporate governance salience to pandemic governance. Fig. 1 Active cases and recoveries March 2020-October 2021 Source Ghana Health Service. Figures 1 and 2 attests to the results-orientedness and the efficacy of concerted stakeholder engaements to pandemic management. After Ghana’s first two cases on 12/3/2020, the deployment of mechanisms comprising; mask wearing, lockdown imposition, mobility restrictions, mass testing and contact tracing among others, proved effective. The intensification thereof from April-May culminated in a decline in active cases, and at the same time, a rise in recoveries - from the month of May-June onwards (Fig. 1). Fig. 2 Regional distribution of COVID-19 cases in Ghana Source Author computation with Ghana Health Service data. Ghana’s success, and the peculiarities surrounding social capital harnessing using institutions to counter the pandemic across the administrative regions (Fig. 3) makes it interesting interrogating state-corporate society synergy via the state capacity perspective. To provide a panoramic view of the situation, the next section is designated to highlighting the COVID-19 in Ghana. Fig. 3 COVID-19 regional distribution map Source Ghana Health Service. Note Active COVID-19 cases in Ghana by Region by 3/10/2021 (left). Cumulative cases of COVID-19 in Ghana by Region by 3/10/2021 (right). Study background COVID-19 and the Government of Ghana’s response. The Noguchi Memorial Institute for Medical Research confirmed the first COVID-19 case in Ghana on March 12, 2020. These were two imported cases from people who had disembarked from Norway and Turkey. Responding to existential risks led to the institution of a series of health protection protocols including a lockdown imposition starting from March 20 – April 20, 2021 (Arkorful, Lugu, Shuliang 2021). Further deepening and strengthening cross-sectoral coordination and collaboration led to the formation of a COVID-19 inter-ministerial presidential task force chaired by the President of Ghana. Given the emergency situation, nationwide commercial activities were suspended; and with the exception of frontline workers defined to include health workers, national security officers, and other essential service providers, socio-economic activities were temporarily halted. With entry and exit boarder closures in force (with effect from 22/03/2020), traveling and transportation activities were suspended. The urgency to institute stringent measures led to the passage of the 2020 Imposition of Restrictions Act (IRA) pursuant to which the President of the Republic of Ghana issued an Executive Instrument (E.I. 64) declaring an emergency. Acting in conformity with Section 169 of the Public Health Act of 2012 Act 851, the Minister of Health declared a public health emergency (Arkorful, Nurudeen et al., 2021). To facilitate tracking and contact tracing, a COVID-19 app was launched on April 12, 2021 (BBC, 2021). Taken together social distancing and mask wearing rules, these legislations helped in strengthening the institution of health protocols directed at controlling COVID-19 upsurge. As of 24/8/2021, Ghana had recorded 423 new cases, 6, 850 active cases, 115, 525 confirmed cases, 107, 693 recoveries and 982 deaths (Ghana Health Service, 2021). On 24/2/2021, Ghana received 600,000 doses of the AstraZeneca vaccine procured through the COVID-19 Vaccines Global Access Facility, known as COVAX - a World Health Organisation-led initiative, in partnership with the Coalition for Epidemic Preparedness Innovations and the Global Alliance for Vaccines and Immunisations (Gavi). Weaving together these initiatives involving varying government actors (i.e., national, subnational etc.) demonstrates institutions indispensability to state capacity harnessing and augmentation towards pandemic management - thereby underscoring multilevel corporate governance interaction relevance, as discussed below. Central-subnational local state capacity salience Tailored according to the stipulations of the Provisional National Defence Council (PNDC) Law 207 and Article 240 of the 1992 Decentralisation and Local Government Law, central-subnational state relations in Ghana is captured in a four-tier structure composed of efficiently interwoven Municipal, Metropolitan and District Assemblies (MMDA’s) headed by Chief Executive appointees of the President of the Republic of Ghana. The appointment of these Executives is subject to two-thirds majority approval of Assembly Persons representating the electoral areas constituting the assembly’s legislature. The MMDA’s membership include 70% elected members (via universal adult suffrage) and 30% nominees of the President. As an extension representing government at the grassroots, MMDAs’ strategic position enables them strike a balance between central and local state interests, via deepening citizens’ planning and decision-making participation (Gyimah-Boadi, 2009). The pertinence of these structures is ingrained in their efficacy to central-local state communication, and policy design and implementation. Not ending there, with the advantage of proximity to the local state and its citizens, MMDA’s find themselves strategically positioned to enhance the distribution of public goods like water, electricity and healthcare whilst enforcing central government’s social, economic, cultural and health policies and programs among others (Arkorful et al., 2021). Ghana’s well decentralised local governance system has over the years facilitated broader multistakeholder engagement for socio-economic development. Because this has been the case, local state structures have not only been the fulcrum around which government revolves, but also, contributed to augmenting and expediting government decision making and implementation capacity. Acting as a communication medium facilitating information flow from central government to subnational states and vice versa, central-subnational state relationship has helped increase information capital, and as such, government’s capacity. In relations to the COVID-19 pandemic in Ghana, a careful observation of engagements confirm a flexible, yet centralised approach - with information and other important directives disseminated from the central government to the grassroots. However, aware of the urgent responses required to maintain social stability whilst confronting the pandemic, the top-down approach assumed a more decentralised dimension involving already-established local state agencies. Evidence to this is the 15/03/2020 presidential directive to the Ministry of Local Government and Rural Development (MLGRD) tasking them to coordinate subnational local state activities to promote hygiene and compliance with COVID-19 protocols. The MLGRD in response, instituted measures to promote sanitation activities across the sixteen administrative regions in Ghana (Asante & Mills, 2020). Central-local relationships has strengthened and unified national COVID-19 policy design and execution. Despite central-local state relations salience, the 2014 Ebola outbreak significantly informed the institution of comprehensive preparedness and response systems, notably, laboratory and isolation unit establishment, contact tracing, screening and temperature checks, and other communication approaches (i.e., print and electronic media) which later provided pointers for pandemic management (Obern, 2020; Antwi-Boasiako et al., 2021). Not only were public health laboratory facilities repurposed for COVID-19 management, the complementary deployment of technology contributed to enhancing outcomes for health governance. Nonetheless, local state incorporation helped garner local citizens and institutional support for COVID-19 policy making and implementation. Though decentralized governments are saddled with policy implementation challenges (Arkorful et al., 2021) often times stemming from resource constraints, with Ghana’s local government Act 462 delimiting fiscal federalism parameters entailing revenue sources like grants and donor supports, internally generated funds (i.e., fines, taxes and rates, licenses etc.) as well as intergovernmental transfers, pandemic governance co-financing and co-production engagements received signifcant impetus. And while the existence of health committees (chaired by MMDA’s Chief Executives, assisted by deputies) responsible for health administration created an avenue for deliberations, the incorporation of religious bodies (i.e., Christians and Muslims), indigenous institutions (i.e., traditional chieftaincy), the Health Directorate, District Coordinating Director and Environmental Health Division, helped promote stakeholder representativeness. In essence, fusing these structures into national and subnational state agencies appreciably expedited decision making and implementation, and at the same time extended some degree of autonomy accordingly. Hence, embedding local structures in central structures, as done under Ghana’s decentralised governance, augmented corporate coordination and integration (Asante & Mills, 2020). The imperativeness of central-subnational state synergy to defining state capacity Corporate governance in Ghana is characterised by top-down central-local relationship. Captured in an elaborate, effective and efficient decentralised local governance system weaving together central-subnational government structures as critical actors as far as socio-economic development is concerned (Arkorful et al., 2021), central and local states are in essence not adversaries competing for power and dominance in a zero sum game; rather, they are complexly interdependent in diverse ways - such that each others stability is dependent on not only the existence, but also, the effectiveness of the other. In this vein, central-subnational relationship is indispensable; in the sense that, whereas central government practically needs local states support to represent its interest and perform policy implementation functions, the local state on the other hand requires the central state to adequately resource and vest them with the necessary powers and functions capable of rendering them functional. These notwithstanding, it is instructive to stress corporate civil society structures as the desiderata to shaping, reshaping and sustaining these relationship forms (Antwi-Boasiako et al., 2021). Evidence to this lies in how the longstanding central-local state relationship has contributed to socio-economic development in Ghana, and the consolidation of democracy therein. Situating these in the state capacity and COVID-19 pandemic context, this review highlights the germaneness of this relationship to state’s resource solicitation and/or mobilisation capacity. Practically in the case of Ghana, with central government under pressure to control the pandemic situation, subnational local states have proven effective in generating public support, trust and confidence for government-led anti pandemic activities. Also, the local state, in concert with the COVID-19 rapid response team dotted around the adminsitrative regions, has been instrumental in enhancing mass testing, treatment and contact tracing. While attributing local states’ efficiency to its proximity to the local settings and citizens, it is important underscoring that their mobilisation, cooperation and social compliance potencies have collectively contributed to defining state capacity (Hanson 2018; Serikbayeva & Oskenbayev 2021). Point of convergence: state capacity and responses to the COVID-19 pandemic In view of the foregoing discussion highlighting the manifestation of state capacity in the Ghanaian context, the study proceeds to designate the next section to discourse extensively on central-subnational state engagements in relations to the various state capacity strands relative to COVID-19. Administrative capacity Within broader administrative context, one strategy featuring prominently in Ghana’s pandemic management remains effective communication. At the outbreak of the pandemic, given the population-wide spread of fear and panic, the search for alternatives to streamline policy actions whilst garnering social support and boosting public confidence became urgent for building social capital for pandemic management. In the “Spread Calm not Fear” campaign, the President of the Republic of Ghana employed communication to allay citizens’ fears. In a profound statement that attracted global plaudits, the President, in an attempt to inspire hope and caution, is quoted to have said; “we know how to bring back the economy to life; what we do not know is how to bring people back to life”. Relative to pandemic management, disseminating relevant, accurate and timely information is acknowledged efficient and effective. This is much so in the case that, important information can be relied upon by the general public to guide them in observing preventive measures and putting up favorable health behaviors (Moon, 2020) like social distance observation and mask wearing. The use of information and communication (by either providing or withholding it from stakeholders) for policy finds affirmation in Hood (1986). On this plane, the Ghanaian government performed creditably by engaging the citizenry and other stakeholders in regular communication, via a series of presidential address (Table 1) held to among other things dispel COVID-19 propaganda, whilst promoting transparency. Interestingly, much as this approach was top-down, it had a distinctive character of flexibility and innovativeness – one that involved various government (i.e., Ministries, Departments and Agencies) and non-government bodies (i.e., community-based, non-governmental organisations and corporate civil societies among others). The engagement of these structures as pandemic management accessories was facilitated coutesy subnational local government establishments (i.e., MMDA’s) whose proximity to local citizens and institutions was enormosly tapped. Table 1 Presidential address on COVID-19 Address Date Time Theme Address 1 11/03/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 2 15/03/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 3 21/03/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 4 27/03/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 5 5/04/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 6 9/04/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 7 19/04/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 8 26/04/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 9 10/05/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 10 31/05/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 11 14/06/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 12 21/06/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Address 13 28/06/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 14 26/07/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 15 16/08/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 16 30/08/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 17 20/08/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 18 18/10/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 19 8/11/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 20 20/12/2020 8:00 pm Measures taken to combat the spread of the Coronavirus Address 21 3/1/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Address 22 17/1/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Address 23 31/1/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Address 24 28/2/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Address 25 16/05/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Address 26 25/7/2021 8:00 pm Measures taken to combat the spread of the Coronavirus Source: Author Construct In actual fact, given the potential debilitating risks of COVID-19 related propaganda to national stability, communication was instrumental in maintaining a certain degree of national stability. Here, with the Government of Ghana prioritising communication as central to fighting the pandemic, the COVID-19 dashboard was created to communicate new and confirmed cases, as well as recoveries. Subsequent to vaccine discovery and populationwide innoculation commencement, Ghana’s government utlised the dashboard for stakeholder communication accordingly. Additionally, communication channels like radio, television and social media platforms (Facebook, twitter etc.) were incorporated as effective information dissemination platforms. Given these measures, Ghanaians were more informed. As such, populationwide misinformation appeared limited, if not nonexistent. To reach out to populations without necessarily leaving anyone behind, the potency of local government structures were exploited for mass communication. And much as the complementary use of central-local state entities for COVID-19 communication and information disclosure aided in allaying fears, leveraging the subnational state’s proximity to local citizens helped in forging trust and galvanising social support for pandemic governance. Policy design and implementation capacity The outbreak of the COVID-19 pandemic in Ghana in March 2020 has engendered innovative results-oriented responses. The declining trends in confirmed cases and deaths (refer to Figs. 1 and 2) in a way confirms the positive outcomes of arrangments put in place. Practically, the inception of innovative institutional governance approaches culminating in the COVID-19 inter-ministerial presidential task force (chaired by the President of the Republic of Ghana-Nana Addo Danquah Akufo Addo) are among measures that has helped in coordinating activities and expediting related decision making procedures and processes. Specifically under this arrangement, various legislations have been enforced. Further, pursuant to the 2012 Public Health Act 851 for instance, the health minister declared the situation a pandemic, subsequent to which the IRA was triggered to impose mobility restrictions (i.e., air, sea and land). It is relevant to clarify at this point that, whiles the Parliament of Ghana (as a lawmaking body) was the fountain of these constitutional provisions, the security agencies making up the executive arm of government were in charge of ensuring compliance – in concert with decentralised state bodies engaged as grassroots policy implementing partners. Moreover, the Communication Ministry and the Presidency’s intermittent televised COVID-19 updates have been critical to creating pandemic policy awareness. The complementary involvement of decentralised local state structures, thus the MLGRD, has fast-tracked local level policy design and implementation. In addition to the local government’s Health Committee enhancing mass testing and contact tracing, the formation of COVID-19 rapid response teams have been helpful to mitigating negative ramifications at the local level. This has been possible, owing to Ghana robust decentralised governance system characterised by strong stakeholder relations between various Ministries, Department and Agencies. The coordination of pandemic activities by reputable entities like the MLGRD and its ancillary bodies has aided not only crowdsourcing and relief items distribution (i.e., masks, hand sanitisers, veronica buckets, food, clothing etc.) to covid-stressed households and individuals in the MMDA’s, but also, ensured creating general policy awareness and compliance, which is generally relevant to state pandemic governance, and more particularly, policy design and implementation capacity. COVID-19 and state’s Coercive capacity Another manifestation of state capacity in Ghana’s COVID-19 pandemic fight is coercive capacity. To start with, the study will at this point appreciate state’s coercive capacity from a constitutional point of view. Precisely, the 1992 Constitution of Ghana vests the President with excessive powers to take decisions deemed to be in the national interest. In this regard, under Ghana’s presidential system of government (featuring monocephalous executive), the President doubles as the Head of State and Government, and by default acts as the Commander-in-Chief of Armed Forces – as highlighted under Article 57 of Chap. 8 of the constitution (Government of Ghana, 1992). Under this provision, the President is empowered to pass Executive Instruments – particularly in times of emergency (Government of Ghana, 1992). Vested with these powers, and acting intra vires, after Ghana’s first two recorded cases of the COVID-19 on 12/3/2020, the President, in an address on 15/3/2020 directed the Attorney General to submit an emergency legislation to Parliament. This was in line with article 21(4) (c) (d) and (e) of the 1992 Constitution. Subsequently, the President directed the Health Minister to declare a state of public health emergency – in line with Sect. 169 of the 2012 Public Health Act 851 (Communications Bureau, 2020). Acting on the executive orders of the president, the Attorney General, within five days period, drafted and presented, under a certificate of urgency, the 2020 Imposition of Restrictions Act 1012, which was later passed by parliament after a third reading. Subsequent to receiving a presidential assent, the Act was published in the gazette on 23/3/2020 and became enforceable (Arkorful, Abdul-Rahaman et al., 2021). Interestingly, the institution of these legal arrangements were not without resistance; they encountered stiffer political opposition, with the National Democratic Congress (NDC) filing a motion against the bill - on grounds that it fell short to merit urgency. The NDC threatened to challenge the decision by seeking interpretation from Ghana’s apex court - the Supreme Court (Mordy, 2020). In other related incidents, other legal luminaries expressed dissatisfaction with the IRA’s content, constitutionality, and the procedures via which it came into force. Moreover, to enhance a population-wide adherence to pandemic protocols, Ghana’s Government deployed a combined team of Police and Military force during, and even briefly after the three week lockdown period (30/3/2020–20/4/2020) in bigger cities like the Greater Accra and Greater Kumasi. Within these period, except essential service providers (i.e., food retailers, medical service providers, water and electricity distributors and retailers etc.), economic activities were halted. To avoid overcrowding in public places, not only were rotational arrangments rolled out to regulate the informal market sector as well as employees of the public Ministries, Departments and Agencies, but also, schools churches and mosques were temporarily closed. Whereas restaurant and bar operators were required to provide services through delivery, funerals, weddings and other forms of social gathering were abruptly suspended. Later after the lift of the qualrantine, public gathering was allowed, but limited to 25 persons. On 5/6/2020, the restriction on religious activities were lifted with mosques and churches allowed to host not more than a hundred population. Appraising and taken together these measures affirm that, though a democratic state, Ghana adopted stringent coercive measures composed of security and legal approaches to enforce pandemic control measures (BBC, 2021). In spite of resorting to these mechanisms, in the COVID-19 fight, the Government of Ghana stuck to democratic tenets like transparency and respect for the individual’s fundamental freedom and human rights, and at the same time, safeguarding individuals’ privacy (Antwi-Boasiako & Nyarkoh, 2021). Taken together, much as the foregoing discourse highlight states’ adherence to constitutional processes and procedures, it also manifests the state’s leveraging of coercive powers for crises era policy formulation and implementation. Social support solicitation and state extractive capacity The top-down bureaucratic structure of responses has helped strategically position the central state whilst bolstering its strength to coordinate, harness potentials and incorporate cross-sectoral efforts towards fighting the COVID-19 pandemic. The pandemic’s spontaneity required immediate measures; and this called for cross stakeholder forces moblisation. Much as this was driven by considerations of adequate representation, it provided an opportunity for coalescing social capital for pandemic management, profound of which is the COVID-19 Alleviation Program (CAP) establishment, meant to among other things, provide social protection against unemployment, and advancing sustenance for small businesses (Ministry of Finance, 2020). Particularly considering the employment of a larger chunk of Ghanaians in the informal sector (accommodating approximately 87% of small businesses) contributing 70% to gross domestic product (Abor & Quartey, 2010), the Government of Ghana rolled out the COVID Alleviation Program Business Support Scheme. Out of the USD 174 million seed money, USD 104 million was government’s contribution. The remaining sum targeted to be disbursed among 230, 000 business establishments from across Ghana were contributions from the ARB Apex Bank (National Board for Small Scale Industries, 2020). With decentralised local structures parlaying their proximity strengths, resources like veronica buckets, masks, sanitisers, and personal protective equipments, among others were crowdsourced from benevolent individuals and organisations within the local states dispersed across the MMDA’s. Moreover, acting on the International Labour Organisation’s (2020) recommendation to provide relief for burdened populations, social protection was extended to populations including “kayayei” – a description for head porters. This was however in cities like Greater Accra and Greater Kumasi – perhaps because of their predominance in the areas. Efforts to strengthen social services informed the provision of a three-month free electricity for vulnerable populations on life line consumption, and a 50% subsidy for residential and commercial accommodation consumers. These were meant to cushion individuals and households (majority of who are in the informal sector) from negative impacts like job losses and income drops, among others. Moreover, with recommendation on frequent handwashing gaining advocacy grounds, “free water” came to be part of the social support package. The Government of Ghana extended these packages to the end of the year 2020. Regardless of the fact that these safety net measures undergirded by good intentions, they were constantly criticised imposing needless strains on an already-stressed public expenditure. Others also described it as a populist gesture of the incumbent government intended to elicit political gains and favour. From the health sector, entities like the University of Ghana, Kumasi Center for Collaborative Research (KCCR) and the Noguchi Memorial Institute for Medical Research, among others were all incorporated into the pandemic fighting efforts. Also, the Veterinary Service Department, Public Health Reference Laboratory, the University of Health and Allied Sciences, Center for Scientific and Industrial Research were also included. To facilitate COVID-19 data handling and real time communication, the University of Ghana’s Geography Department was engaged. Taken together, much as though these demonstrated state pandemic response capacity, they also signified and/or symbolised it’s rallying capacity. From the foregoing illuminations, it is important to underscore that, these engagements do not occur in a vacuum; rather, they are contingent on an appreciable state-corporate society relationship which Ostrom (1997), Arkorful, Basiru et al., (2019) and Arkorful & Lugu (2022) propose as critical to stakeholder co-production and cocreation guided by complementarity principles and ethos. More so, the relevance of state-corporate society synergy in pandemic governance is reiterated by Oh et al., (2020). By garnering support from corporate social entities, the Government of Ghana has in essence, deployed effective and efficient mechanisms tailored to bring the COVID-19 pandemic under control – hence the reputed pandemic governance success. Conclusions The outbreak of the COVID-19 pandemic has excited a plethora of administrative and governance related issues. Part of the discourse has significantly focused on the need to balance various interests to among other things draw cross-sectoral and institutional potentials for crises governance (Farazmand & Danaeefard, 2021). Given polities’ varying approaches to battling the COVID-19, relying on Christensen et al., (2016) and Hanson’s (2018) state capacity proposition (i.e., social support and solicitation, coercive, administrative and), this study reviews Ghana’s approach to pandemic governance, and affirms healthy interagency relations as imperative to stakeholder interest representation and state capacity enhancement and sustenance. Stakeholder relations is imperative to creating vibrant corporate governance spaces for crises management, policy design and implementation, and it is also critical to social capital moblization for crises management and containment. Thus, Ghana’s pandemic governance has affirmed that, instituting appropriate structures and processes is one thing from establishing interagency relations and leveraging inherent strengths towards countervailing not only uncertainties (i.e., pandemics), but also, gratifying social needs and expectations, and at the same time, mitigating challenges. The review highlights engagements that characterised Ghana’s anti COVID-19 efforts. Funding None. Compliance with ethical standards Conflict of interest None. Ethical Approval All procedures performed in this study were reconcilable with the ethical standards of the insituional and/or national research committeee and the 1964 Helsinki declaration and its later amendments or comparable ethical standards. In line with this declaration amended in 2008, study participants were informed about the study purpose. Informed Consent Informed consent was obtained from all study participants. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Antwi-Boasiako J Nyarkoh E Government Communication during the Covid-19 Pandemic; the Case of Ghana International Journal of Public Administration 2021 44 11–12 1039 1040 Antwi-Boasiako J Abbey COA Ogbey P Ofori RA Policy Responses to fight COVID-19; the case of Ghana Revista de Administração Pública 2021 55 122 139 10.1590/0034-761220200507 Arkorful, V. E. (2021). The role of Indigenous traditional institutions in fighting against the COVID-19 pandemic in Ghana. Development in Practice, 1–5 Arkorful, V. E., & Lugu, B. K. (2022). Understanding Rate Evasion Behavior in Local Governance: Application of an Extended Version of the Theory of Planned Behaviour.Public Organization Review,1–20 Arkorful, V. E., Abdul-Rahaman, N., Ibrahim, H. S., & Arkorful, V. A. (2021). Fostering Trust, Transparency, Satisfaction and Participation Amidst COVID-19 corruption: Does the Civil Society Matter?–Evidence from Ghana.Public Organization Review,1–25 Arkorful, V. E., Basiru, I., Anokye, R., Latif, A., Agyei, E. K., Hammond, A. … Abdul-Rahaman, S. (2019). Equitable Access and Inclusiveness in Basic Education: Roadblocks to Sustainable Development Goals. International Journal of Public Administration Arkorful, V. E., Lugu, B. K., & Shuliang, Z. (2021). Unearthing mask waste separation behavior in COVID-19 pandemic period: An empirical evidence from Ghana using an integrated theory of planned behavior and norm activation model.Current Psychology,1–16 Arkorful, V. E., Lugu, B. K., Hammond, A., & Basiru, I. (2021). Decentralization and Citizens’ Participation in Local Governance: Does Trust and Transparency Matter?–An Empirical Study. Forum for development studies (pp. 1–25). Routledge Asante LA Mills RO Exploring the socio-economic impact of COVID-19 pandemic in marketplaces in urban Ghana Africa Spectrum 2020 55 2 170 181 10.1177/0002039720943612 Assan, A., Hussein, H., & Agyeman-Duah, D. N. (2022). COVID-19 lockdown implementation in Ghana: lessons learned and hurdles to overcome.Journal of Public Health Policy,1–11 BBC (2021). Covid-19 cases in Ghana: Police arrest over 200 people sake of dem no wear nose masks in public. Retrieved from: https://www.bbc.com/pidgin/tori-55745151, 10/10/2021 Besley T Persson T State capacity, conflict, and development Econometrica 2010 78 1 1 34 10.3982/ECTA8073 Bowen, G. A. (2009). Document analysis as a qualitative research method. Qualitative Research Journal, 9(2), 27–40. Retrieved from . org/10.3316/QRJ0902027 Bynander, F., & Nohrstedt, D. (2019). Collaborative crisis management: Inter-organizational approaches to extreme events (pp. 1–12). Routledge Diamond, L. (2020). America’s COVID-19 disaster is a setback for democracy. The Atlantic, 16 April. Available at: https://www.theatlantic.com/ideas/archive/2020/04/americas-covid-19-disaster-setbackdemocracy/610102/ Dincecco M Katz G State capacity and long-run economic performance The Economic Journal 2016 126 590 189 218 10.1111/ecoj.12161 Farazmand, A., & Danaeefard, H. (2021). Crisismanship under the Most Severe Sanctions: Lessons Learned from the Iranian Government’s Responses to the COVID-19.International Journal of Public Administration,1–16 Government of Ghana. (1992). Constitution of the Republic of Ghana. Ghana Publishing Corporation. Accra Hale, T., Sam, W., Anna, P., Toby, P., & Beatriz, K. (2020). Oxford COVID-19 Government Response Tracker. Blavatnik School of Government Hanson JK State capacity and the resilience of electoral authoritarianism: Conceptualizing and measuring the institutional underpinnings of autocratic power International Political Science Review 2018 39 1 17 32 10.1177/0192512117702523 Hood C The tools of government 1986 Chatham, NJ Chatham House 978 971 Kraay, A., Kaufmann, D., & Mastruzzi, M. (2010). The Worldwide Governance Indicators. Methodology and Analytical Issues Lin TH Governing natural disasters: state capacity, democracy, and human vulnerability Social Forces 2015 93 3 1267 1300 10.1093/sf/sou104 Mao Y Political institutions, state capacity, and crisis management: A comparison of China and South Korea International Political Science Review 2021 42 3 316 332 10.1177/0192512121994026 Mordy, J. (2020). T. ‘Minority Signals Supreme Court action after rushed Restriction Bill Approval. Retrieved from: https://www.myjoyonline.com/news/national/minority-signals-supreme-court-action-after-rushed-restriction-billapproval/?_gl=1%2A1k959zr%2A_ga%2AS25zZUswS0ctNWxXbGZVUVJXWlpJYlRtRmFvbVpRbE0xekNJTGtVdm9YTHRidEtuQ19tbmtoell2VGxOVWdOVA, on 15/09/2020 Obern, C. (2020). What is the secret behind Ghana’s pandemic success? Retrieved from: https://www.newamerica.org/weekly/whats-secret-behind-ghanas-pandemic-success, on 20/02/2022 Oh, J., Lee, J. K., Schwarz, D., Ratcliffe, H. L., Markuns, J. F., & Hirschhorn, L. R. (2020). National response to COVID-19 in the Republic of Korea and lessons learned for other countries.Health Systems & Reform, 6(1), e1753464 Taylor, A., & Berger, M. (2020). When it comes to coronavirus response, superpowers may need to study smaller nations. on 17 May 2020. The Washington Post. Accessed from https://www.washingtonpost.com/world/2020/05/16/when-it-comes-coronavirus-response-superpowers-may-need-study-smaller-nations/ Wach, E., Ward, R., & Jacimovic, R. (2013). Learning about Qualitative Document Analysis (IDS Practice Paper In Brief;13). Brighton, UK: Institute of Development Studies. Retrieved from https://www.ids.ac.uk/publications/learning-about-qualitativedocument-analysis/ Williams MJ Beyond state capacity: bureaucratic performance, policy implementation and reform Journal of Institutional Economics 2021 17 2 339 357 10.1017/S1744137420000478 Wu X Ramesh M Howlett M Policy capacity: A conceptual framework for understanding policy competences and capabilities Policy and Society 2015 34 3–4 165 171 10.1016/j.polsoc.2015.09.001 Yicai (2020). China Tops World Survey on Epidemic Control Work, Economic Recovery, Yicai Report Shows. Retrieved from: https://www.yicaiglobal.com/news/china-tops-world-survey-on-epidemic-control-work-economic-recovery-yicai-report-shows, on 12/10/2021
PMC009xxxxxx/PMC9005912.txt	==== Front J Med Syst J Med Syst Journal of Medical Systems 0148-5598 1573-689X Springer US New York 1814 10.1007/s10916-022-01814-2 Mobile & Wireless Health Text Messaging as a Communication Modality to Promote Screening Mammography in Low-income African American Women http://orcid.org/0000-0002-2543-8493 Ntiri Shana O. sntiri@som.umaryland.edu 1 Swanson Malia 2 Klyushnenkova Elena N. 1 1 grid.411024.2 0000 0001 2175 4264 Department of Family & Community Medicine, University of Maryland School of Medicine, 29 S. Paca Street, Baltimore, MD 21201 United States 2 grid.411024.2 0000 0001 2175 4264 University of Maryland School of Medicine, 29 S. Paca Street, Baltimore, MD 21201, United States 13 4 2022 2022 46 5 2827 11 2019 27 3 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Introduction Though text messages are increasingly used in health promotion, the current understanding of text message-based interventions to increase screening mammography in low-income African American women is limited. This study aimed to assess the feasibility and acceptability of a text message-based intervention to increase screening mammography in low-income African American women. Materials and Methods A 15-item, self-administered, paper-based survey on cell phone ownership, text messaging practices and preferences for future breast health information was administered to 120 female patients at an urban family medicine office. Descriptive analyses and demographic correlates of text messaging practices and preferences were examined. Results and Discussion The majority of respondents (95%) were cell phone owners of whom 81% reported texting. Prior receipt of a text message from a doctor’s office was reported by 51% of cell phone owners. Mammography appointment reminders were the most desired content for future breast health text messages. Age (≥ 70 years old) was found to have a significant negative relationship with text messaging practices and perceptions. Implications The use of text messages to promote mammography was found to be acceptable in this patient population. In addition to age, variables such as the frequency, timing and subject content of text messages also influence their acceptability. Keywords Breast cancer screening Text message Primary care African American Cancer health disparities issue-copyright-statement© Springer Science+Business Media, LLC, part of Springer Nature 2022 ==== Body pmcIntroduction Breast cancer is the second leading cause of cancer death in American women. Screening mammography reduces breast cancer mortality through early detection [1]. While screening mammography has contributed to overall declines in breast cancer mortality, significant disparities in mortality persist between African American and White women. Although White women are diagnosed with breast cancer more frequently, breast cancer mortality is over 40% higher in African American women in comparison to White women. The higher mortality rates in African American women are likely reflective of multiple factors including access and adherence to breast cancer screening, particularly African American women of lower socioeconomic status [1]. Women insured through Medicaid have a mammography completion rate that is nearly 10% lower than completion rates in women with commercial or private insurance [2, 3]. In addition to lower screening completion rates, women of lower socioeconomic status also have lower 5-year breast cancer survival rates [1]. Patient reminders for breast cancer screening increase adherence to recommended screening [4]. Increasingly, the utility and effectiveness of technology-based strategies is demonstrated in health promotion interventions [5]. Widespread cell phone ownership facilitates access to technology-based strategies within populations that may not have access via computer-based platforms [6, 7]. As cell phone usage has increased, health care providers have implemented text messaging to communicate with their patients, particularly in the form of reminders–this form of communication has been demonstrated as acceptable in certain patient populations [5, 8]. According to the Pew Research Center, as of 2021, 97% of American adults owned a cell phone including those with a household income level at or below $30,000 [9]. African American adults are more likely to own a cell phone and to text more frequently than White adults [9, 10]. Text message cancer control interventional strategies targeted at African American adults are effective [11]. Although multiple studies have demonstrated text messaging interventions successfully increase uptake of screening mammography, there is limited data on screening mammogram text message interventions in African American women [12–16]. The COVID-19 pandemic introduced additional barriers to cancer screening and significantly disrupted face-to-face interactions between health care providers and patients. The CDC reported a 154% increase in telehealth visits in March 2020 in comparison to March 2019 [17]. The increase in technology-based contact between patients and providers can improve health behaviors and clinical outcomes for populations like low-income, African American women [18]. This study sought to conduct an initial examination of the feasibility and acceptability of text message-based interventions to increase screening mammography in low-income African American women. Given the recent shift from conventional in-person visits to telehealth due to the COVID-19 pandemic understanding the utility of technology-based communication platforms within specific populations such as low-income African American women is important. The knowledge gained from the study presented here has the potential to advance the knowledge on technology-based strategies to improve breast cancer screening in low-income African American women. Materials and Methods Potential respondents were identified from the waiting rooms of an urban, academic primary care office. One hundred and twenty-four women were invited to participate, and 120 women consented. Study eligibility criteria were: female, patient at the study site, ≥ 40 years old, English-speaking, and able to complete a survey written on the 8th grade level. The study instrument was a 15-item, self-administered, paper-based survey on cell phone ownership and text messaging preferences and practices. Descriptive analyses were conducted on categorical and numerical data. For categorical variables, chi-square (χ2) or Fisher exact tests were performed to identify the demographic and texting practices associated with acceptance of future text message-based promotion of screening mammography for respondents. Wilcoxon signed rank test was used to compare related ordinal variables. Data were analyzed using SAS 9.3 software (SAS Institute Inc., Cary, NC) and SPSS 22 software (Chicago, IL) [19, 20]. This study was deemed exempt by the University of Maryland at Baltimore’s Institutional Review Board. Results The mean age of respondents was 55 years old with a standard deviation (SD) of 9.2 years. The majority of the study population was African American (88%) and insured, either publicly (66%) or commercially (27%). Most (95%) respondents reported they owned a cell phone (Table 1). Cell phone ownership was significantly associated with age as only 78% of women ≥ 70 years old reported cell phone ownership compared to 90–100% of women from younger age groups (Fisher exact test p < 0.05). While 81% of all cell phone owners reported using their phones for texting, the use was limited to younger women (40–69 years old). The highest rates of texting were among women 40–49 years old at 91%, followed by 87% of women aged 50–59 years old and 65% of women 60–69 years old; none of the women in 70–79 years old group reported using their phones for texting (Fisher exact test p < 0.0001). More than half of cell phone owners reported both sending and receiving ≥ 4 text messages per day (Fig. 1). Respondents reported receiving significantly more text messages than they send (Wilcoxon Signed Rank test p < 0.0001, data not shown). While only 20% of respondents reported sending ≥ 10 text messages per day, 36% of respondents reporting receiving ≥ 10 text messages per day (Fig. 1). Respondents younger than 50 years old reported sending and receiving significantly more text messages per day than respondents who were 50 years and older. Among women ages 40–49 years old, 86% reported receiving more than 4 texts per day, compared to 58% of women 50–59 years old and 38% women 60–69 years old (Fisher exact test p < 0.001). Similarly, 85% of women who were 40–49 years old reported sending more than 4 texts per day compared to 44% women 50–59 years old and 29%of women 60–69 years old (Fisher exact test p < 0.0001).Table 1 Respondent Demographics Table 1: Respondent Demographics (N = 120) N % Age (years old) 40–49 50–59 60–69 70–79 36 45 30 9 30 37.5 25 7.5 Race African American Other Missing 106 13 1 88 11 1 Health Insurance Source Public Commercial No insurance Unspecified/Missing 79 32 2 7 66 27 2 5 Cell Phone Ownership Yes No 114 6 95 5 Fig. 1 Daily frequencies of sending/receiving text messages by cell phone owners (N = 106). Percentages designate the frequency of cell phone owners sending (A) and receiving (B) text messages Fifty-five cell phone owners (51%) had previously received a text message from a doctor’s office. For those who received such messages, the most common types of prior text message from a doctor’s office were: appointment reminders (95%), prescription refill notification (27%) and personalized health messages (9%). Age was not significantly associated with prior receipt of a health care-related text message, or the type of message received (data not shown). More than half of respondents (54%) indicated future willingness to receive a health care related text message from the study site, however 29% of respondents were opposed to receiving a future text message and 17% were neutral. Future willingness to receive a text message and the message type that respondents were willing to receive from the doctor office were not significantly associated with age (data not shown). Prior receipt of a text message from a doctor’s office was positively associated with future willingness to receive a health-related text message (Wilcoxon Signed Rank test p < 0.0001, data not shown). Preferences for content of future health related text message varied by prior history of text message receipt from a doctor’s office. Among respondents willing to receive a personalized health-related message, 73% had previously received a text message from a doctor’s office compared to 27% of those who had not previously received such messages, and 63% of those not willing to receive such a message had never received a prior text message from a doctor’s office (Fisher exact test p < 0.001). Among respondents willing to receive general health-related information, 65% had previously received a text message from a doctor compared to 35% of those who had not (Fisher exact test p < 0.05). In contrast, among respondents who were not willing to receive prescription refill notifications text, 65% had not previously received a text message from a doctor’s office compared to 35% of those who had received such messages in the past (Fisher exact test p < 0.05). Among respondents who were not willing to receive appointment reminders, 82% had not previously received a text message from a doctor’s office compared to 18% of those who had received such messages in the past (Fisher exact test p < 0.01). Respondents indicated specific preferences for both the frequency and timing of future text messages. Among cell phone owners, the most desired frequency for delivery of future health related text messages was monthly (40%), followed by weekly (22%) and several times a month (14%). There were no significant differences in preference for frequency of future text message by age or other demographic factors (data not shown). Most respondents preferred their messages delivered between 12:00 PM – 4:00 PM (27%) followed by 8:00 AM – 12:00 PM time slot (21%). There were age-based differences in preferences for the time of day for delivery of future text messages. Respondents > 70 years old indicated that they were not willing to receive messages in the early morning, while respondents younger than 70 years old were not willing to be contacted in the evening (Fig. 2).Fig. 2 Preferences for Text Message Time of Delivery by Age (N = 100). Respondents could opt for more than one (all that applies) response. Percentages indicate positive responses within the subgroups Preferences for Future Breast Cancer Screening Text Messages The survey also assessed respondents’ willingness to receive a future breast cancer screening related text message from the study site. Respondents were asked to indicate preference for the content of future breast cancer screening related text messages. The most desired content was screening mammogram appointment reminders (60%), mammogram results (42%) and general information on breast cancer (38%). Willingness to receive a future breast cancer screening text message was significantly associated with text message practices. Most respondents (89–95%) who were willing to receive a future breast cancer screening text message were those who reported ever sending text messages (Fig. 3A). A similar trend was observed among respondents who reported ever receiving text messages, although the association was not statistically significant in most cases (Fig. 3B). There was a significant association between responses from women who had previously received a text message from a doctor’s office and whether they were interested in receiving breast cancer screening related text messages in the future. More woman indicated willingness to receive future breast cancer screening text messages if they had previously received a text message from a doctor’s office. Further, preferences for content of future breast cancer screening related text message varied by prior history of text message from a doctor’s office (Fig. 3C). In addition, women who were open to receiving a text message from a doctor’s office were significantly more interested in receiving breast cancer screening related text messages (Fig. 3D). There was no significant correlation between willingness to receive a future breast cancer screening text message and patient’s age (data not shown).Fig. 3 Content Preference for Future Breast Cancer Screening Related Text Message by Prior Receipt of Text Message. Respondents could choose more than one response. N indicates the total number of positive responses. Percentages within the bars indicate positive responses by subgroups. Fisher exact test: p < 0.05, p < 0.01, ***p < 0.0001 Discussion This study assessed the initial feasibility and acceptability of text messages as a modality to promote screening mammography within a primary care patient population of low-income African American women. This study found the use of text messages as a communication modality to be feasible, but with some limitations. Most respondents reported cell phone ownership, and more than half of cell phone owners reported using their phones for text messaging. While there were significant age-based differences in cell phone ownership and text messaging practices, a significant proportion of women 70 years old or older indicated willingness to receive a future screening mammogram text message appointment reminder. Importantly, even those respondents who did not report sending text messages indicated willingness to receive both general health and breast cancer specific text messages. Prior research has shown that text messages are effective at increasing breast cancer screening rates in women due for screening mammography [12–15]. This study demonstrates that text messages are a communication platform that is readily accessible and agreeable to women of age for screening mammography in this patient population. Notably, women’s prior receipt of a health-related text message positively influenced future preferences. Next steps will be identifying the best practice strategies to implement text messaging for the promotion of breast cancer screening in this patient population. Given that several women opposed receiving future health related text messages, an important next step component will be identifying reasons women are opposed to receiving future health related text messages. This information would allow for these concerns to be addressed through the implementation of targeted strategies with the potential to make the communication form more acceptable to this group of women. Such steps will enable the opportunity to assess if the impact of text message-based interventions on the uptake of screening mammography in low-income African American women is similar to what has been seen in other populations. Lastly, it will be important to determine if there is any impact on breast cancer morbidity and mortality as a result of this intervention strategy for this group of women. Prior Experience with Health Care Related Text Messages Prior receipt of health care related text messages was common in this study population. More than half of respondents reported previously receiving a text message from a doctor’s office, primarily in the form of an appointment reminder. Previous experience with text message communication from a health care provider was significantly associated with respondents’ future willingness to receive future health related text messages. This finding suggests that respondents’ prior exposure to health care text messaging was positive or beneficial and respondents anticipated similar outcomes from future health related text messages. Further investigation should be done to determine if the level of willingness to receive health care related text messages changes in women with no prior exposure to health care text messages after receipt of health care related text messages. It is also important to determine if responsiveness to screening mammography text message reminders varies by history of prior receipt of health care related text messaging. Such knowledge would enable for tailoring of text messages according to their health care text message history. Preferences for Future Text Messages The overwhelming preference for the content of health care related text messages was appointment reminders for both general text messages and breast cancer specific messages. Interestingly, respondents did not indicate a strong preference for personalized health care text messages. This finding may be indicative of a desire for in-person or phone communication for individualized information or situations when there is a greater desire for bi-directional communication. Text messages may provide a succinct way to provide more generic health information such as an appointment reminder. Most respondents (77%) indicated a willingness to receive a future breast cancer related text message. Prior research has shown that patient reminders increase completion rates for screening mammography [4]. This study’s finding that of a top preference for mammogram appointment reminders suggests there is a need to for mammography reminders in this population and that text messages are an acceptable strategy to meet that need. It would be informative to assess if women’s preference for text messages changes at the different phases of the breast cancer screening continuum e.g., preparation for mammogram, mammogram results and/or need for follow-up. Such information could drive the development of a tailored text message system to communicate with patients at various points during the screening cascade. The study results did show that women had specific desires for the timing and frequency of text message delivery. Respondents were provided with multiple options to select from for the frequency text messages delivery including on an as needed basis only. Respondents overwhelmingly indicated a preference for receiving text messages at the frequency of once a month or once a week. This finding suggests that respondents have a desire to maintain regular communication from their doctor’s office and that text messages would be an acceptable option to maintain this contact. The survey asked about desired time of delivery using broad time categories. For all respondents younger than 70 years old, the most preferred time of day for text message delivery was 8 am to 4 pm. These findings of definitive preferences for text messages frequency and time of delivery suggest that future interventions may want to assess at baseline patient preferences for text delivery options such as frequency, time of day and other delivery modifications that may differ by age and other population specific variables. Limitations The generalizability of our study is limited by our sample size and the sample’s slight skew to younger women. It is possible that with a larger sample and an increased proportion of women older than 70 years old there would have been greater variation in text messaging practices and preferences. Despite this, our sample did include a broad age range of women who are eligible for screening mammography and therefore provides important preliminary information on this topic. Implications This study is an important first step in understanding the role of text message-based interventions to increase screening mammography in low-income African American women in a primary care setting. Text messages provide a communication modality suitable for the primary care setting that this study found to be largely available among a population of low-income African American women of age for screening mammography. There is a need to identify effective strategies that promote breast cancer screening and improve breast cancer outcomes for low-income African American women [12, 16]. The COVID-19 pandemic changed the way health care providers and patients communicate. Text message-based strategies to improve screening such as the one described here build upon the recent increase in technology-based strategies in health care and importantly enable outreach in the absence of conventional face-to-face interactions between patient and provider. This study identified text messages as a potential interventional strategy to improve breast cancer outcomes in low-income African American women. Further, this strategy may also be applicable in cancer control efforts for other screen-detectable cancers such as cervical and colorectal cancers which also care disproportionately high morbidity and mortality burdens in low-income African American patient populations. Funding None, this study did not receive funding support. Declarations Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This study was deemed exempt by the University of Maryland, Baltimore (UMB) Institutional Review Board. Conflict of Interest Shana O. Ntiri, MD, MPH declares that she has no conflict of interest. Malia Swanson, MD declares that she has no conflict of interest. Elena N. Klyushnenkova, PhD, MSPH declares that she has no conflict of interest. This article is part of the Topical Collection on Mobile & Wireless Health Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. DeSantis CE Ma J Gaudet MM Breast cancer statistics, 2019 CA Cancer J Clin 2019 69 6 438 451 10.3322/caac.21583 31577379 2. National Center for Health Statistics., Health, United States. With Special Feature on Racial and Ethnic Health Disparities. Hyattsville, MD. 2016, 2015. Accessed August 10, 2018, from https://www.ncbi.nlm.nih.gov/books/NBK367640/pdf/Bookshelf_NBK367640.pdf. 3. Harper S Lynch J Meersman SC Breen N Davis WW Reichman MC Trends in Area-Socioeconomic and Race-Ethnic Disparities in Breast Cancer Incidence, Stage at Diagnosis, Screening, Mortality and Survival among Women Ages 50 years and Over (1987–2005) Cancer Epidemiol Biomarkers Prev 2009 18 1 121 131 10.1158/1055-9965.EPI-08-0679 19124489 4. Schmittdiel J McMenamin SB Halpin HA The use of patient and physician reminders for preventive services: results from a National Study of Physician Organizations Preventive Medicine 2004 39 5 1000 1006 10.1016/j.ypmed.2004.04.005 15475035 5. Cole-Lewis H Kershaw T Text Messaging as a Tool for Behavior Change in Disease Prevention and Management Epidemiologic Reviews 2010 32 1 56 69 10.1093/epirev/mxq004 20354039 6. Anderson, M., For Vast Majority of Seniors Who Own One, a Smartphone Equals ‘Freedom.’ Pew Res. Center. 2015. Accessed August 12, 2018, from http://www.pewresearch.org/fact-tank/2015/04/29/seniors-smartphones. 7. Pew Research Center., Technology Device Ownership: Pew. Res. Center. 2015. Accessed August 12, 2018, from http://www.pewinternet.org/2015/10/29/technology-device-ownership-2015/. 8. Nundy, S., Razi, R. R., Dick, J. J., et al., A Text Messaging Intervention to Improve Heart Failure Self-Management After Hospital Discharge in a Largely African-American Population: Before-After Study. J. Med. Int. Res. 15(3):e5, 2013. 9. Pew Research Center., Ownership of Other Devices. Surveys of U.S. adults conducted 2002–2021. 2022. Accessed January 19, 2022, from https://www.pewresearch.org/internet/fact-sheet/mobile/. 10. Lenhart, A., Cell phones and American adults. Pew. Res. Center. 2010. Accessed August 12, 2018, from http://www.pewinternet.org/2010/09/02/cell-phones-and-american-adults/. 11. Le D Holt CL Darlene SR Feasibility and acceptability of SMS text messaging in a prostate cancer educational intervention for African American men Health Informatics Journal 2015 22 4 932 974 10.1177/1460458215598636 26324051 12. Kerrison RS Shukla H Cunningham D Oyebode O Friedman E Text-message reminders increase uptake of routine breast screening appointments: A randomised controlled trial in a hard-to-reach population British Journal of Cancer 2015 112 1005 1010 10.1038/bjc.2015.36 25668008 13. Lakkis NA Atfeh AM El-Zein YR Mahmassani DM Hamadeh GN The effect of two types of sms-texts on the uptake of screening mammogram: A randomized controlled trial Preventive Medicine 2011 53 4–5 325 327 10.1016/j.ypmed.2011.08.013 21871480 14. Arcas MM Buron A Ramis O Esturi M Hernanodes C Macià F Can a mobile phone short message increase participation in breast cancer screening programmes? Rev Calid Asist. 2014 29 4 188 196 10.1016/j.cali.2014.02.003 25002239 15. Vidal C Garcia M Benito L Milà N Binefa G Moreno V Use of Text-Message Reminders to Improve Participation in a Population-Based Breast Cancer Screening Program Journal of Medical Systems 2014 38 9 118 10.1007/s10916-014-0118-x 25073694 16. Coughlin SS Intervention Approaches for Addressing Breast Cancer Disparities among African American Women Annals of Translational Medicine & Epidemiology 2014 1 1 1001 25568890 17. Koonin LM Hoots B Tsang CA Trends in the Use of Telehealth During the Emergence of the COVID-19 Pandemic — United States, January–March 2020 MMWR Morbidity and Mortality Weekly Report 2020 69 43 1595 1599 10.15585/mmwr.mm6943a3 33119561 18. Ganapathy S de Korne DF Chong NK Car J The Role of Text Messaging and Telehealth Messaging Apps Pediatric Clinics of North America 2020 67 4 613 621 10.1016/j.pcl.2020.04.002 32650857 19. SAS Institute Inc. 2009. SAS® 9.4 Statements: Reference. Cary, NC: SAS Institute Inc IBM Corp. Released 2013. 20. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp.
PMC009xxxxxx/PMC9005913.txt	==== Front Environ Earth Sci Environ Earth Sci Environmental Earth Sciences 1866-6280 1866-6299 Springer Berlin Heidelberg Berlin/Heidelberg 35432620 10364 10.1007/s12665-022-10364-2 Original Article Experimental research on deformation failure process of roadway tunnel in fractured rock mass induced by mining excavation Li Guang 123 Ma Fengshan fsma@mail.iggcas.ac.cn 12 Guo Jie 12 Zhao Haijun 12 1 grid.9227.e 0000000119573309 Key Laboratory of Shale Gas and Geoengineering, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, 100029 China 2 grid.9227.e 0000000119573309 Innovation Academy for Earth Science, CAS, Beijing, 100029 China 3 grid.410726.6 0000 0004 1797 8419 University of Chinese Academy of Sciences, Beijing, 100049 China 13 4 2022 2022 81 8 24316 11 2020 19 3 2022 © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Deformation failure of roadways in fractured rock can lead to large-volume collapse and other engineering accidents. Failure mechanisms in fractured rock are complex and poorly understood, so to explore this issue, we simulated fractured rock masses using physical model tests in combination with numerical computations. A set of experimental techniques for roadway excavation under jointed surrounding rock included a mixed pouring–bricking method and a roadway excavation device, which can reproduce the structural characteristics of the prototype and replicate the excavation conditions of the roadway. Stress distribution characteristics of the roadway, from loading to excavation, were obtained based on strain monitoring and image acquisition, and the process of roadway deformation and failure was described in detail. A series of numerical simulations were conducted to investigate the deformation failure mechanisms of roadways under different excavation conditions. Results indicate that the deformation failure modes of roadways including collapse, rock burst, and floor heaving that were similar regardless of depth. Deformation failure modes of the roadway were determined by rock mass structure, and the deformation intensity was determined by geo-stress. Model testing and numerical simulation were consistent; hence, findings provide a theoretical basis and technical guidance for roadway engineering in fractured rock masses. Keywords Roadway deformation failure Jointed rock mass Physical model test PFC2D http://dx.doi.org/10.13039/501100001809 National Natural Science Foundation of China 41831293 41772341 41877274 42072305 Ma Fengshan Guo Jie Zhao Haijun issue-copyright-statement© Springer-Verlag GmbH Germany, part of Springer Nature 2022 ==== Body pmcIntroduction Roadways are the lifeline of successful mining projects because they are critical for personnel operation, infrastructure construction, and ore transportation. With greater mining depth, roadways may deteriorate and result in a series of engineering disasters, such as large deformation of rock mass, long-time rheology of roadway, roof collapse, floor uplift, and rock burst, which become increasingly serious (Ren et al. 2020). It is especially difficult to apply the deformation failure law for fractured rock bodies, and to engineer efficient support modes for safe and cost effective projects. Roadway deformation and stability research mainly rely on field investigation, theoretical analysis, numerical simulation, and physical model tests. However, it is expensive and laborious to conduct a detailed field investigation because there are numerous factors affecting deformation in a complex fractured rock body (Xu et al. 2019). An extreme simplification and a number of assumptions are unavoidable for the prototype before the theoretical analysis, and it is very difficult to describe engineering-scale problems with mathematical models (Tu et al. 2018). In contrast, physical experiments can replicate the whole process from elasticity deformation to destruction under complex conditions. However, numerical techniques that are characterized by economy and high efficiency have made great progress, and the simulation results can be mutually verified with the experimental results (Shreedharan et al. 2016). In recent years, a large number of physical model tests have been conducted to examine the deformation failure laws of roadway under different excavation conditions. Hendron et al. (1972) studied the mechanical behavior of jointed rock roadway under static load. Norman (1981) used a small brittle model to study the deformation failure mechanism of coal mine roadway. Lee and Schubert (2008) conducted experimental research on the deformation failure mechanism of soft rock roadway surface. Gou et al. (2009) built two physical models based on the self-developed model test system and discussed roadway deformation failure under different horizontal stresses. These studies showed that the roof and floor of the roadway are the key supporting parts. He et al. (2009) combined infrared thermal imaging system with physical model test to study the impact of strata dip angle and temperature in the excavation process of mining roadway. Zhang et al. (2010) evaluated deformation characteristics, fracture evolution process, and displacement distribution of an inclined and layered surrounding rock roadway under different stress conditions, and proposed an optimal support method. Huang et al. (2013) investigated the role of soft structural planes in the roadway deformation failure process, and put forward several different failure modes. Li et al. (2015) studied the special working condition of a deep thick coal roadway and revealed the distribution characteristics of surrounding rock displacement and stress in the whole excavation process. Fakhimi et al. (2002) and Wang et al. (2003) used Particle Flow Code (PFC) to simulate deformation failure of circular roadways after excavation. Cai et al. (2007) obtained acoustic emission characteristics of surrounding rock during excavation in large underground engineering projects. An and Tannant (2007) discussed the deformation failure characteristics of roadway under dynamic loading. Wang and Tian (2018) analyzed the failure mechanical mechanism and crack evolution process of surrounding rock mass in coal strata under different conditions. Liu et al. (2019) established a horseshoe-shaped roadway model, compared the distribution of force chain, displacement and strain in the model results, and proposed four zones in the roadway deformation process. However, in the previous studies, most of the research objects were homogeneous structures or layered structures, and there are few studies on the excavation of roadways having multiple groups of cross joints in the surrounding rock. In this case study of the fractured roadway tunnels in Jinchuan nickel mine, China, we perform a physical model test and a numerical simulation calculation with the objectives being to obtain displacement and strain distribution laws, simulate the deformation failure process of fractured roadways, and analyze deformation failure mechanisms of roadways under different excavation conditions. Physical model test A physical model test is an important method for studying complex engineering geological problems, because the physical model can be used to simulate deformation from elasticity to plasticity and replicate the whole destruction process in a laboratory experiment. Simulation prototype Jinchuan is the largest nickel producing mine in China, with thick and deeply buried ore bodies as shown in Fig. 1. The mining area is located in Hexi Corridor, where tectonic movement is intense, ground stress is high, the rock mass is fractured, and joints and fissures are abundant. The average altitude of surface in the mining area is about + 1750 m. At present, the underground infrastructure has reached + 700 m and is buried more than 1000 m. A downward filling mining method is adopted in the Jinchuan Mine. Because of complex engineering geological conditions, roadway repair increased from 3200 m in 1999 to 18,423 m in 2019, which affected the economic benefits of the mine and safety of underground workers (Hui et al. 2019; Lu et al. 2018, 2020).Fig. 1 Engineering geological background of the Jinchuan Mine The dimensions of the horseshoe-shaped roadway were 4.5 × 4.5 m, as shown in Fig. 2. The engineering geological data in the study mine showed that the typical surrounding rock mass was rhombic marble with two sets of intersecting joints. The strike of the two sets of structural planes was nearly perpendicular, the inclination ranged from 30° to 60°, and the rock stratum thickness was about 0.5–1.5 m. The deformation failure in the surrounding rock mass was mainly controlled by the structural plane, and was unstable. Thus, the roadway under this type of surrounding rock mass was selected as the prototype in this test.Fig. 2 Typical roadway section in Jinchuan mining area Similarity relation Physical experiments require a similarity between the model and the prototype. However, because of high demands on materials, equipment, and technology, it is difficult to achieve complete similarity. Therefore, in general, several important indicators are selected according to the purpose of the study. The ratio of the same dimensional quantities in the prototype and model was used as a similarity constant, which is indicated by the letter C. Considering the influence of excavation stress, boundary conditions, and laboratory infrastructures, the dimensions of the physical model were 105 × 105 × 20 cm and the roadway diameter was 15 cm. Thus, the geometry similarity constant Cl is 30, and other basic physical quantities were calculated in accordance with the law of Buckingham π theorem, as shown in Table 1 (Sun et al. 2017).Table 1 Physical and mechanics parameters of the rock and rock mass Physical quantity Similarity relation Similarity constant Geometry (key constant) Cl 30 Density (key constant) Cρ 1.6 Displacement (key constant) CD 30 Poisson’s ratio Cμ = 1 1 Elasticity modulus CE = CρCl 48 Strain Cε = CρCl/CE 1 Stress Cσ = ClCγ 48 Internal friction angle Cφ = 1 1 Cohesion Cc = CρCl 48 Time Ct = Cl(Cρ/CE)1/2 5.48 A crucial part of a large-scale physical model test is quickly and accurately determining the ratio of similar materials. Most of the roadways in the study area are distributed in the marble rock mass on both sides of the ore body, and the physical and mechanical parameters of the intact marble rock and marble rock mass are shown in Table 2. River sand, cement, and gypsum were adopted as the raw materials to reduce the test cost, simplify the test procedures, and fully utilize the properties of raw materials. Then, a series of comparative tests with bone glue ratio and water–paste ratio as variables were conducted (Fig. 3). Experimental results indicate that the ratio A6 (river sand:cement = 4:1) and ratio C6 (river sand:cement:gypsum = 8:1:1) match intact marble and marble rock mass, as shown in Tables 2, 3 (Li et al. 2020a, b, c).Table 2 Physical and mechanics parameters of rock mass and similar material Lithology Type Density (g·cm−3) Tensile strength (MPa) Compressive strength (MPa) Cohesion (MPa) Internal friction angle (°) Elastic moduli (GPa) Poisson’s ratio Intact marble rock Measured value 2.80–3.00 6.90–12.20 96.00–152.00 13.5–22.5 35–45 64–124 0.22–0.32 Design value 1.75–1.88 0.14–0.25 2.00–3.29 0.28–0.47 35–45 0.88–2.58 0.22–0.32 Ratio A6 1.82 0.43 2.62 0.24 27 0.42 0.26 Marble rock mass Measured value 2.60–2.80 0.80–1.40 18.60–32.40 0.70–6.50 25–35 6–12 0.18–0.26 Design value 1.63–1.75 0.02–0.03 0.39–0.68 0.04–0.14 25–35 0.13–0.25 0.18–0.26 Ratio C6 1.75 0.18 1.42 0.14 25 0.26 0.23 Fig. 3 Parts of the samples (Li et al. 2020a, b, c) Table 3 Mechanics parameters of the structural plane Type Cohesion (MPa) Friction angle (°) Structural plane 0.04–0.10 27–29 Model building A physical model test is important for studying roadway stability; however, it is difficult to build similar models for jointed rock masses. Therefore, a set of physical model test methods was designed for roadway excavation under jointed surrounding rock. For roadways excavated in the jointed surrounding rock mass, a mixed pouring–bricking method was designed. Pouring was adopted in the inner ring of the model, and bricking was used in the outer ring, which reflects the structural characteristics of jointed surrounding rocks and improves test efficiency (Fig. 4).Fig. 4 Model section diagram The joint spacing of rhombic rock masses is 0.5–1.5 m and the dip angle is 30°–60° in the study area; thus, the block was designed as 4 × 4 × 20 cm based on the simulation theory. A polypropylene plastic mold, that is light and easy to demold, was made to save cost and realize mass production. After several steps including stirring, stuffing, scraping, stripping, and curing, the production of block is completed, and part of the building blocks is shown in Fig. 5.Fig. 5 Building blocks Figure 6 shows the model building process. The outer surrounding of the model is made of similar material according to the parameters of the marble rock mass, which is compact and uniform. The material can effectively transfer stress and is conducive to observing the fracture development on the surrounding rock mass. The internal block of the model is made of similar material according to the intact marble rock, and the joints in the surrounding rock mass are clearly visible, which can reproduce the structural characteristics of the prototype.Fig. 6 Model building process In mining engineering, roadways roadway have to be excavated under the action of high in-situ stress. However, it is not easy to simulate the excavation under loading in the laboratory test. Thus, a roadway excavation device was independently developed based on embedded mold and spiral traction to form a roadway that would only slightly disturb the model. The embedded mold is made according to the design size of the excavated roadway, and the material is cast iron with sufficient stiffness, as shown in Fig. 7a. A nut was embedded in the middle of the mold to match the mold removal device. Through spiral traction, the embedded mold can be pulled out slowly, which is in accordance with the practice of a step-by-step excavation that only slightly disturbs the model, as shown in Fig. 7b.Fig. 7 Roadway excavation device; a embedded mold and b mold removal device Data collection The experiment was recorded by two digital cameras, one of which took photos at regular intervals, and another one recorded the whole process. Strain gauges were installed on the back of the model to allow for observation of deformation failure on the model front. Four measuring lines including 20 gauges were set on the roof, floor, and two sides, as shown in Fig. 8.Fig. 8 Data collection devices; a design of the strain gauges and b image of the data collection devices Loading mode The model was loaded by a self-developed hydraulic servo comprehensive experimental platform that consisted of three parts: model box, loading system, and control system, as shown in Fig. 9. The loading system was controlled using a computer, with a maximum of 300 kN loading force in the horizontal and vertical directions.Fig. 9 Self-developed hydraulic servo comprehensive experimental platform According to the actual ground stress conditions in the study area, the horizontal and vertical ground stresses were set as 40 MPa and 30 MPa, respectively. Based on the similarity relation, the horizontal force was 175 kN, and the vertical force was 131 kN. Multi-stage loading was adopted, and the interval between the two adjacent loadings was five minutes. After the loading process finished, the force was maintained for about 10 min. When the stresses on each measuring point inside the model tended to be stable and balanced, the roadway was excavated. The loading speed was 0.1 kN/s, as shown in Fig. 10.Fig. 10 Loading curves Physical simulation results analysis Real-time monitoring was conducted during model loading and excavation based on the strain gauges. Four strain measuring lines were arranged on the roof, floor, and two sides of the roadway, with a total of 20 measuring points. The measuring point, R5, on the right side was destroyed during the experiment, so only 19 points were recorded, as shown in Fig. 11.Fig. 11 Strain gauge monitoring data. Data Collection Time: the number of times of data collection When the model is loaded, the strain data of the sensors are collected, as shown in Fig. 11. Strain data were collected for about 120 times in the test, and ‘Data Collection Times’ is the number of times of data collection. The strain gauge was sensitive and fragile and was easily disturbed in the test. Therefore, the shape of each curve, shown in Fig. 11, was irregular. All the points on the model showed compressive strain initially, which was the result of the pre-loading and the model pressure stress. At about the 69th sampling, the surrounding rock of the roadway had deformation space because of the excavation, and the compressive strain changed into tensile strain, which was also the main cause of roadway deformation and failure. The four strain curves indicate that the strain value and deformation trend on the roof and floor were consistent. The maximum compressive strain was about 400, and the maximum tensile strain was about 1200. The pressure strain response time of the measuring points on the roof was earlier, which was induced by top loading. The strain distribution differed for the left and right sides. Responses of the measuring points on the left side were not substantial in the pressure strain stage, with compressive strain value of measuring point L1 being the only one to exceed 200; however, the value in the tensile strain stage was greater than 1200. Measuring points on the right side showed compressive strain, but tensile strain of about 500 was relatively low. The asymmetry of strain was related to the boundary effect of the model. During the pressure strain stage, the outermost measurement point 5 generally responded first, which was determined by the stress propagation path in the model. The tensile strain was higher for measuring points closer to the goaf, which was consistent with the distribution characteristics of the loose zone in the actual engineering (Li et al. 2020a, b, c). Because the roadway was excavated from front to back and the strain gauges were glued on the back face of the model, the front part of the curve fluctuated as a result of excavation disturbance. The entire process was recorded in the test including roadway excavation, deformation, and failure. Semi-circular arch sections were the first to deform on the two straight wall feet and the floor, because they were more prone to stress concentration, as shown in Fig. 12a. Figure 12b indicates that local failure appeared at the two sides and roof. With continuous increase of the deformation, rock blocks fell from the roof, and tensile cracks along the roadway trend were developed on the arch shoulders and springing line, as shown in Fig. 12c and d. Next, the deformation on the roof extended to the deeper zone and the falling blocks were more serious. Collapse of this scale would seriously threaten the safety of construction workers in an actual project. Figure 12e and f showed apparent floor heaving, with serious cracking and swelling of the roadway, which was similar to the phenomenon in the field investigation. At later times, a dislocation occurred under the action of horizontal stress, so that a pointed or peach section was formed and the roadway roof was divided into two parts along a longitudinal crack. Floor heaving gradually affected the deeper rock mass, and an independent wedge uplift was developed on the floor. Fractures on the two sides constantly expanded, and deep surrounding rocks turned over and squeezed inward to the free face, as shown in Fig. 12g. Figure 12h presented the final state of the roadway at the end of the test. The surrounding rock collapse on the roof was serious and peach roof was formed because of rock mass displacement; the floor heaving was obvious, the amount of uplift was large, and the affected zone was wide; several considerable scale tensile cracks were developed on the two sides, which became the breakthroughs of the surrounding rock converged into the excavated area.Fig. 12 Process of the roadway deformation failure; a deformation on the wall foot and floor; b failure appeared at the two sides and roof; c falling blocks from the roof and cracks on the sides; d tensile cracks on the arch shoulders and springing line; e serious collapse; f floor heaving; g development of the deformation failure; and h final state of the roadway Roadways in the research area experienced severe deformation failure under the condition of no support, and the roadways could not be put into normal service with reasonable support methods. In early stage of the excavation, a timely primary support, such as bolt net and spraying, was able to limit the stress concentration deformation and collapse. Then, after a proper pressure relief, a secondary support including bolts, grouting, and rigid support could be adopted to limit the radial displacement and fully mobilize the self-bearing capacity of the surrounding rock mass. Finally, rigid supports were used to restrain creep of surrounding rock mass and ensure long-term stability of the roadway. Numerical simulation Discrete element method (DEM) numerical simulations were performed to further study the deformation failure mechanisms of roadways in fractured rock mass and to verify the results of the physical model test. DEM was suitable for the simulation of rock mass which was anisotropic and nonlinear (Bai and Tu 2020) using PFC2D software (Itasca 2008). PFC2D was selected for this because it represents the rock masses with the following characteristics:Particle combinations interact with each other through boundary contact intrusion. Interaction between the elements can reflect the discontinuity of a rock mass and the characteristics of common occurrence of joints. Iterative calculations are adopted and large displacement and rotation are allowed. Numerical simulation model The numerical simulation model with size of 30 × 30 m was constructed based on the roadway size in Jinchuan mine, as shown in Fig. 13. The excavated roadway was a semi-circular arch with a height of 4.5 m and a width of 4.5 m. There were 39,402 particles with a diameter of 6–10 cm and 70 structural planes whose inclination angle was 45° and spacing was 1.2 m. A parallel bond model (PBM) was selected, which was suitable for the mechanical analysis of rock materials (Liu et al. 2020).Fig. 13 The numerical simulation model The PFC model does not permit direct adoption of the macroscopic mechanical parameters of the rock mass, so a calibration model for micromechanical parameters was necessary. Rock parameters and structural plane parameters were grouped and calibrated in this study. Therefore, uniaxial compression tests were conducted on intact rock specimens and rock specimens with structural planes in the study area. Through repeated simulation calculations, the microscopic parameters shown in Table 4 were obtained, and the experimental comparison was shown in Fig. 14.Table 4 Microscopic parameters in PFC (Li et al. 2021) Type Parameter Magnitude Parameter Magnitude Particles Density (kg/m3) 2500 Young’s modulus of particle (GPa) 20 Minimum particle radius (mm) 60 Ratio of normal to shear stiffness 2 Ratio of maximum to minimum particle radius 1.67 Friction coefficient 0.5 Parallel Bond Young’s modulus of particle (GPa) 20 Cohesion (MPa) 20 Ratio of normal to shear stiffness 1 Internal friction angle (°) 25 Tensile strength (MPa) 20 Bond radius multiplier 1.5 Structural Planes Cohesion (MPa) 0.1 Tensile strength (MPa) 0.1 Internal friction angle (°) 25 Tensile strength 1 Fig. 14 Calibration results (Li et al. 2021); a intact rock and b rock with structural plane Servo loading was applied around the model, and geo-stresses of three different depths were adopted in the numerical calculation, as shown in Table 5.Table 5 Geo-stress applied in PFC Depth (m) σV (MPa) σH (MPa) 550 10 20 750 20 30 1000 30 40 Numerical simulation results analysis Based on the numerical model construction approach introduced above, three roadway models in fractured rock mass were developed under different in situ stress with results of the numerical simulation shown in Fig. 15. For roadways in fractured rock mass, several groups of intersecting structural planes were developed in the surrounding rock, cutting the rock mass into independent blocks, which had the tendency to slide along the joint planes. Deformation failure modes, mainly including collapse, rock burst, and floor heaving, were similar for roadways of various depths. The wedge blocks located on the right side and roof were extremely unstable, which would break away from the deep surrounding rock as soon as the free surface appeared. If the ground stress was low, the block slid toward the goaf along the structure planes. Instead, the block was thrown to develop a rock burst or collapse. Blocks on the left side were relatively stable, the surrounding rock was squeezed into the goaf with the deformation increasing, forming side cracking and rib spalling. Blocks located on the floor were easily uplifted under the action of in-situ stress, which led to multiple or partial floor heaves. Roadway deformation became more violent and rapid with higher ground stress. Thus, the structure of surrounding rock determined the deformation mode and the geo-stress determined the deformation intensity of roadways.Fig. 15 Results of the numerical simulation Displacements on the roof, floor, and two sides of the roadways were recorded, with results shown in Fig. 16.Fig. 16 Displacement–time step curves on important positions of the roadway Total amount and speed of deformation was higher when there was higher ground stress, shown in Fig. 16. Deformations on the left side and floor of the roadway with a buried depth of 550 m were greater, exceeding 20 cm, while the deformations on the right side and roof were lesser, at about 10 cm. Deformation growth rate of each position gradually slowed down and the roadway was stable at the end of the simulation. In the roadway with a depth of 750 m, the deformation on the left side exceeded 75 cm, the right side and floor exceeded 30 cm, and the roof exceeded 15 cm, which increased with varying degrees. Because of the more unstable block combination, the left side had the fastest growth, while the other parts were all individual block movement. Under the buried depth of 1000 m, the deformation law of each position only slightly changed. The deformation on the left side exceeded 100 cm, and the floor and right side were close to 100 cm. Each position rapidly deformed when the simulated depths were 750 m and 1000 m, especially when the buried depth was 1000 m, because most of the roadway section was occupied by surrounding rocks. Conclusion The following conclusions are drawn from a comprehensive comparative analysis of the physical model tests and numerical simulations:A mixed pouring–bricking method most accurately represents the fractured rock failure mechanisms. Pouring was adopted in the inner ring of the model, and bricking was used in the outer ring, which can not only reflect the structural characteristics of jointed surrounding rocks, but also improve the test efficiency. A roadway excavation device (based on an embedded mold) and spiral traction were independently developed to represent the roadway in a model. The physical model based on the roadway in the Jinchuan mining area can truly reproduce the structural characteristics of the prototype and fully replicate the excavation conditions of the roadway, which proves the higher practicability and efficiency of the method. This provides a technical reference for the design and production of similar model experiments. Deformation failure processes of the fractured rock mass roadways with semi-circular arch section were replicated. The deformation initially occurred on the two straight wall feet and the floor. Then, failure appeared at the two sides, blocks fell from the roof, and tensile cracks developed along the roadway trend with continuous increase of the deformation. Additionally, the floor of the roadway cracked and swelled, and the floor gradually heaved, which affected the deeper rock mass. Finally, a peach roof was formed and several considerable scale tensile cracks developed on the two sides, which became the breakthroughs of the surrounding rock converging into the excavated zone. Deformation failure modes of roadways were similar regardless of depths, which mainly included collapse, rock burst, and floor heaving. Roadway deformation became more violent and rapid with the increase of ground stress. The structure of surrounding rock determined the deformation mode and the geo-stress determined the deformation intensity of roadways. Roadway construction in complex geological environments, such as fractured rock masses, requires support technology. A timely first support was needed to prevent initial collapse and rock burst, but a rigid support to ensure the long-term stability of the roadway was also needed. This study performs experimental research on the deformation failure processes of roadway tunnels in fractured rock masses that are induced by mining excavation. Although several meaningful observations are made, there are still some limitations, as follows. (1) Because the drying time of the physical model test was so long (due to the constraints of the COVID-19 pandemic), the bond strength between blocks was high and resulted in the experiment failing to fully reflect the discontinuity action. (2) The structure plane parameters were not accurately calibrated in the numerical model, so the structural planes were simplified and differed from the actual rock mass structure. In the future, the model can be optimized according to the actual observed joint fracture distribution to truly reflect the deformation failure characteristics of roadway surrounding rock. (3) Both physical and simulation tests were two-dimensional representations of the system, ignoring the influence of one horizontal in-situ stress, and three-dimensional tests should be supplemented in future studies. Acknowledgements The research received support from the National Natural Science Foundation of China (Grant Nos. 42072305, 41877274, and 41831293). We appreciate the kind support. Funding All authors agree to submit the paper to this journal. The authors declare that the supporting source had no such involvement. Declarations Conflict of interest The authors declare no conflict of interest. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References An BQ Tannant DD Discrete element method contact model for dynamic simulation of inelastic rock impact Comput Geosci 2007 33 4 513 521 10.1016/j.cageo.2006.07.006 Bai QS Tu SH Numerical observations of the failure of a laminated and jointed roof and the effective of different support schemes: a case study Environ Earth Sci 2020 79 202 10.1007/s12665-020-08935-2 Cai M Kaiser PK Morioka H Minami M Maejima T Tasaka Y FLAC/PFC coupled numerical simulation of AE in large-scale underground excavations Int J RockMech Min Sci 2007 44 4 550 564 10.1016/j.ijrmms.200(6)09.013 Fakhimi A Carvalho F Ishida T Labuz JF Simulation of failure around a circular opening in rock Int J Rock MechMin Sci 2002 39 4 507 515 10.1016/S1365-1609(02)00041-2 He MC Gong WL Li DJ Physical modeling of failure process of the excavation in horizontal strata based on IR thermography Min Sci Technol 2009 19 06 689 698 10.1016/S1674-5264(09)60128-9 Hendron AJ, Paul Engeling, Aiyer AK et al (1972) Geomechanical model study of the behavior of underground openings in rock subjected to static loads(report 3)-tests on lined openings in jointed and intact rock. Illinois Univ at Urbana Dept of Civil Engineering Huang F Zhu HH Xu QW The effect of weak interlayer on the failure pattern of rock mass around tunnel-Scaled model tests and numerical analysis Tunn Undergr Space Technol 2013 35 207 218 10.1016/j.tust.2012.06.014 Hui X Ma FS Zhao HJ Xu JM Monitoring and statistical analysis of mine subsidence at three metal mines in China Bull Eng Geol Environ 2019 78 3983 4001 10.1007/s10064-018-1367-6 Itasca Consulting Group PFC2D (Particle Flow Code in 2dimensions) users guide 2008 Minneapolis Itasca Lee YZ Schubert W Determination of the length for tunnel excavation in weak rock Tunnel Underground Space Technol 2008 23 221 231 10.1016/j.tust.2007.04.001 Li SC Wang Q Wang HT Model test study on surrounding rock deformation and failure mechanisms of deep roadways with thick top coal. Tunnell Underground Space Technol 2015 47 52 60 10.1016/j.tust.2014.12.013 Li G Ma FS Liu G A strain-softening constitutive model of heterogeneous rock mass considering statistical damage and its application in numerical modeling of deep roadways Sustainability 2019 11 2399 10.3390/su11082399 Li G Ma FS Guo J Study on deformation failure mechanism and support technology of deep soft rock roadway Eng Geol 2020 264 105262 10.1016/j.enggeo.2019.105262 Li G Ma FS Guo J Deformation characteristics and control method of kilometer-depth roadways in a nickel mine: a case study Appl Sci 2020 10 3937 10.3390/app10113937 Li G Ma FS Guo J Experimental study on similar materials ratio used in large scale engineering model test J Northeastern Univ (natural Science) 2020 41 11 1653 1660 10.12068/j.issn.1005-3026.2020.11.021 Li G Ma FS Guo J Case study of roadway deformation failure mechanisms: field investigation and numerical simulation Energies 2021 14 1032 10.3390/en14041032 Liu WR Wang X Li CM Numerical study of damage evolution law of coal mineroadway by particle flow code (PFC) model Geotech Geol Eng 2019 37 2883 2891 10.1007/s10706-019-00803-6 Liu SQ Ma FS Zhao HJ Numerical analysis on the mechanism of hydraulic fracture behaviour in heterogeneous reservoir under the stress perturbation J Nat Gas Sci Eng 2020 78 103277 10.1016/j.jngse.2020.103277 Lu R Fs Ma Guo J Zhao HJ Monitoring and analysis of ground subsidence and backfill stress distribution in Jinchuan Mine, China Curr Sci 2018 115 10 1970 1977 10.18520/cs/v115/i10/1970-1977 Lu R Fs Ma Guo J Monitoring and analysis of stress and deformation features of boundary part of backfill in metal mine Sustainability 2020 12 733 10.3390/su12020733 Norman Brook (1981) Small scale brittle model studies of mine roadway deformation. In: Farmer IW (eds) Developments in geotechnical engineering, vol. 32. Elsevier, pp 184–189. 10.1016/B978-0-444-42086-2.50035-1 Pf G Zhang ZP Wei SJ Physical simulation test of damage character of surrounding rock under different levels of the horizontal stress J China Coal Soc 2009 10.13225/j.cnki.jccs.2009.10.011 Ren FQ Chang Y He MC A systematic analysis method for rock failure mechanism under stress unloading conditions: a case of rock burst Environ Earth Sci 2020 79 370 10.1007/s12665-020-09111-2 Shreedharan S Kulatilake PHSW Discontinuum–equivalent continuum analysis of the stability of tunnels in a deep coal mine using the distinct element method Rock Mech Rock Eng 2016 49 5 1903 1922 10.1007/s00603-015-0885-9 Sun XM Chen F Mc He Physical modeling of floor heave for the deep-buried roadway excavated in ten degree inclined strata using infrared thermal imaging technology Tunnel Underground Space Technol 2017 63 228 243 10.1016/j.tust.2016.12.018 Tu HS Tu SH Wang C Mechanical analysis of a vertical-wall, semicircular-arch roadway and a repair technique using double-shell support Environ Earth Sci 2018 77 509 10.1007/s12665-018-7680-3 Wang X Tian L Mechanical and crack evolution characteristics of coal–rock under different fracture-hole conditions: a numerical study based on particle flow code Environ Earth Sci 2018 77 8 297 10.1007/s12665-018-7486-3 Wang C Tannant DD Lilly PA Numerical analysis of the stability of heavily jointed rock slopes using PFC2D Int J Rock Mech Min Sci 2003 40 3 415 424 10.1016/S1365-1609(03)00004-2 Xu YL Pan KR Zhang H Investigation of key techniques on floor roadway support under the impacts of superimposed mining: theoretical analysis and field study Environ Earth Sci 2019 78 436 10.1007/s12665-019-8431-9 Zhang MJ Gao JH Wei SY Similarity simulation study of failure characteristics of surrounding rocks of tilted strata roadway Rock Mech Eng 2010 29 1 3259 3326
PMC009xxxxxx/PMC9005914.txt	==== Front Eur J Dev Res Eur J Dev Res The European Journal of Development Research 0957-8811 1743-9728 Palgrave Macmillan UK London 35431467 529 10.1057/s41287-022-00529-x Impact Evaluation Article Effects of Mobile Money Education on Mobile Money Usage: Evidence from Ghana http://orcid.org/0000-0003-2095-6862 Apiors Emmanuel Kwablah emmanuel.apiors@s.k.u-tokyo.ac.jp Suzuki Aya ayaszk@k.u-tokyo.ac.jp grid.26999.3d 0000 0001 2151 536X Department of International Studies, The University of Tokyo, Kashiwa, 277-0882 Japan 13 4 2022 2023 35 3 715742 21 12 2021 © European Association of Development Research and Training Institutes (EADI) 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. This study conducts a randomised control trial to offer a technical workshop and examine whether providing information about the full range of services on the mobile money platform would increase mobile money usage, by taking a case of the Ashanti Region, Ghana. We find a significant positive impact of mobile money education on the recent usage of mobile money for transactions. However, no significant evidence of the workshop was found on new mobile money account ownership, or on the share of transactions transmitted through mobile money. Furthermore, weak and volatile outcomes were observed as impacts on remittances after the interventions. We discuss potential reasons behind the weak effects found. Supplementary Information The online version contains supplementary material available at 10.1057/s41287-022-00529-x. Résumé Cette étude consiste en un essai randomisé contrôlé qui propose un atelier technique et cherche à savoir si le fait de fournir des informations à propos de la gamme complète de services offerts par la plateforme d'argent mobile permettrait d’augmenter l'utilisation de la monnaie mobile. Pour ce faire, l’étude s’appuie sur le cas de la région d'Ashanti, au Ghana. Nous constatons que l'éducation liée à l'argent mobile a un impact positif significatif sur l'utilisation d'argent mobile pour des transactions récentes. Cependant, nous n'avons trouvé aucune preuve significative que l'atelier a eu un impact sur l’ouverture de nouveaux comptes d'argent mobile, ni sur la proportion des transactions réalisées grâce à de l’argent mobile. En outre, en termes d’impact, des résultats faibles et instables ont été observés en matière d’envoi de fonds après les interventions. Nous discutons des raisons potentielles pouvant éclairer la faiblesse des effets relevés. Keywords Mobile money education Technology adoption Financial literacy Financial inclusion Payments Ghana Africa JEL Classification C93 D14 D83 G21 O16 O32 O33 KDDI Foundation (JP)issue-copyright-statement© European Association of Development Research and Training Institutes (EADI) 2023 ==== Body pmcIntroduction Mobile money has received attention in recent years as a financial inclusion tool, particularly in Sub-Saharan Africa (Suri et al. 2012; Karlan et al. 2017). Access to mobile money increases savings and reduces vulnerability to shocks (Jack and Suri 2014); increases consumption by increasing remittance receipts (Munyegera and Matsumoto 2016; Suri and Jack 2016); increases remittances sent and received (Suri et al. 2012); and increases investment in education, micro- and small businesses, and savings (Apiors and Suzuki 2018). However, the adoption and usage of mobile money have been uneven across the developing world; studies on the intensity of its usage have mostly been concentrated around Kenya (Suárez 2016) and neighbouring East African countries. For example, while mobile money account ownership is still low among mobile phone users in Ghana, a wide gap exists between registered mobile money accounts and active mobile money accounts (Ghana 2017), thus indicates high dormancy and low usage. According to Arun and Kamath (2015), the adoption of mobile money does not automatically lead to usage because not everyone who owns a payment product uses it for inflows. Therefore, usage needs proactive pursuance to reflect the benefits derived from it. Batista and Vicente (2020) have explored mobile money adoption patterns and impacts on welfare outcomes in rural Mozambique, focussing on the newly introduced mKesh. In a field experiment, they organised community meetings to disseminate information about mKesh, and distributed leaflets about mKesh. They found that mKesh availability resulted in mobile money services adoption, although adoption decreased over time. Particularly, mKesh adoption increased remittances received, consumption, and reduced vulnerability to hunger, but reduced agricultural investments. However, these studies have been limited to strictly rural populations and non-mobile money user samples and have not been tested on a more general sample. Two arguments surround the low usage of financial services. First, financial services are too expensive to provide (Beck et al. 2007), especially in rural areas, and for low-income people. The second, and our primary focus is low financial literacy, a barrier which may be caused by an inadequate supply of information, or users’ inability to understand their usage. To use a mobile money platform independently, users need to have technical information regarding the usage of mobile money applications, and basic financial literacy, owing to the diverse financial products offered through mobile money services. Moreover, many mobile money kiosks (where users can receive or transfer mobile money) are located widely even across rural areas in developing countries. However, official mobile telecommunication shops, which consumers can access when they face problems or have questions, are limited to urban areas, and are not easily accessible for the majority of consumers. It is likely that consumers lack access to the information necessary to use mobile money effectively. Evidence suggests that the significant barrier of low financial literacy on the demand side of financial services can be overcome through well-designed and targeted training. However, existing studies are limited and report mixed results (Hastings et al. 2013). Cole et al. (2011) found that financial education had moderate to no effect on the likelihood of opening an account. Fernandes et al. (2014) found that financial education explained only 0.1% of financial behaviour, and Collins (2013) observed a lack of impact of financial education on savings behaviour. However, Bruhn et al. (2016) found that monetary incentives may convince individuals to attend financial education programmes. Other studies (Mitchell and Lusardi 2011; Lusardi and Tufano 2015; Miller et al. 2015) found strong evidence that the comprehension of financial concepts and improved financial decisions are positively related, and translate into household welfare. However, most financial education studies are not specific to mobile money usage. Essentially, mobile money service users need to possess more than basic financial literacy, as practical or legal knowledge that can protect from fraud is crucial. We examine whether providing information about the full range of services on the mobile money platform will increase user participation. For an empirical case, our study builds on the study of Batista and Vicente (2020) by examining the effects of detailed information of mobile money on both users and non-users, in the context of urban, peri-urban, and rural communities in Ghana. Specifically, considering the ubiquitous nature of mobile money agent networks nationwide, we examined how mobile money education workshop enhances mobile money account ownership, recent use of mobile money for a transaction, and the share of financial transactions conducted through mobile money. Through a randomised control trial (RCT) experiment, we provided information about mobile money in a workshop by inviting 216 randomly selected people out of a sample of 557. Using two rounds of data collected in a panel, we examined the effects of the workshop on the financial behaviour of mobile phone users. We employed fixed effect (FE) and fixed effects instrumental variable (FEIV) estimation methods to correct for possible endogeneity of workshop participation. We found no significant treatment effect of mobile money education on mobile money account ownership. However, we observed strong positive impacts on recent transactions undertaken using mobile money after the workshop. This means that the education provided a better understanding of the benefits of using mobile money, hence, participants utilised the services, although they were not eager to open mobile money accounts. Conversely, after the education, mixed and weakly significant outcomes were observed as impacts on the remittances sent and received. This weak significance may be explained by the awareness of a 1% transaction fee charged to every amount sent or withdrawn from a mobile wallet. Heterogeneous analyses revealed joint significance effects of location and household size on mobile money activities. This study contributes to experimental evidence regarding the effects of mobile money education on customers’ financial transactions. The participants’ recent usage of mobile money for financial transactions increased after the workshop, indicating that limited knowledge on the use of this service was likely the reason for low usage in Ghana. While mobile money was well known among the sample many participants were unaware of the variety of services provided with mobile money. The service industry is relatively new in Sub-Saharan Africa, and, by the nature of the industry, the development of customer services is important, for example providing necessary technical information, or easy access to service desks to solve problems that customers may have. Our findings shed some light on the need to improve the information provision to customers, in order to enhance mobile money adoption. The remainder of this paper is structured as follows. “Study Context and Hypotheses” section explains the study context and hypotheses, “Data and RCT” section describes data and the experiment, “Estimation Strategy” section enumerates the estimation strategy, while “Main Results” section presents the main results. “Mechanism and Discussion” section discusses the results, and the conclusion is presented in “Conclusion” section. Study Context and Hypotheses Mobile Money Trend in Ghana Mobile money is an electronic payment and banking concept using subscriber identification module (SIM) cards in mobile phones (The World Bank 2012), and it is aimed at bringing financial services to the unbanked (Tobbin and Kuwornu 2011). In this study, mobile money is electronic cash backed by an equivalent amount of Bank of Ghana notes and coins, stored using the SIM card in a mobile phone as an identifier (Ghana 2017). The storage of value function leads to quarterly interest payments (1.5–7%) to customers based on balances in their wallets. For user’s convenience, mobile money wallets may be linked to personal bank accounts, to provide access to a variety of financial services (GSMA 2010) designed to meet the needs of the economically vulnerable or the unbanked (Page et al. 2013). Mobile voice call subscription (a proxy for both mobile analogue and smartphones ownership) has been on a sharp rise since 2012 (Fig. 1). As mobile phone subscription grew at an average annual rate of 9.09% of projected population between 2012 and 2016, registered mobile money accounts and active mobile money accounts grew at 13.77% and 7.0%, respectively. Note that citizens are free to own multiple—mobile phones, mobile money accounts, and bank accounts, thus the absolute figures give a clearer picture. Also, although, people in Ghana use both analogue and smart phones, the National Communication Authority is yet to provide statistics on their proportions. In 2013, there were approximately four million mobile money accounts, fewer than the approximately seven million bank accounts. However, by December 2017, the number of mobile money accounts increased to approximately 24 million, twice higher than bank accounts (National 2018). Conversely, the number of active mobile money accounts rose steadily from 2012 and plateaued at approximately the same figure as that of bank accounts countrywide by December 2017. This shows that while registered mobile money accounts increased rapidly, active mobile money accounts trailed at an unequal pace.Fig. 1 Mobile money accounts vs. formal bank accounts in Ghana. Source Authors drawing from Bank of Ghana data 2018 Information and Transaction Costs In developed countries, telecom companies sell and subscribe consumers to mobile telephony and data altogether. In Ghana, however, mobile phones, mobile network and data subscriptions, and mobile money accounts are three independent products, offered separately by various actors. While mobile phones can easily be purchased from retail shops, mobile network and data subscriptions, and mobile money accounts are only available through mobile telecommunication companies. To own a mobile phone, one must purchase it from the free market. To access a network’s services, one must subscribe by purchasing a SIM card of a preferred network, the network operator then registers and activates the SIM. Similarly, to own a mobile money account, one must consult the respective telecom company’s centre to register and activate a SIM that is already subscribed to the same network’s telephony. Unlike in developed countries, where consumers receive packaged information about mobile phone acquisition and usage subscription contracts, in Ghana, no specific packaged information is given to consumers when they purchase mobile phones or subscribe to telephony networks, although consumers may receive oral information about on-going subscription promotions. Concerning mobile money accounts, the basic information provided is the importance of consumers keeping their generated personal identification number (pin) as a secrete, that the platform is used to send and receive cash. Once a consumer requests mobile account ownership, their knowledge about mobile money is assumed. To access further information, consumers are directed to telecom companies’ general complaint helplines, or to one of the few customer complaints centres in the cities and big towns. Consumers’ awareness and knowledge about mobile money is generally low, as they lack opportunities to obtain the necessary information. The main form of dissemination campaign done to promote mobile money is a short 30–60 s radio and television advertisements (Online Appendix A) on the remittance function, but in-depth knowledge about the related products offered by the mobile money platform is low. The adverts duration did not change overtime, but the content for each company changed based on new products available to consumers, for example, connecting to bank accounts, withdrawing money using automated teller machines (ATMs), receiving international remittance, and interoperability. About 78% of our baseline sample had advert as their source of first time information about mobile money. Only 5% received their first information from operator text message, 10% from mobile money vendors and 7% from other sources. Although the short adverts and codes direct prospective consumers to the nearest mobile money vendor for further information, these vendors are usually busy performing only sending and receiving transactions for users. For instance, our baseline survey shows that many people lack detailed knowledge about the full range of products, apart from sending and receiving money, although there are more products and services to enable consumers to be financially included, other than remittances. At baseline, 68.4% of all mobile money account holders did so with the reason of either receiving or sending remittances. Only 11% and 9% had reasons for saving and investment, and facilitation of day-to -day living, respectively, while the remaining 11.6% stated other non-specified reasons for owning mobile money accounts. Seventy-five percent (75%) of non-mobile money account holders at baseline indicated reasons such as not knowing much about the product and not expecting to receive money from anyone or do not see the need. The remaining 25% stated personal reasons for not owning mobile money accounts. Inadequate information centres in consumer neighbourhoods fuelled the lack of access to information. Users pay a two-way transaction cost of 1%. When users transfers funds from their mobile wallet, they pay 1% of the amount transferred. When a receiver goes to cash the amount received at a vendor, 1% of the amount withdrawn is paid. When an account holder sends money to a non-account holder, either the sender or receiver pays a 3% transaction charge. Hypotheses Given the above situations, we examine whether providing more information on mobile money usage promotes people’s use of mobile money ex post. In particular, we investigate the impacts of our intervention on mobile money account ownership, recent transaction, and share of various transactions relative to total financial transactions made. Specific hypotheses are as follows. Hypothesis 1 (H1) Mobile money education contributes to increased mobile money account ownership. Hypothesis 2 (H2) Mobile money education contributes to increased usage of mobile money for a recent transaction. Hypothesis 3 (H3) Mobile money education contributes to an increase in the share of payments, remittances, and savings conducted through mobile money. Mobile money accounts are not necessary when users want to receive or send money through mobile money platforms, but they are necessary to save on the platform. Thus, H1 will show whether providing information can deepen the peoples’ use of mobile money. While H2 will show us the general impacts on the use of mobile money, H3 will show which type of transaction is affected by information dissemination and by how much, relative to the total amount of transactions conducted including the amount transacted by traditional channels. Data and RCT Data The analyses combined data from two surveys. We conducted the baseline survey in September 2016 and February 2017, and randomised the sample into treatment and control groups. The intervention was implemented through a workshop in March 2017. A follow-up survey was conducted 6 months after the intervention (Fig. 2). The sample composed individuals, rather than households. We used stratified random sampling approach to select respondents. First, all ten regions of Ghana were considered, based on the 2010 Population and Housing Census. We selected the Ashanti Region because it’s most populous and represented rural, urban, and peri-urban communities. One district each was selected from urban [Kumasi Metropolis (100% urban)], peri-urban [Obuasi Municipality (61% urban)], and rural [Amansie West District (96% rural)].Fig. 2 Study timeline In each district, we stratified communities using population and level of commercial activities. Using financial institutions, nature and scale of commercial activities (Online Appendix B) we selected two communities, one of high commercial activity and another which is low, from which respondents were randomly sampled. While we acknowledge that it is ideal to have a list of residents and pick respondents randomly from the list, such lists are not available. Thus we had our enumerators enter the communities and talk to people randomly within a radius of 3 km from the station.1 To examine the impact of the intervention on both users and non-users, mobile money user status was used to stratify the sample within communities. To obtain user or non-user categories, randomly approached prospective participants were asked whether or not they were users. A mobile money user was any adult who had ever used mobile money to conduct a transaction, at least once, before the baseline survey. A non-user was any adult who had never used mobile money to do any transaction before the baseline survey. The prospective participants were informed about the study, and as with most surveys, participation in our interview was voluntary. Non-participation rate was 10%.2 The target sample comprised adults who were at least 18 years and owned a mobile phone. Precaution was taken to ensure that gender was balanced among the participants surveyed. A mobile money account holder described an adult who had registered a personal phone number as a wallet to conduct any self-initiated mobile money transaction. Note that one does not need to hold a mobile money account to be a user. However, certain features on the platform are only available for mobile money account holders, such as savings on and the purchase of government investment bonds (TBILL4ALL). The informal sector employs about 80% of the labour force, therefore participants were mainly informal sector workers engaged in retail and other self-employments. In total, 557 participants were enrolled in the baseline survey, 388 users and 169 non-users. Table 1 shows the number of respondents surveyed per location. The end-line survey was completed by 81% of the baseline sample (Table 2).Table 1 Study participants by location. Source Authors’ survey Location User (1) Non-user (2) Total (3) Urban 130 60 190 Peri-urban 120 60 180 Rural 138 49 187 Total 388 169 557 Table 2 Participants per survey round. Source Authors’ survey Survey rounds Survey sample (1) Attrition (2) Baseline 557 – Endline 456 18.1% Randomisation From each six communities, 36 respondents were randomly selected and invited to attend the education workshop. Of these 36 respondents, half were mobile money users, and the other half, non-users. In total, 216 randomly selected respondents formed the treatment group, and the rest (341) formed a control group. Table 3 indicates the number of observations randomised into the treatment group (invited and participated) versus the control group.Table 3 Randomisation and take-up (treated). Source Authors’ field experiment User (1) Non-user (2) Total (3) Treated Invited (participated) 108 (55) 108 (51) 216 (106) Controlled 280 61 341 Total 557 Approximately half of the invitees attended the workshop (Table 3). The low take-up rate was partly due to respondents being sceptical about the workshop and information relating to mobile money. During the intervention, frequent incidences of fraud were reported by mobile money users. Nevertheless, the take-up rate was very similar between the user and non-user groups. To avoid non-invited respondents at the workshop, during the baseline survey, all respondents were informed that only randomly selected respondents would be invited for the workshop, that the non-selected would receive the information after the study. Randomly selected participants were informed through a phone call—1 month, 1 week, a day before the workshop, and on the day of the workshop; they were asked to keep their selection a secret, and not to invite anyone when coming to the workshop. Therefore, although it was anticipated that some non-invitees might show up at the workshop, especially in the rural areas, no non-invitees attended the workshops. The RCT The RCT was conducted in collaboration with telecom companies that offer mobile money products in Ghana; AIRTEL, MTN, and VODAFONE. This study liaised with officials of these companies to design an education programme about the unique product offerings each company has on its platform. Apart from the packaging, SIM cards, and short codes, the core product (electronic cash) is the same. Therefore, it has been possible since 2018 to transfer the store of value (core product) from one company’s SIM to another. The three companies promoted the same product and the different packages (telecom specific) it comes in (Fig. 3). Photos 1 and 2 display the outline of the content taught by officials from MTN and VODAFONE, respectively. The participants were educated on AIRTEL Money, MTN Mobile Money, and VODAFONE Cash. Each company had representatives in each community. Each representative was supervised by the same regional heads and the research team, during the delivery of the prepared content at each location.Fig. 3 Mobile money account ownership. Source Authors’ drawing from field data Photo 1 Image of the outline taught by official from VODAFONE Photo 2 Image of the outline taught by MTN A two-hour workshop was organised in each of the six communities. The content taught participants about electronic cash. It informed about product kinds available on the mobile money platform; how to use the products regularly, how participants can access the products on personal phones, and how to maintain vigilance to prevent fraud. The telecom officials answered questions raised by attendees, mobile money accounts were opened only for new customers who were willing to do so after they had understood how it worked and participants were allowed to practice the codes and processes on their mobile phones using their own money. Fraud incidences and the various forms it takes on the platform were discussed. Officials taught customers how to prevent fraud and explained the transaction charges associated with services on the mobile money platform. Participants evaluated the effectiveness of the workshop, after. Overall, the workshop was effective; 60% of participants reported that the workshop content ‘was just right’, 30% reported that ‘it was easy’, and 8% said ‘it was too easy to understand’. However, 2% reported that ‘it was difficult to understand’. Fifty-six percent of the participants said that ‘it is beneficial’, and 44% said, ‘It will be helpful’. The telecom officials evaluated the workshop and said “it was an eye-opener”, “it gave us insight into customers’ awareness of mobile money”. Estimation Strategy Treatment Effects The analyses of workshop impacts on the use of mobile money for daily activities, focussed on mobile money account ownership, payments sent and received, remittances sent and received, savings, and the share of these transactions through mobile money. In a simple form, the aim is to estimate the coefficient, β3 in the following equation based, on the panel data:1 Yit=β0+β1Treati+β2time1+β3Treati×time1+β5Xit′+γi+δt+εit, where Yit is the outcome variable of interest for participant i at month t; Treati is a dummy variable, equal to 1 if the person is “treated” and 0 otherwise; timet indicates the follow-up survey period; Xit′ is a set of socioeconomic characteristics, γi is an individual fixed effect; δt is a time fixed effect; and εit is an error term. We examine both the “intention-to-treat (ITT)” effects and the “treatment-effect-on-the-treated (TOT)”. ITT shows the causal impact of offering the treatment (being “invited”) on participants’ financial behaviours and is used often by policymakers as it offers the average effects of being invited. TOT shows the causal impact of the actual exposure to treatment on participants’ financial behaviour and is useful to the evaluator (Ravallion 2007; Glennerster and Takavarasha 2013). Random invitation status was used as the Treati variable in estimating ITT, whereas actual participation status to the workshop was used as the Treati variable for TOT. Both models were run by FE and FEIV estimations. To estimate TOT, it is necessary to consider the potential endogeneity of Treati. Especially, workshop attendance was very low among the treatment group, indicating that invitees to the workshop may have self-selected themselves to attend. The effects of any unobserved systematic differences between the participants and non-participants among the invited will be captured in the error terms. This would violate one of the necessary assumptions for ordinary least squares (OLS), which states that the covariance between independent variables and an error term should be zero. To correct for any such endogeneity, the random invitation to the workshop (equal to 1 if randomised into a treatment group, and 0 otherwise) was used to instrument the actual participation status to estimate the FEIV. While we report the FE and FEIV results, the random effect, pooled OLS, and ANCOVA specifications were also estimated. The results were not very different. Heterogeneous Effects To examine whether specific characteristics of respondents affect the impacts of the workshop, the heterogeneous impacts focussed on differences due to gender, bank account ownership, household size, rural and urban dummies. Gender may affect people’s financial decisions. Females are more likely to participate in mobile money compared to males (Apiors and Suzuki 2018). Hence, how gender and workshop information jointly affected usage may be useful for policy. Bank accounts can either complement or compete with mobile money accounts ownership. Intuitively, individuals with larger household sizes are likely to have higher household expenses; hence are more likely to transmit higher amounts through mobile money. Rural communities may pose a challenge for cost-effective distribution of financial access (Beck et al. 2007). A user’s location may either promote or inhibit mobile money adoption and usage. Attrition Analyses Attrition within the sample was low (18.1%) (Table 2), however, we examined whether this attrition posed validity issues to the analyses (Glennerster and Takavarasha 2013). We investigate whether attrition systematically affected the outcome variables of interest, using the baseline sample in Eq. (2):2 Y=∝0+β1T+β2Attrit+β3T∗Attrit+ε, where Y is a primary outcome variable at the baseline, T is an indicator for the treatment group, and Attrit is an indicator for attrition. β3-β2 was shown to be statistically insignificant for each baseline variable using an F-test. Main Results Socioeconomic Characteristic of Respondents Table 4 shows the t-test results at baseline, indicating how treatment and control groups compared before the intervention. The treatment group comprises the invitees. In column (3), the treatment and control groups did not vary significantly in most of the socioeconomic variables, except for the years of education, suggesting that the randomisation was relatively successful. The average years of education was 8.5 years for the treatment group, a little lower than 9.3 years for the control group.Table 4 Socioeconomic variables’ summary at baseline. Source Authors’ survey Variables Treatment (N = 216) (1) Control (N = 341) (2) p value of difference (3) Age (years) 32.296 (12.15) 32.935 (12.42) 0.551 Education (years) 8.47 (4.5) 9.32 (3.86) 0.017* Gender (1 = male) 0.51 (0.5) 0.52 (0.5) 0.79 Married (1 = yes) 0.43 (0.49) 0.43 (0.49) 0.94 Household size 3.56 (2.2) 3.36 (2.19) 0.299 Non-household dependents 0.44 (0.64) 0.47 (0.65) 0.548 Employment status 0.87 (0.33) 0.85 (0.35) 0.51 Major sources of first information about mobile money Television 0.29 (0.45) 0.30 (0.46) 0.73 Radio 0.24 (0.44) 0.26 (0.43) 0.64 Friends 0.19 (0.40) 0.22 (0.41) 0.45 Vendor 0.09 (0.29) 0.10 (0.30) 0.75 Standard deviations in parentheses p < 0.1 First, we examine participant’s financial activities, using bank account ownership, awareness, and the use of mobile money features. At baseline, 54% of the sample owned bank accounts, of which 70% deposited and withdrew at least once a month. Approximately 73% of no bank account holders attributed it to low income. All respondents were aware of mobile money. However, only 63.4% had ever used mobile money to receive cash, and 61% had ever used mobile money to send cash. Only 22.5%, 7%, 5%, and 0% had ever used mobile money to save, take or repay a loan, pay utility bills, and purchase treasury bills, respectively. Thirty-seven percent of the sample had mobile money accounts, of which 90%, 7%, 6%, and 5% were with MTN, AIRTEL, TIGO, and VODAFONE, respectively. Some participants had multiple accounts with multiple service providers. Table 5 shows how financial activities compared between the treatment and control groups. Although the time between the intervention and the survey was 6 months, we used outcomes that cover the last 12 months as some of the expenditures may be annual rather than half-year base and payment timing may be different across respondents. Thus restricting the time may result in biassed answers. Besides mobile money account ownership, payments sent in the last 30 days and remittances and gifts received in the last 12 months, all other variables did not differ significantly between the treatment and control groups. This suggests that the randomisation worked moderately well regarding the outcome behaviours. The treatment group had a lower mobile money account ownership when compared to the control group. This was significant at the 1% level.Table 5 Financial activities’ summary at baseline. Source Authors’ survey Variables Treatment (N = 216) Control (N = 341) p value of difference (1) (2) (3) Ownership of Bank Account 0.54 (0.5) 0.56 (0.49) 0.157 Ownership of Mobile Money Account 0.287 (0.45) 0.436 (0.49) 0.0004* Mobile Money Sent (All transactions, last 12 months(GH¢)) 72.96 (328.44) 469.91 (3562.98) 0.103 Mobile Money Received (all transactions last 12 months(GH¢)) 201.04 (1294.83) 300.05 (2795.54) 0.625 Payments Sent (last 30 days(GH¢) 1410.09 (1619.14) 1846.75 (2957.36) 0.047* Payments Sent (last 12 months(GH¢)) 14,276.53 (57,632.24) 12,006.28 (17,544.98) 0.497 Respondents’ Income (last 12 months(GH¢)) 9002.5 (16,912.68) 7729.9 (10,901.81) 0.28 Respondents Income (last 30 days(GH¢)) 689.34 (2,409.87) 639.96 (1505.33) 0.766 Remittance and Gifts Sent (last 12 months(GH¢)) 227.56 (462.51) 251.24 (488.05) 0.56 Remittance and Gifts Received (last 12 months(GH¢)) 190.87 (409.24) 318.01 (589.53) 0.0057 Investment in Micro-Enterprise, Land, and Buildings (GH¢) 5499.07 (55,096.62) 1997.47 (6099.78) 0.245 Total savings (GH¢) 892.26 (2865.2) 1363.86 (9023.07) 0.456 Total Financial Assets (GH¢) 1376.11 (2834.22) 1438.78 (4213.55) 0.847 Household Total Physical Assets (GH¢) 1,122,518 (15,000,000) 668,523 (5,026,467) 0.605 Consumption (last 12 months in (GH¢)) 4912.45 (5646.4) 5088.2 (7234.26) 0.761 Notes Except “Mobile Money Sent” and “Mobile Money Received”, all other variables indicate the amount transacted disregarding the methods (Mobile Money, bank, or cash) In 2016, the basic minimum wage in the Ghanaian formal sector was GH¢8.00, GH¢40.00, and GH¢160.00 for daily, weekly, and monthly minimum incomes, respectively Standard deviations in parentheses p < 0.05, p < 0.1 Payments sent is the total value of cash-outs made for goods and services received. At baseline, the treatment group made an average payments of GH¢1410.09 in the last 30 days and GH¢14,276.53 in the last 12 months. However, the control group made average payments of GH¢1846.75 in the last 30 days and GH¢12,006.28 in the last 12 months. Of these payments, the treatment group sent an average of GH¢72.96 through mobile while the control group sent GH¢469.91. The treatment group received average payments of GH¢204.04 through mobile money, while the control group received GH¢300.05. Remittances sent in the last 30 days did not vary significantly; however, remittances received in the last 12 months were lower for the treatment group’s customers (GH¢190.87) compared to those in the control group (GH¢318.01). Impact on Account Ownership and Usage Figure 3 indicates how mobile money account ownership increased across the first 6 months after the intervention, for both treatment and control groups. At baseline, only 30% of the workshop attendees own mobile money accounts. However, after the intervention, 60% of the workshop attendees became new mobile money account holders. Among the control group, mobile money account ownership increased from 35% at baseline to 65% after the intervention. Table 6 presents estimations of ITT and TOT of mobile money education on participants’ likelihood of opening new mobile money accounts. The results (Table 6, column (1)) show no significant ITT effect of mobile money education on new account ownership. Similarly, no significant TOT impact is seen on account ownership, although a positive sign is observed (column (3)). Thus, the first hypothesis “mobile money education contributes to mobile money account ownership” is not supported. The coefficients of the Endline are positive and significant across models, indicating that, after the workshop, the number of people who opened mobile money accounts increased, both for the treated and control groups.Table 6 Mobile money (MM) account ownership Variables ITT TOT FE FE FEIV (1) (2) (3) InviteEndline 0.02 (0.05) ParticipateEndline − 0.03 (0.05) 0.09 (0.23) Endline 0.317 (0.028) 0.329* (0.026) 0.304* (0.056) Observations 912 912 912 R2 [Overall] 0.335 0.335 [0.090] Number of PID (N) 456 456 456 “Participate” is instrumented with random invitation status in FEIV model. All models include log financial assets, log physical assets, and constant Robust standard errors in parentheses p < 0.01 The “recent usage” refers to the incidence of using mobile money to transact at least one financial activity at least once in the last 30 days. Column 1 (Table 7) shows that workshop invitees increased their probability of using mobile money for a recent activity after the workshop, and this was statistically significant at 1%. Further, a strong positive TOT impact of workshop participation was observed on the recent use of mobile money for transactions. The coefficient of the Endline was negative and significant, implying that, generally, respondents decreased their usage of mobile money at Endline. However, the treatment of inviting participants to the workshop compensated for the decline in the probability of use by − 0.728 + 0.513.Table 7 Impact on recent use of MM for transactions Variables ITT TOT FE FE FEIV (1) (2) (3) InviteEndline 0.513*** (0.08) ParticipateEndline 0.376 (0.10) 2.537* (0.57) Endline − 0.728* (0.040) − 0.621* (0.041) − 1.103* (0.140) Observations 912 912 912 R2 [Overall] 0.36 0.318 [0.061] Number of PID (N) 456 456 456 “Participate” is instrumented with random invitation status in FEIV model Recent use is the incidence of using mobile money to transact at least one financial activity for at least once in the last 30 days All models include log financial assets, log physical assets, and constant Robust standard errors in parentheses p < 0.01 In both columns (2) and (3), the coefficients of Participate × Endline are positive and statistically significant at 1%. The FEIV results show that, at the Endline, the probability of usage was generally lower than at the baseline (the negative coefficient at the Endline). However, the coefficient of Participate × Endline compensates for this lower usage, suggesting that the workshop participants increased the probability of recent use by 143% (− 1.103 + 2.537). At the 1% significance level, workshop participation increased the recent usage of mobile money for a transaction by 143% within 6 months after the intervention. These results support the second hypothesis: “mobile money education contributes to increased usage of mobile money for a recent transaction”. Impact on Financial Transaction To examine the third hypothesis, the ITT and TOT of mobile money education on (1) payments (sent and received; Table 8), (2) remittances (sent and received; Table 9), and (3) savings for the last 12 months were examined. Note that for each outcome, the first three columns use the log of total amounts transacted, disregarding the transaction methods, i.e. including transfers through banks, as the dependent variables. The last three columns examined the share of these transactions conducted through mobile money to test Hypothesis 3.Table 8 Impact on payments ITT TOT ITT TOT FE FE FEIV FE FE FEIV Panel A Log payments sent Share of payments sent through MM (1) (2) (3) (4) (5) (6) Invite x Endline 0.004 0.003 (0.10) (0.01) Participate x Endline 0.12 0.02 − 0.01 0.01 (0.13) (0.51) (0.02) (0.06) Endline − 0.027 − 0.052 − 0.03 0.024 0.028* 0.022 (0.063) (0.055) (0.124) (0.008) (0.008) (0.016) Observations 912 912 912 912 912 912 R2 [Overall] 0.058 0.06 [0.180] 0.037 0.039 [0.005] Number of PID (N) 456 456 456 456 456 456 ITT TOT ITT TOT FE FE FEIV FE FE FEIV Panel B Log payments received Share of payments received through MM (7) (8) (9) (10) (11) (12) Invite x Endline 0.20 − 0.06 (0.24) (0.08) Participate x Endline 0.40 1.01 0.10 − 0.33 (0.26) (1.27) (0.13) (0.53) Endline − 0.468* − 0.482* − 0.618 0.188 0.140** 0.239* (0.171) (0.143) (0.311) (0.088) (0.062) (0.131) Observations 911 911 911 878 878 878 R2 [Overall] 0.09 0.092 [0.123] 0.028 0.029 [0.004] Number of PID (N) 456 456 456 456 456 456 “Participate” is instrumented with random invitation status in FEIV models All models include log of financial assets, log of physical assets, and constant Robust standard errors in parentheses *p < 0.01, p < 0.05, p < 0.1 Table 9 Impact on remittances ITT TOT ITT TOT FE FE FEIV FE FE FEIV Panel A Log remittances sent Share of remittances sent through MM (1) (2) (3) (4) (5) (6) Invite x Endline 0.434 − 0.0415 (0.295) (0.050) Participate x Endline 0.825* 2.15 0.04 − 0.30 (0.33) (1.47) (0.06) (0.38) Endline 1.105* 1.082* 0.787 0.194* 0.171* 0.240* (0.182) (0.167) (0.360) (0.030) (0.028) (0.082) Observations 912 912 912 680 680 680 R2 [Overall] 0.152 0.159 [0.054] 0.18 0.18 [0.045] Number of PID (N) 456 456 456 415 415 415 ITT TOT ITT TOT FE FE FEIV FE FE FEIV Panel B Log remittances received Share of remittances received through MM (7) (8) (9) (10) (11) (12) Invite x Endline − 0.09 − 0.05 (0.27) (0.09) Participate x Endline 0.512* − 0.46 0.07 − 0.23 (0.29) (1.37) (0.07) (0.46) Endline 1.175* 1.026* 1.243* 0.175* 0.146* 0.205 (0.178) (0.163) (0.336) (0.062) (0.054) (0.103) Observations 912 912 912 555 555 555 R2 [Overall] 0.13 0.134 [0.053] 0.068 0.069 [0.015] Number of PID (N) 456 456 456 377 377 377 “Participate” is instrumented with random invitation status in FEIV models. All models include log financial assets, log physical assets, and constant Robust standard errors in parentheses *p < 0.01, p < 0.05, p < 0.1 Tables 8 and 9 are presented in the same format, for easy comprehension. Columns (1), (4), (7), and (10) show the ITT, estimated using random invitation status as the treatment. Columns (2), (5), (8), and (11) estimate the FE by using the participation status as the treatment, whereas columns (3), (6), (9), and (12) use FEIV estimation, by using participation status as a treatment and instrumenting it with random invitation status, to correct for endogeneity. The ITT results indicate no significant effect of invitation to the workshop on the payments sent, nor on the share of payments sent through mobile money (Table 8, Panel A). Generally, the payments sent among the sample did not change; however, the share of payments sent through mobile money increased above the baseline values. Similarly, no significant treatment effect of mobile money education was observed on the payments received, or on the share of payments received through mobile money (Table 8, Panel B). The general sample decreased their total amount of payments relative to the baseline values. However, both the treatment and control groups increased their shares of payments received through mobile money. Concerning remittances sent, while no significant ITT impact was observed on the remittances sent, the FE model shows a positive TOT that faded after instrumentation (Table 9, Panel A). There were no significant TOT effects on the share of remittances sent through mobile money; however, the general sample increased both the remittances sent and the share of remittances sent through mobile money. Likewise, models of remittances received show similar results (Table 9, Panel B). The FE model marked a 10% significance of TOT for the remittances received; however, the significance fades after instrumentation, and no significant TOT is seen on the share of remittances received through mobile money. Conversely, relative to baseline values, both the treatment and controlled group samples had received more remittances and increased their share of remittances received through mobile money during the Endline survey. Regarding savings, ITT results showed no significant impact. The effects on the log of savings and share of savings both showed no significant impact; however, the general study sample increased the amount of savings relative to the baseline value at the Endline; but, this was not necessarily achieved through mobile money. Results of Heterogeneous Effects Table 10 shows the results of heterogeneous effects on the outcome variables. Bank account ownership and gender showed no significant joint effect on the outcome variables. A negative significant joint effect was observed between workshop participation and urban dwellers on the share of payments sent through mobile money (column (1). This suggests that the effect of the workshop attendance on the share of payments sent through mobile money was 10.4% less for urban dwellers than for rural dwellers. Concerning the share of savings conducted through mobile money, a significant negative effect was observed when household size interacted with workshop participation. Column 4 (Table 10) shows that larger households experienced a 50.4% less impact on the share of savings through mobile money, and this is significant at 5%. Workshop attendance had positive effects on the share of savings conducted through mobile money, but this effect was greater for smaller households.Table 10 Heterogeneous analyses on the share of transaction through MM Variables Shares of: Payments sent through MM Payments received through MM Remittances sent through MM Savings through MM A = urban A = bank account A = rural A = household size A = Male (1) (2) (3) (4) (5) Participate x Endline 0.052 − 0.697 − 0.213 3.029* 1.287 (0.071) (1.016) (0.412) (1.453) (1.297) Participate x Endline x A − 0.104* 0.538 − 0.722 − 0.504** 0.121 (0.060) (0.957) (0.443) (0.199) (1.017) Endline 0.021 0.238* 0.278* − 0.449 − 0.475 (0.016) 0.538 (0.103) (0.306) (0.316) Observations 912 878 680 524 524 R2 [Overall] [0.007] [0.004] [0.025] [0.044] [0.0003] Number of PID (N) 456 456 415 351 351 Joint significance χ2 2.99 0.32 2.66 4.17 0.01 p value 0.083 0.574 0.103 0.041 0.905 All models are FEIV and “Participate” is instrumented with random invitation status All models include log financial assets and log physical assets and constant Robust standard errors in parentheses p < 0.01, p < 0.05, p < 0.1 Attrition Results The OLS and IV estimates (Table 11) show that none of the coefficients on Attrition x Participate were statistically significant for the primary outcome variables. This means that attrition in this study was independent of potential outcomes, and did not cause bias in the RCT or estimation results (Gerber and Green 2012).Table 11 Attrition across main outcome variables Variables Payments Remittances Savings Sent Received Sent Received (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) OLS IV OLS IV OLS IV OLS IV OLS IV Participate − 0.108 − 0.376 − 197.2* − 1117 − 13.19 64.52 − 74.24* − 359.7** 60.93 − 2654 (0.123) (0.538) (67.74) (1207) (13.27) (87.71) (21.48) (162.5) (670.9) (3661) Attrition 0.003 0.078 320.9 − 727.2 23.48 28.65 − 26.07 − 137.3* − 429.7 − 1011 (0.105) (0.253) (403.4) (567.3) (20.76) (41.23) (26.52) (76.40) (540.4) (1721) Attrition × participate − 0.213 − 2.858 − 219.3 5664 − 35.09 488.6 23.34 777.2 − 731.6 − 419.2 (0.400) (2.846) (421.4) (6384) (23.12) (464.0) (26.57) (859.8) (697.9) (19,370) Constant 8.855* 8.921* 224.2*** 579.2* 24.80* 2.180 76.97* 148.8* 1283 1936 (0.057) (0.141) (63.25) (317.3) (8.516) (23.06) (21.42) (42.73) (511.8) (962.7) F 0.24 0.93 0.44 0.88 1.88 0.86 0.87 0.99 0.07 0.00 Prob > F 0.622 0.335 0.508 0.349 0.171 0.354 0.352 0.320 0.796 0.977 Observations 557 557 557 557 557 557 557 557 557 557 R2 0.003 0.005 0.006 0.009 0.001 For IV models, “Participate” was instrumented with random invitation status Robust standard errors in parentheses p < 0.01, p < 0.05, p < 0.1 Mechanism and Discussion Overall, the results of recent usage of mobile money for transactions are consistent with the suggestions that well-designed targeted training may increase financial literacy (Behrman et al. 2012; Drexler et al. 2014). The treatment group’s response to the education provided, shown through an increase in the usage for transactions, highlights the importance of knowledge acquired during the workshop. Therefore, how did the intervention catalyse recent incidences of using mobile money for transactions? Possibly, the education improved trainees’ understanding of mobile money and its benefits. As participants improved their knowledge of mobile money, their confidence and trust increased, thereby increasing usage. The puzzle is participants’ reluctance to open new mobile money accounts, which would enable them to take advantage of the more useful features of the platform, for example, savings which may improve their livelihood (Klapper et al. 2016). Although workshop attendees opened new mobile money accounts, they did so at a similar rate to the control group. Contrary to expectations, this finding aligns with literature that suggests little to no impact of financial education on account opening (Cole et al. 2011). Though a spillover of information about mobile money to the control group could be one reason for this observation, the widespread occurrence of any spillover is likely slim, because of the strict design and implementation protocols followed. Alternatively, field data suggest that the reasons respondents stated for not having a mobile money accounts were consistent with the main reason why people may want to own accounts. For instance, ‘not expecting to receive money from anyone’ was a major reason stated for not owning mobile money accounts. Moreover, this reflects the goals stated by respondents who opened new mobile accounts; 82.9% of the existing users at baseline had the goal of using their mobile money account to ‘receive’ cash. Likewise, 85% of the non-users who became new account holders after the intervention stated to ‘receive’ cash as their main goal. These suggest cash receipt expectations remain a significant driver of mobile money account ownership. Qualitatively, evidence confirms this, as 80% of workshop attendees became new account holders after the intervention with the primary goal of receiving cash. This reason closely aligns with our data and the situation on the ground. Concerning financial activities, the swing between weak significance to no treatment effect on financial activities through mobile money implies that, although mobile money education could have had a positive outlook for financial activities through mobile money, this effect was weak and volatile. These findings align with other financial education studies, which observed weak unstable impacts (Collins 2013; Fernandes et al. 2014). We found that workshop increased the recent use of mobile money among users (Table 7), but it did not have significant impacts on the transacted amounts or share of transaction through mobile money (Tables 8, 9). Several reasons may account for the volatile or no treatment effect observed. First, probably the one-time two-hours training was inadequate. Second, our Endline survey was too early to capture the sufficient effects on the full transaction of the past 12 months. We did find that recent use (within 30 days) increased for the workshop participants, but this behavioural change was not enough to observe the effect of the workshop on the financial transactions of a longer term. A longer time frame might be necessary to observe the change. Third, every payment transaction on a mobile money account attracts a 1% charge. Consequently, as knowledge on mobile money improved, consumers became more aware of the 1% transaction charges when payment is sent or received through mobile money. Such charges are non-existing for cash payment, bank account deposits, and withdrawals at banks or ATMs. When customers make smaller units of transactions, the charge is less noticeable; however, it becomes substantial when the transaction amounts increase. Therefore, the charges may act as a disincentive to transmit more substantial amounts of payments through mobile money. To confirm this assertion, during the lockdown periods of COVID-19, the 1% charge was waved for all transactions of GH¢ 100 and below, and this contributed to a 4.3% increase in active mobile money accounts, new activations, and higher person-to-person transactions on MTN. The heterogeneous effect results provide more insights into interaction terms that explain changes that may occur. Relative to rural dwellers, urban workshop participants decreased their share of payments sent through mobile money by 10.4%. Furthermore, larger householders who attended the workshop decreased their share of savings held in mobile money by 50.4%. Conclusion We examined the impact of mobile money education on individuals’ mobile money participation. Through an RCT and two survey rounds in a panel, we estimated the treatment effect of mobile money education on mobile money usage. Overall, the intervention promoted the recent usage of mobile money for transactions among workshop participants. However, weak significant treatment effects were found on the remittances sent and received by attendees. These weak effects faded after instrumentation with the randomisation variable. Concerning the share of financial activities conducted through mobile money, although a positive outlook was observed, no significant treatment effect was observed. Examining the qualitative data, these weak to no treatment effects can likely be attributed to either possible spillover effects between the treated and control groups (as mobile money participation increased for both groups in general over time) or to the 1% transaction charge incurred on every transaction on mobile money, which became more evident to the workshop participants. Heterogeneous impact analyses revealed significant joint effects of the workshop and location, and household size on the share of financial activities transmitted through mobile money. Urban dwellers experienced lesser impact from the workshop regarding payments sent through mobile money, and larger households decreased their savings held via mobile money. Policies directed towards educating people about financial products on mobile money platforms will constitute reasonable attempts to improve participation through the incidence of current use, with moderate impacts on users sending and receiving remittances. It will encourage people to use mobile money products with understanding, thereby maximising their potential benefits from mobile money usage, while promoting financial literacy and sustainable development goals. Concerning further work, this study was conducted in only one region of Ghana, hence limited by ample sample size for strong statistical power. However, the results are robust, and may be used as a starting point for a more extensive study to observe the short- and long-term impacts of mobile money education. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 191 kb) Acknowledgements We appreciate the generosity of KDDI foundation for the grant for this study. Thank you KDDI. We thank Settor Amediku and Stephen Sasu-Yeboah of Bank of Ghana; Berthold Gadagbui and George Ofosu Boateng of Ecobank; Adu Danso, Confidence Amegashitsi, Yaw Antwi-Boasiako, and Kwame Amoako Agyeman of VODAFONE Cash; Alexander Oti Boateng and Isaac Kwadwo Bediakor of AIRTEL Money; Eli, Hini, Gabriel Agana, Steven Asare, Charity Darko, Richmond Darfar, Anthony Kweinin, Aboagye Mizhack, and Deladem of MTN Mobile Money; Bernard Frimpong and Alexander Alordeppey and the survey team for their support during fieldwork. Declarations Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest. 1 Enumerator’s characteristics may affect what kind of person s/he chooses to approach for a survey or from whom s/he receives a consent to interview. If there is such a tendency, it may create bias in our sample. To prevent such bias, our enumerators were not recruited locally but from another city (i.e. no social connection that may lead to bias). Enumerators were selected from a pool of applicants and were trained for five days to learn about the research and the protocols and ethics in conducting surveys. They all had first degrees and previous experience of working for similar projects. The researchers were also present in the field during the survey to supervise them and maintain consistency in the quality of their work. We also checked whether the respondents’ characteristics differ importantly depending on enumerators and found that the difference was minimum. 2 We checked the non-response rate at the end of each survey day and there was no systematic difference across days, enumerators, nor locations. The survey team visited these communities as a team, so each location was surveyed by all the enumerators to minimise bias across locations. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Apiors EK Suzuki A Mobile money, individuals’ payments, remittances, and investments: Evidence from the Ashanti Region, Ghana Sustainability 2018 10 5 1409 10.3390/su10051409 Arun T Kamath R Financial inclusion: Policies and practices IIMB Management Review 2015 27 4 267 287 10.1016/J.IIMB.2015.09.004 Bank of Ghana. 2017. Impact of mobile money on the payment system in Ghana: an econometric analysis. https://www.bog.gov.gh/privatecontent/Public_Notices/ImpactofMobileMoneyonthePaymentSystemsinGhana.pdf. Batista, C., and P.C. Vicente. 2020. Is mobile money changing rural Africa ? Evidence from a field experiment. NOVAFRICA Working Paper, 1805(August), pp. 1–62. www.gsma.com. Accessed 31 Dec 2020. Beck T Demirgüç-Kunt A Levine R Finance, inequality and the poor Journal of Economic Growth 2007 12 1 27 49 10.1007/s10887-007-9010-6 Behrman JR Mitchell OS Soo CK Bravo D How financial literacy affects household wealth accumulation American Economic Review 2012 10.1257/aer.102.3.300 Bruhn M Leao LS Legovini A Marchetti R Zia B The impact of high school financial education: Evidence from a large- scale evaluation in Brazil American Economic Journal: Applied Economics 2016 8 4 256 295 10.1257/app.20150149 Cole S Sampson T Zia B Prices or knowledge ? What drives demand for financial services in emerging markets? The Journal of Finance 2011 66 6 1933 1967 10.1111/j.1540-6261.2011.01696.x Collins JM The impacts of mandatory financial education: Evidence from a randomized field study Journal of Economic Behavior and Organization 2013 95 146 158 10.1016/j.jebo.2012.08.011 Drexler A Fischer G Schoar A Keeping it simple: Financial literacy and rules of thumb American Economic Journal: Applied Economics 2014 6 2 1 31 10.1257/app.6.2.1 25485039 Fernandes D Lynch JG Netemeyer RG Financial literacy, financial education, and downstream financial behaviors Management Science 2014 60 8 1861 1883 10.1287/mnsc.2013.1849 Gerber A Green D Field experiments: Design, analysis, and interpretation 2012 New York W. W. Norton Ghana, B. of. 2017. Payment systems Department Bank of Ghana payment systems oversight Annual report. https://www.bog.gov.gh/privatecontent/PaymentSystems/PaymentSystemsAnnualReport2017.pdf. Accessed 5 July 2018. Glennerster R Takavarasha K Running randomized evaluations: A practical guide 2013 Princeton Princeton University Press GSMA. 2010. Mobile money definitions. https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2012/06/mobilemoneydefinitionsnomarks56.pdf. Accessed 5 March 2019. Hastings JS Madrian BC Skimmyhorn WL Financial literacy financial education, and economic outcomes Annual Review of Economics 2013 10.1146/annurev-economics-082312-125807 Jack W Suri T Risk sharing and transactions costs: Evidence from Kenya’s mobile money revolution American Economic Review 2014 104 1 183 223 10.1257/aer.104.1.183 Karlan DS Dupas P Robinson J Ubfal D Banking the unbanked? Evidence from three countries SSRN Electronic Journal 2017 10.2139/ssrn.2815092 Klapper, L., M. El-Zoghbi, and J. Hess. 2016. Achieving the sustainable development goals—the role of financial inclusion, p. 20. www.cgap.org. Accessed 19 Feb 2018. Lusardi A Tufano P Debt literacy, financial experiences, and overindebtedness Journal of Pension Economics and Finance 2015 14 4 332 368 10.1017/S1474747215000232 Miller M Reichelstein J Silas C Zia B Can you help someone become financially capable? A meta-analysis of the literature World Bank Research Observer 2015 30 2 220 246 10.1093/wbro/lkv009 Mitchell OS Lusardi A Financial literacy and planning: Implications for retirement well-being Financial Literacy: Implications for Retirement Security and the Financial Marketplace 2011 10.1093/acprof:oso/9780199696819.003.0002 Munyegera GK Matsumoto T Mobile Money, remittances, and household welfare: Panel evidence from rural Uganda World Development 2016 79 25101002 127 137 10.1016/j.worlddev.2015.11.006 National, C.A. 2018. MNOs voice subscription trends. https://www.nca.org.gh/assets/Uploads/Voice-Oct-Dec-2018.pdf. Accessed 27 Feb 2019. Page, M., M. Molina, G. Jones, and D. Makarov. 2013. The Mobile Economy 2013, p. 100. https://www.gsma.com/newsroom/wp-content/uploads/2013/12/GSMA-Mobile-Economy-2013.pdf. Accessed 24 Feb 2018. Payment System Statistics 1. Ghana Interbank Settlement (RTGS). 2010. https://www.bog.gov.gh/privatecontent/PaymentSystems/PaymentSystemStatistics-December2018.pdf. Accessed 30 Jan 2019. Ravallion M Chapter 59 evaluating anti-poverty programs Handbook of Development Economics 2007 4 3787 3846 10.1016/S1573-4471(07)04059-4 Suárez SL Poor people׳s money: The politics of mobile money in Mexico and Kenya Telecommunications Policy 2016 40 10–11 945 955 10.1016/j.telpol.2016.03.001 Suri T Jack W The long-run poverty and gender impacts of mobile money Science 2016 354 6317 1288 1292 10.1126/science.aah5309 27940873 Suri T Jack W Stoker TM Documenting the birth of a financial economy Proceedings of the National Academy of Sciences 2012 109 26 10257 10262 10.1073/pnas.1115843109 The World Bank. 2012. Maximizing Mobile: The World Bank. Information and Communications for Development 2012. https://siteresources.worldbank.org/EXTINFORMATIONANDCOMMUNICATIONANDTECHNOLOGIES/Resources/IC4D-2012-Report.pdf. Accessed 6 March 2019. Tobbin P Kuwornu JK Adoption of mobile Money Transfer Technology: Structural equation modeling approach European Journal of Business and Management 2011 3 7 59 78
PMC009xxxxxx/PMC9005918.txt	==== Front SOPHIA Sophia 0038-1527 1873-930X Springer Netherlands Dordrecht 920 10.1007/s11841-022-00920-5 Article Luce Irigaray’s Philosophy of the Child and Philosophical Thinking for a New Era http://orcid.org/0000-0001-6976-8104 Thorgeirsdottir Sigridur sigrthor@hi.is grid.14013.37 0000 0004 0640 0021 Department of Philosophy, Gimli, University of Iceland, IS-101 Reykjavik, Iceland 13 4 2022 116 16 3 2022 © The Author(s), under exclusive licence to Springer Nature B.V. 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. In her book To be Born (2017), Luce Irigaray offers a novel philosophy of the child. Instead of viewing the child as a bearer of rights and in need of adequate care as is common in contemporary philosophies of childhood, Irigaray presents the child as a metaphor of a new human being which represents natural belonging. The rearticulation of the human has been ongoing in Irigaray’s philosophy from its beginnings with its efforts to give voice to the excluded, silenced, repressed feminine. Irigaray’s phenomenological restructuring of subjectivity in her philosophy of sexuate difference is taken to a new level with her philosophy of the child. Her conception of the child is interpreted here in light of the experiential and affective turn within phenomenology and cognitive sciences about philosophical thinking as embodied and embedded thinking for a new era. Irigaray sheds light on the silencing and repressing of the child within us in an effort to enable us as adult beings to think from and with it. Philosophical thinking needs to be more consciously connected with the embodied sources of thought that are already present in early infancy and continue to be present in adult thinking as neglected or repressed experiential and affective layers of thought. Irigaray’s philosophy of the child is a basis for a methodology of embodied philosophical thinking such as has been developed within Claire Petitmengin’s microphenomenology and within Eugene Gendlin’s methodology of philosophical thinking from the felt sense. Keywords Philosophy of embodied thinking Philosophy of the child The experiential and affective turn Touch ==== Body pmcChildren have in recent years become important voices in current global, political debates, perhaps most notably regarding issues of climate change and education for girls with Greta Thunberg and Malala Yousafzai as the best known poster children for such interventions. These interventions are very much in the spirit of The United Nations Convention on the Rights of the Child, setting out the civil, political, economic, social, and cultural rights of every child, and emphasizing the rights of children to express their opinions and be listened to. The child is in times of the UN Declaration primarily seen as a bearer of human rights that protect childhood as a means to grow and mature regardless of a child’s gender, race, ability, or other variables. In the current debates, children like Malala Yousafzai and Greta Thunberg have been viewed as speaking truth to power, raising awareness, and mobilizing protest, rather than as being directly politically involved. There is a reason for protecting children from direct involvement in politics against the history of examples of the abuse of children for political purposes such as in totalitarian states where children are mobilized in political youth movements or in the atrocious cases of child soldiers and war-related sexual abuse of children. Without being directly engaged in politics, children should nevertheless be able to express themselves publicly and be listened to because their perspective is essential to many political issues. Greta Thunberg has compared her own activism with the child in H.C. Anderson’s tale of the emperor’s new clothes where the little boy speaks out loud the obvious truth that the adults do not want to see or admit. This idea echoes Christian ideas about innocent children who speak truth because they are not politically strategic in their thinking and they do not speak out of a position of power. Political efforts therefore aim to enable children to be such voices that speak up and to educate them in ways that will allow them to become active and responsible citizens who will ultimately contribute to a better world. In that sense children are from a human rights perspective primarily seen as beings that need to be fostered, schooled, and cared for in ways that protect them as children and prepare them for adulthood. Even if we consider ourselves to be enlightened about children and their needs, childhood is still under massive assault. Greta Thunberg began the school strike for climate so that present day children can have a future rather than disasters or an ecological collapse. Malala Yousafzai has been fighting against the repression of girls in parts of the world where they are not allowed to or able to be educated. The global sexual abuse of children and youth is becoming more apparent and visible in times of #Metoo. Social media platforms are causing harm to kids, like former Facebook employee Frances Haugen testified before a U.S. Senate committee. ‘Facebook knows that they are leading young users to anorexia content,’ Haugen said, and added that despite the company’s claims that Instagram can help connect kids who may feel isolated, the rates of suicide and depression among teenagers are on the rise.1 I would like to argue that Irigaray’s philosophy of childhood offers a different approach to only listening to the voices of children that speak out on behalf of children, although it is vital to take them seriously. She enters the discussion about children and childhood at a different level. If we really want to be able to hear what children are experiencing, one vital precondition is that we are able to listen to the child within us. The child within us has a philosophical-epistemological meaning for Irigaray that is to be distinguished from an everyday psychological understanding of the term of the ‘inner child.’ A richer understanding of how infants and young children sense and relate to others and the world, allows to enable them better to flourish. What is needed to improve the lives of children globally (in addition to socio-economic justices that secure safety, health, and flourishing) is a better understanding of how children ‘tick’ and how they sense and think. A richer understanding of the child as a relational, embodied, vulnerable, and dependent person offers better preconditions for upbringing and education, like many of the contributors to the collection of articles on topics of Irigaray’s book To be Born (Irigaray 2017, TBB) discuss (Irigaray et al., 2019, Towards a New Human Being (TNHB).) If children are better taken care of and better understood as a special kind at home, in families, and in society at large, the possibility of having more content adults in the future is greater. The focus of my reading of Irigaray’s philosophy of the child goes however in a different direction, as I will argue that attending to the child within is not only about protecting childhood from damage. Attending to the child within is for us, as philosophically thinking beings, about connecting with levels of thought that are developed in infancy and childhood and that have been neglected in our epistemological understanding of philosophical thinking and knowing. This aspect needs to be understood much more precisely than merely as childlike wonder and spontaneity which often have a ring of naivety to it. It is also not about cultivating childlike innocence of children dating back to Jesus’ praising of it. Like I will argue, children’s innocence in this epistemological context means for Irigaray an opening of space for embodied thinking, and as such Irigaray’s idea is a reformulation of the phenomenological epoché, or bracketing, as a process of setting aside assumptions and beliefs:For it to be achieved, a place must be set, a sort of clearing of innocence where the not yet happened can be welcomed, heard, and in which it can germinate from a virgin space continuously won back. In this way, our flesh, our being, become revived and fertilized towards a new blossoming. (TBB 96) We ourselves as adult thinkers have to take the first steps in this direction by opening ourselves up to our lived experience. For this reason, the body is for Irigaray the bridge ‘from past humanity to a new humanity’ (TBB 85). In To Be Born Irigaray widens the traditional epistemological framework by incorporating childlike knowing and relating into it. Childlike forms of knowing and understanding open several new horizons at the same time within epistemology and phenomenology. The political implications of this approach are vast, among others the overcoming of one-sided mental hierarchies and exclusions that have made Western conceptions of cognition narrow and disconnected from being in touch with sensitive environments and embodied experience. A Continuity Interrupted By introducing the child as a metaphor for a new method of philosophical thinking may at first sound like offering a conception of philosophy leading to regression and some form of neo-naivity. We become thinking beings precisely by being able to learn to articulate, verbalize, differentiate, distinguish, and reflect the sense certainty that characterizes infant perception according to Hegel’s phenomenology of the itinerary of human consciousness from infancy to maturity as a journey towards the concept and abstract philosophical reflection. On this journey it is necessary in Hegel’s view to break ‘the child’s self-will and thereby eradicate his purely natural and sensuous self’ (Elements of the Philosophy of Right, §174 of the “Ethical Life”). Even though we judge this statement now as representative of an authoritarian pedagogical style of the early 18th century, there would for Irigaray still be truth to it in a broader sense than that of a mere tyrannical rearing style of the past. There is a disconnection from our natural belonging that Irigaray views as a core feature of our contemporary predicament in our relationship to ourselves as human beings, to others and to the earth. We are born into the world, into such relations, but between ‘our original experience of being in relations and its so-called cultural working-out, a continuity has been interrupted so that the modes of meeting which are proposed to us, and even imposed on us, do not ensure the cultivation of the first physical emotions and excitement’ (TBB 66). In To be Born Irigaray proceeds to uncover this continuity and the legacy of dualistic, metaphysical thinking in abstract and representational modes of thought that interrupt or obstruct it. Her philosophy of the child is hence not political in terms of human rights and social justice in a narrow sense. Her philosophy of the child is an examination of how we violate and repress rather than allow and cultivate the connection to our natural belonging that we have immediate access to in infancy and a close connection with in childhood. This level of our relation to others and the world is not a stage that we outgrow as we develop in our perception and thinking, but a sensory and embodied level of thought that remains within us, ready to be reconnected with in our adult ways of thinking. From that perspective, the reason for our disconnection in our relations to ourselves, others, and the world is a disconnection to what is ‘closest to us: our own lived experience’ (Petitmengin, 2021, 172). Like the micro-phenomenologist Claire Petitmengin argues, our current way of life and the ecological disaster it is bringing about has in a very basic sense to do with this disconnection. She describes it further as being a mode in which we are ‘cut off from ourselves, from what vibrates and lives within us, and this disconnection has catastrophic consequences in all areas of human existence’ (Petitmengin, 2021, 172). Retrieving contact with our experience ‘is the precondition that would allow us to regain our lucidity, our dignity, and the courage to change our model of society’ (Petitmengin, 2021, 172). Irigaray discusses how we are born as sensuous selves but learn to repress our natural belonging in a culture that is still permeated by an outdated dualistic metaphysics of body and mind. To be Born is therefore about the child within us, the child that was born into a culture that represses their natural belonging, and the possible rebirth or reawakening of that part of the inner child in ourselves. Her philosophy of the child is therefore no less about adults, and in a more specific sense about philosophically thinking adults and in that sense it is about the birth of a philosophy that is attuned to phusis, to our natural belonging, out of the body of the child that is within us. This idea of philosophical thinking is by no means restricted to academic philosophy. Philosophical thinking holds a prime place for Irigaray as a discourse of discourses. But in the context of her philosophy of the child, it becomes evident that she introduces a new method of embodied thinking that signals a new era of being human, and in that sense it is a way of thinking that is a possibility for all of us. Connecting Back to Nietzsche Irigaray explicitly states that her philosophy of the child carries forward Nietzsche’s philosophy of the child as a new beginning for culture as presented in his book Thus spoke Zarathustra.2 For both Nietzsche and Irigaray, our culture calls for a new type of human being and a new way of thinking that replaces the old man of the West (Irigaray et al., 2019). Traditional concepts of the human being have in recent decades been much under attack for being anthropocentric, sexist, racist, ableist, excluding groups in positons of ‘others’, categorizing attributes associated with them as non-human or less than human. The rethinking of the human that Irigaray undertakes with her philosophy of the child does neither entail a return to any traditional version of humanism nor does it imply a total abandonment of it. Like Judith Butler writes, ‘the category of the “human” retains within itself’ the workings of powers that have conditioned our understaning of the human ‘as part of its historicity’ but ‘the history of the category is not over, and the “human” is not captured once and for all’ (Butler, 2004, 12). For critics of traditional, exclusive notions of the human, its rearticulation begins ‘at the point where the excluded speak to and from such a category’ (Butler, 2004, 13). A rearticulation of the human has been ongoing in Irigaray’s philosophy from its beginnings with its efforts to give voice to the excluded, silenced, repressed feminine, rooted in a denial of our maternal origin in predominant strands of Western philosophy.3 With her philosophy of the child, Irigaray sheds light on the silencing and repressing of the child within us in an effort to enable us as adult beings to think from and with it. Her approach is not psychological or therapeutic in the sense of connecting with the individual inner child to help us heal a trauma the child may have suffered in infancy, childhood, and adolescence. In spite of her training in psychoanalysis, Irigaray’s approach in TBB does not consist in addressing the psychological needs of the child that have not or unsufficiently been met. For Irigaray, the point of connection with Nietzsche’s call for a new human being (which he, according to Irigaray, wrongly named the Übermensch) is that he understood how traditional conceptions of thinking and knowing are cut off from the real, leading us to practice thinking that disconnects it from sensible perception (TBB 10). For Nietzsche the child is a metaphor or symbol for a human being that is body and soul, implying that cognition is embodied and connected to lived experience. The metaphor of the child is meant to be inspirational for reenchanting sterile, abstract philosophical thinking that has become disembodied and disconnected. The child symbolizes for Nietzsche how we need to liberate us from moralistic views that condemn the body and do not acknowledge it as part of philosophical thinking. In that sense the child stands for how we need to learn to become beginners again in philosophy, to think freshly as embodied and embedded beings rather than being disembodied in our thinking and thus lost in abstraction. With her carrying forward of Nietzsche’s philosophy of the child, Irigaray opens a new perspective on the figure of the child in Nietzsche’s philosophy. The embodied child not only interacts in perception and in movement with the world and is a figure for criticizing disembodied, dualistic epistemological notions of cognitive neutrality and objectivity. Moreover, Irigaray takes Nietzsche’s conception of the child as a symbol for embodied thinking further by accentuating more explicitly embedded thinking as part of it. As embedded beings we are of the earth and intertwined with all living things. We are also interactive with the environments we are situated in be it a house or an online meeting room, although Irigaray does not address that directly. Her understanding of the new human being is nevertheless critical of a transhumanistic understanding of the human as embedded in a technological, cybernetic environment because her philosophy of the child contains an appeal to cultivate embodied knowing that is needed to protect us from being overly dominated by artificial intelligence. Touch, how the infant touches its way through the world is therefore central to the conception of the child in To be Born. The concepts of touch as touching and being touched as well as the concept of self-affection are key to this philosophy of the child as metaphor for philosophical thinking that is attuned to the real. With her philosophy of thinking that touches, Irigaray develops Merleau-Ponty’s phenomenology of touch further in her efforts to modify and complement the traditional vision-orientation of philosophical thinking, which, as Elizabeth Grosz has discussed, is less embodied than touch as a base of thinking and knowing (Grosz, 1994). The Affective and Experiential Turn in Phenomenology: the Touch and Being Touched The phenomenology of touch is central to the experiential and affective turn within philosophical epistemology. As part of this turn, Irigaray’s philosophy of the child is a basis for a methodology of embodied philosophical thinking. Irigaray has not elaborated a concise methodology of embodied thinking such as the micro-phenomenological interview-method of accessing lived experience as a source for philosophical and scientific thinking. There is nevertheless a great affinity between Irigaray’s descriptions of accessing experience with basic assumptions of micro-phenomenology. Her descriptions of self-affective embodied thinking invite further more to be read in light of Eugene Gendlin’s philosophy of the felt sense and his focusing-based methodology of connecting with the ‘felt sense’ as a felt meaning of an issue or a thought4:Focusing is not an invitation to drop thinking and just feel. That would leave our feelings unchanged. Focusing begins with that odd and little known ‘felt sense,’ and then we think verbally, logically, or with image forms—but in such a way that the felt sense shifts. When there is a body shift, we sense that our usual kind of thinking has come together with body-mind, and has succeeded in letting body-mind move a step. (Gendlin, 1982, 57)5 The goal of my interpretation is to think Irigaray’s philosophy of the child as a theory of embodied philosophical thinking further by examining and discussing how her descriptions of embodied thinking can be made more explicit by viewing them in light of basic assumptions of Petitmengin’s and Gendlin’s methodologies (Gendlin, 2004), but both these pioneering methodologies have roots in phenomenology like Irigaray’s philosophy does. With her idea of the touch, Irigaray’s philosophy of embodied thinking can be situated within new phenomenology. If Husserl, as a major founder of phenomenology, defined its task with his call for going ‘back to the things themselves’ (Husserl, 2001, 168), his focus was not so much on the experience of the phenomenon itself as the transcendental conditions for the experience of it. Later phenomenologists like Merleau-Ponty, Herman Schmitz, and Luce Irigaray have elaborated further what it means that phenomenology is a project and a methodology to discover or rediscover things and phenomena of real life. The basic assumption of phenomenology from Husserl to later phenomenologists is that we have lost sight of phenomena because of how we have been conditioned to perceive things in certain ways by objectifying ways of knowing and technological forms of life. The phenomenological method is therefore a kind of escavation or an accessing of experience of beings that as embodied and embedded sense and feel things and are affected and touched by them. It is primarily Merleau-Ponty’s phenomenology of touch that is a point of Irigaray’s departure for her conception of the child as a self-affective being that interacts through touch with others and the world. Merleau-Ponty’s famous description in his Phenomenology of Perception of the hand that touches the other hand illustrates how the knowing subject and the known object are to be seen as intertwined rather than as separate. Irigaray’s widening of Merleau-Ponty’s conception of touch is partly based on a critique of what she views as his mechanical idea of sensory perception by a ‘Sentient in general before a sensible in general.’ In her view, he neglects the specificity of what happens in the perception in addition to ‘the mediation of the sensory perceptions’ (TBB 26). With her own metaphors of the lips that touch each other, Irigaray extends Merleau-Ponty’s phenomenon of the hands that touch each other by emphasizing introceptive perception of the inner and the outer (Irigaray, 1980). The lips represent a cognitive eroticism that is ‘more or less internal and porous in relation to the outside world, to the other.’ The lips are a metaphor for a morphology which can ‘close while remaining open,’ requiring ‘open structures and meanings which can conform to a living growth’ (Irigaray, 2020, 32; Fuchs & de Jaegher, 2009).6 Touch in the context of an embodied, embedded, enactive, and extended cognition (Clark & Chalmers, 1998) is much wider than mere haptic touch which is also that aspect of touch that has most to do with control and possible abuse. In times of the COVID-19 pandemic, touch is for example associated with contagion, infection, and disease. In line with a wider idea of touch within the embodied conception of knowing discussed here, we are touched in thinking when some thought resonates within something in us. We can touch others with our words, in a way that can hurt them, console them, soothe them, or enlighten them. Thoughts have a meaning for us when they strike a chord in us. Touch is an embodied sense (in addition to the five senses) that Ratcliffe describes as background touch or a background sense of belonging to the world (Ratcliffe, 2013). The concept of touch refers therefore to our sense of being situated in a world in a subjective way. Finding oneself touched is a kind of first wave of an inner dialogue with oneself, of being ‘two in one’ like Hannah Arendt describes philosophical thinking (Arendt, 1981). Irigaray therefore wants to draw attention to the ‘relation between our two different beings’ which is largely neglected in our culture (TBB 84). We are moved by something that motivates us to think about it. Coming to oneself in thinking is also a precondition for entering into a philosophical dialogue with another subject. Embodied philosophical thinking implies a way of dialoging philosophically in a different way. In her book Conversations Irigaray states that ‘[e]ntering into dialogue requires us to use a language which touches, which involves sensibility, which preserves the role of the other in the constitution of meaning’ (Irigaray, 2008, 33; Irigaray et al., 2019). The infant is important for the notion of touch because babies learn about the world through touching and later putting things in the mouth, and therefore the senses of touch and hearing are developed earlier than vision. Cognitive sciences findings on embodied cognition have indeed shown the limits of a vision-centric approach to knowledge in our philosophical tradition (Damasio, 2021; Varela et al., 1991). The emphasis on vision in traditional ideas about cognition and thinking has contributed to upholding a strict distinction between perception that is directed outwards to an object and an internally directed perception of the body, like Matthew Ratcliffe argues (Ratcliffe, 2008). Irigaray is aware of that when she writes that ‘the parameters which rule over our traditional logic –visibility, face to face or representattion– are no longer really helpful, and it is the way of getting in touch itself which remains inconceivable’ (TBB 84). Embodied philosophical thinking is a path that is ‘more inspired and paved by listening and touch than by watchful eyes’ (TBB 83). As an inner dialogue, embodied philosophical thinking is hence an inner listening to a felt sense for an issue. Gendlin describes how this approach can be worked out methodologically as a way of mediating between a felt meaning and a verbal articulation of it: There is a new method here, not only for personal concerns but also for theory and science. Logical thinking stays within whatever ‘conceptual boxes’ it starts with. It has only the different, competing interpretations, assumptions, viewpoints—and one must stay within one of these. When felt sense is the touchstone, one can try out all kinds of different concepts without being locked into any one set. This is what scientists (now rarely) do when they come up with something new after living with a problem for a long time. Rather than using concepts only, one can return to one’s un-split felt sense of whatever one is working on. (Gendlin, 1982, 57) Connecting with the Child Within Discussing philosophical thinking from the affective-experiential perspective of infancy and childhood means presenting a different way of connecting and orientating oneself in thinking in-with-about the world. Irigaray’s notion of the child is informed by early childhood theories of how the infant acquires a sense of self in relation with itself, others, and the world. That does not entail disqualifying adult ways of thinking and philosophizing or proposing that maturity and wisdom are not useful ingredients of philosophical thinking. They obviously are and always will be. Nor does this mean that abstract, logical, representational thinking is redundant. We will continue to think philosophically in patterns, structures and in line with logical rules, and the goal of philosophy will continue to be to offer comprehensive, conceptual clarification. The experiential affective turn should also not be understood as a return to some precultural, natural origin but Irigaray’s use of the term of ‘origin’ may invite misunderstanding it in such a way. Like Gendlin rightly points out, the human individual as an embodied being is not conceivable ‘apart from culture’ because man’s ‘animals functions are culturally patterned. The individual self develops out of an interpersonal, linguistic and social matrix. The individual is cultural, social and interpersonal before he is an individual’ (Gendlin, 1967, 141; Thorgeirsdottir & Karlsdottir, 2020). Yet, like Gendlin and Irigaray both argue, the felt, embodied, experienced, affective, tacit, dimensions of knowing get lost due to traditional notions about the split between the ‘subjective’ and the ‘objective.’ The subjective and the objective are intertwined layers of cognition that are rooted in a situated and embodied context of living in today’s world (Schoeller & Thorgeirsdottir, 2019). Connecting with embodied layers of thinking takes us out of an auto-pilot mode of thinking that Irigaray claims exhausts our vitality for ‘want of language which tells and cultivates life’ (TBB 11, 52). Life, so Irigaray, can ‘exist and develop only from an unrepresented,’ and hence authentic thinking comes from a felt place of meaning within ourselves, our embodied situatedness that allows us to articulate something in a fresh way (TBB 92). The cognitive developmental stages we go through in childhood on our way to adulthood cannot be understood as stages that dissolve when a stage is surpassed and a new stage is entered, like some theories of cognitive and psychological development have it. (Gheaus et al., 2019) It is rather so that each stage remains operative as an integrated level in our thinking. In light of that, the training of philosophical thinking as abstract, logical, and representational thinking lacks cultivating a connection with lived experience. Any original, fresh thinking has some direct reference to lived experience of its creator. It would for example be hard to claim, if not counterintuitive, that this particular and novel philosophy of the child has nothing to do with Irigaray’s own childhood experiences. Yet we have come to pretend and even assume that this source of thinking, the thinker’s intuition and deep rooted sense for a topic, is more or less irrelevant. With her conception of the child, Irigaray portrays a way of thinking that is different from mainstream notions of the detached, disembodied, and disconnected way of abstract and representational thinking. Philosophical thinking needs to be envigorated by being connected more consciously with the embodied sources of thought that are already present in early infancy and continue to be present in adult thinking as neglected or repressed experiential and affective layers of our thinking. This does not only entail a return to sensory modalities of thought. Being conscious is not the same as sensing like Damasio has argued based on his neuroscientific research into the interworkings of feeling and knowing (Damasio, 2021). The nervous systems are basic to the development of feelings, and feelings, like Damasio argues, open the way to consciousness. Such findings of neurosciences about how the mind is embodied and embedded display the need to develop better methodologies of training thinking on the basis of such findings. We need no less than to ‘restructure human subjectivity’ like Irigaray claims (TNHB, xviii). To Come to Ourselves in Thinking Irigaray’s phenomenological rehabilitation and restructuring of subjectivity in her philosophy of sexuate difference is taken to a new level with her philosophy of the child, by pluralizing and individualizing difference. Although the individual is socially and culturally formed, there is a particular core to every individual that becomes apparent right at birth. This core is not only the uniqueness of every being in terms of situatedness but also a will to live and a capacity for self-determination. This feature is the condition that makes every new individual born capable of adding a fresh and a unique perspective to the world. For that reason, Irigaray points out how the newborn child is an autonomous being. This may strike odd because we are used to view autonomy as something mature and a result of proper education and upbringing. Yet Irigaray opens her book To be Born with a notion of autonomy that is more primal as a kind of life force and a kind of wonder:Whatever the unknown factors of our conception, we have wanted to be born. Our existence cannot be the outcome of a mere chance, and our will to live clearly manifested itself at the time of our birth. We were the ones who determined its moment. We were also the ones who gave birth to ourselves through our first breathing. In spite of the long dependence of the little human on others for its survival, it gave life to itself to come into the world, and it gave life to itself alone. Even if it has been conceived by two and it began its human existence in the body of an other, it is the one who, alone, decided to come into the universe of the living. (TBB 1) Obviously, the precise time of birth is mystery that no science has been able to predict. A normal birth usually happens in the 38th to 42nd week of pregnancy, but no midwife or doctor has the means to predict precisely when a child will be born. Irigaray thus presents a speculative hypothesis about a scientific enigma by claiming that it is the child alone that decides when it enters the world. She also claims that the child’s first breathing outside the maternal body is the first sign of its autonomous human potential. Irigaray points out something novel here, namely how the newborn has to decide when they will embark on a dangerous journey to be born. Education and upbringing should therefore center on enabling this individual freedom to be oneself, in addition to fulfilling the needs of the child for nurture and care. Irigaray’s conception of the child as a desiring and autonomous living being is crucial to her philosophy of the child as a model of embodied transformative thinking. If Hegel—to refer to his philosophy again as a contrast—illustrated the human driving force for freedom and social progress with his model of the battle of the master and the slave, Irigaray poses the child which is born as the ‘young hero’ taking a risk and fighting for life and freedom (TBB 8). Giving birth has for the most part of human history been high risk for birthgiving women as well although thinkers of risk and battles of life and death like Hegel were blind to it. Irigaray takes this idea of risk that child takes at birth into philosophy because it is about thinking for oneself, to become oneself in thinking, and becoming ‘oneself requires as much heroism as being born’ (TBB 42). If we do not only want to repeat what others have said, combine positions that others have come up with but really think our own thoughts, we need to connect with experiential and affective sources of our thought for they are the knowledge and wisdom we have gathered from early on and make us who we are. We are a living process, a continuous becoming. Every person has a unique perspective on the world, and therefore any newborn, like Hannah Arendt also pointed out with her philosophy of natality, may be someone who comes up with something new and important for the world (Arendt, 1958, 8-9). Philosophers, like everybody else, must attempt to be and become themselves in order to connect with the source of their own thoughts. For Irigaray, self-affection is the royal road to accessing one’s own thought. The term self-affection has nothing to do with narcissism or auto-eroticism in a narrow sense but be described as the ‘felt sense’ as defined by Eugene Gendlin:A felt sense is not a mental experience but a physical one. Physical. A bodily awareness of a situation or person or event. An internal aura that encompasses everything you feel and know about the given subject at a given time— encompasses it and communicates it to you all at once rather than detail by detail. Think of it as a taste, if you like, or a great musical chord that makes you feel a powerful impact, a big round unclear feeling. A felt sense doesn't come to you in the form of thoughts or words or other separate units, but as a single (though often puzzling and very complex) bodily feeling. (Gendlin, 1982, 15) This kind of sensing into as a form of deeper thinking begins with a meditative step that Irigaray defines as re-touching. She refers to the morphology of the body, of the two lips that touch each other, and imagery of Buddhist meditation of fingers touching each other, eyelids closing and touching each other. Silence is also a precondition for this form of being with the neglected layers of our thoughts. In To Be Born we find this description which explains what she means by self-affection: Contemplating Buddha in meditation can lead us to glimpse what it is about. The matter consists of calmly staying in oneself, being silent, preferably with one’s eyes closed, trying to perceive and concentrate in this way one’s own inner energy. To succeed in this, I suggest focussing, at least in the first instance, one’s attention on the perception of one’s lips, one’s hands and one’s eyelids touching one another. Such a gesture—that I call ‘re-touch’—contributes to realizing what our limits are and the thresholds between the inside and the outside of the space that is ours, something which favors a repose in ourselves. It is possible to teach children how to practice self-affection in order to help them to develop, while remaining themselves, from their own energy and will so that they can ensure in this way an inner centring. (TBB, 17) Situated and Felt Knowing The meditative state Irigaray describes is more than just a mindfulness exercise in breathing and calming the mind. For Irigaray this is an entry point for inventing a path in our thinking, ‘a path in the opening of a “not that”, “not there”, “not yet”, “not knowable,” “not appropriable” ... . While advancing, we must continuously make room ... not only for imagining or representing what appears – as our tradition has taught us – but also within ourselves – what our logic did not teach us. We have thus to invent the path’ (TBB 81). Embarking on this path is like an opening of a door to a room where one is free to think and make sense for oneself, and where we allow thoughts to arise, and welcome them like a child, as something that is part of oneself but yet different and other. One approaches the thought with a friendly, non-judgemental attitude that allows it to form and show itself, like Gendlin describes it:A felt sense is usually not just there, it must form. You have to know how to let it form by attending inside your body. When it comes, it is at first unclear, fuzzy. By certain steps it can come into focus and also change. A felt sense is the body's sense of a particular problem or situation. A felt sense is not an emotion. We recognize emotions. We know when we are angry, or sad, or glad. A felt sense is something you do not at first recognize— it is vague and murky. It feels meaningful, but not known. It is a bodysense of meaning. When you learn how to focus, you will discover that the body finding its own way provides its own answers to many of your problems. (Gendlin, 1982, 7) We as contemporary philosophers need to emancipate ourselves in thinking. We belong to a malecentric tradition and culture of philosophy that determine our academic profession, its institutional styles as well as the content and basic concepts of philosophy (Thorgeirsdottir, 2020). Irigaray therefore rightly asks: ‘Why does our culture constrain us to hold a discourse about a presumed objectivity of the world without taking into account our own objectivity, including at the level of moods, feelings, sensitive life’ (TBNH, 251). Feminist epistemologists like Sandra Harding and Donna Haraway have discussed for decades the need for extended objectivity (Haraway, 1988). Such objectivity is about acknowledging how one’s own cognitive perspective is situated and positioned. Social context and tradition are part and parcel of how knowledge emerges and is produced. We are situated beings because we are embodied beings born in a place and a time. Yet the kind of situated knowledge I propose here with the help of Irigaray, Nietzsche, Gendlin, Petitmengin, developmental psychology and cognitive sciences takes the situatedness and the perspectival nature of knowledge from social situatedness of gender, class, and other sociological norms and variables deeper into the embodied layers of human beings, into feelings, moods, and sensitive life. Haraway characterizes this as a thinking and making with, as sym-poiesis (Haraway, 2016). We are socially constructed by norms, values, ideas, social structures, conditions, and goals and at the same time we are subjective, experiential beings with a unique perspective on the world because we are all differentially located, situated, and conditioned. This felt situatedness is what gives us our individuality and allows us to come closer to ourselves, not as a narcissistic move, but as a move that increases plurality and deepens universality in the world. The more self we are in thinking, the better we are understood by others. And the closer we come to our own thinking, the better we can understand others’ thinking. Irigaray describes this as a closeness to oneself which can be perceived ‘thanks to a distance from oneself and a distance from the other, two distances which cannot be mistaken for one another’ (TBB 69). But how to connect with oneself? Let me illustrate it again with the example of the conscious act of breathing which is for Irigaray crucial to the kind of beings we are. For Irigaray, transcendence, as the basic movement of philosophical thinking, is initiated by touching ground with oneself in a self-affective gesture of breathing. Breathing also expands the conception of the self because there is no clear demarcation between us as bodies and the wider environment we are located in. Air enters our bodies, and it exits our bodies through the nostrils showing how the body-environment is one as a continuum. There is an affective attunement in the type of an inner or intersubjective dialogue that harnesses and nurtures a felt level. The notion of affective attunement comes from research into early infant development that shows how infants learn and adapt through affective attunement between them and their mothers or primary caretakers (Stern, 1985). This happens on a pre-reflective level where mother and child attune their internal rhythms through gestures, sounds, and caressing. ‘This rhythmic synchronization, which enables the resonance or tuning of two interior universes, is the basis of affective intersubjectivity’ (Petitmengin, 2007, 66). Being touched is not a one-way street of how society and culture influence and form us as thinkers. Being touched is at least a two-way street because there is always something within us that resonates with what touches us; otherwise, we would not be touched if there were not something within us that makes us receptive towards it. Interaction always comes first. The moment of interactive resonating that Hartmut Rosa has discussed in his philosophy of resonance is a basic way of connecting with any phenomena (Rosa, 2019). Concluding Remarks In the context of the philosophy of child as a model of philosophical thinking, childlike wonder is traditionally emphasized. The child is seen as prefiguring the philosophical wonder that has since ancient times been seen as what ignites philosophical thinking. The philosophy of embodied philosophical thinking that I introduce with this interpretation of Irigaray’s philosophy of the child is situated earlier than in the moment when our philosophical eyes get wide open with wonder at something that puzzles us, amazes, or appalls us. It is the level of thinking that precedes problematizing thinking of puzzlement and the value judgements involved in being amazed or feeling appalled. That type of feelings of wonder are more cognitive than being moved or stirred by something. There is less value judgement in emotion than in feeling which is a more reflective and cognitive level of being touched. Embodied thinking has also as its source a level prior to feelings and emotions, and that is the level of affect. Affects are prepersonal and precognitive intensities. Irigaray describes the internal or the intersubjective dialogue with her linguistic concept of the middle voice as a way of articulating affects as kind of natural rhythms:The middle voice ... allows us to be in harmony, or to part from an immediate communion, with natural rhythms, and even with the other. It builds a sort of place in which we can dwell, which does not amount to a confinement into the ‘house of language’ of Heidegger, but is an opportunity for us to inhabit ourselves—our body, our heart, our soul or spirit, being the elements supplying matter and form(s) to such dwelling that the middle voice tries to express with words. In this way, it removes our affects from a mere instinctive or impulsive economy, and makes our body speak, which then affects itself, is moved, unites with itself, before any separation between subjectivity and objectivity. (TBB 49-50) With my interpretation I have argued that a politics that strives for creating a better world for children calls for more than adults listening to and responding responsibly and with appropriate measures and actions to children who voice their concerns and needs. In order to be able to resonate with what children say and express, adults need to access their own felt layers of subjective experience. The reason is not because that allows us to think like a child for that is not the goal with this theory of embodied philosophical knowing presented here. Embodied ways of thinking are always part and parcel of our thinking and understanding but an explicit and conscious accessing of lived experience has been neglected in predominant ways of thinking, and it is being threatened increasingly with disembodied forms knowing that culminate in artificial intelligence. The reason that Malala Yousafzai and Greta Thunberg have been heard is not only because of what they say about what they sense and experience. It is no less because what they say resonates with something in those that are listening. Resonating does not always imply that one thinks in unison with what is being said, but it always means that one is moved, challenged, and even disturbed by it, prompting one to listen further into the issue at hand, within oneself and by informing oneself. Such a form of inner listening while listening to the other is characterized by a deferral of judgement in an effort to understand where the other is thinking from, out of what situation and from what kind of a felt sense. The political implications of embodied thinking and embodied listening are vast, although they have not been spelled out here in any detail. That would also not be in line with Irigaray’s conception of the new human being because it cannot be a prescriptive category for how the new human being should think and act politically. Her philosophical conception of the child rather uncovers conditions to connect more deeply with oneself, others, and the environment, and that can, in the long run, change how we discuss and behave in the sphere of the political. 1 Becky Upham, “Facebook comes under fire after whistleblower and leaked documents reveal negative impact on girls”, Everyday Health, October 9, 2021. https://www.everydayhealth.com/public-health/facebook-comes-under-fire-after-whistleblower-and-leaked-documents-reveal-negative-impact-on-young-girls/ 2 For a comparative analysis of Nietzsche’s and Irigaray’s understanding of a new human being see Mitcheson, Katrina, “On Nietzsche and pregnancy: The beginning of the genesis of a new human being”, in L. Irigaray, M. O'Brien, & C. Hadjioannou (Eds.), Towards a New Human Being, Cham, 2019, 199-220. 3 In To be Born Irigaray also discusses embodied thinking from the perspective of sexuate difference, especially in the last chapters on love and giving birth to each other. As I focus on the philosophy of the child in my interpretation here, there is not space to discuss the sexuate aspects of it. 4 I thank Steinunn Hreinsdottir, a fellow reasearcher in the international research project Embodied Critical Thinking (www.ect.hi.is and www.trainingect.com), for pointing out to me the affinity of self-affective embodied thinking with Gendlin’s conception of the felt sense. 5 Gendlin developed with Mary Hendricks a methodology of embodied philosophical thinking on the basis of focusing that they called Thinking at the Edge. Thinking at the Edge is a methodology to be used in philosophy and scientific and scholarly research. See E.T. Gendlin, “Introduction to ‘Thinking at the Edge’” The Folio, 19(1), 1-8, http://previous.focusing.org/gendlin/docs/gol_2160.html 6 Irigaray’s philosophy of interactive knowing could be developed further in the direction of recent research into enactive intersubjectivity. See Fuchs, Thomas, and Hanne de Jaegher, “Enactive Intersubjectivity: Participatory Sense-making and Mutual Incorporation." Phenomenology and the Cognitive Sciences 2009, 8:4, 465-86.i Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Arendt H The human conditon 1958 The University of Chicago Press Arendt H The life of the mind 1981 Harcourt Butler J Undoing gender 2004 Routledge Clark A Chalmers D The extended mind Analysis 1998 58: 1 7 19 10.1093/analys/58.1.7 Damasio A Feeling and knowing. Making minds conscious 2021 Panteon Fuchs T de Jaegher H Enactive intersubjectivity: Participatory sense-making and mutual incorporation Phenomenology and the Cognitive Sciences 2009 8 4 465 486 10.1007/s11097-009-9136-4 Gendlin ET Neurosis and human nature in the experiential method of thought and therapy Humanitas 1967 3 2 139 152 Gendlin ET Focusing 1982 Bantam Books Gendlin ET Introduction to 'Thinking at the edge The Folio 2004 19 1 1 8 Gheaus A Calder G De Wispelaere J The Routledge handbook of the philosophy of childhood and children 2019 Routledge Grosz E Volatile bodies: Toward a corporeal feminism 1994 Indiana University Press Haraway D Situated knowledges. The science question in feminism Feminist Studies 1988 14 3 575 599 10.2307/3178066 Haraway D Staying with the trouble: Making kin in the Chthulucene 2016 Duke University Press Husserl, E.. (2001 [1900/1901]). Logical investigations (J. N. Findlay, Trans. & D. Moran, Ed.), 2nd ed., 2 vols. Routledge. Irigaray, L. (1980). When our lips speak together (L. Irigaray & C. Burke, Trans.). Signs, 6(1), 69–79. Irigaray L Conversations 2008 Bloomsbury Irigaray L To be Born 2017 Palgrave Macmillan/Springer Nature Irigaray L Schwab GM How can we achieve women’s liberation? Thinking life with Luce Irigaray 2020 SUNY Press 25 36 Irigaray L O’Brien M Hadjioannou C Towards a new human being 2019 Palgrave Macmillan/Springer Nature Petitmengin C Towards the source of thoughts: The gestural and transmodal dimension of lived experience Journal of Consciousness Studies 2007 14 3 54 82 Petitmengin C Anchoring in lived experience as an act of resistance Constructivist Foundations 2021 16 2 171 180 Ratcliffe M Touch and situatedness International Journal of Philosophical Studies 2008 16 3 299 322 10.1080/09672550802110827 Ratcliffe M Radman Z Touch and the sense of reality The hand, an organ of the mind: What the manual tells the mental 2013 MIT Press 131 158 Rosa, H. (2019). Resonance: A sociology of our relationship to the world (J. Wagner, Trans.) Polity. Schoeller D Thorgeirsdottir S Embodied critical thinking: The experiential turn and its transformative aspects PhiloSophia 2019 9 1 92 109 10.1353/phi.2019.0015 Stern D The interpersonal world of the infant: A view from psychoanalysis and developmental psychology 1985 Basic Books Thorgeirsdottir S Shame, vulnerability and philosophical thinking Sophia 2020 59 1 5 17 10.1007/s11841-020-00773-w Thorgeirsdottir S Karlsdottir E Schwab GM Nature, culture and sexuate difference in Luce Irigaray’s pluralist model of embodied life Thinking Life with Luce Irigaray 2020 SUNY Press 99 117 Varela FJ Thompson E Rosch E The embodied mind: Cognitive science and human experience 1991 MIT Press
PMC009xxxxxx/PMC9005919.txt	==== Front AIDS Behav AIDS Behav AIDS and Behavior 1090-7165 1573-3254 Springer US New York 35416594 3662 10.1007/s10461-022-03662-0 Original Paper Risk of Severe COVID-19 Disease and the Pandemic’s Impact on Service Utilization Among a Longitudinal Cohort of Persons with HIV-Washington, DC http://orcid.org/0000-0003-3201-3845 Monroe Anne K. amonroe@gwu.edu 1 Xiao Jiayang 2 Greenberg Alan E. 1 Levy Matt E. 123 Temprosa Marinella 2 Resnik Jenna B. 1 Castel Amanda D. 1 The DC Cohort Executive CommitteeD’Angelo Lawrence Rakhmanina Natella Kharfen Michael Serlin Michael Kumar Princy Bhandaru Vinay Bezabeh Tsedenia Grover-Fairchild Nisha Mele Lisa Reamer Susan Sapozhnikova Alla Strylewicz Greg Temprosa Marinella 2 Xiao Kevin Byrne Morgan Castel Amanda Greenberg Alan Jaurretche Maria Kulie Paige Monroe Anne Peterson James Stewart Bianca Wilbourn Brittany Ma Yan Akselrod Hana Gajjala Jhansi L. Rana Sohail Horberg Michael Fernandez Ricardo Hebou Annick Dieffenbach Carl Masur Henry Bordon Jose Teferi Gebeyehu Benator Debra Ruiz Maria Elena Abbott Stephen 1 grid.253615.6 0000 0004 1936 9510 Department of Epidemiology, Milken Institute School of Public Health, The George Washington University, 950 New Hampshire Avenue, NW, Washington, DC 20052 USA 2 grid.253615.6 0000 0004 1936 9510 Department of Biostatistics, Milken Institute School of Public Health, The George Washington University, Washington, DC USA 3 grid.280561.8 0000 0000 9270 6633 Westat, Rockville, MD USA 13 4 2022 111 18 3 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. People with HIV (PWH) have a high burden of medical comorbidities, potentially putting them at increased risk for severe COVID-19. Additionally, during the COVID-19 pandemic, HIV care delivery has been restructured and the impact on HIV outcomes is unknown. The objectives of this study were first, to examine the risk of severe COVID-19 among PWH, using a definition incorporating clinical risk factors, and second, to examine the pandemic’s impact on HIV care. We used data from the DC Cohort, a large cohort of people receiving HIV care in Washington, DC. We found that a high proportion of participants across all age groups qualified as increased (58%) or high risk (34%) for severe COVID-19. Between 2019 and 2020, encounters increased (17.7%, increasing to 23.5% of active DC Cohort participants had an encounter) while laboratory utilization decreased (14.4%, decreasing to 11.4% of active DC Cohort participants had an HIV RNA test performed). Implications of our work include the importance of protecting vulnerable people with HIV from acquiring COVID-19 and potentially manifesting severe complications through strategies including vaccination. Additionally, acknowledging that HIV service delivery will likely be changed long-term by the pandemic, adaptation is required to ensure continued progress towards 90-90-90 goals. Supplementary Information The online version contains supplementary material available at 10.1007/s10461-022-03662-0. Keywords HIV COVID-19 Comorbidities Utilization HIV RNA http://dx.doi.org/10.13039/100000060 National Institute of Allergy and Infectious Diseases 1R24AI152598-01 ==== Body pmcIntroduction COVID-19 has had a severe global and national impact with over 464 million infections and 6 million deaths as of March 17, 2022 [1]. In the United States, the pandemic has highlighted the fragility of our public health and medical infrastructures and has unveiled many longstanding health disparities as reflected by the excessively high rates of COVID-19 among persons of color and those of lower socioeconomic status [2, 3]. COVID-19 also disproportionately impacts those who are over 65 years old and have underlying medical conditions. Our understanding of the impact of COVID-19 on people with HIV (PWH) continues to evolve [4]. Studies examining the impact of HIV on COVID-19 outcomes have been mixed, with some showing similar outcomes for people with and without HIV [5–8] and others showing worse outcomes for PWH [9, 10] or more specifically for PWH with low CD4 cell counts [11]. It remains unclear whether HIV itself or the comorbidities in PWH contribute more to risk of severe outcomes for COVID-19 [12]. Certainly, in the general population, presence of certain medical comorbidities portends worse outcomes from COVID-19. These medical comorbidities include cardiovascular disease, diabetes mellitus, hypertension, and chronic lung disease, among others [13–15]. Racial and income level disparities in the distribution of comorbidities that are most likely to be associated with severe COVID-19 have also been noted [16]. Medical comorbidities are even more common among PWH, especially older PWH [17–20], when compared with the general population. Many PWH exhibit at least one other high risk condition, and this comorbidity, or multimorbidity if they exhibit multiple other risk factors, increases their chance of having severe COVID-19 [21]. In Washington, DC, the first COVID-19 case was reported on March 7, 2020 [22]; since then there have been over 134,000 cases and 1319 deaths as of February 28, 2022 [23]. In a densely populated city where 57.5% of the population are racial/ethnic minorities, 54.7% of the population has at least one chronic medical condition [24] and there are an estimated 12,300 PWH [25], the city has become one of the national hotspots for COVID-19 infection [26]. At the same time that PWH living in this area are grappling with the impacts of COVID-19 on daily life and their potential increased health risk from COVID-19, they must continue to maintain their HIV-related care, medication-taking behavior, and viral suppression. Given this background, we used data from the DC Cohort to assess the impact of COVID-19 among HIV outpatients in Washington, DC during the first year of the pandemic. Specifically, we studied the proportion at high risk of severe COVID-19 disease and assessed the impact of the COVID-19 pandemic on HIV service utilization. Methods Study Setting This was a secondary data analysis of DC Cohort data. The DC Cohort is a clinical cohort of people receiving HIV care at 15 outpatient clinical sites in Washington, DC. Following informed consent at the 14 out of 15 DC Cohort sites where separate consent to participate in the DC Cohort is required, participants in the DC Cohort have their demographic and clinical data electronically and manually abstracted from the electronic health record (EHR) and entered into a centralized database on a monthly basis. Data are periodically linked to DC Department of Health HIV/AIDS, Hepatitis, STD, TB Administration (HAHSTA) data [27, 28]. All study procedures are approved by the George Washington University Institutional Review Board (IRB). Inclusion Criteria Included in this analysis were participants who were DC residents greater than 18 years old and with “active” study status at select DC Cohort sites as of June 30, 2019 and as of June 30, 2020. Active study status in each of those time intervals was defined as having an encounter in the 18 months prior to June 30, 2019 or in the 18 months prior to June 30, 2020. Sites that were included had utilization data (i.e. HIV encounter date or HIV viral load test date) available from both March–June 2019 and March–June 2020 (9 out of 15 clinic sites). Patients did not have to be active at both time points; these were separate samples used to compare 2019 to 2020. Descriptive statistics were calculated using the sample active on June 30, 2019. The sample in 2019 was used to characterize risk of severe COVID-19. Outcome: Risk of Severe COVID-19 We classified participants into three categories to indicate risk of severe COVID-19 disease, adapted from Banerjee et al. [21], which was a general population study and therefore did not include CD4 count as a risk factor. The high-risk category included individuals with BMI > 40 m/kg2, cardiovascular disease, diabetes and renal disease. Increased risk included BMI > 30 m/kg2, hypertension, respiratory disease, history of transplant, liver disease, autoimmune/rheumatologic disease, cancer, any smoking history, substance abuse, asthma, hemoglobin disorder and HIV with immunosuppression (CD4 < 200 cells/mm3 or not on ART). Low risk had none of the conditions. Additional details of this categorization, including ICD 9/10 codes, are displayed in Supplemental Table 1. Low risk participants had none of the conditions displayed in Supplemental Table 1, while increased risk and high-risk participants had at least one of the increased risk or high-risk conditions, respectively. We described the proportion of participants meeting the definition of increased risk and high risk, and the proportion in each risk category overall and by age group. Outcome: Utilization We also examined the impact of the pandemic on encounters and HIV-related laboratory testing. An encounter could be either an in-person or remote encounter (audio or video). The proportion of active participants with an encounter during a given month in 2019 was compared with the proportion of active participants with an encounter during the same month in 2020, after the start of the pandemic. Additionally, the proportion of active patients with an HIV RNA test during a given month in 2019 was compared with the proportion of active patients with an HIV RNA test during the same month in 2020. HIV RNA results are results available in the EHR for that patient, regardless of location (e.g., clinic vs outside lab provider) performed. Finally, we examined the proportion of participants who were virally suppressed (HIV RNA < 50 copies/mL) on the last assessment from 2019 compared with the proportion virally suppressed on the last assessment from 2020. Chi-square testing was used to compare proportions. Predictor Variables At enrollment into the DC cohort, and during the annual update process, the following variables used in this analysis were collected by manual abstraction from the medical record: Age, race, gender, housing, employment, smoking, and substance use disorder. The following are electronically abstracted from the medical record: antiretroviral (ART) prescription, HIV RNA, and CD4 cell count. Results As shown in Table 1, of the 3584 participants, the median age was 53 years and 1242 (34.7%) were aged 51–60. Most were male (64.2%) and Non-Hispanic Black (81.0%). A significant proportion were unemployed or disabled (38.5%). The participants had a high median recent CD4 count (643 cells/mm3). A small proportion (8.6%) had last HIV RNA > 200 copies/ml. A high proportion (58.8%) had ever smoked. Comorbidities and elevated BMI were common. Many participants were either at increased (58.3%) or high (33.8%) risk for severe COVID-19 if they were to become infected.Table 1 Characteristics of actively enrolled DC residents as of June 30, 2019, DC Cohort, (n = 3584) Variable N % Age (median, IQR) 53 (43–60) 19–30 160 4.5 31–40 568 15.8 41–50 782 21.8 51–60 1242 34.7 61–70 704 19.6 71–80 118 3.3 81 + 10 0.2 Gender Cisgender male 2300 64.2 Cisgender female 1189 33.2 Transgender female (assigned male at birth) 88 2.5 Transgender male (assigned female at birth) 7 0.2 Unknown 0 0.0 Race/ethnicity Non-Hispanic black 2904 81.0 Non-Hispanic white 348 9.7 Hispanic 203 5.7 Other/unknown 129 2.6 Unstably housed/homeless 400 11.2 Unemployed/disabled 1380 38.5 Recent CD4 count (last year only)a (median, IQR) 643 (445–883) Recent HIV viral load (last year only)a, copies/mL (median, IQR) UD (UD-30) Recent HIV viral load > 200 copies/mL (last year only)a 308 8.6 Not on ART 159 4.44 Ever smoker 2108 58.8 Substance use disorder 520 14.5 Chronic kidney disease/ESRD/hemodialysis 428 11.9 Liver disease 237 6.6 Cancer 318 8.9 Hypertensiona 2115 59.0 Diabetesa 642 17.9 BMI 30 + (weight last year only)a kg/m2 995 27.8 BMI 40 + (weight last year only)a kg/m2 228 6.4 Cardiovascular disease 337 9.4 Respiratory disease 686 19.1 Asthma 599 16.7 History of transplant 0 0.0 Autoimmune/rheumatologic disease 97 2.7 Hemoglobin disorder 7 0.2 Dementia 26 0.7 Risk for severe COVID-19 Low risk 283 7.9 Increased risk 2090 58.3 High risk 1211 33.8 ESRD end stage renal disease, ART antiretroviral therapy, BMI body mass index, UD undetectable aDue to missing values for clinical/laboratory results in the last year (Jul 2018–Jun 2019), not all participants could be evaluated for all conditions: BMI (n = 3133 with weight in last year and height ever); blood pressure (n = 3,198); glucose (n = 3158); HbA1c (n = 800), CD4 count (n = 2848), HIV viral load (n = 2822) Table 2 and Fig. 1 show the distribution of increased/high risk of severe COVID-19 disease stratified by age group. The vast majority of DC Cohort participants at all ages fell into the increased or high risk of severe COVID-19 groups. The highest number of high-risk participants was among the 51–60 year olds (N = 447) and the highest proportion of high risk participants was among those older than 80 years of age (80%).Table 2 Distribution of increased/high risk of severe COVID-19 disease stratified by age group, DC Cohort, 6/3/2019 (n = 3584) Age Total in age group Low risk Row % Increased risk High risk Row % N n n Row % n 19–30 160 35 21.9 114 71.3 11 6.9 31–40 568 89 15.7 384 67.6 95 16.7 41–50 782 72 9.2 496 63.4 214 27.4 51–60 1242 65 5.2 730 58.8 447 35.9 61–70 704 20 2.8 319 45.3 365 51.8 71–80 118 2 1.7 45 38.1 71 60.2 81+ 10 0 0 2 20.0 8 80.0 Overall 3584 283 7.9 2090 58.3 1211 33.8 Fig. 1 Risk group of DC Cohort participants by age, N = 3584 Figure 2 and Table 3 show utilization (encounters) and lab tests in 2019 and 2020. Comparing monthly encounters between 2019 and 2020, the proportion of participants with at least 1 encounter was similar in March and April but was significantly higher in June (23.5% vs. 17.7%, p < 0.0001). Figure 3 shows the proportion of encounters that were remote by month. Two-thirds (65.1%) of the encounters in March to June 2020 were remote encounters. Additionally, the proportion of individuals with at least one HIV RNA test was significantly lower in April, May and June 2020 compared to April, May and June 2019 (as an example, 15.2% of the cohort had an HIV RNA test in 2019 while 3.6% of the cohort had an HIV RNA test in 2020, p < 0.0001).The proportion of participants with an undetectable viral load (< 50 copies/ml) declined from 79.4% in 2019 to 76.2% in 2020 (p = 0.04).Fig. 2 Proportion of active patients who had an HIV-related encounter or HIV RNA tests performed, March–June 2019 and March–June 2020, DC Cohort Table 3 Utilization and labs, 3/1/19–6/30/19 and 3/1/20–6/30/20 Utilization/labs March 2019 April 2019 May 2019 June 2019 Number of active participants at end of month 3490 3523 3550 3584 Number of unique participants with > 1 encounter 596 682 697 635 % with > 1 encounter 17.1% 19.4% 19.6% 17.7% Number of encounters, all participants 634 722 769 702 Number of unique participants with HIV RNA test 368 537 603 517 % with > 1 HIV RNA test 10.5% 15.2% 17.0% 14.4% Number of HIV RNA tests, all participants 570 751 816 702 Number of unique participants with CD4 test 350 537 609 505 Number of CD4 tests, all participants 552 736 810 682 Utilization/labs March 2020 April 2020 May 2020 June 2020 Number of active participants at end of month 3452 3466 3485 3498 Number of unique participants with > 1 encounter 636 698 740 823 % with > 1 encounter 19.8% 20.1% 212.2% 23.5% Number of encounters, all participants 686 872 893 963 Number of unique participants with HIV RNA test 381 124 255 399 % with > 1 HIV RNA test 11.0% 3.5% 7.3% 11.4% Number of HIV RNA tests, all participants 392 131 259 411 Number of unique participants with CD4 test 370 114 253 380 Number of CD4 tests, all participants 401 131 280 419 Utilization/labs 3/1/19–6/30/19 3/1/20–6/30/20 Number of active participants at end of interval 3584 3498 Number of unique participants with > 1 encounter 2610 1997 Number of encounters, all participants 2834 3414 Number of participants with HIV RNA test 2025 1082 Number of HIV RNA tests, all participants 2839 1352 Number of unique participants with CD4 test 2001 1063 Number of CD4 tests, all participants 2780 1193 Proportion with last viral load in interval suppressed (< 50 copies/mL) 79.4% 76.2% The two proportions with last viral load in interval suppressed between 2019 and 2020 are not equal. The p value is 0.04 and statistically significant Fig. 3 Encounter type by month, 2019–2020 Tables 4 and 5 examine the association between demographic and clinical characteristics and having an encounter in the interval between 1 March 2020 and 30 June 2020 (Table 4) or having an HIV RNA test performed in the interval between 1 March 2020 and 30 June 2020 (Table 5). As shown in Table 4, cisgender females were more likely to have an encounter [aOR 1.36 (1.17, 1.58)], as were all other races compared to non-Hispanic white patients (aOR ranging from 1.67 to 2.45, all statistically significant). Individuals who were at high risk for COVID-19 were most likely to have an encounter [aOR 1.39 (1.05, 1.84)].As shown in Table 5, there was no significant difference by gender in having an HIV RNA test. Similarly to the results shown for having an encounter in Table 4, all races other than non Hispanic white were more likely to have an HIV RNA test. Those who were unstably housed were less likely to have an HIV RNA test, while those who were unemployed or disabled were more likely. There was a borderline p-value for an inverse association between being at increased risk for severe COVID-19 and having an HIV RNA test (p = 0.048).Table 4 Factors associated with having an encounter, 3/1/2020–6/30/2020, DC Cohort Variable OR (95% CI) (univariate analysis) aOR (95% CI) (multivariate analysis) Age (per 5 year increase) 1.01 (0.98, 1.04) 1 (0.97, 1.04) Gender Cisgender male (Ref) 1 1 Cisgender female 1.51 (1.31, 1.75)* 1.36 (1.17, 1.58)* Transgender female (assigned male at birth) 1.20 (0.79, 1.86) 1.13 (0.74, 1.76) Transgender male (assigned female at birth) 2.29 (0.49, 16.0) 2.08 (0.44, 14.7) Race/ethnicity Non-Hispanic Black 2.29 (1.83, 2.89)* 1.93 (1.52, 2.46)* Non-Hispanic white (Ref) 1 1 Hispanic 2.53 (1.77, 3.61)* 2.45 (1.71, 3.51)* Other/unknown 1.87 (1.24, 2.82)* 1.67 (1.10, 2.53)* Unstably housed/homeless [Yes vs. No (Ref)] 1.05 (0.85, 1.29) 1.01 (0.81, 1.25) Unemployed/disabled [(Yes vs. No (Ref)] 1.27 (1.11, 1.46)* 1.10 (0.96, 1.27) Risk for severe COVID-19 Low risk (ref) 1 1 Increased risk 1.25 (0.98, 1.61) 1.15 (0.89, 1.48) High risk 1.62 (1.25, 2.10)* 1.39 (1.05, 1.84)* p < 0.05; p < 0.01; *p < 0.001 Table 5 Factors associated with having an HIV RNA test performed, 3/1/2020–6/30/2020, DC Cohort Variable OR (95% CI) (univariate analysis) aOR (95% CI) (multivariate analysis) Age (per 5 year increase) 1.01 (0.98, 1.05) 1.01 (0.98, 1.05) Gender Cisgender male (Ref) 1 1 Cisgender female 1.18 (1.01, 1.37) 1.12 (0.96, 1.31) Transgender female (assigned male at birth) 0.86 (0.52, 1.38) 0.89 (0.53, 1.43) Transgender male (assigned female at birth) 0.41 (0.02, 2.38) 0.42 (0.02, 2.54) Race/ethnicity Non-Hispanic Black 1.46 (1.13, 1.91)* 1.34 (1.02, 1.78)* Non-Hispanic white (Ref) 1 1 Hispanic 2.09 (1.43, 3.05)* 2.10 (1.43, 3.07)* Other/unknown 1.56 (0.99, 2.43)* 1.48 (0.93, 2.31) Unstably housed/homeless [Yes vs. No (Ref)] 0.73 (0.57, 0.92)* 0.73 (0.57, 0.93)* Unemployed/disabled [Yes vs. No (Ref)] 1.33 (1.15, 1.54)* 1.29 (1.11, 1.51)* Risk for severe COVID-19 Low risk (ref) 1 1 Increased risk 0.80 (0.61,1.04) 0.76 (0.58,1.00)* High risk 0.93 (0.71, 1.23) 0.84 (0.63, 1.13) p < 0.05; p < 0.01; **p < 0.001 Discussion We demonstrated that a large proportion of PWH receiving HIV care in Washington, DC are at high risk for severe COVID-19 if they acquire SARS-CoV-2 infection. Additionally, we found that while HIV-related encounters were higher in March–June 2020 compared with the same months in the previous year, HIV RNA test utilization was lower. Among those who did receive an HIV RNA test, the proportion who were undetectable was lower. Our study participants have a mean age of 53 years, and most are on antiretroviral therapy (ART), are virally suppressed, and have CD4 cell counts over 500 cells/mm3. While they may be at lower risk for complications from COVID-19 based on these parameters and having relatively stable HIV disease, most are persons of color and are living with at least one condition known to increase the risk of complications and mortality from COVID-19 [29]. Our findings were similar to an analysis in the general U.S. population showing that a high proportion of Americans (75.4%) are at increased risk for severe COVID-19 given the high prevalence of comorbidities such as obesity and hypertension [30]. These findings highlight that regardless of whether a person is living with HIV, we need to protect those most vulnerable. This involves providing resources including access to medical care for non-COVID-19—related issues, providing insurance to those in need, and providing locations to properly isolate and/or quarantine if needed. In evaluating clinical utilization among our sample, encounters were stable or higher, which is encouraging news. At the height of the COVID-19 pandemic, the ability to receive in-person services was extremely difficult and impractical [31]. With social distancing measures and other COVID-19-related restrictions, in-person clinic encounters were greatly restricted, and the transition to telehealth and a virtual environment posed some new obstacles and opportunities. Despite those challenges, there has been a significant increase in phone and/or video consultations with care providers, and clinic websites and social media have been updated to reflect current protocols and offer assistance when needed [31]. In general, many HIV clinics had historically high no-show rates pre-pandemic [32] and the increasing use of phone/telehealth encounters may have actually been beneficial in increasing visit adherence. This may be particularly true for those patients with challenges in obtaining transportation, child care, or time off from work [33] or individuals for whom the experience of coming to clinic is stigmatizing and reminds them of their HIV status [34]. However, individuals with less reliable access to phone and/or data may not reap the same benefits and telehealth potentially could worsen disparities [35]. In addition, the clinical impact of not having face-to-face encounters is unknown. Also, the use of preventive services has declined during COVID-19 [36] which may be particularly detrimental to people with HIV who are at higher risk of cardiovascular disease, cancer, and cancer death [37–39]. An analysis of data from the HIV Outpatient Study (HOPS) found an increase in telemedicine visits in 2020, however, determined that total encounters had not rebounded to pre-pandemic levels by September 2020 [40]. However, we found that a higher proportion of active patients had visits in May and June 2020 compared to May and June 2020. The explanation for this difference is not clear. Clinics in DC may have made extra effort to increase encounters in May/June 2020. As clinics envision the ways they will provide care in the future, engagement in care may be enhanced by ensuring the ongoing availability of both in-person and telehealth visits. Our results showed that lab monitoring, which is an important part of HIV care, was lower during the pandemic. People who are adherent and in good general health may not need as frequent lab monitoring. However, those who are potentially most vulnerable to poor health outcomes may do worse with less frequent lab monitoring. Our finding of a lower proportion of undetectable results in individuals receiving an HIV RNA test in 2020 compared with 2019 has several potential explanations. It may represent challenges obtaining ART due to pandemic-related service interruptions, resulting in less suppression. Or, it may represent that clinicians advised clinically stable patients to delay laboratory testing. In examining factors associated with using telehealth or having an HIV RNA test, we found that women were more likely to have an encounter, as were individuals who were not of white race. We also showed that the individuals at highest risk for severe COVID-19 were more likely to have an encounter in March–June 2020 and less likely to have an HIV RNA. The significance of these findings is unclear. Generally, women are more likely to seek medical care [41]. Perhaps women in our sample desired the support of their clinic provider or were more worried about COVID and wanted to engage in medical care to discuss their concerns with their provider. With individuals of non-white race being more vulnerable to COVID infection, hospitalization, and death [42, 43], perhaps clinics or providers were making additional effort to engage those patients. Finally, it is interesting that those at increased risk for severe COVID-19 were more likely to have an encounter and less likely to have a HIV RNA test. This may indicate that physicians were advising their most medically fragile patients to stay home if their HIV was stable or medically fragile patients choosing to stay home because of their concern for being exposed to COVID. Our findings represent a preliminary investigation of how HIV care delivery has been impacted by the COVID pandemic and what those changes may mean for HIV care outcomes. Additional studies that have examined these issues have shown that the pandemic may impact maintenance of viral suppression due to decreased access to clinical services through in-person visits [44]. Our findings were similar to those of Mayer et al. in Boston area HIV clinic [45] with a surge in telehealth and similar viral suppression and differed somewhat from a Midwestern HIV clinic [46] where viral suppression was stable and retention in care decreased. These findings were presented in a review of the impact of COVID 19 [47] which concluded that telehealth offers many opportunities and simultaneously presents many challenges, particularly for the most vulnerable patients. It remains to be seen if people who always did well in HIV care (i.e., were always retained and suppressed) will continue to do well in HIV care, whatever the delivery modality, and people whose needs were not well met by our medical care delivery system prior to the pandemic will continue to struggle to remain engaged and suppressed. We can use DC Cohort data to examine these outcomes over time in future analyses. The major limitation of our analysis is that we have not yet linked risk of severe COVID-19 to occurrence of severe COVID-19 and/or change in behavior to mitigate COVID-19 acquisition risk. However, we have a major survey underway that will address COVID-19 incidence (self-reported, and electronic health record confirmed) as well as the adoption of risk mitigation behaviors, the social and mental health impacts on PWH, and the use of telehealth. Another limitation is that this analysis included 9 out of 15 DC Cohort clinics based on data availability. Although all DC Cohort clinics were not included, the clinics with data available represented a mix of clinics located in academic centers and community-based clinics. However, these results may not be fully generalizable to all people receiving HIV care in Washington, DC. Another limitation is that we did not collect information on whether HIV RNA test data came from the primary clinical site or an outside lab site. If the primary clinical site was the only place that a patient could get labs, and it was closed, that could explain lack of viral load testing. We are conducting a site survey to capture information about COVID-related service changes and this issue could potentially be addressed in future research. A major benefit of this survey is that we will be able to link diagnosis and other clinical data to patient-reported survey data. A strength of our study is the large sample and the availability of utilization data with a relatively short turnaround, giving us the ability to examine the impact of COVID-19 on clinical utilization in the short-term. In future analyses, we will use subsequent data to evaluate changes beyond the immediate start of the pandemic (March–June 2020). This will allow us to determine whether these patterns hold throughout 2020 and into 2021, or if they were temporary. In summary, we found high risk for severe disease among DC Cohort participants and declines in laboratory utilization despite increases in HIV-related contacts. There are several important implications of these findings. First, because many people with HIV have a high risk of severe disease they should be a high priority group for the SARS-CoV-2 vaccine. All those providing health care and wraparound services to PWH should emphasize the importance of vaccination. Our findings additionally reinforce the importance of ensuring continuity of care and consistent access to ARTs and HIV care so that we do not lose momentum and the gains we have made to date in curbing the HIV epidemic in DC and nationally in achieving the 90-90-90 goals. The conduct of systematic studies to further understand interactions between COVID-19 and HIV are needed. These will be facilitated by using existing longitudinal cohorts such as the DC Cohort and enable researchers to assess the incidence of COVID-19 among PWH, as well as its longer- term impacts on the care continuum. Finally, given that we found an impact of the pandemic on service provision for PWH, future studies could assess the impact of the pandemic on services such as STI prevention as well as look at whether the rates of new HIV diagnoses increased and whether those newly diagnosed were able to get started on ART in a timely manner. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 16 KB) Acknowledgements Data in this manuscript were collected by the DC Cohort Study Group with investigators and research staff located at: Children's National Medical Center Adolescent (Lawrence D’Angelo) and Pediatric (Natella Rakhmanina) clinics; The Senior Deputy Director of the DC Department of Health HAHSTA (Michael Kharfen); Family and Medical Counseling Service (Michael Serlin); Georgetown University (Princy Kumar); The George Washington University Biostatistics Center (Vinay Bhandaru, Tsedenia Bezabeh, Nisha Grover-Fairchild, Lisa Mele, Susan Reamer, Alla Sapozhnikova, Greg Strylewicz, Marinella Temprosa, and Kevin Xiao); The George Washington University Department of Epidemiology (Morgan Byrne, Amanda Castel, Alan Greenberg, Maria Jaurretche, Paige Kulie, Anne Monroe, James Peterson, Bianca Stewart, and Brittany Wilbourn) and Department of Biostatistics and Bioinformatics (Yan Ma); The George Washington University Medical Faculty Associates (Hana Akselrod); Howard University Adult Infectious Disease Clinic (Jhansi L. Gajjala) and Pediatric Clinic (Sohail Rana); Kaiser Permanente Mid-Atlantic States (Michael Horberg); La Clinica Del Pueblo (Ricardo Fernandez); MetroHealth (Annick Hebou); National Institutes of Health (Carl Dieffenbach, Henry Masur); Washington Health Institute, formerly Providence Hospital (Jose Bordon); Unity Health Care (Gebeyehu Teferi); Veterans Affairs Medical Center (Debra Benator); Washington Hospital Center (Maria Elena Ruiz); and Whitman-Walker Institute (Stephen Abbott). Author Contributions ADC, MEL, and AKM designed the study, MEL, JX, MT, AKM processed the data, performed the analysis, and designed the tables and figures. AKM, JX, AEG, MEL, MT, ADC aided in interpreting the results. AKM took the lead in writing the manuscript with assistance from JBR. All authors provided critical feedback and helped shape the research, analysis and manuscript. Funding National Institute of Allergy and Infectious Diseases, 1R24AI152598-01. Data Availability Deidentified data available upon written request. Code Availability Analytic code available upon written request. Declarations Conflict of interest The authors have no relevant financial or non-financial interests to disclose. Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the George Washington University Institutional Review Board (IRB 071029). Consent to Participate Informed consent was obtained from all individual participants included in the study. Consent for Publication N/A. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References 1. Johns Hopkins Coronavirus Resource Center. COVID-19 dashboard by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (JHU). https://coronavirus.jhu.edu/map.html. Accessed 17 Mar 2022. 2. Mackey K Ayers CK Kondo KK Saha S Advani SM Young S Racial and ethnic disparities in COVID-19–related infections, hospitalizations, and deaths Ann Intern Med 2020 10.7326/m20-6306 33253040 3. Seligman B Ferranna M Bloom DE Social determinants of mortality from COVID-19: a simulation study using NHANES PLoS Med 2021 18 1 13 4. Lesko C Bengtson A HIV and COVID-19: intersecting epidemics with many unknowns Am J Epidemiol 2021 190 10 16 10.1093/aje/kwaa158 32696057 5. Blanco J Ambrosioni J Garcia F Martinez E Soriano A Mallolas J COVID-19 in patients with HIV: clinical case series Lancet HIV 2020 10.1016/S2352-3018(20)30111-9 32416770 6. Gervasoni C Meraviglia P Riva A Giacomelli A Oreni L Minisci D Clinical features and outcomes of patients with human immunodeficiency virus with COVID-19 Clin Infect Dis 2020 10.1093/cid/ciaa579 32459828 7. Sigel K Swartz T Golden E Covid-19 and people with HIV infection: outcomes for hospitalized patients in New York city Clin Infect Dis 2020 10.1093/cid/ciaa880 32594164 8. Karmen-Tuohy S Carlucci P Zervou F Zacharioudakis I Rebick G Klein E Outcomes among HIV-positive patients hospitalized with COVID-19 J Acquir Immune Defic Syndr 2020 395 10229 9. Del Amo J Polo R Moreno S Incidence and severity of COVID-19 in HIV-positive persons receiving antiretroviral therapy Ann Intern Med 2020 10.7326/M20-3689 32589451 10. Bhaskaran K Rentsch CT MacKenna B Schultze A Mehrkar A Bates CJ HIV infection and COVID-19 death: a population-based cohort analysis of UK primary care data and linked national death registrations within the OpenSAFELY platform Lancet HIV 2021 8 e24 e32 10.1016/S2352-3018(20)30305-2 33316211 11. Dandachi D Geiger G Montgomery MW Karmen-Tuohy S Golzy M Antar AAR Characteristics, comorbidities, and outcomes in a multicenter registry of patients with human immunodeficiency virus and coronavirus disease 2019 Clin Infect Dis 2020 65212 1 9 12. Collins LF Clinical characteristics, comorbidities and outcomes among persons with HIV hospitalized with coronavirus disease 2019 in Atlanta Georgia AIDS 2020 34 1789 10.1097/QAD.0000000000002632 32675581 13. Cunningham JW Vaduganathan M Claggett BL Jering KS Bhatt AS Rosenthal N Clinical outcomes in young US adults hospitalized with COVID-19 JAMA Intern Med 2020 10.1001/jamainternmed.2020.5313 32955549 14. Petrilli CM Jones SA Yang J Rajagopalan H O’Donnell L Chernyak Y Factors associated with hospital admission and critical illness among 5279 people with coronavirus disease 2019 in New York City: prospective cohort study BMJ 2020 10.1136/bmj.m1966 32444366 15. Williamson EJ Walker AJ Bhaskaran K Bacon S Bates C Morton CE Factors associated with COVID-19-related death using OpenSAFELY Nature 2020 584 430 436 10.1038/s41586-020-2521-4 32640463 16. Weiser JK Tie Y Beer L Neblett Fanfair R Shouse RL Racial/ethnic and income disparities in the prevalence of comorbidities that are associated with risk for severe COVID-19 among adults receiving HIV care, United States, 2014–2019 J Acquir Immune Defic Syndr 2021 86 3 297 304 10.1097/QAI.0000000000002592 33351530 17. Levy ME Greenberg AE Hart R Powers Happ L Hadigan C Castel A High burden of metabolic comorbidities in a citywide cohort of HIV outpatients: evolving health care needs of people aging with HIV in Washington, DC HIV Med 2017 18 724 735 10.1111/hiv.12516 28503912 18. Collins LF Armstrong WS What it means to age with HIV infection: years gained are not comorbidity free JAMA Netw Open 2020 3 e208023 10.1001/jamanetworkopen.2020.8023 32539147 19. Gallant J Hsue PY Shreay S Meyer N Comorbidities among US patients with prevalent HIV infection—a trend analysis J Infect Dis 2017 216 1525 1533 10.1093/infdis/jix238 29253205 20. Mayer KH Loo S Crawford PM Crane HM Leo M Denouden P Excess clinical comorbidity among HIV-infected patients accessing primary care in US community health centers Public Health Rep 2018 133 109 118 10.1177/0033354917748670 29262289 21. Banerjee A Pasea L Harris S Estimating excess 1-year mortality associated with the COVID-19 pandemic according to underlying conditions and age: a population-based cohort study Lancet 2020 395 1715 10.1016/S0140-6736(20)30854-0 32405103 22. The New York Times. Tracking Coronavirus in Washington, D.C.: latest map and case count. https://www.nytimes.com/interactive/2020/us/washington-dc-coronavirus-cases.html. Accessed 17 Mar 2022. 23. DC Health. COVID-19 Surveillance. https://coronavirus.dc.gov/data. Accessed 17 Mar 2022. 24. Partnership to fight chronic disease. What is the impact of chronic disease on Washington, D.C.? https://www.fightchronicdisease.org/sites/default/files/download/PFCD_DC_Factsheet_FINAL1.pdf. Accessed 17 Mar 2022. 25. DC Health. Annual Epidemiology & Surveillance Report. https://dchealth.dc.gov/sites/default/files/dc/sites/doh/publication/attachments/2021%20Annual%20Surveillance%20Report_final_3.29.pdf. Accessed 17 Mar 2022. 26. Washington Post. White House says D.C. region among worst in country, as summer closures continue. https://www.washingtonpost.com/local/white-house-says-dc-region-among-worst-in-country-as-summer-closures-continue/2020/05/22/31e4cc8c-9c3a-11ea-ac72-3841fcc9b35f_story.html. Accessed 22 May 2020. 27. Greenberg AE Hays H Castel AD Subramanian T Happ LP Jaurretche M Development of a large urban longitudinal HIV clinical cohort using a web-based platform to merge electronically and manually abstracted data from disparate medical record systems: technical challenges and innovative solutions J Am Med Inform Assoc 2016 23 635 643 10.1093/jamia/ocv176 26721732 28. Castel AD Terzian A Opoku J Happ LP Younes N Kharfen M Defining care patterns and outcomes among persons living with HIV in Washington, DC: linkage of clinical cohort and surveillance data JMIR Public Health Surveill 2018 4 e23 10.2196/publichealth.9221 29549065 29. Centers for Disease Control. People at increased risk and other people who need to take extra precautions. 2021. https://www.cdc.gov/coronavirus/2019-ncov/need-extra-precautions/people-at-higher-risk.html. Accessed 3 Jan 2021 30. Ajufo E Rao S Navar AM Pandey A Ayers CR Khera A U.S. population at increased risk of severe illness from COVID-19 Am J Prev Cardiol 2021 6 100156 10.1016/j.ajpc.2021.100156 33615285 31. Pinto R Park S COVID-19 pandemic disrupts HIV continuum of care and prevention: implications for research and practice concerning community-based organizations and frontline providers AIDS Behav 2020 24 1 4 10.1007/s10461-020-02893-3 30903450 32. Adams JA Whiteman K McGraw S Reducing missed appointments for patients with HIV: an evidence-based approach J Nurs Care Qual 2020 35 165 170 10.1097/NCQ.0000000000000434 31464846 33. Yehia BR Stewart L Momplaisir F Mody A Holtzman CW Jacobs LM Barriers and facilitators to patient retention in HIV care BMC Infect Dis 2015 15 1 10 10.1186/s12879-015-0990-0 25567701 34. Vanable PA Carey MP Blair DC Littlewood RA Impact of HIV-related stigma on health behaviors and psychological adjustment among HIV-positive men and women AIDS Behav 2006 10 473 482 10.1007/s10461-006-9099-1 16604295 35. Velasquez D Mehrotra A Ensuring the growth of telehealth during COVID-19 does not exacerbate disparities in care Health Aff Blog 2020 10.1377/hblog20200505.591306 36. Martin K, Kurowski D, Given P, Kennedy K, Clayton E. The impact of COVID-19 on the use of preventive healthcare. 2021. https://healthcostinstitute.org/hcci-research/the-impact-of-covid-19-on-the-use-of-preventive-health-care. Accessed 1 Mar 2021 37. Coghill AE Pfeiffer RM Shiels MS Engels EA Excess mortality among HIV-infected individuals with cancer in the United States Cancer Epidemiol Biomarkers Prev 2017 26 1027 1033 10.1158/1055-9965.EPI-16-0964 28619832 38. Alonso A Barnes AE Guest JL Shah A Shao IY Marconi V HIV infection and incidence of cardiovascular diseases: an analysis of a large healthcare database J Am Heart Assoc 2019 10.1161/JAHA.119.012241 31830872 39. Engels EA Biggar RJ Hall HI Cross H Crutchfield A Finch JL Cancer risk in people infected with human immunodeficiency virus in the United States Int J Cancer 2008 123 187 194 10.1002/ijc.23487 18435450 40. Tedaldi EM Hou Q Armon C Palella FJ Li J Simoncini G HIV Ambulatory care during COVID-19 pandemic in US: visits and viral load testing Topics Antivir Med 2021 2021 294 41. Brett KM and Burt CW. Utilization of ambulatory medical care by women; United States, 1997-1998. (2001). Accessed from https://www.cdc.gov/nchs/data/series/sr_13/sr13_149.pdf 42. Price-Haywood EG Burton J Fort D Seoane L Hospitalization and mortality among black patients and white patients with Covid-19 N Engl J Med 2020 382 26 2534 2543 10.1056/NEJMsa2011686 32459916 43. CDC. Risk for COVID-19 Infection, Hospitalization, and death by race/ethnicity https://www.cdc.gov/coronavirus/2019-ncov/covid-data/investigations-discovery/hospitalization-death-by-race-ethnicity.html. Accessed 31 Oct 2021 44. Spinelli MA Hickey MD Glidden DV Nguyen JQ Oskarsson JJ Havlir D Gandhi M Viral suppression rates in a safety-net HIV clinic in San Francisco destabilized during COVID-19 AIDS 2020 34 15 2328 2331 10.1097/QAD.0000000000002677 32910069 45. Mayer KH Levine K Grasso C Multani A Gonzalez A Biello K 541. Rapid migration to telemedicine in a Boston Community Health Center is associated with maintenance of effective engagement in HIV care Open Forum Infect Dis 2020 7 Suppl 1 S337 S338 10.1093/ofid/ofaa439.735 46. Fadul N 112. A quality management project of a midwestern academic HIV clinic operation during COVID-19: implementation strategy and preliminary outcomes Open Forum Infect Dis 2020 7 S184 S185 10.1093/ofid/ofaa439.422 47. Budak JZ Scott JD Dhanireddy S Wood BR The impact of COVID-19 on HIV care provided via telemedicine-past, present, and future Curr HIV/AIDS Rep 2021 18 2 98 104 10.1007/s11904-021-00543-4 33616811
PMC009xxxxxx/PMC9006065.txt	==== Front Women Birth Women Birth Women and Birth 1871-5192 1878-1799 Published by Elsevier Ltd on behalf of Australian College of Midwives. S1871-5192(21)00044-5 10.1016/j.wombi.2021.03.008 Article Perspectives of pregnant women during the COVID-19 pandemic: A qualitative study Atmuri Kiran ab* Sarkar Mahbub c Obudu Efe b Kumar Arunaz a a Department of Obstetrics and Gynaecology, Monash University, Victoria, Australia b Women’s Health Unit, Peninsula Health, Victoria, Australia c Monash Centre for Scholarship in Health Education, Faculty of Medicine, Nursing and Health Sciences, Monash University, Victoria, Australia ⁎ Corresponding author at: Peninsula Clinical School, Monash University, Academic Centre, Level 3, PO Box 52, Frankston, 3199, Victoria, Australia. 15 3 2021 5 2022 15 3 2021 35 3 280288 13 10 2020 21 2 2021 10 3 2021 © 2021 Published by Elsevier Ltd on behalf of Australian College of Midwives. 2021 Since January 2020 Elsevier has created a COVID-19 resource centre with free information in English and Mandarin on the novel coronavirus COVID-19. The COVID-19 resource centre is hosted on Elsevier Connect, the company's public news and information website. Elsevier hereby grants permission to make all its COVID-19-related research that is available on the COVID-19 resource centre - including this research content - immediately available in PubMed Central and other publicly funded repositories, such as the WHO COVID database with rights for unrestricted research re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for free by Elsevier for as long as the COVID-19 resource centre remains active. Background The current COVID-19 pandemic has been shown to have profound effects on pregnant women globally, particularly, on their psycho-social wellbeing. Despite this, there has been limited qualitative inquiry into the experiences of pregnant women during the pandemic. Aim This original research aimed to study the perspectives of pregnant women in Australia in relation to the impact of the COVID-19 pandemic on their pregnancy experience. Methods A qualitative descriptive study design with semi-structured interviews was adopted. The study was performed in Melbourne, Australia. A total of fifteen interviews were conducted. Data was analysed thematically to develop major themes and subthemes. Results A total of four major themes were developed: support for a positive experience, impact on preparedness in pregnancy and beyond, facing uncertainty of a pandemic, and retaining resilience and optimism. Conclusions The COVID-19 pandemic has affected the experience of pregnant women with potential to compromise their psycho-social wellbeing. The major themes identified in this study offer insight to organisations to develop woman-centred care during the pandemic and optimise the psycho-social wellbeing of pregnant women. Keywords COVID-19 Pregnancy Birth Australia Perspectives Experiences ==== Body pmc Statement of significance Problem or issue There is limited qualitative inquiry, particularly in the Australian context, on the impact of the COVID-19 pandemic on the experiences of pregnant women. What is already known There is emerging research indicating that the COVID-19 pandemic is having a negative impact on the perinatal mental health and perception of social wellbeing in pregnant women. What this study adds This qualitative descriptive study provides insights into the perspectives of pregnant women in regards to their care during the COVID-19 pandemic. Women responded with concerns surrounding support, uncertainty and preparedness, balanced with views of resilience and optimism. This information will assist organisations in continued focus on woman-centred care during the pandemic. 1 Introduction The coronavirus disease 2019 (COVID-19) pandemic is a public health emergency that is currently active with a global mortality of more than 2.3 million and there continues to be an increase in the number of infections and deaths each day [1]. Our understanding of COVID-19 and its implications to pregnant women continues to evolve. The indirect impacts of the wider COVID-19 pandemic on pregnant women may be substantial [2]. Physical distancing and infection control policies have altered how health care has been provided for pregnant women, such as reduced face-to-face appointments, increased use of telehealth and limitations in support people during pregnancy and the peri-partum period [3]. Learning from past infectious disease outbreaks, healthcare providers propose that these changes in health care delivery may challenge the psychological wellbeing of pregnant women [[4], [5], [6]]. The effects of the COVID-19 pandemic on mental wellbeing may be heightened by existing mental health illness, social isolation, reduced social supports, disruption to normal routines and misinformation [7,8]. Similar findings have been reported during previous infectious disease outbreaks such as severe acute respiratory syndrome (SARS), Ebola virus and more recently, Zika [9]. An early COVID-19 study in Italy which used anxiety and stress-related assessments in pregnant women found more than half of pregnant women self-rated the psychological impact of the pandemic as severe and two thirds more anxious than normal [10]. Emerging studies from regions with high case numbers of COVID-19, such as Italy, China and North America, have consolidated these findings and shown that, compared to prior to the pandemic, pregnant women are suffering from significantly increased rates of depressive symptoms and anxiety disorders [[10], [11], [12], [13], [14], [15]]. The psychological effects have been found to be disproportionality affecting women with social vulnerabilities, existing mental health illness and those from minority ethnic groups [7,12]. Despite findings of compromised psychological wellbeing, there is limited understanding of the experiences and needs of pregnant women living through the COVID-19 pandemic. Understanding people’s experience is a complex task. It requires an understanding of how people perceive a certain situation (cognition), how they react under that situation (behaviour) and how they are affected by external influences (environmental effects) [16]. Considering pregnant women’s experience, they are not only affected by pregnancy but also by external influences that may have impacted their lives during the pandemic such as restrictions imposed on physical interactions [17]. In turn, these influences may affect women’s behaviour on how they respond to the challenges or adapt to situations. In reverse, women’s behaviour may further impact their self-perception of wellbeing and their external environment, hence, fitting the dynamic nature of cognition, behaviour and environmental effects [17,18]. Teti et al. [19] advocated for qualitative studies during the COVID-19 pandemic to help capture people’s emotional responses to the pandemic and its social implications to help better understand the assumptions in quantitative epidemiology models and improve the management of the pandemic. In 2019, the Australian Department of Health advised utilising woman-reported experiences as a matter of priority to improve the quality of maternity care [20]. With a view to help in the development of woman-centred care and minimise the psycho-social impact of the COVID-19 pandemic, this original research was conducted to understand the perspectives of pregnant women in an Australian public health setting. It also aimed to explore how women are responding to challenges presented by COVID-19 in relation to their pregnancy experience and receiving maternity care during the pandemic. 2 Methods 2.1 Design This study was underpinned by a social constructionist view, which acknowledges multiple interpretations of reality as individuals make sense of their experiences through social interactions and the surrounding environment [21]. Aligning with this view, we employed a qualitative descriptive research design [22]. Qualitative descriptive design is grounded in the principles of naturalistic inquiry representing the view that reality exists within various contexts that are dynamic and perceived differently by people; therefore, reality is multiple and subjective [23]. The goal of qualitative descriptive research is to present a comprehensive descriptive summary of the experiences and perceptions of a group of people, without abstract rendering of data [22]. In the current research, this translates into understanding the experiences of pregnant women in a unique context: the COVID-19 pandemic. Following the dominant tradition of data collection methods being used in qualitative descriptive research [22], we employed semi-structured interviews to understand the experiences of pregnant women during the pandemic. 2.2 Setting The study was performed in an outer metropolitan area of Melbourne, Australia. The study was based at a secondary level public hospital where approximately 3000 babies are born each year. 2.3 Participants and recruitment Participants were women with a pregnancy of any gestation booked and receiving antenatal care at the hospital. The inclusion criterion was being currently pregnant. There were no exclusion criteria for participants. In order to minimise any coercion, the authors were not involved in recruitment of participants; instead, participants were recruited by midwives who were not involved in the other aspects of the research. Midwives who care for women in the outpatient clinic at the hospital were briefed on the project with a verbal presentation and printed information. They recruited women face-to-face through convenience sampling. Potential participants were provided with written information about the research project and a consent form. The participants were contacted by a member of the research team (KA) by telephone to arrange an interview. In total, twenty-four women were approached to participate in the study of which 21 provided written consent. Three women had birthed prior to scheduled interview, one woman withdrew consent to participate and two women were not able to be contacted. Eventually, 15 women took part in the interview. The initial two interviews served as pilot interviews. The socio-demographic details of the participating women are listed in Table 1 . Ethnicity, highest qualification, marital status, home ownership and annual household income data was self-reported by women. Overall, the mean age of the participating group of women was 31 years (range, 20–36). There were 10 primigravida and 5 multigravida women. The average gestational age was 30 weeks (range, 19–36). Eleven women identified as Caucasian, 2 women as East Asian and 2 women as South Asian. Twelve women had partners, either married or in de facto relationships, and three women were single. Two of the fifteen women were unemployed.Table 1 Socio-demographic characteristics of participants. Table 1Initials Age Parity Gestation (weeks) Ethnicity Highest qualification Marital status Home ownership Annual household income (AUD) aAM 33 Primigravida 29 Caucasian Certificate Single Own $50−100,000 aCL 20 Primigravida 31 Caucasian High school Single Renting $50−100,000 EW 39 Multigravida 35 Caucasian Bachelor Married Own $150−200,000 JR 33 Primigravida 32 East Asian High school Married Own $50−100,000 TB 33 Primigravida 32 Caucasian Certificate Single b $50−100,000 FR 28 Primigravida 36 Caucasian Masters Married Own $50−100,000 JD 26 Primigravida 28 Caucasian Bachelor De facto Own $150−200,000 KC 36 Primigravida 34 Caucasian Bachelor Married Renting $150−200,000 NM 36 Primigravida 36 Caucasian Masters Married Own $100−150,000 SS 33 Multigravida 24 South Asian b Married b b BS 29 Primigravida 29 Caucasian Certificate De facto Rent $100−150,000 MN 26 Multigravida 22 Caucasian Diploma De facto Own $150−200,000 FM 29 Primigravida 19 Caucasian Bachelor Married Own $200−250,000 CC 23 Multigravida 29 East Asian Diploma De facto Renting $100−150,000 VB 35 Multigravida 28 South Asian Masters Married Own $50−100,000 a The interviews with the women, AM and CL, served as pilot interviews. b Information not provided. 2.4 Ethics The study received ethical approval from the hospital Human Research Ethics Committee on May 27, 2020 (HREC ref. number LNR/64473PH-2020) as a low-risk project. Participating women gave informed and free written consent to take part in the study. Confidentiality was assured by following local institutional policy on data management. 2.5 Data collection Recruitment and interviews with women were completed between 1st June and 19th June, 2020. At the time of data collection, the metropolitan region in the state of Victoria entered stay-at-home restrictions with key mandates being face-covering, closure of all non-essential businesses and limited travel permitted. Physical distancing restrictions and telehealth models of maternity care required that no face-to-face interaction was permitted apart from necessary clinical care. Accordingly, all interviews were conducted either through telephone or video conference application based on participant preference. All interviews were conducted by the author, KA, who contacted participants one week after initial recruitment to confirm consent and arrange the interview time. This initial contact allowed participants to reconsider participation in the study (if necessary) and become familiar with author KA which helped build initial rapport between the interviewer and interviewees. As women were in their homes during the stay-at-home restrictions, it was possible that non-participants (e.g. women’s family members) were present during the interview. De-identified audio-only recordings were made. To ensure data integrity, the interviews were transcribed verbatim through an independent professional transcription service approved by the University. Author, KA, listened to all recordings and checked the transcripts for accuracy. Personal details were de-identified in the transcripts. Fifteen interviews were conducted, with a total of 4 hours and 36 minutes of transcribed data. Each interview lasted between 9 min and 28 min (mean 17 min). We felt our sample achieved sufficient information power given our focused aim for this paper (i.e. pregnant women’s perspectives on the impact of the COVID-19 pandemic on their pregnancy experience), our tight sample specificity (i.e. pregnant women in Australia), the high-quality dialogue in the interview, and our focused team-based analysis strategy [24]. An interview guide was developed and refined with two pilot interviews. The date and time of the interview was noted at the commencement of recording, together with participants’ basic socio-demographic details. The interview started with an open-ended question: “How have you been feeling during this COVID-19 pandemic?”. This broad and non-threatening question allowed rapport-building early in the interview. As the interviews were semi-structured, the direction of the interview was guided by the participant’s answers. The interview guide covered topics such as the impact of the pandemic on participants’ pregnancy experience and birth preferences, concerns in relation to COVID-19, and questions on health care delivery (e.g. impact of changes in care, knowledge of changes in care and precautions taken by the hospital). Finally, participants were invited to express how they could be best supported by the health care service during the pandemic. 2.6 Data analysis The transcripts were read multiple times to ensure familiarity with, and to develop a deeper understanding of the data. The data was analysed thematically using reflexive inductive coding method following Braun and Clarke [25]. Initially, the authors, AK and KA, independently coded the data inductively, not pre-determined by any pre-existing theories. After coding, prominent themes and subthemes were identified. The authors, AK and KA, discussed the results to identify themes in agreement; this process took several rounds of analysis and was undertaken by telephone and online meetings. These meetings were used to exercise reflexivity, where the authors examined their positioning within the research. After the initial coding was done and the themes identified, these were all checked by the third author, MS, across all the transcripts. Further online meetings were organised between the three authors to compare, contrast and negotiate our interpretations of the data and discuss the interpretations of the findings in light of the research literature. This approach helped to maximise the credibility of the analysis and enhance the rigor of the study. 2.7 Team reflexivity Reflexivity is an important aspect of qualitative research, allowing readers to assess the credibility of the data analysis by understanding the positionality of the authors within the research [26]. Our team of four was diverse in terms of research experience, and academic/clinical backgrounds. KA, the primary author, is an obstetrician with two years of experience at the research site. AK is a senior obstetrician and an academic health education researcher with over eight years of experience in qualitative research. MS is an academic health education researcher with over ten years of experience in qualitative research. EO is an obstetrician with over five years of experience at the research site. Both KA and EO are entry-level qualitative researchers and value the role of qualitative approach in understanding complex phenomena. Diversity within the author team supported more rigorous data interpretation and researcher triangulation with team members contributing different perspectives and insights into the data analysis and reporting. 3 Results Four overarching themes were developed through data analysis. The themes and subthemes are listed in Table 2 and their connections represented diagrammatically in Fig. 1 .Table 2 Themes and sub-themes. Table 2Theme Sub-themes Support for a positive experience Seeking inclusion of partner or support person Value for peer and intergenerational support Re-assurance from healthcare professionals Impact on preparedness in pregnancy and beyond Changes to birthing and parenting education Concerns around early discharge Facing uncertainty of a pandemic Disruption to the “normal” pregnancy experience Perceived risk of acquiring infection Timely access to essential goods and health care services Concerns with telehealth Retaining resilience and optimism Information seeking and solution focused Unintended positive effects of social restrictions Fig. 1 Connections between themes and sub-themes represented diagrammatically. Fig. 1 3.1 Support for a positive experience Women described that support during the COVID-19 pandemic was negatively affected by physical-distancing restrictions set by both the community and hospital. 3.1.1 Seeking inclusion of partner or support person Women expressed that they valued the role of a partner or support person in antenatal visits and advocated for their inclusion.“One of my coping mechanisms is having my partner there to hear the same things I am hearing because I kind of shut down sometimes when I get too upset and I don’t listen to everything. So, it’s always good to have that second person listening…and walking out with strength of unity.” BS Women thought that their partner’s experience during pregnancy had been adversely impacted, mainly as they were not able to co-attend hospital appointments.“I feel like he’s missing out on a lot of the experience as well… I can show him the photo, but when I get to see the actual movement and things like that, I know that would affect him. I know he does get upset. I feel like he probably has questions too, like obviously pregnancy’s even more unknown for him as well, and I feel like he’s missing out on being able to ask those questions.” JD Some women desired a support person beyond their partners during labour and birth, with a few intending on a home birth in order to access the support they desired for a positive birth environment.“A support person such as a doula…was incredibly important because she supports you through the birth and…holds the space for you. It was to the point that we did consider a home birth.” EW Women expressed concern about having limited support postnatally in hospital, particularly, if a difficult recovery was foreseen.“If I have a birth that requires a bit more intervention…that requires a hospital stay, then I do think that would affect us…that could have quite a big impact on visitors and support.” EW 3.1.2 Value for peer and intergenerational support Many women described that they valued peer support from other mothers during pregnancy and post-partum, together with intergenerational support from parents.”You can’t attend some yoga or birth classes so you can’t meet some other mums as well. You can’t ask advice or experience from them. Because of this, I have to do it by myself.” CC “The older generation have more experience on what babies need or what they feel… with my other two [children]…they knew exactly what may make them feel better whereas I have to obviously be on phone and ask, ‘What do you do?’…I don’t know, I will just learn as I go, I guess.” VB In their responses, women explored the potential impact of social isolation and access to support on their mental wellbeing.“I hope I won't be going through with depression because of less people around. Sometimes you just need family support or even a friend’s support, just to help you out a little bit. Because your first time…you don’t really have any idea about anything.” CC Some multigravida women desired access to additional support to help with childcare.“Last time when my daughter was born, I invited my in-laws and my family from overseas. It’s going to be too hard for me because my daughter she just started at school. My husband, he’s working. I have to look after my daughter and the new baby as well. It’s going to affect my life.” SS 3.1.3 Re-assurance from healthcare professionals The re-assurance provided by healthcare professionals was valued. Women gave appreciation to the care given by healthcare providers and acknowledged the initiative taken by healthcare organisations with infection control policies. This re-assurance helped address uncertainties for some women, particularly in relation to timely access to health care services and risk of acquiring infection.“I had an issue at one point and then I called up the maternity ward and they were very comforting in the sense that they were like, come on in…They did the full follow through and that was really good, you didn't feel like you can't go to your hospital.” FR “They're very careful with everything. Like asking where have I been…and checking the temperatures. Even if I go to the ultrasound, they always have a social distance…It's actually pretty good. They are taking extra care for their patients. They're just making everyone safe…you know, so this Coronavirus won't spread. Which is understandable actually.” CC As a whole, women advocated for increased support during pregnancy, birth and post-partum, both in the hospital and community settings. 3.2 Impact on preparedness in pregnancy and beyond 3.2.1 Changes to birthing and parenting education Women perceived that their preparedness for birth and motherhood was negatively impacted by the cancellation of face-to-face birth and parenting education. This created a sentiment of uncertainty in women as they reflected on birth and the post-partum period.“I wish that I knew, had a picture in my head, of what I was going to be walking into… I guess there’s a little bit of anxiety about getting lost and just, yeah, the idea of not knowing is a little disappointing” NM “The birthing classes were cancelled… so I have looked up the information online and I don't know if that’s everything that was there… it's hard to know when you're doing the research yourself as opposed to in a class.” KC Cancellation of educational initiatives was seen as a missed learning opportunity for couples. One woman described her concern in relation to parentcraft.“If I could have my boyfriend there so he could get taught how to do things like bath the baby…and swaddle the baby – I think that side of things will affect [us] after pregnancy.” CL 3.2.2 Concerns around early discharge Some women viewed the potential for early discharge as a risk to confidence in the early post-partum period, particularly, where additional support was limited in the context of community stay-at-home restrictions.“I was a little bit worried about being sent home early… because this is my first baby. I’m also going home on my own… I don't have a partner to help me or to help me look after me or the baby or anything… and even just breastfeeding - so that was a bit of a worry.” TB Some showed forethought and anticipated the implications of needing early unplanned medical assistance at home. A lack of re-assurance surrounding this concern created uncertainty and worry.“What if something happens with me not straight away but it could be a lot of medical challenges and obviously I don’t want newborn to be exposed to situation where it may get affected.” VB 3.3 Facing uncertainty of a pandemic Women expressed concerns related to various uncertainties surrounding the COVID-19 pandemic. 3.3.1 Disruption to the “normal” pregnancy experience Women described the COVID-19 pandemic as having an overall negative impact on the experience of pregnancy, with most describing that they missed out on a “normal” pregnancy experience. Not only did social restrictions limit the valued involvement of partner, family and peer supports, but women also expressed disappointment that they were not able to plan for and engage in traditional pregnancy rituals to share their pregnancy journey with family and friends.“I don't think many people have even seen me showing… you miss all that part of pregnancy a bit because we’re being locked down. Likewise, with work, we do video conferences, and you can’t really tell I’m pregnant. I guess you kind of miss a bit of the hubbub about being pregnant, the excitement.” KC “Am I going to have a baby shower? Who knows? The people that I’m catching up with now, it’s the first time I’ve seen them since they found out I was pregnant and now I have a bump.” FM There was apprehension about the possibility that their partner or support person could be absent at birth or post-partum.“I always have a slight fear when I go into the hospital…that my husband – he’ll get a temperature at delivery and won’t be able to join me.” NM 3.3.2 Perceived risk of acquiring infection Concern was expressed by women about the COVID-19 infection itself and the potential impact on their unborn baby or newborn, particularly, in the early stages of the pandemic when less was known about the novel virus.“I think at the start there was so little understanding of how it could affect pregnant women. And I was hearing awful things where it wouldn’t affect the child and then it could affect the child. A lack of understanding around risks was probably a little bit anxious for me.” FM Women were vigilant about the risk of acquiring infection in either the community, hospital or work settings.“You do worry. You are concerned that everyone's doing the right thing, can I go to the shops, am I putting the baby at risk?” FR “I’m giving birth in a hospital where sick people are. And it’s probably – if people have COVID – they would be in the hospital. Just being in that same environment with a newborn is definitely a bit daunting.” MN Some women had ceased their jobs or modified workplace duties to reduce the risk of acquiring the infection.“I actually stopped going to work… I’m a kindergarten teacher and it was just recommended that given the limited research on what would happen.” EW 3.3.3 Timely access to newborn goods and healthcare services Women anticipated the implications of restrictions on access to essential maternity and newborn goods.“You’ve got people stockpiling stuff… and you think am I actually going to be able to get the resources I need… I couldn’t sleep so I ended up just going and buying a pack of newborn nappies.” EW Clarity was sought on whether women could have timely access to healthcare during the pandemic.“I just want to know, if anything happened to me that I could come to the hospital straightaway and they can treat me straight away.” SS 3.3.4 Concerns with telehealth Women acknowledged the role of telehealth in minimising the risk of COVID-19 transmission but they did perceive telehealth as a compromise of their pregnancy experience, with some describing it as impersonal due to limited physical connection with clinicians and care feeling rushed.“Sometimes I’ve walked out and thought, ‘Oh, I meant to ask that’, but I didn’t really get time because I felt a bit of pressure to hurry up.” NM “You don't have that physical connection with someone or just being, knowing that they can physically see you and assess you. I had a miscarriage only a few months before I actually fell pregnant again… I just feel like that being delayed (the physical) it just made me more anxious.” JD Re-assurance was sought in relation to telehealth and its ability to maintain safe clinical decision-making.“I would say, if it was closer towards my due date, it would be very concerning for me… am I getting the amount of scans that I need to or the amount of check-ups that I need to in person? Have they seen enough of me or got enough information to be able to make a good call around the stage?” FM 3.4 Retaining resilience and optimism 3.4.1 Information seeking and solution focused Most women identified gaps in communication and were information seeking. In relation to birthing options, women sought clarification on any limitations to water birth and comfort options such as water immersion.“I’d like to have a water birth and I don’t know if that’s still possible. I don’t know if water births are allowed again? Or being in the water during labour?” EW Many women desired official communication from the hospital about the various uncertainties of the pandemic, such as the risk of acquiring the infection and what precautions were being taken to reduce this risk.“You do feel a little bit stressed in there, and I think probably one thing that maybe could be improved is just that extra information of what you are doing with the COVID stuff in terms of precautions, what it's going to look like when I come in to have bubs, just what to expect.” TB With the cancellation of face-to-face education and hospital tours, women sought guidance on alternate sources of information to prepare for birth. They also requested clarification on the impact on delivery of postnatal services in the community.“What happens when you come in? Because obviously there’s not been hospital tours… I wouldn't even know where to go. A video online on the website or something that you can go on and get a tour may be helpful, starting from outside so you know where you're going.” TB “Attending maternal child health appointments – I don’t know if they’re over the phone or face to face… As well moving forward I’m really keen for my mental health to be part of a mother’s group and I’m concerned that those mother’s groups might not be happening.” BS Women were solution focused and proposed greater utilisation of the internet medium for online video hospital tours and support groups.“Maybe setting up something like a support group, maybe an online support group… for mums that have babies in a pandemic that probably do feel isolated.” MN 3.4.2 Unintended positive effects of social restrictions Some women expressed that there had been unexpected positive benefits arising from the restrictions due to flexible workplace arrangements and the opportunity to physically rest at home.“It’s been really good, in that, I got to work from home instead of going into the office, so I’ve been able to, yeah, take a rest when I need a rest and pick my own hours.” NM 4 Discussion The study provides insight into the experiences and narratives of pregnant women during the COVID-19 pandemic. Women responded with concerns surrounding support, preparedness and uncertainty in the pandemic. Yet, women pro-actively sought information and solutions, and acknowledged the unintended benefits of social restrictions. The themes identified in the study indicate that the experiences of pregnant women were framed by not only COVID-19 as an infection but by their responses to physical-distancing and other restrictions. Women sought support to achieve a positive experience in pregnancy, birth and post-partum. Support people have been cited as a protective factor against fear in pregnancy during the COVID-19 pandemic [11,27]. For partners, they may have a sense of grief and loss if wanting to participate in the pregnancy and this has the potential to affect how fathers connect to their baby during the pregnancy [4]. Some women were left to choose between partners and doula birth support, who often serve a health advocate role complementary to family support [28]. Some considered home birth to be able to access an additional birth assistant, a trend seen in other countries during the pandemic [29]. Programs that prioritise doula support, such as in-house accredited doulas or “virtual” doulas may help alleviate fear in women [4,30]. Additional support was also desired by multigravida women who have increased caregiving responsibilities. Disruption to childcare is a risk factor for psychological distress; mothers with young children during the current COVID-19 pandemic have been shown to be at significantly higher risk for developing clinically-relevant anxiety [15,31]. Women desired the support from their peers and their own mothers, which has been shown to transfer confidence, parenting skills and provide re-assurance [32]. Online support groups and telehealth support programs can create group cohesion, and have been shown to reduce isolation and anxiety, feelings of loneliness and increase maternal positive emotions and feelings of calm [14,33]. Women described that the changes in service delivery had negatively affected their preparedness for pregnancy and beyond. A North American study performed during the COVID-19 pandemic coined the term “preparedness stress”, which affected nearly one in three pregnant women and women in this study who perceived alterations in their care were at higher risk of stress [12]. Peer and parental support, the focus of the first theme, was described in the study by women as important for the sense of preparedness. In addition, online resources for women on hospital-specific websites, including virtual tours and online classroom perinatal education, may help offset preparedness stress [14]. Women in our study were solution-focused and searched for such resources. Concern surrounding early discharges could be similarly alleviated through an increase in compensatory earlier virtual, either online or telehealth, follow-up with midwifery, lactation consultants and allied health clinicians after discharge [30]. Women experienced uncertainty in the unfolding pandemic. Women respond positively when care-providers acknowledge that pregnancy in a pandemic is not what was expected and that feelings of anxiety and sadness are normal [5]. Physical distancing has meant that women have had to forgo “normal pregnancy” rituals and baby showers, which can affect a woman’s mental wellbeing as they traditionally strengthen family and support networks, self-efficacy, and a family’s connection to heritage and culture [34]. The distress that these changes can cause has been noted in another Australian thematic analysis, which found that when women have attempted to celebrate their pregnancy during the pandemic, they may experience ‘guilt-tampered happiness’, or a paradox of guilt due to the contrast of happiness of pregnancy and the dire community situation [35]. Women also expressed uncertainty regarding telehealth in our study. Telehealth at the study site did not utilise video-conferencing, a tool that has been shown to improve the patient experience [5]. Traditionally, telehealth is known to be beneficial in reducing travel requirements, overcoming childcare barriers and for accessible psycho-social follow-up [5]. Prior to the pandemic, a successful low-risk antenatal model in North America combined telehealth together with virtual midwifery visits, online portal for queries and an online community forum for women moderated by midwives [36]. Whilst telehealth holds potential, it may not be suitable for women at risk of intimate partner violence and women with English as a second language [30]. The development of a successful telehealth service depends on an internal audit process, such as the Consolidated Framework for Implementation Research (CFIR) rapid cycle evaluation process, which focuses on culture and seeks feedback from key stakeholders [37]. Despite the challenges the pandemic has posed to pregnant women, women in our study identified knowledge gaps, were information seeking and solution focused. Failure to address these needs can amplify feelings of uncertainty, particularly the perceived risk of acquiring the infection and associated concern [8]. A cross-sectional study of pregnant women in Wuhan found that those with higher knowledge scores related to COVID were significantly less likely to have anxiety symptoms [38]. An ongoing North American study has found that 40% of pregnant mothers reported their provider had not spoken to them directly about COVID-19 [4]. Open communication channels with knowledge dissemination can help women identify false information, particularly, within social media. It is recommended that health organisations first determine the knowledge needs, literacy level and preferred communication medium of their women [5]. Studies have found geographical variation in preferred modes of communication during the pandemic, with women in China trusting information received directly from midwives whilst women in Italy accessed information online and through social media [8]. Our study highlighted areas that could be addressed include infection control and safety policies for women and their partners, analgesia options particularly nitrous gas and water immersion, restrictions for support people in labour and post-partum, and impact to postnatal support services such as home visits by the maternal child health nurse. When the major themes are considered together, it is evident that the mental wellbeing of pregnant women is at risk during the COVID-19 pandemic. Experiencing deficits in support, feeling unprepared and dealing with emotions associated with uncertainty present as challenges to the resilience of mothers. This is consistent with other studies exploring the responses of pregnant women to the pandemic which have similarly found a dominance in negative emotions and psychological constructs [7,35]. Previous epidemics have proven unfavourable psychological outcomes, and anxiety symptoms are an independent risk factor for abnormal obstetric and developmental outcomes, such as impaired bonding and pre-term birth [6]. Some women may even start to develop negative feelings about their own pregnancies [9]. A study of women giving birth in Italy during the pandemic found that almost a third of women described post-traumatic stress symptoms [39]. Following from this, some advocate for a “trauma-informed approach” to maternity care during the pandemic, with advocacy for support persons, preservation of shared-decision making and choice, and exploring women’s pregnancy experiences [5]. Healthcare organisations should endeavour to assess the mental health status of their pregnant women to help initiate appropriate care. New pandemic-relevant assessment tools are being developed and validated such as COVID-ASSESS which incorporates existing mental health tools [40]. Women identified as high-risk may be chosen for intensified telehealth follow-up post-partum and to connect them to virtual communities and support programs [32,35]. The findings of this study may provide health care providers with information on the social, cognitive, and emotional constructs associated with pregnant women’s experiences and their behavioural response to the COVID-19 pandemic. This understanding may assist maternity care providers in ensuring changes to the delivery of health care maintain a positive pregnancy experience for women. Acknowledgement of women’s perspectives in official health communications may help strengthen relationships between maternity units and their pregnant women. During the COVID-19 pandemic, women’s experiences in pregnancy were shaped by not only their personal factors but also by external influences such as community and hospital-level restrictions. Women continued to be self-reflecting, self-regulating and pro-active. The strength of this study is in its design capturing the experiences of pregnant women as described by themselves. However, the study does have limitations. Convenience sampling limits the extrapolation of the findings to wider populations. The community environment at the research location site may influence interviewee’s attitudes. At the time of the study, the region was witnessing a steady rise in community transmissions and a plan to increase physical distancing restrictions. This may have an influence on the participants’ communicating anxiety, uncertainty and to some extent a feeling of negativism. A dominance of negative constructs has been similarly found in other studies assessing the impact of the pandemic on pregnant women [10,12,40]. Commentary regarding the partner’s perspectives were not from partner’s themselves which limits its generalisability. The women were from a Western setting and a predominantly Caucasian background; the findings may vary in other settings. Psycho-social effects of a pandemic have the potential to disproportionally harm women with social vulnerabilities and women from minority and under-resourced communities [12]. Exploring this was considered beyond the scope of our study. 5 Conclusion This study adds to our deeper understanding of the perspectives of pregnant women during the current COVID-19 pandemic. As we grapple with the COVID-19 infection itself, the learnings from the study can be applied by organisations to deliver woman-centred care. The findings of this study highlight women’s experiences are centred around concerns regarding support, preparedness, and uncertainty, balanced with a positive introspection and desire for solutions. Directly addressing the concerns expressed by women may not only have a positive influence on their pregnancy experience but also enhance long-term psycho-social wellbeing. Conflicts of interest None declared. Ethical statement This study received ethical approval and support from the Peninsula Health Human Research Ethics Committee on May 27, 2020 (HREC ref. number LNR/64473PH-2020). Funding None declared. Author agreement In line with the Author Agreement, the authors confirm that:• The article is our original work. • Has not received prior publication and is not under consideration for publication elsewhere. • All authors have seen and approved the manuscript being submitted. • The authors abide by the copyright terms and conditions of Elsevier and the Australian College of Midwives. CRediT authorship contribution statement Kiran Atmuri: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing. Mahbub Sarkar: Methodology, Formal analysis, Investigation, Data curation, Writing - review & editing. Efe Obudu: Conceptualization, Project administration. Arunaz Kumar: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Supervision, Project administration. Acknowledgements We are grateful to the midwifery staff for recruitment and to women for sharing their experiences. ==== Refs References 1 World Health Organization Coronavirus Disease (COVID-2019) Situational Report 2020 (Accessed 13 February 2020) https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/ 2 Homer C.S.E. Leisher S.H. Aggarwal N. Counting stillbirths and COVID 19-there has never been a more urgent time Lancet Glob. Health 9 1 2020 e10 e11 33212029 3 Royal Australian and New Zealand College of Obstetricians & Gynaecologists (RANZCOG) COVID-19: Clinical Resources for Health Care Workers 2020 (Accessed 18 September 2020) https://ranzcog.edu.au/statements-guidelines/covid-19-statement/clinical-resources-for-doctors 4 Diamond R.M. Brown K.S. Miranda J. Impact of COVID-19 on the perinatal period through a biopsychosocial systemic framework Contemp. Fam. Ther. 2020 1 12 5 Choi K.R. Records K. Low L.K. Promotion of maternal-infant mental health and trauma-informed care during the COVID-19 pandemic J. Obstet. Gynecol. Neonatal Nurs. 49 5 2020 409 415 6 Mappa I. Distefano F.A. Rizzo G. Effects of coronavirus 19 pandemic on maternal anxiety during pregnancy: a prospectic observational study J. Perinat. Med. 48 6 2020 545 550 32598320 7 Ravaldi C. Wilson A. Ricca V. Homer C. Vannacci A. Pregnant women voice their concerns and birth expectations during the COVID-19 pandemic in Italy Women Birth 2020 S1871-5192(20)30280-8 8 Lee T.Y. Zhong Y. Zhou J. He X. Kong R. Ji J. The outbreak of coronavirus disease in China: risk perceptions, knowledge, and information sources among prenatal and postnatal women Women Birth 2020 S1871-5192(20)30249-3 9 Shefaly Shorey V.C. Lessons from past epidemics and pandemics and a way forward for pregnant women, midwives and nurses during COVID-19 and beyond: a meta-synthesis Midwifery 90 2020 10 Saccone G. Florio A. Aiello F. Psychological impact of coronavirus disease 2019 in pregnant women Am. J. Obstet. Gynecol. 223 2 2020 293 295 32387321 11 Mei H. Li N. Li J. Impact of the COVID-19 pandemic on mental health in pregnancy women: results from two cohort studies in China BMC Public Health 2020 10.21203/rs.3.rs-42153/v1 12 Preis H. Mahaffey B. Heiselman C. Lobel M. Vulnerability and resilience to pandemic-related stress among U.S. women pregnant at the start of the COVID-19 pandemic Soc. Sci. Med. 266 2020 13 Aghababaei S. Bashirian S. Soltanian A. Perceived risk and protective behaviors regarding COVID-19 among Iranian pregnant women Middle East Fertil. Soc. J. 25 1 2020 29 32963467 14 Jago C.A. Singh S.S. Moretti F. Coronavirus disease 2019 (COVID-19) and pregnancy: combating isolation to improve outcomes Obstet. Gynecol. 136 1 2020 33 36 32384386 15 Moyer C.A. Compton S.D. Kaselitz E. Muzik M. Pregnancy-related anxiety during COVID-19: a nationwide survey of 2,740 pregnant women Arch. Womens Ment. Health 23 6 2020 757 765 32989598 16 Bandura A. Social Foundations of Thought and Action: a Social Cognitive Theory 1986 Prentice-Hall Englewood Cliffs, NJ 17 Whitaker K.M. Hung P. Alberg A.J. Hair N.L. Liu J. Variations in health behaviors among pregnant women during the COVID-19 pandemic Midwifery 95 2021 102929 33508485 18 Khoury J.E. Atkinson L. Bennett T. Jack S.M. Gonzalez A. COVID-19 and mental health during pregnancy: the importance of cognitive appraisal and social support J. Affect. Disord. 282 2021 1161 1169 33601691 19 Teti M. Schatz E. Liebenberg L. Methods in the time of COVID-19: the vital role of qualitative inquiries Int. J. Qual. Methods 19 2020 1 5 20 Australian Department of Health Woman-Centred Care: Strategic Directions for Australian Maternity Services 2019 21 Burr V. Social Constructionism 3rd ed. 2015 Routledge London 22 Doyle L. McCabe C. Keogh B. Brady A. McCann M. An overview of the qualitative descriptive design within nursing research J. Res. Nurs. 25 5 2020 443 455 34394658 23 Colorafi K.J. Evans B. Qualitative descriptive methods in health science research HERD 9 4 2016 16 25 24 Malterud K.S.V. Guassora A.D. Sample size in qualitative interview studies: guided by information power Qual. Health Res. 26 2016 1753 1760 26613970 25 Braun V. Clarke V. Using thematic analysis in psychology Qual. Res. Psychol. 3 2 2006 77 101 26 Berger R. Now i see it, now i don’t: researcher’s position and reflexivity in qualitative research Qual. Res. 15 2 2015 219 234 27 Hartman S. The importance of antenatal partner support J. Womens Health (Larchmt) 25 7 2016 659 661 26981842 28 Gruber K.J. Cupito S.H. Dobson C.F. Impact of doulas on healthy birth outcomes J. Perinat. Educ. 22 1 2013 49 58 24381478 29 Coxon K. Turienzo C.F. Kweekel L. The impact of the coronavirus (COVID-19) pandemic on maternity care in Europe Midwifery 88 2020 102779 32600862 30 Hermann A. Fitelson E.M. Bergink V. Meeting maternal mental health needs during the COVID-19 pandemic JAMA Psychiatry 78 2 2020 123 124 31 Cameron E.E. Joyce K.M. Delaquis C.P. Reynolds K. Protudjer J.L.P. Roos L.E. Maternal psychological distress & mental health service use during the COVID-19 pandemic J. Affect. Disord. 276 2020 765 774 32736186 32 Morison S. Hauck Y. Percival P. McMurray A. Constructing a home birth environment through assuming control Midwifery 14 4 1998 233 241 10076318 33 Shorey S. Chee C.Y.I. Ng E.D. Lau Y. Dennis C.L. Chan Y.H. Evaluation of a technology-based peer-support intervention program for preventing postnatal depression (Part 1): randomized controlled trial J. Med. Internet Res. 21 8 2019 e12410 34 Kumari A. Ranjan P. Sharma K.A. Impact of COVID-19 on psychosocial functioning of peripartum women: a qualitative study comprising focus group discussions and in-depth interviews Int. J. Gynaecol. Obstet. 152 3 2020 321 327 33305351 35 Chivers B.R. Garad R.M. Boyle J.A. Skouteris H. Teede H.J. Harrison C.L. Perinatal distress during COVID-19: thematic analysis of an online parenting forum J. Med. Internet Res. 22 9 2020 e22002 36 de Mooij M.J.M. Hodny R.L. O’Neil D.A. OB nest: reimagining low-risk prenatal care Mayo Clin. Proc. 93 4 2018 458 466 29545005 37 Fryer K. Delgado A. Foti T. Reid C.N. Marshall J. Implementation of obstetric telehealth during COVID-19 and beyond Matern. Child Health J. 24 9 2020 1104 1110 32564248 38 Ding W. Lu J. Zhou Y. Wei W. Zhou Z. Chen M. Knowledge, attitudes, practices, and influencing factors of anxiety among pregnant women in Wuhan during the outbreak of COVID-19: a cross-sectional study BMC Pregnancy Childbirth 21 1 2021 80 33494723 39 Luca Ostacoli S.C. Bevilacqua Federica Berchialla Paola Bovetti Marialuisa Carosso Andrea Roberto Malandrone Francesca Carletto Sara Benedetto Chiara Psychosocial factors associated with postpartum psychological distress during the Covid-19 pandemic: a cross-sectional study BMC Pregnancy Childbirth 20 1 2020 703 33208115 40 Ravaldi C. Ricca V. Wilson A. Homer C. Vannacci A. Previous psychopathology predicted severe COVID-19 concern, anxiety and PTSD symptoms in pregnant women during lockdown in Italy Arch. Womens Ment. Health November (20) 2020 1 4
PMC009xxxxxx/PMC9006071.txt	==== Front Environ Sci Pollut Res Int Environ Sci Pollut Res Int Environmental Science and Pollution Research International 0944-1344 1614-7499 Springer Berlin Heidelberg Berlin/Heidelberg 20010 10.1007/s11356-022-20010-w Review Article An outlook on the development of renewable energy, policy measures to reshape the current energy mix, and how to achieve sustainable economic growth in the post COVID-19 era http://orcid.org/0000-0001-5103-4639 Farhan Bashir Muhammad farhan.paks89@gmail.com farhan.paks@csu.edu.cn 1 Sadiq Muhammad msadiq_110@yahoo.com msadiq@csu.edu.cn 1 Talbi Besma besmatalbi@yahoo.fr 2 Shahzad Luqman luqmanshahzad1975@gmail.com 3 Adnan Bashir Muhammad adnanbashir2034@gmail.com 4 1 grid.216417.7 0000 0001 0379 7164 Business School, Central South University, (410083), Changsha, Hunan People’s Republic of China 2 grid.419508.1 0000 0001 2295 3249 Polytechnic School of Tunisia, University of Carthage, Tunis, Tunisia 3 grid.12981.33 0000 0001 2360 039X Department of Business Administration (SYSBS), Sun Yat-Sen University, Guangzhou, Guangdong China 4 grid.263488.3 0000 0001 0472 9649 College of Economics, Shenzhen University, Shenzhen, Guangdong People’s Republic of China Responsible Editor: Roula Inglesi-Lotz 13 4 2022 112 24 1 2022 28 3 2022 © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Currently, COVID-19 due to emergence of various variants shows no signs of slowing down and has affected every aspect of life with significant negative impact on economic and energy structures around the world. As a result, the governments around the world have introduced policy responses to help businesses and industrial units to overcome the consequences of compliance with COVID-19 strategies. Our analysis indicates that global energy sector is one of the most severely affected industries as energy price mechanisms, energy demand, and energy supply have shown great uncertainty under these unprecedented economic and social changes. In this regard, we provide brief overview of demand, supply, and pricing structure of energy products as well as policy mechanisms to provide better outlook about how industrial sector can cope with energy consumption in the post pandemic era. We further propose changes in the existing policy mechanisms so that transition towards renewable energy sources under different environmental agreements can be achieved. Moreover, as a reference, we outline major challenges and policy recommendations to ease energy transition from fossil fuels to environmental friendly energy mix. Keywords COVID-19 Energy demand Energy supply Clean energy Sustainable economic growth Renewable energy ==== Body pmcIntroduction In recent years, the world has experienced a number of health pandemics. The World Health Organization (WHO) defines a pandemic as the spread of a new disease which adversely affects a significant portion of the world population (Arslan et al. 2021; Badr et al. 2020). Some pandemics which were lethal in recent centuries are cholera (1817–1823) which originated from the Indian sub-continent and resulted in the deaths of millions of people. In recent decades, AIDS and HIV first appeared in 1976 and still affect a significant portion of the population in developing economies with 36 million confirmed deaths. SARS, another global pandemic, is a severe acute respiratory syndrome and is caused by one of 7 coronaviruses that can infect humans. This pandemic originated from China in 2003 and infected more than 8000 people around the world. Ebola is another example, which was first detected in 1976 but continues to ravage the African continent and has killed approximately 11,325 people (Sun et al. 2020). However, the most recent pandemic known as the COVID-19 has given grave consequences to the whole world. This coronavirus is a single-stranded positive-sense enveloped RNA virus whose genome consists primarily of nearly thirty thousand nucleotides with a genome size from 26 to 36 kb. Its transcription and reproduction be contingent on a combination of two replicate genes encoding pp1a and pp1ab polyproteins (Bilal et al. 2020a). The coronavirus whole-genome level corresponds to a 96% bat coronavirus (Zhou et al. 2020) and is referred to as coronavirus as its club-shaped spikes forming under an electron microscope at the top of a crown-like structure (Fofana et al. 2020). The COVID-19 was declared as a global pandemic on March 11, 2020, by the WHO (Bashir et al. 2020a), and since then, there have been many cases of high fatality rates worldwide and adverse global economic impact (Bashir et al. 2020b). As of March 21, 2022, there were 469,713,452 confirmed cases, including 6,074,560 fatalities (WHO 2021). In addition to massive loss of life, the global economy is estimated to shrink by 6%, and at least 300 million people became unemployed (Qaid et al. 2022; Bashir et al. 2020c). The COVID-19 is a “communicable” disease, and its dissemination worldwide is entirely associated with community transmission rather than airborne transmission. Although, in recent months, there have been operational vaccines to stop COVID-19 transmission (Aktar et al. 2020; Bashir et al. 2020d), till this day, lockdowns, i.e., prohibiting overseas travel, social distancing, mass quarantine, and region-wise lockdowns, are the widely used preventive measures to reduce the transmission of the COVID19 (Bilal et al. 2020b). The implementation of these preventive measures has sent seismic waves across global energy markets and resulted in an unprecedented 20% reduction in global energy expenditures. This also allows us to confirm that energy markets have been most significantly impacted by the lockdown measures as there is a significant drop in demand for oil and energy products (Bilal et al. 2021a; Gautam 2000b). Unsurprisingly, the relatively low level of prices propelled both supply shock and policy war between major oil and fossil energy producers, i.e., Russia and Saudi Arabia, as the oil market was hugely affected by demand shocks by the COVID-19 pandemic. Yet, the dramatic effect of COVID-19 on energy is not homogenous, and the consequences are very different for both sustainable and fossil fuel energy use. OPEC Secretary-General Mohammad Barkindo stated, “Covid-19 is an invisible beast that seems to affect everything in its course,” and it has up-ended market supply and demand dynamics for the energy market. Currently, the energy sector needs to address two major challenges: first, arrest the health emergency problems from COVID-19 faced by all industries; and second, the low demand and low oil price scenario, because financial liabilities and profits need to be resolved in the near future. Not only that, from the perspective of the energy sector, many other things have occurred during the spread of COVID-19, which affected the usage of the energy itself (Bashir et al. 2020e; Akrofi & Antwi 2020). For example, the current state of geopolitics has evolved and made a significant impact to spark the oil crisis, as OPEC + countries tended to manufacture and overwhelm extra supplies on the world market, possibly forcing other oil-producing regions to leave the market. Additionally, the full spectrum of the COVID-19 for the energy sector is evolving and is not easy to foresee. The current situation requires changes in policies to help minimize the detrimental effects on the energy sector. Compared with energy use and the COVID-19, the association between the COVID-19 pandemic and the environment has been extensively discussed (Bilal et al. 2021b; Akrofi & Antwi 2020). A significant number of studies have explored the overall environmental effects of COVID-19 (Aktar et al. 2020; Wang & Su 2020), while other studies have explored the association between specific environmental features and the COVID-19 outbreak, i.e., weather, temperature, government policy, pollution, and air quality (Muhammad et al. 2020; Baloch et al. 2020; Bannister-Tyrrell et al. 2020; Wu et al. 2020; Tosepu et al. 2020; Bilal et al. 2021c). Nevertheless, there is an intrinsic and definite association between environment and energy usage, which has resulted in unprecedented shocks in the energy sector and, therefore, would have environmental consequences, especially towards climate change (Akrofi & Antwi 2020). While the political consensus about climate change remains unresolved, recent decades have highlighted that environmental policies need to be reformed to protect the planet, and the continuous dependence on fossil energy remains as the focal concern (Talbi et al. 2020; D’Adamo et al. 2020). Keeping this in mind, more significant research is required to document and track the changes and impact on the energy sector with a focus on the occurrence of key moments when the global phenomenon in the form of the COVID-19 has induced drastic shifts in the pre-existing patterns so that expected outcomes become obsolete or suddenly change (Fareed et al. 2021; Drezner 2020). Keeping this in mind, our aim is to review various aspects of the COVID-19, how it impacts the energy sector, and provide an overview of the relevant concerns to provide policy suggestions to benefit policymakers, researchers, and the environmental quality. The current study addresses the following key research questions: (1) What are the ramifications of the COVID-19 pandemic for global energy demand and supply? (2) Analyze the shift in global energy prices from the COVID-19 outbreak. (3) How global energy politics has been impacting by the novel coronavirus? (4) How have policymakers responded to sustain the energy transition? Current research will provide significant implications for environmental stakeholders, policymakers, and scientists to design and implement a roadmap to curtail the adverse effects of the COVID-19 pandemic by addressing key issues being addressed by the energy sector. This issue is investigated by analyzing the most interrelated and critical elements of the ecosystem. Figure 1 represents the overall picture of the COVID-19 pandemic and what it means for the energy sector, a theme that will be further illustrated in detail in the subsequent sections.Fig. 1 COVID-19 and the energy industries COVID-19 and world energy demand In recent decades, global economy has been fundamentally transformed due to globalization and emphasis on the rise of technology, higher capital expenditures, and free-market business models (Bashir et al. 2015; Bashir 2018). However, it has also led to environmental degradation from higher volumes of energy consumption (Xia et al. 2022; Bashir et al. 2021e; Shahbaz et al. 2016). Hence, global energy demand, the essential element in eco-social development, shows no sign of slowing down especially in developing economies. Currently, global energy demand has risen to 13.9 Gtoe (Gigatons of oil equivalent) per annum in 2018, with 2.9% growth every year since 2010 (BP 2019). The largest increase in energy consumption has been recorded in Asian economies, especially India and China (Hussain et al. 2021). Figure 2 records the volumes for the ten highest energy consumption countries during the period of 2010–2019. Existing statistics indicate that China consumes 25% of world energy, which almost accounts for half of world energy consumption when the USA and India have been included (Wei et al. 2020).Fig. 2 Energy consumption breakdown by country (Mtoe) for the period of 2000–2019. Source: Global Energy Statistical Yearbook, 2020 Energy demand is another indicator to highlight the movement of global resources. Currently, the demand for energy sources is highest in Asian developing economies, with an expected increase of 3.7% per year. It is further estimated that energy consumption in Asia will double in the next 20 years and will be responsible for 65% of total energy demand in developing economies (Bashir et al. 2021a; Vaka et al. 2020). However, it is estimated that demand for energy resources in other developing economies will be slower than in Asian economies, but still, it is projected to surpass the world average (Table 1). As industrial units have a significant role in energy demand and supply, the immense increase in the expected level of energy consumption in Asia is also critical to region’s future production of energy resources (Bashir et al. 2021b). Fossil fuel consumption, i.e., natural gas, coal, and oil, will also lead to a significant increase in CO2 emissions worldwide (Vaka et al. 2020). Furthermore, the increase in CO2 emissions is also expected to contribute to unavoidable environmental damage from climate change and environmental degradation.Table 1 World total marketed energy consumption by region and fuel, 1990–2030 (quadrillion Btu) Region 1990 2004 2010 2020 2030 OECD North America 100.8 120.9 130.3 145.1 161.6 OECD Europe 69.9 81.1 84.1 86.1 89.2 OECD Asia 26.6 37.8 39.9 43.9 47.2 Non-OECD Europe and Eurasia 67.2 49.7 54.7 64.4 71.5 Non-OECD Asia 47.5 99.9 131.0 178.8 227.6 Near East 11.3 21.1 26.3 32.6 38.2 Africa 9.5 13.7 16.9 21.2 24.9 Central and South America 14.5 22.5 27.7 34.8 41.4 Total OECD 197.4 239.8 254.4 275.1 298.0 Total Non-OECD 150.0 206.9 256.6 331.9 403.5 Total Sources Oil 136.2 168.2 183.9 210.6 238.9 Natural gas 75.2 103.4 120.6 147.0 170.4 Coal 89.4 114.5 136.4 167.2 199.1 Other 26.2 33.2 40.4 46.5 53.5 Source: International Energy Agency The introduction of renewable energy through environmental policies has been preferred by policymakers to limit the damage from economic activities (IEA 2019). The abundance of resources like biomass, sun, and wind allows massive production of renewable energy to have little or no further GHG emissions (Bashir et al. 2021c). Although RE has significant potential to reshape the current energy mix, but the adoption of renewable energy has been relatively slow in developing economies (Malahayati 2020). Currently, the share of renewable energy sources in the global energy supply is 27%, with the rest of 73% coming from fossil fuels (Global Energy Statistical Yearbook 2020). The share of RE in 2019 within the energy mix also significantly varies across different economic regions, 3.8% in the Middle East, 17.6% in commonwealth countries, 20.2% in Africa, 23.7% in Asia, 23.9% in North America, 27.4% in Pacific, 38.7% in Europe, and 58.1% in Latin America (Fig. 3).Fig. 3 % Share of Renewables in Electricity Production, 1990–2019. Source: Global Energy Statistical Yearbook, 2020 The recent surge of COVID-19 has disrupted the energy structure, and as a result, there has been a growing focus on energy self-dependence through renewable energy (Anderson et al. 2020). Additionally, lockdowns and travel restrictions have also created a significant impact on demand for national energy profile such as residential electricity consumption has increased significantly by 40% as the majority of the population comply with movement orders but the lower electricity consumption in the commercial and industrial sector has far more significant socioeconomic impact (Broom 2020). Electricity demand fell by 2.5% globally in 2020. The USA is a key illustration as electricity demand fell by 5.7% during the lockdown measures (Elavarasan et al. 2020). Although there was a significant increase in demand for electricity by the residential sector (World Economic Forum 2020), economic sector reported a negative impact with 15%, 8.6%, and 6.6% lower electricity demand in the transportation, industrial, and commercial sectors. Another example, India, with the second-most reported cases from the COVID-19 pandemic (Worldometers 2020) experienced 20–40% reduction in different states (Aruga et al. 2020). The implementation of lockdown measures in China also reported similar findings where electricity demand by the industrial sector was 8% lower in comparison to the previous year. The lower demand for energy sources in recent months has led to the contraction of centralized fuel-based power generation (Sadiq et al. 2022; Bashir et al. 2021d), which has emphasized on increasing the share of renewable energy consumption across the globe (see Fig. 4). Countries with stringent environmental policies have introduced and implemented practices to increase the share of clean energy (Sultan et al. 2021). Although natural gas remains the leading source of power generation, renewable energy has surpassed coal-based power plants to generate electricity in recent months. The introduction of lockdown measures in India also resulted in similar outcomes where the share between renewable energy and coal has narrowed as the proportion of coal remains below 70% in the electricity generation, which is in line with the proportion of low-carbon-based electricity. Electricity demand trends recovered in late May 2020, while their seasonal supply was reflected by the increasing share of renewables in the energy mix. The electricity demand recovered after May 2020, where the share of renewable energy sources in the energy mix was higher and remained above 20%, with wind and hydro as the main sources of clean energy. Although coal consumption has increased in recent months in China, the share of clean energy has significantly improved in recent months with significant growth in hydroelectricity (Bashir et al. 2021e). Overall, the renewable energy sector has performed reasonably well and currently accounts for 30% of global energy production (Badr et al. 2020).Fig. 4 Electricity generation mix (Source: IEA 2019) COVID-19 and world energy supply COVID-19 has not only triggered a financial and public health crisis, it has also had a detrimental effect on the global energy supply. The consequences have been observable in key energy suppliers, i.e., the USA, where “force majeure” notifications have been issued to vendors about delays in the energy production (Bashir et al. 2021f). Similarly, state-owned energy suppliers have announced their inability to fulfill natural gas contracts as they cannot import the required fuel. Similarly, Spain and Italy have suspended production lines in recent months. Similar forecasts have been predicted that the renewable energy sector has possible negative implications that are worrying. Another key factor is the development of green technology, which had shown encouraging signs before the COVID-19 crisis. The Internal Energy Agency’s annual MCED (monitoring clean energy development) reports at least six, i.e., mass transit and electric cars, of possible forty-six indicators were on track to achieve sustainability targets under UN SDGs (IEA 2019). Another twenty-four indicators showed significant improvement, while the remaining sixteen indicators require further reforms. Although the renewable energy sector has not shown a significant decline since the COVID-19 outbreak, IEA estimates that the renewable energy sector will contract by around 13%, with the breakdown of supply chains and limitations on construction activities having a significant impact (Cherp & Jewell 2020). IEA also reports that overall energy expenditures declined by 20% in 2020 alone (Bashir et al. 2022a). Additionally, construction work to set up renewable energy technology has slowed down significantly. With the slackening production of solar panels in Australia, China, and India up to 17–48%, solar installations declined as expected in 2020 (Singh 2020). The disruptions of solar photovoltaic power have interrupted manufacturing in Italy, Spain, and other leading producers as well (Bashir et al. 2022a), with consumption of material in solar arrays and solar panels negatively impacting Singapore, South Korea, Thailand, Malaysia, and Vietnam (Vaka et al. 2020) with the production of solar panels which is expected to decrease considerably in the USA, who imports 90% of raw material required. South Africa is also expected to undergo a similar experience where mega-solar projects have been shelved due to disruptions in the supply of the photovoltaic (PV) systems from China (Bashir et al. 2021g; Carrington et al. 2020) which generates around 70% of raw material with Chinese firms working in Southeast Asia which account for another 10–15% of the global share. The implementation of lockdown measures in different parts of China has led to a 16% lower demand for solar panels as China suspended or restricted the production capacity of the solar industry (Vaka et al. 2020). As the WHO declared the COVID-19 as a global pandemic, it impacted not only the solar industry, but also had an adverse impact over associated renewable sectors, i.e., smart grids, battery technologies, and renewable power. This has led to crisis projections in the wind industry (Bashir 2022). LM Wind power and Siemens Gamesa in Spain announced that the manufacturing units have halted the production of blade wind turbine plants. Likewise, the COVID-19 outbreak has disrupted the construction of 100 wind turbines in Scotland. A report published by the International Renewable Energy Agency (IRENA) indicated that the slower than expected growth in renewable and environmental technology had slowed the progress towards global sustainable goals as material shortages, production shutdown, and higher costs in the renewable energy sector being key attributes (ICIS 2020). Hence, circumstances like these have led to potential conflict from policymakers as more and more countries are exploring to carry on RETs or, despite the ecological indicators, resort back to fossil fuel consumption (Bashir et al. 2022a). COVID-19 and world energy price Due to sheer impact, COVID-19 is classified as the biggest threat since World War II (Gautam, 2020a). The dramatic impact on energy consumption is evident as supply and demand will have a significant impact on the energy mix in the industrial sector (Shahbaz et al. 2021). The COVID-19 pandemic in a number of ways has changed the energy price trends as profiles, and historical usage patterns have been dramatically altered. In recent years, the oil and gas industry has deteriorated due to unstable global oil prices; since then, the global fossil prices have evolved at a slightly lower level and hence faced a vicious dual shock in the first few months of the pandemic (Thiéry 2020). The declaration of COVID-19 as a global pandemic coincided with a huge reduction in global demand for fossil fuels; consumption of crude oil was reduced by at least 30% compared to its peak and is expected to be 8% lower than 2020 (Ma et al. 2021a). On the other hand, the disagreement about oil prices between Saudi Arabia and Russia has also led to fluctuations in domestic West Texas Intermediate (WTI) crude oil and international Brent crude oil (Fig. 5). This is due to the combination of these two factors (see Fig. 5).Fig. 5 Changes in energy price (%). Source: Thomson Reuters Datastream, 2020 In recent months, the global crude oil prices for WTI fell by a record 22.30 USD/BBL (WTI), which accounts for 67% reduction; and global crude oil for Brent fell by 22.36 USD/BBL, which is a 65% decline compared to the reference year 2019. The most significant shock came from a drastic drop in transportation, residential, industrial, and commercial usage (Ma et al. 2020a, b). This has contributed to additional pressure on balancing demand and supply to regulate oil prices more efficiently. Also, demand for crude oil dropped by 9.3 mb/d in 2020 and there is a projection that it would further reduce by 29 million barrels/day (mb/d) as the OPEC + agreement came into effect and production fell elsewhere. The total stockpile of oil in OECD countries reached 42.4 Mb in March 2020, where the storage of crude oil rose to 103.1 Mb. Also, the drop in gas prices further intensified in 2020 as compared to 2019 as the base year, which was powered by lower demand and still relatively high production. The gas prices are particularly poor in Europe due to excessive supply in the pre-pandemic era. The natural gas prices dropped by 16% in 2018–2019 alone, and this pattern continued in the post-COVID-19 era, with gas prices reaching 8.25 EUR/MWhth in 2020 compared to 18.02 EUR/MWhth in 2018 and 15.20 EUR/MWhth in 2019. This indicated that the decline in gas prices was exacerbated by the COVID-19 crisis in comparison to the crisis-free previous year (Farooq et al. 2022a, b). The trend of declining has also been observed in the electricity and coal sectors as well, although the reduction in coal prices is rather difficult as its prices were already lower than other energy sources. There has been a strong drop in coal prices since 2019, and this trend has persisted since 2019 as the coal prices were reduced to USD 47.40/MT in 2020 from 2019 (70.70 USD/MT). Our analysis also allows us to infer that a lower impact on the coal sector can be attributed to several factors which kept the coal prices under stress (i.e., ETS certificate prices, lower gas prices). Energy prices in European countries continue to decline due to lower electricity demand with 47% less demand than the base year of 2019. The International Energy Agency (IEA) indicates that not only persistent lower oil prices affect fossil fuels but also have a significant impact on the renewable energy sector, with economic shutdowns slowing the transition towards adoption of renewable energy (Bashir et al. 2022b). COVID-19 and energy geo-politics of the world Generally, politics is a reference towards the formation of bureaucratic structure and formal and informal working of political institutions. Energy politics is a multi-scalar concept that requires cooperation between regional, international, and sub-national players (Van de Graaf & Sovacool 2020). The emergence of globalization has meant that international politics and energy are profoundly intertwined: international politics is significantly influenced by the essence of energy, i.e., the extent to which it becomes sustainable, but energy also exerts a significant impact on politics (Kuzemko 2019). In the global political and economic system, the degree of energy politics is interconnected with the world energy market (Fig. 6), where stability in the oil exports or changes in the production capacity has the ability to affect the global market. This observation means that the economic threats from the COVID-19 will affect global economies as a whole, and it is essential to formulate cohesive policy instruments to overcome these challenges as they will also influence global energy politics (Ma et al. 2021b). Historical shocks in price and demand for energy products have established far-reaching indications of changes in the global energy economy, and the current crisis is no exception from this.Fig. 6 Interconnected world oil market On March 9, 2020, the global oil prices, as COVID-19 worsened, witnessed the most significant decline in three decades, which occurred due to OPEC being unable to arrive at an agreement with Russia. OPEC proposed to decrease supply in the global market to sustain price levels by limiting global supply by 1.5 million barrels per day due to lack of demand and suggested that non-OPEC members and Russia also implement similar measures. According to Arezki and Fan (2020), OPEC countries supply 1 Mbd while non-OPEC countries supply for 0.5 Mbd. However, Russia showed reluctance to do so as oil and gas exports account for 40% of national income. However, it showed a reluctant willingness to decrease oil production by 10% from May 2020 (Jefferson 2020), but an immediate reversal from the proposal resulted in oil prices being reduced by a further 10%. Figure 7 below shows world oil consumption, prices, and OPEC oil production and supply adjustment. It is evident from Fig. 7a (changes in oil consumption) that between 2009 and 2017, the demand for crude oil consistently grew but has been declining since then and is expected to decrease further, mainly due to the COVID-19 pandemic and lack of industrial activities. Figure 7 b (crude oil prices in 2020) further illustrates that since January 2020, west Texas intermediate and Brent crude oil prices are under pressure. Figre 7c (OPEC oil production and supply adjustments) illustrates that in recent years, the focus of OPEC is to balance the oil market by making adjustments in the production capacities. Lastly, Fig. 7d (dwindling share of world oil market) provide details about crude oil market share by Russia, Saudi Arabia and United States, who produce most crude oil globally.Fig. 7 World oil consumption, crude oil prices, OPEC production and supply adjustments, and share of the world oil market In order to maintain political dominance over the global energy supply, Saudi Arabia increased its daily oil production from 9.7 million barrels a day to more than 10 million barrels a day during February and March 2020 (Jefferson 2020) and further increased the daily production to more than 12 million barrels a day. This caused political controversy as to why the previous decision was to retain production levels by highlighting uncertainty in oil prices from COVID-19. Other reports also asserted that such changes were implemented to harm the other oil exporters such as Russia by low prices as Russia, far more dependent on oil exports for national income, between 2016 and 2019 added sixty oil production fields to generate nearly 1 million barrels per day (Jefferson 2020). Such political moves indicate that a few issues need to be addressed to implement substantial cuts in oil production (Farooq et al. 2022a, b; Johnson & Standish 2020). First, what is the actual size of global oil demand is under consideration, with estimates ranging from 6 million barrels/day to 10 million barrels/day (Johnson & Standish 2020). Second, whether against Saudi Arabia or Russia, another major oil producer, the USA needs to instruct its oil and gas sector to prioritize which areas require investment choices, especially in current times when oil prices have been hit by the global pandemic. This has led to non-compliance with earlier agreements to reduce global energy supply so that oil prices do not fall any further. Barring political disputes to curb energy supply and overcome price issues, major energy-exporting countries have been implanting significant reforms to overcome economic losses. Meanwhile, energy importers are willing to further import to take advantage of the current low energy prices (Egan 2020), although unstable energy prices are expected to influence infrastructure projects especially in Middle East counties that rely heavily on energy exports to finance public infrastructure initiatives. Keeping this in mind, the political response to overcome the adverse impacts of COVID-19 on the energy process needs to be more innovative, strategic, and sustainable in the near future (Hepburn et al. 2020). Discussion and post-COVID-19 policy recommendations The COVID-19 pandemic has adversely impacted many markets, with the energy sector being most significantly impacted as global energy demand plunged in recent months and led to a significant drop in economic and industrial activities. According to Bloomberg statistics, the low oil prices resulted in coal becoming relatively expensive than other energy sources, which further led to less coal consumption in the USA and Europe to take advantage of cheaper renewable energy (Biao et al. 2018; Holder 2020). Although global demand for renewable energy sources has increased by 1% in the last fiscal year, the disruption in the global supply chain means that the renewable energy sector has downgraded its projects for electric vehicles, battery, solar, and wind sector. This also means that the global fight to combat climate change has been disrupted by the COVID-19 pandemic. Hence, we provide significant changes in policy recommendations as most versatile energy policies are needed to be adopted to navigate energy transition. Figure 8 highlights key policy challenges and policy recommendations from the viewpoint of three distinct policy horizons: immediate response to the COVID-19 crisis is to shield public health (short-term), when economic recovery is stimulated (mid-term), and when trade-off occurs between economic recovery and climate change (long-term).Fig. 8 Key challenges and recommendations for energy policymakers towards renewable energy transition in response to COVID-19 Given the status of the current pandemic, public health infrastructures must become sustainable to properly function during future pandemics. This change, up to some extent, is guided by the issues faced by policymakers, who need to determine policy shifts to ensure the transition towards clean energy sources (Bashir et al. 2021b; Ma et al. 2021c). In these situations, taking a firm stand against the regulatory structures or negative counterattacks and implementing a safeguard mechanism to ensure transition towards renewable energy must be prioritized. Hence, immediate exemptions instead of systemic reforms must resolve the adverse economic impacts from the COVID-19. Hence, instead of preferring “quick wins” from relief packages, policy changes must be aimed to shape industrial and economic activities in a way that they become compatible with climate agreements, i.e., Kyoto Protocol and Paris Agreement. Governments around the world have introduced stimulus measures to overcome economic damage by the COVID-19 as the financial outcomes of the COVID-19 will have a significant bearing on the macroeconomic environment, which is why energy transformation towards sustainable energy sources from political and macroeconomic perspectives remains a significant challenge. In this regard, investments from the private sector to encourage energy transition are critical, which must be accompanied by lower borrowing costs from the conventional banking sector. On the other hand, the continuous fluctuations in oil prices can lead to developments in fuel-efficient technologies. Keeping this discussion in mind, further policy changes must be aimed at linking economic stimulus packages with low-carbon pathways through pricing reforms, which can provide incentives to accelerate the transition towards cleaner energy sources, higher investments in energy infrastructure, and lower the financial burden towards energy transition. Lastly, recent interaction between economic recovery and climate change indicates that energy transition towards cleaner energy is significantly influenced by macroeconomic shocks. As emphasized by the current crisis, instead of investments in new infrastructure, if policy changes only focus on easing current socioeconomic circumstances, the current pandemic will continue to hinder investments in green energy technologies. Such delays will exacerbate the risks of climate change, especially in developing economies. Hence, the key issue faced by policymakers is to integrate policy shifts with long-term energy transition policies. In this regard, the implementation of hybrid instruments, i.e., allocation of funds for infrastructure development, implementation of additional taxation, and elimination of subsidies, can be effective. Also, the outbreak of COVID-19 has highlighted the deficiencies in political response towards public health crisis, and it is indicated that the post-COVID era will mainly focus on economic recovery rather than politics. The shifts in oil price following the declaration of COVID-19 as a global pandemic has highlighted the significance of “producer economies” to put them on a sustainable path, particularly if “green recovery” leads to an earlier peak in demand for fossil fuels. Failure to accommodate such changes will create tension and uncertainty among global economies both within and between nation-states. Within this context, we suggest that the determination of geopolitical aspects for trade pacts and sustainable energy agreements is critical in the future, while not neglecting the need to retain the democratic nature of global economies’ policy formulations. Conclusion In recent months, the global energy industry is under constant duress due to the emergence of the COVID-19. The current study has reviewed the impact of the COVID-19 on the energy industry. Although the short-term impact from COVID-19 remains unclear, i.e., demand for energy products fell unexpectedly as well as fluctuating prices, and they correlate with industrial activities. According to the International Energy Agency, the future growth of the renewable sector is under threat in the long run, while medium-term investments in the energy sector are also expected to decrease. Lastly, our analysis indicates that global and geo-political tensions can significantly impact the recovery of the energy sector, which is why the current study recommends three guiding principles to facilitate future energy transition and prevent global climate change: (1) there should be higher restrictions on overspending, especially in the shirt-term; (2) policy goals need to be adjusted to avail economic opportunities during the energy transition; and (3) introduce policy mechanism to ensure consistent energy flow from the renewable energy sector. Furthermore, as COVID-19 is still evolving, especially in developing economies, there is a need for further research to analyze the extensive impact of the current pandemic over energy consumption energy resources and mitigation strategies. Long-term policy rationalizations need to be given further considerations within the energy policy frameworks across the nation/region. Also, further research can analyze the effectiveness of revising energy-related sustainable development goals or how existing energy goals can be affected by the introduction of policy designs proposed by the current study. Author contribution Muhammad Farhan Bashir: Conceptualization, writing the original draft, revision Besma Talbi: Data analysis Muhammad Adnan Bashir: Review Muhammad Sadiq: Methodology Luqman Shahzad: Revision Data availability Data and relevant materials will be available from the corresponding author through email. Declarations Ethical approval Not applicable. Consent to participate Not applicable. Consent to publish Not applicable. Competing interests The authors declare no competing interests. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Akrofi MM Antwi SH COVID-19 energy sector responses in Africa: a review of preliminary government interventions Energy Res Soc Sci 2020 68 101681 10.1016/j.erss.2020.101681 32839700 Aktar MA Alam MM Al-Amin AQ Global economic crisis, energy use, CO2 emissions, and policy roadmap amid COVID-19 Sustain Prod Consum 2020 26 770 781 10.1016/j.spc.2020.12.029 33786357 Anderson RM Heesterbeek H Klinkenberg D Hollingsworth TD How will country-based mitigation measures influence the course of the COVID-19 epidemic? Lancet 2020 395 10228 931e934 10.1016/S0140-6736(20)30567-5 32164834 Arezki R, Fan RY (2020) Oil price wars in a time of COVID-19. Available at: https://voxeu.org/article/oil-price-wars-time-covid-19. Accessed 2 Jan 2022 Arslan H, Bilal, Bashir MF (2021) Contemporary research on spillover effects of COVID-19 in stock markets. A systematic and bibliometric review. Sci Forum:1–14. 10.3390/ECERPH-3-09103 Aruga K Islam M Jannat A Effects of COVID-19 on Indian energy consumption Sustainability 2020 12 14 5616 10.3390/su12145616 Badr HS Du H Marshall M Dong E Squire MM Gardner LM Association between mobility patterns and COVID-19 transmission in the USA: a mathematical modelling study Lancet Infect Dis 2020 20 11 1247 1254 10.1016/S1473-3099(20)30553-3 32621869 Baloch, M. A., Khan, S. U. D., Ulucak, Z. Ş., & Ahmad, A. (2020). Analyzing the relationship between poverty, income inequality, and CO2 emission in Sub-Saharan African countries. Sci Total Environ Bannister-Tyrrell, M., Meyer, A., Faverjon, C., & Cameron, A. (2020). Preliminary evidence that higher temperatures are associated with lower incidence of COVID-19, for cases reported globally up to 29th February 2020. medRxiv Bashir MF (2018) A class of multi-attribute auction transformed into single-attribute auction on Margin Bid. Bus Eco J, 9(352), 2. https://www.hilarispublisher.com/abstract/a-class-of-multiattribute-auction-transformed-into-singleattribute-auction-on-margin-bid-32782.html Bashir MF Oil price shocks, stock market returns, and volatility spillovers: a bibliometric analysis and its implications Environ Sci Pollut Res 2022 10.1007/s11356-021-18314-4 Bashir MF, Shahzad U, Latif S, Bashir M (2015) The Nexus between Economic Indicators and Economic Growth in Brazil. Nexus, 13(1). https://iiste.org/Journals/index.php/JRDM/article/view/26653/27303 Bashir MF Ma B Bilal KB Bashir MA Tan D Bashir M Correlation between climate indicators and COVID-19 pandemic in New York, USA Sci Total Environ 2020 728 138835 10.1016/j.scitotenv.2020.138835 32334162 Bashir MF Ma BJ Bilal Komal B Bashir MA Farooq TH Iqbal N Bashir M Correlation between environmental pollution indicators and COVID-19 pandemic: a brief study in Californian context Environ Res 2020 187 109652 10.1016/j.envres.2020.109652 32405084 Bashir MF Ma B Shahzad L A brief review of socio-economic and environmental impact of Covid-19 Air Qual Atmos Health 2020 13 12 1403 1409 10.1007/s11869-020-00894-8 32837620 Bashir MF, Benjiang MA, Shahzad L (2020d) Breve revisión del impacto socioeconómico y ambiental de la COVID-19. LECTURA SUCESIVA, 16. http://temas.cult.cu/articulos-academicos/breve-revision-impacto-socioeconomico-ambiental-covid-19/ Bashir MF Ma B Shahbaz M Jiao Z The nexus between environmental tax and carbon emissions with the roles of environmental technology and financial development PLoS ONE 2020 15 11 e0242412 10.1371/journal.pone.0242412 33237920 Bashir MF Ma B Shahbaz M Shahzad U Vo XV Unveiling the heterogeneous impacts of environmental taxes on energy consumption and energy intensity: empirical evidence from OECD countries Energy 2021 226 120366 10.1016/j.energy.2021.120366 Bashir MF Ma B Bilal Komal B Bashir MA Analysis of environmental taxes publications: a bibliometric and systematic literature review Environ Sci Pollut Res 2021 28 20700 20716 10.1007/s11356-020-12123-x Bashir MF Ma B Bashir MA Bilal Shahzad L Scientific data-driven evaluation of academic publications on environmental Kuznets curve Environ Sci Pollut Res 2021 28 16982 16999 10.1007/s11356-021-13110-6 Bashir MF, Ma B, Shahzad L, Liu B, Ruan Q (2021d) China’s quest for economic dominance and energy consumption: can Asian economies provide natural resources for the success of One Belt One Road? Manage Decis Econ, 2021d. 42(3) 570–587. 10.1002/mde.3255 Bashir MA Sheng B Farooq MU Bashir MF Shahzad U The role of macroeconomic and institutional factors in foreign direct investment and economic growth: empirical evidence in the context of emerging economies Global Local Econ Rev 2021 24 2 67 Bashir MF, Ma B, Bashir MA, Radulescu M, Shahzad U (2021e) Investigating the role of environmental taxes and regulations for renewable energy consumption: evidence from developed economies. Econ Res-EkonIstraživanja, 1-2310.1080/1331677X.2021.1962383 Bashir MF Ma B Qin Y Bashir MA Evaluation of One Belt One Road publications: a bibliometric and literature review analysis Environ Sci Pollut Res 2021 28 37016 37030 10.1007/s11356-021-14621-y Bashir MF Benjiang MA Hussain HI Shahbaz M Koca K Shahzadi I Evaluating environmental commitments to COP21 and the role of economic complexity, renewable energy, financial development, urbanization, and energy innovation: empirical evidence from the RCEP countries Renew Energy 2022 184 541 550 10.1016/j.renene.2021.11.102 Bashir MF, Ma B, Xia W, Shahzad U, Radulescu M (2022b) Do economic openness and institutional quality influence environmental patents? Empirical evidence from South Asia. Environ Eng Manag J 21(1):49–61 Biao L, Benjiang M, Bashir MF (2018) Mechanism design of multi-attribute reverse auction on margin bid. 4th International Conference on Education Technology, Management and Humanities Science. 1–6. https://www.atlantis-press.com/proceedings/etmhs-18/25893331 Bilal Bashir MF Benghoul M Numan U Shakoor A Komal B Bashir MA Bashir M Tan D Environmental pollution and COVID-19 outbreak: insights from Germany Air Qual Atmos Health 2020 13 11 1385 1394 10.1007/s11869-020-00893-9 Bilal Latif F Bashir MF Komal B Tan D Role of electronic media in mitigating the psychological impacts of novel coronavirus (COVID-19) Psychiatry Res 2020 289 113041 10.1016/j.psychres.2020.113041 Bilal, Bashir MF, Shahzad K, Komal B, Bashir MA, Bashir M, Tan D, Fatima T, Numan U (2021a) Environmental quality, climate indicators, and COVID-19 pandemic: insights from top 10 most affected states of the USA. Environ Sci Pollut Res Int:1–10. 10.1007/s11356-021-12646-x Bilal Bashir MF Komal B Benghoul M Bashir MA Tan D Nexus between the COVID-19 dynamics and environmental pollution indicators in South America Risk Mana Healthcare Policy 2021 14 67 10.2147/RMHP.S290153 Bilal, Bashir MF, Shahzad A, Komal B, Bashir MA, Tan D (2021c) Nexus between temperature and COVID-19 pandemic: a meta-analysis. Sci Forum:1–5. 10.3390/ECERPH-3-09098 BP (2019) BP Statistical Review of World Energy (68th edition) Broom D (2020) These 3 charts show what COVID-19 has done to global energy demand. The World Economic Forum COVID Action Platform. Available at https://www.weforum.org/agenda/2020/08/covid19-change-energy-electricity-uselockdowns-falling-demand/. Accessed 1 Jan 2022 Carrington D, Ambrose J, Taylor M (2020) Will the coronavirus kill the oil industry and help save the climate? The Guardian International Edition. Available at: https://www.theguardian.com/environment/2020/apr/01/the-fossil-fuel-industry-isbroken-will-a-cleaner-climate-be-the-result. Accessed 1 Jan 2022 Cherp, A., & Jewell, J. (2020). COVID-19 weakens both sides in the battle between coal and renewables. Behavioural and Social Sciences. Available at: https://socialsciences.nature.com/posts/66644-by-disrupting-technology-diffusion-andsupply-chains-covid-19-may-harm-renewables-more-than-coal-but-still-weaken-coallock-in-in-developing-countries D'Adamo I Gastaldi M Morone P The post COVID-19 green recovery in practice: assessing the profitability of a policy proposal on residential photovoltaic plants Energy Policy 2020 147 111910 10.1016/j.enpol.2020.111910 32989340 Drezner DW The song remains the same: international relations after COVID-19 Int Organ 2020 74 Supplement 1 18 Egan, M (2020) Oil prices turned negative. Hundreds of US oil companies could go bankrupt. CNN Business, New York, USA Elavarasan RM, Shafiullah GM, Kannadasan R, Mudgal V, Arif MT, Jamal T, ... & Subramaniam U (2020) COVID-19: Impact analysis and recommendations for power sector operation. Appl Energy, 279, 115739 Fareed Z Bashir MF Bilal Saleem S Investigating the co-movement nexus between air quality, temperature, and COVID-19 in California: implications for public health Front Public Health 2021 9 815248 10.3389/fpubh.2021.815248 35004602 Farooq TH Xincheng X Shakoor A Rashid MHU Bashir MF Nawaz MF Kumar U Shahzad SM Yan W Spatial distribution of carbon dynamics and nutrient enrichment capacity in different layers and tree tissues of Castanopsis eyeri natural forest ecosystem Environ Sci Pollut Res 2022 29 7 10250 10262 10.1007/s11356-021-16400-1 Farooq U Nasir A Bilal Bashir MF The COVID-19 pandemic and stock market performance of transportation and travel services firms: a cross-country study Econ Res-Ekon Istraživanja. 2022 1 1 1 17 10.1080/1331677X.2022.2053784 Fofana NK Latif F Sarfraz S Bashir MF Komal B Fear and agony of the pandemic leading to stress and mental illness: an emerging crisis in the novel coronavirus (COVID-19) outbreak Psychiatry Res 2020 291 113230 10.1016/j.psychres.2020.113230 32593067 Gautam S The influence of COVID-19 on air quality in India: A boon or inutile Bull Environ Contam Toxicol 2020 10.1007/s00128-020-02877-y Gautam, S. (2020b). COVID-19: air pollution remains low as people stay at home. Air Quality, Atmosphere, & Health, 1–5. Global Energy Statistical yearbook (2020). Available at: https://yearbook.enerdata.net/totalenergy/world-consumption-statistics.html. Accessed 7 Jan 2022 Hepburn, C., O’Callaghan, B., Stern, N., Stiglitz, J., & Zenghelis, D. (2020). Will COVID-19 fiscal recovery packages accelerate or retard progress on climate change? Oxford Smith School of Enterprise and the Environment, Working Paper 20–02. Holder M (2020) Coronavirus: falling power demand is impacting clean energy. Available at: https://www.greenbiz.com/article/coronavirus-falling-power-demand-impactingclean-energy. Accessed 2 Jan 2022 Hussain M Bashir MF Shahzad U Do foreign direct investments help to bolster economic growth? New insights from Asian and Middle East economies World J Entrep Manag Sustain Dev 2021 17 1 62 84 10.1108/WJEMSD-10-2019-0085 IEA (2019) Southeast Asia Energy Outlook. https://www.iea.org/reports/southeast-asia-energy-outlook-2019 ICIS (2020) Coronavirus impact on energy markets. Available at: https://www.icis.com/explore/resources/news/2020/03/19/10482507/topic-pagecoronavirus-impact-on-energy-markets. Accessed 1 Jan 2022 Jefferson M A crude future? COVID-19s challenges for oil demand, supply and prices Energy Res Soc Sci 2020 68 101669 10.1016/j.erss.2020.101669 32839697 Johnson K, Standish R (2020) Trump’s promised oil deal still eludes big producers as prices dive again. Available at: https://foreignpolicy.com/2020/04/06/trump-promisedoil-deal-saudi-russia-price-war/. Accessed 2 Jan 2022 Kuzemko C Re-scaling IPE: local government, sustainable energy and change Rev Int Polit Econ 2019 26 1 80 103 10.1080/09692290.2018.1527239 Ma B Zhou Z Bashir MF Huang Y A multi-attribute reverse auction model on margin bidding Asia-Pacific J Oper Res 2020 37 06 2050032 10.1142/S0217595920500323 Ma BJ Ye JY Huang YJ Bashir MF Research of two-period insurance contract model with a low compensation period under adverse selection Manag Decis Econ 2020 41 3 293 307 10.1002/mde.3100 Ma B Wang Y Zhou Z Lai Y Zhou Z Bashir MF Can controlling family involvement promote firms to fulfill environmental responsibilities?—Evidence from China Manag Decis Econ 2021 43 2 569 592 10.1002/mde.3403 Ma B Tang Q Qin Y Bashir MF Policyholder cluster divergence based differential premium in diabetes insurance Manag Decis Econ 2021 42 7 1793 1807 10.1002/mde.3345 Ma B Zhang Y Qin Y Bashir MF Optimal insurance contract design with “No-claim Bonus and Coverage Upper Bound” under moral hazard Expert Syst Appl 2021 178 115050 10.1016/j.eswa.2021.115050 Malahayati M Achieving renewable energies utilization target in South-East Asia: progress, challenges, and recommendations Electr J 2020 33 5 106736 Muhammad S Long X Salman M COVID-19 pandemic and environmental pollution: a blessing in disguise? Sci Total Environ 2020 728 138820 10.1016/j.scitotenv.2020.138820 32334164 Qaid A Bashir MF Remaz Ossen D Shahzad K Long-term statistical assessment of meteorological indicators and COVID-19 outbreak in hot and arid climate, Bahrain Environ Sci Pollut Res 2022 29 1 1106 1116 10.1007/s11356-021-15433-w Sadiq M, Shinwari R, Usman M, Ozturk I, Maghyereh AI (2022) Linking nuclear energy, human development and carbon emission in BRICS region: do external debt and financial globalization protect the environment? Nucl Eng Technol:1–17. 10.1016/j.net.2022.03.024 Shahbaz M Mallick H Mahalik MK Sadorsky P The role of globalization on the recent evolution of energy demand in India: implications for sustainable development Energy Econ 2016 55 52 68 10.1016/j.eneco.2016.01.013 Singh RK India Renewables Projects Set for Slowdown on Lockdown Bloomberg Green 2020 New York United States Shahbaz M, Bashir MF, Bashir MA, Shahzad L (2021) A bibliometric analysis and systematic literature review of tourism-environmental degradation nexus. Environ Sci Pollut Res 1-1710.1007/s11356-021-14798-2 Sultan N, Mohamed N, Bashir MA, Bashir MF (2021) The anti‐money laundering and counter financing of terrorism policy in Pakistan: is it truly combating or just a high‐level desk work bureaucracy?. Journal of Public Affairs, e2731. 10.1002/pa.2731 Sun P Lu X Xu C Sun W Pan B Understanding of COVID-19 based on current evidence J Med Virol 2020 92 6 548 551 10.1002/jmv.25722 32096567 Talbi, B., Jebli, M. B., Bashir, M. F., & Shahzad, U. (2020). Does economic progress and electricity price induce electricity demand: a new appraisal in context of Tunisia. Journal of Public Affairs, e2379. https://onlinelibrary.wiley.com/doi/full/10.1002/pa.2379 Thiéry F (2020) Oil and gas sector: COVID-19 will affect oil and gas prices even in the medium to long term. Available at: https://www.credendo.com/nl/node/8491OIL. Accessed 1 Jan 2022 Tosepu R Gunawan J Effendy DS Lestari H Bahar H Asfian P Correlation between weather and Covid-19 pandemic in Jakarta, Indonesia Sci Total Environ 2020 725 138436 10.1016/j.scitotenv.2020.138436 32298883 Vaka M Walvekar R Rasheed AK Khalid M A review on Malaysia’s solar energy pathway towards carbon-neutral Malaysia beyond Covid’19 pandemic J Clean Prod 2020 273 122834 10.1016/j.jclepro.2020.122834 32834565 Van de Graaf T, Sovacool BK (2020) Global Energy Politics. John Wiley & Sons Wang, Q., & Su, M. (2020). A preliminary assessment of the impact of COVID-19 on environment–a case study of China. Sci Total Environ, 138915. Wei C Löschel A Managi S Recent advances in energy demand research in China China Econ Rev 2020 63 101517 10.1016/j.chieco.2020.101517 WHO (2021) Rolling updates on coronavirus disease (COVID-19). Available at: https://www.who.int/emergencies/diseases/novel-coronavirus-2019/events-as-theyhappen. Accessed 21 Jan 2022 World Economic Forum (2020) COVID-19: America hasn’t used this little energy in 16 years. Available at: https://www.weforum.org/agenda/2020/04/united-states-eneregyelectricity-power-coronavirus-covid19/. Accessed 1 Jan 2022 Worldometers (2020) World population by country. Available at: https://www.worldometers.info/world-population/. Accessed 1 Jan 2022 Wu Y, Jing W, Liu J, Ma Q, Yuan J, Wang Y, ... & Liu M (2020) Effects of temperature and humidity on the daily new cases and new deaths of COVID-19 in 166 countries. Sci Total Environ, 729, 139051. Xia W Apergis N Bashir MF Ghosh S Doğan B Shahzad U Investigating the role of globalization, and energy consumption for environmental externalities: empirical evidence from developed and developing economies Renewable Energy 2022 183 219 228 10.1016/j.renene.2021.10.084 Zhou P Yang XL Wang XG Hu B Zhang L Zhang W A pneumonia outbreak associated with a new coronavirus of probable bat origin Nature 2020 579 270 273 10.1038/s41586-020-2012-7 32015507
PMC009xxxxxx/PMC9006072.txt	==== Front Curr Psychol Curr Psychol Current Psychology (New Brunswick, N.j.) 1046-1310 1936-4733 Springer US New York 35431524 3089 10.1007/s12144-022-03089-9 Article Should I buy or not? Revisiting the concept and measurement of panic buying Cham Tat-Huei jaysoncham@gmail.com 1 Cheng Boon-Liat boonliatc@sunway.edu.my 2 Lee Yoon-Heng leeyh@utar.edu.my 3 http://orcid.org/0000-0001-8440-9564 Cheah Jun-Hwa jackycheahjh@gmail.com 4 1 grid.444472.5 0000 0004 1756 3061 UCSI Graduate Business School, UCSI University, Kuala Lumpur, Malaysia 2 grid.430718.9 0000 0001 0585 5508 Sunway University Business School, Sunway University, Petaling Jaya, Selangor Malaysia 3 grid.412261.2 0000 0004 1798 283X Faculty of Accountancy and Management, Universiti Tunku Abdul Rahman (UTAR), Kajang, Selangor Malaysia 4 grid.11142.37 0000 0001 2231 800X School of Business and Economics, Universiti Putra Malaysia (UPM), Serdang, Selangor Malaysia 13 4 2022 121 4 4 2022 © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022 This article is made available via the PMC Open Access Subset for unrestricted research re-use and secondary analysis in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the World Health Organization (WHO) declaration of COVID-19 as a global pandemic. Following various precautionary measures as executed by the government to curb the transmission of COVID-19, erratic changes in the form of temporary lockdowns and movement restrictions have created an emergency phenomenon—panic buying. While such consequence has emerged as a timely and relevant topic, reviewed literature indicate an apparent oversight for portraying panic buying through the perspectives of impulsive and compulsive consumptions. Given the gap in the association between panic buying and consumers’ emotional aspects within the context of the COVID-19 pandemic, this study aspires to develop a contemporary measurement that accurately defines panic buying as a research variable. A combined methodology was hereby adopted, with the employment of qualitative inquiries towards the scale development of panic buying. Following this, quantitative data as collected from a total sample of 600 respondents through an online survey was analysed via both SPSS and AMOS statistical software towards scale assessment and hypothesis testing. Obtained findings uncovered the direct significance of both personal (fear, perceived risk, and perceived scarcity) and social (word-of-mouth and social media) factors on panic buying during the pandemic, whilst having indirect significance on the ensuing post-purchase regret. Impulsivity was further confirmed to exert a substantial moderating impact on the correlation between panic consumption and post-purchase emotional distress. Implications of the study are ultimately discussed. Keywords Consumer behaviour COVID-19 Impulsivity Panic buying Personal factors Social factors ==== Body pmcIntroduction The coronavirus outbreak was announced as a public health emergency of international concern (PHEIC) by the World Health Organisation (WHO) as of January 2020. Outstripping previous outbursts of swine flu in 2009, polio in 2014, Ebola in Western Africa in the same year, and the Zika virus from 2015 to 2016, the new virus was being declared as a worldwide pandemic in March of 2020 due to its proliferating infection rate at the global scale. Its severity was essentially realized following a whopping 55 million reported international infections with more than 1.33 million associated casualties (New Straits Times, 2020). Amid the continuous scientific odyssey in search of medical resolutions, multiple nations practised intense vigilance for a possible third wave of the Covid-19 outbreak by the end of 2020. Given the scale of this public hazard, regulatory measures with the like of social distancing and mandatory national lockdowns were further enforced by governmental and authorised parties across countries in fulfilling the purpose of virus containment. Likewise, the Malaysian government has required its public to continuously observe social distancing, update their status via the tracking app (i.e., MySejahtera) on a timely basis, mandate the wearing of face masks in public, avoid mass gathering and crowded locations, without overlooking the importance of personal hygiene. The sudden invasion of the unavoidable outbreak has, therefore, preceded disrupted regularities via necessary changes like remote working, online education, social transformation, economic reformation as well as virtual communication (Cham et al., 2021a, b). Notable mainstream and social media outlets have concurrently demonstrated market unpreparedness in the face of pre-pandemic existential threats by observable norms of panic buying. Stockpiling of common items ranging from canned food, pasta, rice, flour, and yeast to hand sanitisers and other hygienic items like toilet paper subsequently befallen upon market perceptions on an extended movement restriction order. Nevertheless, such impulsive consumption was apparently short-lifted, with consumers regaining their pre-pandemic levels of consumption upon reaching situational familiarity (Forani, 2020). Such transformative endeavours, thus, raised ambiguities in market perceptions, with manifesting fundamental questions of: 1) whether panic buying is just a transient behaviour, without underlying alteration to the stability of consumers’ buying psychology; and 2) the likelihood in which pandemic-driven shopping habits and changes as introduced by business operators would persist in the post-pandemic scenario. From the academic perspective, panic buying has received immense attention in various scholarly disciplines, such as consumer psychology (Kaur & Malik, 2020), marketing (He & Harris, 2020), and economics (Yoshizaki et al., 2020). However, preliminary reviewed literature have confirmed the new-fangled nature of most related studies, which left significant gaps concerning the operationalisation of the proposed construct to be filled. In particular, limited studies have put forward feasible measurements for panic buying as a research variable (Ahmad & Murad, 2020; Islam et al., 2021; Lins & Aquino, 2020; Tan et al., 2022). Moreover, the role of social media as a potential driver for panic buying during crisis moments remains relatively unexplored, despite its recognition as a primary channel of modern communication (Islam et al., 2021; Naeem, 2020a). These inadequacies signify an explorative gap that warrants further attention. Acknowledging the disastrous health-related crisis plaguing the world, the current study, therefore, sought to propose a theoretical framework aimed towards developing a usable measurement scale for consumers’ panic buying particularly in the current pandemic context. It is anticipated that the new measurement scale will be able to address the shortcomings of the existing measurement scale for panic buying, while allowing gauging the variable in a more conclusive manner. The importance of this topic is rooted in the behavioural shift caused by disruptive or destructive incidents. In this regard, the impact of the pandemic is evident in both societal unease and anxiety that traversed geographical boundaries (Forbes, 2012), consequently interrupting consumers’ cognitive rationality. The resulting impulsive consumption and extreme panic buying are motivated by their need to retain perceived dominance in that aspect of life (Wang et al., 2020). With this, the recent groundwork has highlighted that the reasons behind such behavioural irregularity are individual perceptions of existing threats, resource deficiencies, future uncertainties, coping mechanisms, and societal pressure (Yuen et al., 2020). A contrasting outlook was offered by Loxton et al. (2020), who emphasised the catalytic influence of herd mentality or social factors in encountering distressing incidents, as well as the repercussions of communication media. This context clearly exemplifies the case of “scarcity heuristics”, where certain ordinary products (e.g., toilet paper) gain elevated value following intensified anxiety-driven demands within the marketplace. In this altered environment, it is undeniable that varying results would be derived based on consumers’ changing perceptions in different unfavourable situations. As such, the empirical aspect of panic buying was addressed in the current study by examining the impact of consumers’ personal (fear of COVID-19, perceived risk associated with COVID-19, and perceived scarcity) and social (word-of-mouth communication and social media communication) factors on their panic consumption. Naeem (2020a) highlighted the evident gap in terms of the limited theoretical understanding of the impact of social media on panic buying at a global scale, which motivated the incorporation of social media as an independent variable in the context of crises. Findings as obtained from this study would provide answers to the uncertain nature of panic buying during the COVID-19 pandemic, thereby shaping a better understanding of the market and benefitting both the retailing and the supply chain management fields. In addition, given the research gaps, developing a relevant measurement scale for panic buying and validating it in the proposed model are expected to contribute valuable insights to the consumer behaviour literature. The remainder of this paper presents the literature review, the methodology used, and the analysis results. Finally, the findings and their implications are discussed. Literature Review and Hypotheses Development Drawing from the lens of compensatory control theory (CCT), the paradigm has reported that individuals strive to have a perceived sense of personal control over their environment as a means of believing that the surrounding is within their control and the world is meaningfully structured (Sullivan et al., 2012). That is to say that individuals are able to assuage any threats if they believe that things are well in-hand (Kay & Eibach, 2013). Since its inception, CCT has been employed in various research settings due to its capability in explaining individuals’ reaction towards uncertainties (Barnes et al., 2021). Facing an ambiguous marketing setting, CCT has put forward that consumers with lower perceived control tend to seek a feeling of control by buying utilitarian goods in any event of uncertainties (Arafat et al., 2021; Barnes et al., 2021). Alternatively, explanation is given by such paradigm within the organisational setting on the direct association between anxiety and productivity through increased personnel investment, intuitive efforts as well as physical and cognitive expenses; followed by the attainment of performance consistency by mean of reduced targets at the absence of additional costs (Hockey, 1997). In lieu of other adverse circumstances, CCT is specified towards addressing the consequential influence of diminished governance or perceived control towards lowered tolerance and heightened pessimism on dubious situations (Ma & Kay, 2017). Further devised by Eysenck et al. (2007), consistency is secured under conditions of stress and anxiety through the undertaking of precautionary endeavours in the forms of increased behavioural and capital investments. The statement seemingly reflected impulsive preparations amidst disruptive adversities with the like of a pandemic outbreak. Such defence mechanism fundamentally prevails as the consumers’ immediate solution to a specified problem in the event of uncertainties (Arafat et al., 2021; Lu et al, 2022). Hence, this theory helps to explain why the situation of panic buying has been apparent during the pandemic period. Grounded on the foundation of CCT, examined framework within this study was, therefore, set to explore the fundamentals of panic buying and its antecedents. According to Twedt (1965), the marketplace resembles as a living laboratory that indirectly accounts for the behavioural complexity among its consumers (Simonson et al., 2001). Case in point, the recent ravages of the COVID-19 pandemic resulted in unexpected forays, such as the worldwide stockpiling of toilet papers, a surging interest in bread making as well as an unforeseen scramble to get hold of a bicycle (Andrew, 2020; Chakravarty, 2020). Such behavioural displacement is commonly born from fear and anxiety, whilst prompted an increase in consumption (Lins & Aquino, 2020). Labelled as panic buying, this practice is commonly associated with unusually large consumption or an unexpectedly varied range of products purchased prior to, during, or after a perceived disaster or perceived product scarcity (Yuen et al., 2020); in turn, contributing undesirable repercussions to the aspects of social wellbeing, environmental systems, and supply chains. As revealed by the UN International Strategy for Disaster Reduction (2009), dissimilar purchasing habits would emerge in accordance to diverging nature of the experienced crises. On this note, studies have repeatedly captured trends of panic buying in various disastrous occurrences, while concurrently attracted great interest among consumer behaviour scholars (Arafat et al., 2020a, b; Loxton et al., 2020). However, systematic operationalisation of panic buying as a measurable variable remains underdeveloped, with requiring additional empirical evidence to provide a more conclusive outlook on the construct. Past studies have often compared panic buying to impulsive and compulsive consumption patterns (e.g., Islam et al., 2021; Naeem, 2020b). These claims might not be entirely appropriate in explaining the exact meaning of panic buying, as they have neglected the emotional aspects of consumers. Drawing from the transactional theory of stress and coping by Lazarus and Folkman (1984), panic buying is closely associated with consumers’ perceptions of stress and anxiety in emergency situations, which provokes excessive consumption. Altered spending behaviours amid the COVID-19 pandemic have, therefore, been driven by consumers’ fear of unknown threats (Song et al., 2020), stress-relief coping behaviour (Loxton et al., 2020), and risk (Addo et al., 2020; Arafat et al., 2020a, b). Essentially, these elements posit panic buying as consumers’ attempt to recapture a sense of control against the collective circumstance of increased uncertainty, higher risk, and heightened anxiety. Yuen et al. (2020) understood panic buying as an individual’s fear-limiting stress- and anxiety-coping mechanism to counter their psychological perceptions of ambiguity and uncertainty. Thus, scholars appear concur that panic buying is, in fact, built on the foundation of stress, anxiety, and excessive buying experienced by consumers during times of emergency or uncertainty. Some scholars have separated panic buying into casual pursuits (wants) and critical requirements (needs). The former has often been investigated during regular periods where impulsivity acts as an initial motivator of compulsive buying, along with irrationality and self-governance (Williams & Grisham, 2012). Liu et al. (2018) suggested that individual attributes like materialism, egoism, and mentality are often underlying predictors of consumption in an anxiety-filled environment. This is further worsened by the triggering effect of promotional and marketing efforts by retailers, minimising consumers’ efforts in the decision-making process. On the other hand, the second school of thought (i.e., panic buying for ‘need’) argues that behaviour reflects necessity over mere desire. For instance, Arafat et al., (2020a, b) suggested that necessities are manifested in the forms of: 1) perceived scarcity, 2) sense of control, 3) sense of security, 4) collective mentality, and 5) risk assessment. In these situations, despite unchanged personal preferences and product inclinations, greater consumption of specific items are observed based on consumers’ situational appraisal (Martin-Neuninger & Ruby, 2020). Indeed, as panic consumption parallels the pandemic’s severity, materialism and herd mentality have proven to be substantial influencers (Jin et al., 2020). Similarly, widespread fear in times of crisis are cultivated through mutual interactions, leading to a vicious cycle of supply shortage (Loxton et al., 2020). The cumulated outlook of Wang and Hao (2020) views both product hoarding and panic buying behaviours as separate consequences of irrational considerations, with the former relating to resource availability and the contagiousness of the pandemic and the latter relating to negative emotions and herd mentality. Understanding the determinants of panic buying solely through consumers’ cognitive irrationality (e.g., compulsive consumption) would, once again, be overly simplified. Such findings have yet to address the significance of virtual information, which has overshadowed offline communication in generating and intensifying crisis-associated perceptions in the marketplace (Li et al., 2020a, b; Naeem, 2020b). Behavioural displacement during the COVID-19 outbreak has also been observed through the lens of perceived risk (Addo et al., 2020; Arafat et al., 2020a, b), threat-related uncertainties (Song et al., 2020), coping behaviours towards stress-relief, social influences (Loxton et al., 2020), and the technological influence of social media (Li et al., 2020a, b). In response to these factors, panic buying is potentially executed as a fear-limiting, anxiety-coping mechanism to face the unknown (Yuen et al., 2020). Nonetheless, ambiguity remains in terms of the variable’s operationalisation and determinants beyond mere impulsivity in times of a global pandemic, by which the current research framework sought to explore. Factors that Influence Panic Buying Personal Factors Fear From the psychological perspective, fear is a perceptual and subjective concept (Teachman et al., 2008) that possesses the ability to shape one’s action and behaviour (Maddux & Rogers, 1983). According to Laros and Steenkamp (2005), fear has been identified to be among the emotional sensations that influence the behavioural endeavours of fellow consumers. Moreover, Armfield (2006) argued that fear is a behavioural reaction triggered by individualistic perceptions of uncontrollability, unpredictability, uncertainty, and hazard presented by a specific stimulus. In the current context, such emotional and physical distress arises from the COVID-19 pandemic. In the face of threats and adversity, fear prevails as an evolutionary adaptation that emotionally increases the degree of survivability (Kunimatsu & Marsee, 2012). It is comparable to individuals’ fight-or-flight mechanism upon undergoing emotional and cognitive efforts (i.e., the best route of survivability based on the current level of consideration). Understandably, the accelerated pace of the Coronavirus contagion has incited a corresponding level of fear and anxiety at the international scale, disrupting societal homeostasis in multiple ways (Liu et al., 2020; Sorokowski et al., 2020). Upon being declared a global pandemic, observed norms then followed the theory of fear appeal, wherein both the perceived level of risk and comfort-seeking intention heightened. Emotional unease in the form of stress propelled faster decision-making at this time, with the panic buying of apparent necessities (e.g., food) seen as an endeavour of survival assurance (Jezewska-Zychowicz et al., 2020). Instances in stockpiling of groceries (Hall, 2020), personal protective equipment (Addo et al., 2020), and toilet papers (Hall, 2020) have been solid exemplifications of stress relief in response to the current crisis. Other scholars have further proposed that the potential consequences of lockdowns include not only attempted evasion and frustration of isolation, but also abrupt monetary consumption as a proactive approach in crisis management (Taylor et al., 2020). Another notable implication of fear is reflected in changes in social behaviour, as consumers embrace conformity towards contemporary conventions or norms (Song et al., 2020). This is because long-term familiarity can reduce stress and anxiety. However, this notion was rebutted by Wang et al. (2020), who proposed insignificant changes in stress levels throughout the pandemic when considerations are placed on medical confidence, perceived survivability, risk of contraction, availability of health-related knowledge, and individual precautionary attempts. In fact, while some scholars have showcased a positive correlation between perceived stress and compulsive consumption (Zheng et al., 2020), others have not found evidence of such a relationship (Lee & Yi, 2008). In this case, environmental factors would seemingly emerge as cumulative stimuli that provoke planned purchases, overshadowing the gravity of impulsivity. On another note, despite their causal link, both the concepts of fear and panic have been investigated by researchers as parallel variables in transforming consumers’ market intentions and behaviours (Aydınlıoğlu & Gencer, 2020; Shorey et al., 2020). Balancing self-sufficiency, self-preservation, and social conformity, fear is a potential antecedent that acts beyond psychological aspects to include environmental considerations as well (e.g., personal preparedness). Thus, based on the role of fear in shaping compulsive behaviours via panic consumption, as well as the potential correlation between panic buying and fear, we hypothesised that:H1: Fear has a significant impact on panic buying during the COVID-19 outbreak. Perceived Risk Perceived risk has continuously gained significance within consumer research as a powerful influencer of consumers’ behaviour and decision-making process (Mitchell, 1992). Theory suggests that this concept is linked to a higher degree of uncertainty experienced by individuals (Shimp & Bearden, 1982) within the decision-making process. Hoque and Alam (2018) realised that such uncertainty directly depends on the trustworthiness of both information sources and desired products, pending situational apprehension and behavioural endeavours. As such, perceived risk is identical to fear and anxiety in prioritising consumers’ cognitive orientation (i.e., concerns about the crisis) in their consumption decisions (Guzmán-González et al., 2020). The present study, therefore, defines perceived risk as an individual’s perception of exposure to potential uncertainties and health hazards amidst the COVID-19 pandemic. The overwhelming concern about being infected explains the establishment of common standards of procedures (SoPs) across multiple nations to contain the pandemic. Treatment measures aside, preparedness yet again plays a key role, with studies by Nazione et al. (2021) and Naeem (2020a) demonstrating that concern about the pandemic’s severity is a significant antecedent of preventive practices and behavioural displacement (e.g., frequent hand washing and social distancing measures). In direct consistency to such claim, Avery and Park (2021) regarded risk-reducing practices as well-planned decisions made through the assessment of available information during disastrous moments. Related literature on perceived risk has also noted prior crisis-focused preparation and survivability as a direct predictor of consumption behaviour in a crisis (Crockford, 2018). Notable attention should, thus, be brought to the concept of individual preparedness, which accounts for a sustainable decision-making process. Several findings have manifested a behavioural shift among individuals, who naturally embrace survivalist actions as a precautionary response in events of anxiety or crises (Campbell et al., 2019; Mills, 2018). In fact, investigations undertaken by Clemens et al. (2020) and Herman et al. (2020) have shown the positive impact of perceived pandemic risk on consumers’ product selection, and thus, their panic consumption. Yet, these findings barely argued the relevance of risk perception for planned purchases, potentially overlooking consumers’ consideration of risk reduction. Suciu (2020) highlighted those obvious behavioural changes in relation to uncertainty-driven anxiety is typically indisputable. Seeking to confirm the role of uncertainty-driven perceived risk in promoting planned buying during times of disaster, it was hypothesised that:H2: Perceived risk has a significant impact on panic buying during the COVID-19 outbreak. Perceived Scarcity As understood from the work of Gupta and Gentry (2016), perceived scarcity is the perception of product shortage experienced by consumers in a particular situation or circumstance; in this case, this situation is the COVID-19 pandemic. Scarcity is often directly driven by both human (i.e., fluctuations in both market supply and demand) and environmental (i.e., uncontrollable characteristic changes within the supply chain) factors (Gupta & Gentry, 2019). Recognising the unprecedented shifts in both market demand and supply capacity, economic dynamics ensure readjustments in scarcity-driven pricing, along with a supply-intensive mechanism (Bryan et al., 2018). In the current context, scarcity stems from the underestimation of the surge in market demand due to sudden interventions brought about by the global pandemic. Dependent on diverse types of scarcity, consumers’ decision-making process is then correspondingly affected (Hamilton et al., 2019). Economics literature has outlined perceived scarcity as an influential factor in an individual’s economic behaviour (Gupta & Gentry, 2019; Slack et al., 2020). It is known to affect consumers’ cognition, as it shapes comparable responses by generating a sense of urgency (Camargo et al., 2020; Yuen et al., 2020). Considered a negative state of psychological wellbeing, scarcity perceptions have been shown to influence short-term consumption, specifically via the act of unplanned buying (Nazri et al., 2021). While features like reliability, product quality, and attainable value remain crucial in the decision-making process, the sense of urgency triggered by both item and temporal scarcities would undoubtedly motivate behavioural compulsion (Mou & Shin, 2018). Interestingly, a different perspective states that perceived scarcity is directly correlated to acceptable risk (Liang et al., 2020). Since consumers’ panic buying is undertaken to cope with stress during the pandemic, this study investigated both perceived risk and perceived scarcity as concepts that have potential inverse relationships (i.e., perceived scarcity increases one’s willingness to take risk). Fundamentally, panic consumption due to perceptions of market supply shortage has been the result of media communication, or rather, miscommunication (Arafat et al., 2020a, b; Islam et al., 2021). Nichols (2012) argued that relative changes in both perceived scarcity and the ‘empty shelf’ scenario led to competitive arousal among consumers when making successful purchases. This phenomenon is fairly self-explanatory, considering that scarcity topples the societal order of equal distribution in the face of crises and disasters to prioritise survivability (Mannelli, 2020). The concept has, therefore, shown to factor both deliberate consideration (through available information) and psychological satisfaction (stress). Regardless, earlier research have consistently reported perceived scarcity as a significant antecedent to consumers’ hoarding behaviour, urgency to buy (Gupta & Gentry, 2019), and compulsive buying (Wu et al., 2021). The current study, hence, examined the extensive role of perceived scarcity in generating panic buying during the outbreak of COVID-19. The third hypothesis was postulated as:H3: Perceived scarcity has a significant impact on panic buying during the COVID-19 outbreak. Social Factors Word-of-Mouth Communication Word-of-mouth (WOM) communication has constructively achieved recognition as an imperative means of information transfer among practitioners and academics alike (Dellarocas, 2003). Kotler (2006) defined WOM as an individual’s personal communication with close acquaintances (family and friends) about their consumption decisions under specific circumstances. More often than not, decisions are weighted against the recommendations of family, relatives, and friends. This established reference point then undermines the potential risk and uncertainty associated with an otherwise challenging decision (Cheung & Thadani, 2012). Rationality is, thus, grounded in trust, where the source of information is deemed more inclusive, reliable, and objective (Tucker, 2011). The significance of WOM is highly recognised in the areas of brand communication (Andrei et al., 2017), product preference (Marchand et al., 2017), and mindful consumption decisions (Parsad et al., 2019). In light of consumers’ behavioural authenticity, crisis situations that demand quick decision-making benefit from WOM, as information that is comparatively credible, empathetic, trustworthy, relevant, and reliable can be attained under a rushed timeframe (Porter & Golan, 2006). The generation of WOM is contributed to by aspects like actual experience, encountered quality, and value perception (Mukerjee, 2018); as such, WOM possesses extensive empirical relevance as a predictor of consumers’ behaviour. This is supported by Hu et al. (2019), who found that consumers’ unplanned buying is a direct consequence of both emotional support and information exchange from trustworthy social groups. Evidently, WOM has a direct impact on consumers’ compulsive consumption (Khorrami et al., 2015). Therefore, the adoption of WOM might see increased normality when there is a greater sense of urgency. Alternatively, the framework of Hidayanto et al. (2017) posits that WOM indirectly influences purchase intention by increasing the urge for additional information search and creating a sense of dependency. Particularly in disastrous moments, WOM rivals governmental communication in shaping both underlying perceptions and protective measures to curb the global pandemic (Yasir et al., 2020). Nevertheless, while WOM prevails as a powerful advertising tool that ensures customers’ participation and consumption intention (Mukerjee, 2018), its significance in disaster-related communication has not been well-established. It can be presumed that situational transparency determines the reliability and credibility of an information source (Ataguba & Ataguba, 2020), wherein conversation and information exchange are essential for consumers’ mental wellbeing (Parsad et al., 2019). Yet, evidence on the direct influence of WOM on behavioural displacement, specifically panic consumption, remains underwhelming. The fourth hypothesis was, thus, put forth as:H4: Word-of-mouth communication has a significant impact on panic buying during the COVID-19 outbreak. Social Media Communication According to Kaplan and Haenlein (2010), social media is defined as “a group of Internet-based applications that builds on the technological and the ideological foundations of Web 2.0, which allows the creation and the exchange of user-generated content” (p. 61). In congruence with this statement, social platforms like Facebook, Instagram, YouTube, Tumblr, WeChat, WhatsApp, Pinterest, LinkedIn, and Twitter have allowed users to interact with their acquaintances and exchange their views via self-generated content across these various domains (Cham et al., 2022a; Cheah et al., 2019). Be it to search for news (through news portals on social media) or to obtain information from virtual contacts, the existence of these platforms has indisputably changed the way individuals consume knowledge, communicate, and exchange information (Cham et al., 2020, 2021a, b; Hosen et al., 2021; Irshad et al., 2020; Singh & Chakrabarti, 2020). Moreover, social networks have also been reported to change consumers’ buying patterns and habits (Eger et al., 2021; Liu et al., 2021). Practically speaking, these mediums do not only act as robust marketing tools, but have also emerged as useful sources of information during both emergency and catastrophic events (Eckert et al., 2018; Finau et al., 2018). In fact, social media is a key source of crisis communication when other sources are deemed slow and lagging; in this manner, it ensures the timely dissemination of appropriate messages to aid proactive social reactions (Eriksson, 2018). Ranging from the spread of rumours and relief updates during the Haiti earthquake (Muralidharan et al., 2011), to the proliferation of information about the SARS outbreak (Tai & Sun, 2007), social media has frequently proven itself as a tool that allows far-reaching communicative efficiency. Such robustness has, yet again, been rekindled during the recent COVID-19 outbreak, granting the capability for systematic management of the pandemic (Goel & Gupta, 2020). However, the study by Depoux et al. (2020) recognised the tendency for miscommunication across social media platforms, with negative implications arising from untrue information related to a particular crisis. It is indeed well-acknowledged that despite being useful in transferring real-time updates on a global pandemic, social media has also contributed to the deterioration of mental wellbeing among residents by disseminating anxiety and fear (Ahmad & Murad, 2020). In turn, these negative perceptions generate unintended compulsive or panic buying when consumers are exposed to product-related information via sellers’ social portals (Naeem, 2020b). Aside from disseminating actual information, social media essentially holds the possibility to transform both authorised messages and public apprehension of a disastrous situation into miscommunication and exaggeration, which provokes stockpiling practices during the COVID-19 pandemic (Liu et al., 2021; Naeem, 2020b). Thus, a balance between situational preparedness and anxiety (about shortage, uncertainty, and infection) cannot be ignored. Based on the above discussion, it is anticipated that social media communication has a significant influence on behavioural displacement in the investigated context. The fifth hypothesis was postulated as:H5: Social media communication has a significant impact on panic buying during the COVID-19 outbreak. Post-Purchase Regret As defined by Zeelenberg et al. (1998), post-purchase regret is an emotional state encountered upon perceived miscalculation in a particular purchase, which buyers aspire to revise and progress in a distinct, acceptable fashion. In a study on herd mentality, impulsive consumption induced by the worry of being excluded was found to directly provoke post-purchase regret (Karapinar et al., 2019). Additionally, impulsive buying and post-purchase behaviour are known to be directly correlated (Islam et al., 2021). Monetary consumption based on negative emotions like anxiety and sentimental distress further contribute to the generation of regret due to abrupt decision-making, changed necessities, and opportunity cost, which are regulated by unplanned intuition and attention seeking (Sokić et al., 2020). Ozer and Gultekin (2015) also argued that product-based satisfaction ensues impulsive buying and affects consumers’ pre- and post-purchase emotional states, despite unplanned consumption being a direct consequence of impulsivity. When such consumption fails to deliver the expected experience, resulting guilt potentially strengthens impulsive behaviours through acts of constructive criticism and cognitive withdrawal (Cornish, 2020). In contrast, Sarwar et al. (2020) have proposed that regret from unwarranted spending, collective mentality, and dismissed opportunity is a key factor that sabotages repurchase intention. The possibility of this recurring and disconfirming pattern, thus, holds extensive value as an outcome of panic buying. The concept of post-purchase regret has gained attention within the field of brand-based satisfaction (Davvetas & Diamantopoulos, 2018), but has rarely been acknowledged in the context of panic behaviour. Other scholars have even focused on fashion consumption to validate the importance of reckless decision-making in subsequent remorse (Grigsby et al., 2020). Actions of impulsivity are viewed as a demonstration of luxury, where one’s wants override his/her needs in influencing consumption decisions. The subsequent emotions of guilt and repentance emerge from overlooking the exigency of the purchased products. Redirected to crisis or disaster situations, Prentice et al. (2020) proposed panic buying as a two-edged blade having both consequences of perceived security in view of heightened preparedness and regret due to unplanned spending. Where consumers’ needs are concerned, Grigsby et al. (2020) explained how unplanned purchasing leads to subsequent emotional distress through the cognitive response of foresight, which extends the benefits of purchased goods to future satisfaction over short-term gratification (e.g., if I don’t use it now, I will have it for a later need). Therefore, post-purchase regret can be seen in a situational limelight, under which perceived needs exceed temperamental desires. Since this area has been potentially neglected, the sixth hypothesis was stipulated as:H6: Panic buying during the COVID-19 outbreak has a significant impact on post-purchase regret. The Moderating Effect of Impulsivity In consumer research, the concept of impulsivity has been defined as an individual’s implicit inclination which directly encourages his or her tendency to react instantly without cautious planning and genuine consideration (Beatty & Ferrell, 1998). Impulsivity among consumers can be examined through three diverse perspectives, namely psychological impulsivity, behavioural impulsivity, and process impulsivity (Huang & Kuo, 2012). Psychological impulsivity is manifested by consumers when they have the urge for monetary consumption. This motivates the act of stockpiling among consumers who otherwise withhold from such practices (i.e., sudden change in consumption behaviour) (Rook, 1987). Behavioural impulsivity is manifested when consumers make unusual or poor choices through unplanned purchases that deviate from rationality. Process impulsivity refers to the pattern exhibited by an individual during their decision-making process that varies based on a particular information search pattern and situational assessment. Being closely associated with consumption behaviour, impulsivity has been shown to act with sentimental compulsion and limited cognitive regulation in promoting compulsive practices (Williams & Grisham, 2012). Similar results were obtained by Wu et al. (2021), who also shed light on the potential formation of post-purchase regret through compulsive behaviour. In particular, a higher degree of impulsivity was found to stimulate an increased level of regret following compulsive purchases. A multitude of studies have also supported the impact of impulsivity on post-consumption regret (Mahmood et al., 2019). However, Lin et al. (2009) failed to establish the significance of post-purchase negativity (regret and guilt) in the display of glee and pleasure following decisions made under shallow consideration. Nonetheless, impulsivity has been explored as a moderator between sensibility and regret and the decision-making process, despite its direct influence on consumers’ negative perceptions following unplanned consumption (Sokić et al., 2020). It has also been frequently used to moderate the effects of adverse situations on compulsive behaviours (e.g., gaming addiction, episodic drinking) (Hu et al., 2017; Kaltenegger et al., 2019). Lim et al. (2020) further underlined the role of impulsivity in manoeuvring both consumers’ ease of consumption (a positive outcome) and product return (a negative outcome). In the context of crisis and disaster, regular consumption is often overshadowed by panic buying due to herd mentality, social media communication, receptive anxiety, and the cost and availability of required necessities (Gazali, 2020). In this situation, the evident consequence of unplanned purchases exploits impulsive decision-making. Unplanned purchases are shaped upon three main subsets of impulsivity, namely a sense of necessity, indecisiveness, and unclear expectations (Billieux et al., 2008). Yet, its position as a moderating factor between both unplanned consumption and post-purchase regret, specifically during crisis moments, has been underexplored. To fill this gap, it was hypothesised that:H7: Impulsivity moderates the relationship between panic buying during the COVID-19 outbreak and post-purchase regret. Theoretical framework of the current study is presented in Fig. 1.Fig. 1 The research model Research Methodology Questionnaire Development Fear of the COVID-19 pandemic was operationalised using a seven-item scale adapted from Ahorsu et al. (2020) and Mertens et al. (2020), which reflects the level of fear and worry experienced by respondents during the pandemic. The measurement scale for perceived risk associated with the COVID-19 pandemic was adapted from past studies (e.g., Imai et al., 2005; Koh et al., 2005) as well. This construct was operationalised using a six-item scale that assesses respondents’ overall perception of risk associated with COVID-19. Subsequently, items for perceived scarcity were modified from the work of Gupta and Gentry (2016, 2019) to measure respondents’ perception of the scarcity of groceries when shopping during the COVID-19 Movement Control Order (MCO). Next, WOM communication was operationalised as a five-item scale modified from Yangui and Hajtaïeb El Aoud’s (2015) study to indicate the influence of family and friends on respondents’ perception of COVID-19 and the stockout situation during the MCO. Items for social media communication were adapted from Yangui and Hajtaïeb El Aoud (2015) as well, to measure respondents’ use of and dependency on social media to learn about COVID-19 and the stockout situation during the MCO. Consumers’ impulsivity was measured based on three items adapted from the existing literature (Aragoncillo & Orus, 2018; Li et al., 2020a, b), to gauge the level of impulsivity of consumption among consumers during COVID-19. Gupta and Gentry’s (2019) four-item scale was employed to measure post-purchase regret after panic buying to operationalise respondents’ sense of regret after a grocery shopping trip during the MCO. Since measurements of panic buying remain limited to date (Islam et al., 2021; Lins & Aquino, 2020), a scale was developed based on the suggestions of Churchill (1979) and DeVellis (2003). First, we identified the possible dimensions and items relevant to the face value of panic buying. This was done by reviewing the existing literature in the marketing and psychology domains. Second, the foundation of panic buying derived from the literature review was further explored via qualitative inquiries. Specifically, focus group discussions were conducted to confirm the dimensions and items found in the literature and to generate items for the new scale. As many as 25 consumers were carefully selected for two focus groups. The focus group interviews successfully generated 21 items to capture the domain of panic buying. These items were subsequently reviewed by a panel of 10 experts comprising industry specialists, marketing researchers, and academics. Following the above process, all panel members were requested to review the representativeness, clarity, and relevance of the 21 items for panic buying. Their feedback resulted in the exclusion of three items, leaving 18 items for the scale. These 18 items were then pilot tested with a sample of 60 consumers, the results of which revealed that one item was unclear. This item was subsequently dropped from the scale, yielding a final 17-item scale for panic buying. The above processes established face validity for the panic buying scale. However, as suggested by DeVellis (2003), all 17 items were subjected to item purification processes (i.e., exploratory factor analysis and confirmatory factor analysis), which are presented in the data analysis section. Sampling and Data Collection The target respondents for this study were individuals who had purchased groceries during the MCO period in Malaysia from March 2020 to May 2020. In view of the movement restriction imposed by the Malaysian government during the MCO period, self-administered survey questionnaires were distributed to the respondents via social media and email. A total of 600 respondents were selected with the use of purposive sampling technique. To obtain reliable and justifiable responses, four criteria were imposed to determine eligible respondents. The criteria were: (1) the respondent must be a Malaysian; (2) he or she is responsible for buying groceries during the MCO period; (3) he or she must have a minimum of one active social media account; and (4) he or she is over 18 years of age. Only individuals who fulfilled all four requirements were qualified to be respondents in the present study. Of the returned questionnaires, only 547 were usable for data analysis, as 53 were excluded due to incomplete responses or doubtful response patterns (i.e., straight-lining and/or diagonal-lining responses). The data cleaning process to assess outliers, normality, and missing values further indicated that 25 observations had to be dropped from the data. As a result, the final data from 522 responses was retained for further analysis. A sample size of 522 is capable of representing a big population and was considered sufficient for the present study (Saunders et al., 2012). Moreover, from a statistical point of view, the minimum sample size required for the present study fulfilled the suggestion of Cohen (1988). Likewise, based on the suggestion from Faul et al. (2009), the outcome of the GPower software confirmed that the sample size of 522 satisfied the minimum requirement of 132 samples at the 95% power level with an effect size of 0.15. Based on the evidence above, it can be affirmed that the final sample size of 522 was considered adequate for this study. Common Method Bias According to Podsakoff and Organ (1986), common method bias (CMB) is a methodological issue associated with bias in the estimation of constructs’ relationships, often due to the use of a single method in data collection. Artificial inflation from CMB influences the relationships among the constructs, which affects the validity and reliability of the measures (MacKenzie & Podsakoff, 2012; Podsakoff & Organ, 1986). To address CMB, MacKenzie and Podsakoff (2012) recommended that both procedural and statistical remedies should be applied. For the procedural remedy, we followed the suggestions in the literature (e.g., MacKenzie & Podsakoff, 2012; Podsakoff et al., 2012) by: (1) including detailed research information in the questionnaire’s cover sheet and (2) conducting a pre-test and pilot-test for the questionnaire. These steps were performed to alleviate any uncertainty associated with the questionnaire. As for the statistical remedy, Harman’s Single Factor test was conducted to address the issue of CMB. According to past literature (e.g., Malhotra et al., 2006), CMB is not an issue if (1) the first factor derived from the factor analysis has a variance less than 40 percent, and (2) the hypothesised model (with all the items modelled as a single factor) is not fit. In the present study, the first factor in the factor analysis had a variance of 29.45 percent (< 40%) and the hypothesised model of single factors was not fit; thus, CMB was not a problem in the present study. Sample Characteristics Table 1 presents the demographic characteristics of the respondents who participated in this study. The respondents consisted of 54.2 percent of females while the rest were male. A majority of the respondents were married (79.7%) and held a bachelor’s degree (63.6%). In addition, most of the respondents worked in executive/managerial positions (29.9%) and production/manufacturing positions (24.5%). In terms of shopping for groceries, many of them went shopping between four to six times per week and a majority of them spent RM 2001 to RM 4000 on groceries per month.Table 1 Respondents’ demographic profile Variables Descriptions Percentage Gender Female 54.2% Male 45.8% Marital status Married 79.7% Single 18.0% Divorced 1.1% Widowed 0.8% Others 0.4% Educational level Primary school 1.1% Secondary school 1.7% Diploma/higher diploma 13.2% Bachelor’s degree 63.7% Master’s degree 18.8% Doctorate degree 1.5% Employment Professional position 14.6% Production/Manufacturing position 24.5% Business Proprietors/Self-employed 17.8% Executive/Managerial position 29.9% Clerical/Administrative/Secretarial 7.3% Retiree/Not in the work force 4.6% Unemployed 1.1% Others 0.2% Weekly frequency of grocery shopping 1—3 times 36.2% 4—6 times 52.3% 7 – 9 times 9.0% More than 10 times 2.5% Expenses on groceries per month Less than RM 2000 14.2% RM 2001to RM 4000 47.1% RM 4001 to RM 6000 18.8% RM 6001 to RM 8000 12.1% RM 8001 to RM 10000 6.1% RM 10001 and above 1.7% Data Analysis and Findings Following the procedures suggested by Churchill (1979), an iterative scale purification procedure was conducted to develop a parsimonious scale for panic buying. According to Kim et al. (2012), the iterative scale purification procedure for developing a new measurement scale is commonly conducted via the use of item-to-total correlations analysis, reliability analysis, exploratory factor analysis, and confirmatory factor analysis. Items that are poorly corelated (r < 0.4) to the total score should be eliminated from a construct (Kim et al., 2012). The outcome of the item-to-total correlations analysis in the present study indicated that all the correlation values between items and the sum of their scores were above 0.40. Thus, none of the items in the panic buying construct were removed. Moreover, the reliability score (Cronbach’s alpha value) of the panic buying construct was 0.821, suggesting the scale’s high reliability. To address scale validity, the data was split into two sub-samples (each with 261 cases) for the purpose of exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). Exploratory Factor Analysis EFA was performed in this study using sample 1 (n = 261) to identify and explain the underlying dimensions of panic buying. EFA is assessed based on Bartlett’s test of sphericity, Kaiser–Meyer–Olkin (KMO), Eigenvalues, and measure of sampling adequacy. Based on the results (see Table 2), Bartlett’s Test (X2 = 3164.41) was significant (P-value < 0.001) and the value of KMO was above 0.50, indicating that the data was appropriate for an EFA. Subsequently, a principal component analysis through the Varimax approach revealed that the 17 items of panic buying can be further categorised into three multi-dimensional factors: anxiety, stress, and excessive buying. The three factors were identified and extracted from the construct of panic buying with eigenvalues of 6.210, 1.550, and 1.113 respectively, which exceeded the threshold value of 1.0. Moreover, the three-factor solution derived in the present study accounted for 60.46 percent of the variance.Table 2 Results of exploratory factor analysis (Sample 1) Code Items Anxiety Stress Excessive buying Anxiety (eigenvalue = 6.210; Variation = 39.814%; Cronbach’s Alpha = 0.848) AX1 I felt tensed when thinking or buying groceries 0.608 AX2 I felt pressured when thinking or buying groceries 0.700 AX3 I felt upset when thinking or buying groceries 0.712 AX4 I was worried about the possible shortage of groceries 0.763 AX5 I am worried I do not have sufficient groceries 0.713 AX6 I felt frightened when thinking of buying groceries 0.652 AX7 I felt confused with the rules imposed 0.522 AX8 I felt nervous when thinking or buying groceries* 0.417 Stress (eigenvalue = 1.550; Variation = 10.688%; Cronbach’s Alpha = 0.817) ST1 I felt things were not going my way when buying groceries 0.761 ST2 I have no control over my selection when buying groceries 0.832 ST3 I am not confident with my selection when buying groceries 0.674 ST4 I encountered difficulties in selecting groceries 0.633 ST5 I find myself confused with the selection of groceries that are available to me* 0.558 Excessive buying (eigenvalue = 1.113; Variation = 9.958%; Cronbach’s Alpha = 0.780) EB1 I bought more groceries than usual 0.836 EB2 I made more unplanned purchases than I needed to 0.718 EB3 I gave less consideration to the amount of groceries purchased 0.769 represents item that has been dropped from further analysis In addition, past studies have highlighted that items with low factor loadings (< 0.60) or cross-loadings should be dropped from the factor analysis (Hair et al., 2019; Kim et al., 2012). Referring to this guideline, two items from anxiety and one item from stress were dropped from further analyses. After the deletion of these items, reliability analysis (Cronbach’s alpha) was conducted on the three factors, reporting values well above the recommended value of 0.70 (anxiety = 0.848, stress = 0.817, and excessive buying = 0.780); therefore, all the factors underlying panic buying were reliable. Having completed the EFA, the following section presents results of the CFA for the measurement model. Confirmatory Factor Analysis CFA was employed to further examine the validity of the measurement items for all the constructs using sample 2 (n = 261). CFA is commonly used to examine the model fit of the measurement model and to address constructs’ convergent and discriminant validity. According to Hair et al. (2019), a model is considered to be fit when variance-based software (e.g., AMOS) reports: (1) normed Chi-square (χ2/df) less than 3.0; (2) root mean square error of approximation (RMSEA) less than 0.08; (3) goodness of fit (GFI) greater than 0.90; (4) parsimony normed fit index (PNFI) greater than 0.50; and (5) Tucker-Lewis index (TLI) above 0.90. In the present study, the CFA results indicated that the model had good fit (χ2/df = 1.597, GFI = 0.905, RMSEA = 0.034, PNFI = 0.809, and TLI = 0.951). Convergent validity for the constructs was assessed based on Hair et al.’s (2019) proposition that convergent validity is established if factor loadings for construct items are equal to or greater than 0.60, the average variance extracted (AVE) for each individual construct is larger than 0.50, and the composite reliability for each individual construct is greater than 0.70. The outcome of the CFA, as reported in Table 3, confirmed that all the measurement items for each respective construct had factor loadings higher than 0.60, while both AVE and composite reliability values for all the constructs were above the recommended value of 0.50 and 0.70, respectively. These findings suggested that the convergent validity of the data was achieved.Table 3 Results of convergent and discriminant validity (Sample 2) Items FL AVE CR MSV 1 2 3 4 5 6 7 8 PANICa 16 0.646 – 0.777 0.518 0.763 0.184 0.720b SCAR 7 0.683 – 0.781 0.504 0.876 0.177 0.421c 0.710 FEAR 7 0.615 – 0.828 0.533 0.888 0.184 0.429 0.220 0.730 RISK 6 0.663 – 0.803 0.513 0.840 0.031 0.175 0.035 0.089 0.717 SMEDIA 5 0.622 – 0.893 0.577 0.870 0.060 0.213 0.019 0.054 0.000 0.760 WOM 5 0.721 – 0.808 0.608 0.886 0.141 0.376 0.211 0.199 0.106 0.160 0.780 REGRET 4 0.731 – 0.808 0.551 0.829 0.128 0.346 0.210 0.078 0.076 0.097 0.230 0.742 IMPUL 3 0.671 – 0.826 0.563 0.793 0.128 0.242 0.185 0.069 -0.036 0.244 0.299 0.358 0.750 PANIC Panic buying, SCAR Perceived scarcity, FEAR Fear of the COVID-19 pandemic, RISK Perceived risk associated with the COVID-19 pandemic, SMEDIA Social media communication, WOM Word-of-Mouth communication, REGRET Anticipated post-purchase regret, IMPUL Consumers’ impulsivity, FL Factor loadings, AVE Average variance extracted, CR Composite reliability, MSV Maximum shared variance aSecond order construct bThe diagonal entries (in italics and bold) represent the squared root average variance extracted by the construct cThe off-diagonal entries represent the variance shared between constructs The discriminant validity of the constructs was assessed based on guidelines in the literature (e.g., Fornell & Larcker, 1981; Hair et al., 2019). According to Fornell and Larcker (1981), discriminant validity is evaluated by comparing (1) the variance of the constructs with the square roots of AVE and (2) AVE with Maximum-Shared-Squared-Variance (MSV). Discriminant validity is established if the variance shared between any two constructs is lower than the squared root of AVE for each construct, and if the MSV values for all constructs are smaller than AVE. As shown in Table 3, the values of the squared AVE (diagonal entries in italic and bold) were greater than the values of the correlation (off-diagonal entries), whereas the values of AVE for all the constructs were greater than their respective MSV. As such, discriminant validity was established for all the constructs in this study. Structural Model and Hypothesis Testing The present study employed the structural equation modelling (SEM) technique to examine the structural model and test the hypotheses. According to Hair et al. (2019), SEM is a powerful statistical technique capable of examining the strengths of individual causal paths proposed in the hypotheses. The analysis indicated that the structural model, as presented in Fig. 1, was found to have good fit (χ2/df = 1.621, GFI = 0.908, RMSEA = 0.035, CFI = 0.956, PNFI = 0.822, and TLI = 0.952). Next, the results of the path analysis, presented in Table 4, showed that all the direct hypotheses in this study (H1 to H6) were significant. It was revealed that fear of the COVID-19 pandemic (β = 0.290, p < 0.001), perceived risk associated with the COVID-19 pandemic (β = 0.124, p < 0.05), perceived scarcity (β = 0.311, p < 0.001), WOM (β = 0.225, p < 0.001), and social media (β = 0.159, p < 0.001) have a significant influence on consumers’ panic buying of groceries during the MCO imposed in Malaysia. Additionally, consumers’ panic buying was found to have a significant impact on their regret after a panic purchase (β = 0.356, p < 0.001).Table 4 Results of path analysis Hypothesised path Standardised estimate (β) Critical ratio Hypothesis H1: Fear of COVID-19 pandemic → Panic buying 0.290 5.728* Yes H2: PRisk → Panic buying 0.124 2.676* Yes H3: Perceived scarcity → Panic buying 0.311 6.189 Yes H4: WOM → Panic buying 0.225 4.614 Yes H5: SMEDIA → Panic buying 0.159 3.444 Yes H6: Panic buying → Anticipated post-purchase regret 0.356 6.517 Yes PRisk Perceived risk associated with the COVID-19 pandemic, SMEDIA Social media communication, WOM Word-of-Mouth communication ** and * denote significant at 99% and 95% confidence level respectively The moderating effect of consumers’ impulsivity on the relationship between panic buying and post-panic-purchase regret was assessed using SPSS PROCESS macro, developed by Hayes (2013). Hayes (2013) argued that a moderation effect exists if the interaction term is significant in the regression analysis generated by the PROCESS macro. As indicated in Table 5, the results of the regression test showed that consumers’ impulsive consumption (β = 0.154, p < 0.001) moderates the relationship between panic buying and regret after a panic purchase. Subsequently, the moderating effect of consumers’ impulsivity was plotted in a graph (Cham et al., 2022b; Cheah et al., 2020), shown in Fig. 2. The interpolation line shows consumers with greater impulsivity in grocery purchasing during the MCO compared to those with lower impulsivity. Specifically, the results indicated that panic buying is more strongly associated with post-purchase regret for consumers who purchased groceries impulsively. In view of this finding above, we concluded that H7 was supported.Table 5 Results of moderation analysis β SE t-value LLCI ULCI Outcome variable = Anticipated post-purchase regret, R2 = 0.156, F = 31.904 Constant 4.645 0.024 195.951* 4.598 4.691 Panic buying 0.317 0.058 5.459* 0.203 0.430 Impulse consumption 0.245 0.036 6.763* 0.174 0.316 Interaction 0.154 0.0345 4.410* 0.085 0.222 Interaction Panic purchase X Impulse consumption, β Standardised beta, SE Standard Error, LLCI Low limit confidence interval, ULCI Upper limit confidence interval * = p < 0.001, Bootstrap sample size = 5000 Fig. 2 The moderation effect of impulsive consumption on the relationship between panic buying and anticipated regret Discussion With the current study being set to examine the impact of personal and social factors on panic buying, and the ensuing influence on post-purchase regret by moderation of impulsivity amidst the COVID-19 pandemic, obtained results have primarily demonstrated the significance of personal factors, enclosing the components of fear, perceived risk and perceived scarcity, towards anxiety-driven consumptions. Mirroring the reported findings by Clemens et al. (2020) and Laato et al. (2020), predictive ability of such personal factors is vigorously underscored on the instantaneous consumption urges of necessities during period of uncertainty. The occurrence is understandably generated through consumers’ active judgment of a situation’s severity regarding the perceived scarcity of common items, future concerns, unfavourable emotions and societal pressure (Yuen et al., 2020). Following the transformative lifestyles as experienced by individuals throughout ravage of the global pandemic, regulatory enforcements with the like of movement control and severe healthcare risk have inevitably encouraged unanticipated consumption patterns, whist reflecting their perceptions towards seriousness of the situation. On an extreme, consumption and loyalty are prompted by the yearn for social companionship through elevated degrees of emotional attachment, approval and public knowledge (Addo et al., 2020). On another extreme, sense of relief is sought for aroused anxiety and worries as generated from communicated information on market scarcity by mean of impulsive purchases and stockpiling (Guo et al., 2017). The extent of emotional discomfort as developed from the consumers’ personalized dissemination of encountered adversity would, therefore, act as a gauge to their unplanned shopping frenzies. On similar note, significance of social factors, comprised of WOM and social media towards panic purchase has also been confirmed within the current study. Supported by both Depoux et al. (2020) and Ahmad and Murad (2020), shared information via social platforms possesses considerable influence in shaping consumers’ perceptions regarding such adversity in view of its constant escalation which overshadowed severity of the actual situation. Whereas, such discovery has overturned the proposition by Naeem (2020a) on the associations between social media information, pandemic-oriented beliefs and health-related precautions. Beyond negative cognitions one holds regarding uncertain circumstances, their unplanned consumptions are further maneuverer by societal pressures, communicated news and virtual persuasiveness. The findings fundamentally complete the disposition of Yuen et al. (2020) for their sole emphasis on the significant repercussion of personal factors. Both buzzes and social media communication, therefore, complement autoreactive communications in promoting precautionary and stockpiling attempts facing the global pandemic without regarding their legitimacy and integrity (Liu et al., 2021; Yasir et al., 2020). Behavioural adoption ensues following generated sense of urgency the endorsed communications. Reflectively addressed within the norm of social collectivism (i.e., I need toilet paper when everyone needs toilet paper, vice versa), an integrated outlook is congruently proposed between both personal and social factors. Aggregated relevance of individual perceptions and societal persuasions concerning the pandemic, thus, fulfilled the initial objective of this study in empowering increased understanding of the marketplace during interim of uncertainty, whilst proposing a counterargument towards the contrasted outlooks by both Loxton et al. (2020) and Yuen et al. (2020) on the lack of convergence between both uncertainty-oriented and inclusivity-based panic consumptions, Subsequent findings then disclosed emotional distress in an anticipated feeling of regret as the direct resultant of panic consumption. Similar phenomenon has been exemplified in the study by Saleh’s (2012) with indicating monetary consumption which invested minimal consideration effort as a proven antecedent to post-purchase regret. In the preference for individual centrism over social concerns, impulsive consumption due to the feeling of anxiety has shown to entail regrets as a result of emotional distress and depression (Gallagher et al., 2017). Such association prevails as an uncanny reflection of the current context, following the contradictory aftereffects of unplanned purchases and stockpiling of common items without sufficient informational-based assessment on severity of the adversity (e.g., overstocking of easily expired goods or better financial options during the pandemic interval). With Wang et al. (2019) highlighting the positive correlation between impulsive consumption and post-purchase evaluation, the degree of the intelligence one invested to rationalise the purchase during such limited interim would seemingly prevail as the determinant which diverged regrets and satisfaction. However, finite information concerning the situation further limits the ability for thorough financial rationalisation. Whilst the “bandwagon effect” remains intact alongside personalized judgment of the scenario, consumers’ cognitive preparations in experiencing such negative emotions from their haphazard consumptions would be of utmost critical. Obtained results further acknowledged impulsivity as a substantiated moderator in bridging panic consumption to post-purchase regret. Such association is comparable to the study by M’Barek and Gharbi (2011) alongside other regulatory factors including temporal outlook, product confidence, perfectionism, perceived satisfaction, level of uncertainty avoidance, and market demographics within the area of consumption-oriented regret. Unlike previous studies which essentially examined impulsive consumption in the independent position (Fenton‐O'Creevy et al., 2018; Santini et al., 2019), significance of the variable in the moderating role has validated its importance in channelling unfavourable judgment concerning a specified purchase. Such can be explained through the exposition by Sokić et al. (2020) with constructed influence of impulsivity on the relationship between emotional expression and regret being directly emanated by shortfalls in both expected outcomes and the failure for extensive consideration of other existing alternatives. Integrating the underlying reasoning between direct correlation of panic purchase and post-purchase distress, level of impulsivity, thus, stands as a gauge of the latter (i.e., the more unplanned I am at the purchase, the more regretted I am). Embracing both the society’s bandwagon and self-evaluated consumption blueprint, this, yet again, relies heavily on the amount of efforts invested towards consideration of the purchase pending actual behavioural endeavour during period of disastrous conditions. Theoretical Implications The current research model has illustrated consumers’ decision-making process in its totality. Investigations of panic consumption have proven it to be different from regular consumption. As in the former, consumers seemingly possess a weaker foothold in both exploratory endeavours and the assessment of available alternatives. Rather, disruptions from the pandemic entail dependency on existing, yet, limited situational information, with finite resources (e.g., time) available to thoroughly evaluate existing choices and engage in well-thought consumption. Presented findings, therefore, illustrate direct congruence to the founded paradigm of CCT, with having compensatory behaviours in forms of unplanned purchases as the direct consequent of emotional distress entailed by circumstances of vagueness and uncertainty (Arafat et al., 2021; Barnes et al., 2021). Facing disruptive input to an otherwise homeostasis condition, explored situation mirrors the unavoidable plight as acknowledged by Eysenck et al. (2007) with requiring preparatory and precautionary measures to counteract an encountered cognitive discomfort. In this regard, this study has empirically emphasised the role of both personal and social factors in buyers’ product purchase decisions during a crisis. Moreover, it has divulged the impact of impulsivity on post-purchase reviews. As there is a tendency for changed purchasing decisions upon receiving additional information during emergency situations, intention is less convincing compared to actual consumption behaviour. In this case, the sequential flow of the decision-making process merely gains partial support for both habitual consumption (i.e., accustomed consumption patterns) and panic buying, awarding less importance to both information search and evaluation of alternatives. Revisiting the study’s second objective for unearthing potential measurement variables of panic consumptions, alternative assessments regarding consumptions amid similar situations, thus, driven noteworthy adoption of the currently confirmed personal and social components. Another theoretical implication pertains to the urgent nature of panic purchasing behaviour based on Maslow’s hierarchy of needs, which shows that people are motivated to pursue higher needs along the hierarchical pyramid (Loxton et al., 2020). Spending on necessities over leisure (needs outweigh wants) is a phenomenon described by Lester (2013) as the fulfilment of components most important for survivability. This explains consumers’ urge for panic buying during a crisis to gratify their basic needs (physiological and safety needs) while setting aside psychological and self-fulfilment needs. Gupta and Gentry (2019) also elaborated that the impulsivity to possess items perceived as scarce is prompted by vulnerability, uncertainty, and reduced control of some aspects of life, which drives people to regain feelings of security, readiness, and satisfaction. This study’s findings thereby contribute to the application of Maslow’s hierarchy by underscoring consumers’ sense of desperation amidst the crisis due to personal and social factors. The obtained knowledge illustrates the nature of panic consumption, particularly under circumstances where consumers’ perceptions are formed through personal and surrounding influences rather than the actual situation. Practical Implications Interpretations of panic buying indicate it is more psychological than behavioural. With this being said, Hall et al. (2020) showcased a shift in consumption following the COVID-19 outbreak, exhibiting increased demand for lasting commodities and less service-based expenditure. In view of crisis preparation, short-term stockpiling purchases surfaced as a mirror trend upon realising the severity of the outbreak, which results in peaked short-term consumption (Arafat et al., 2020a, b). Assuming that consumers’ overreactions during crises is solely a result of impulsivity would have been misleading. Rather, there remain aspects of rationality prior to actual consumption, despite the process being relatively reckless. A possible explanation is given by Chen et al. (2020) in terms of containment (i.e., avoiding the potential worsening of the pandemic) following decreased consumption corresponding to gravity of the outbreak, with observable increase in spending prior and after the MCO. Situation-based consumption is understandably driven by a combination of social communication and individual factors to make necessary preparatory purchases. Personal factors aside, well-managed communication through existing touchpoints would hold substantial weight in forming cognitive perceptions of the situation (either to reflect or distort reality), thereby indirectly moulding consumers’ outlook of panic buying. This study, thus, sheds light on relevance of the “bandwagon effect”, where the perceived severity of a crisis can be manipulated through received messages; therefore, extensive administrative efforts must be invested to manage communication during a crisis. On this note, attention should be given to anticipated regret as a direct consequence of panic consumption. Gallagher et al. (2017) outlined that though buyers’ considerations of their own wellbeing outweigh that of the society, their anxiety-driven panic buying is followed by regret due to emotional distress and depression. Both perceptions of crisis severity and post-purchase regret are negative cognitions, for which panic buying has emerged as a means of relief (Yuen et al., 2020). However, it has also revealed consumers’ greater tolerance of consumption-related regret due to unplanned purchases. Moreover, marketing efforts undertaken by organisations to gain a market following are meaningless under the circumstance of panic buying. The need for necessities over luxuries to fulfil physiological and safety needs entail temporary neglect of other needs within Maslow’s hierarchy (Lester, 2013). In other words, products would serve identical purposes regardless of the purchased brand. Nevertheless, motivations to achieve other needs in the hierarchy would potentially regenerate post stabilisation of the situation, thereby offering a solid forecast for massive recovery and revenge spending. Ultimately, blame can be assigned to environmental uncertainties. As stated by Li et al., 2020a, b), perceived control, actual situational severity, and personalised materialism are significant predictors of unplanned purchases during the outbreak. Recognising panic buying as a means of assurance, an increased sense of control would entail greater considerations towards monetary spending, further reducing impulsivity. Often, cases of misguided consumption can also be resolved through the availability of continual directives (Lehmann et al., 2019). The importance of source credibility, proficiency, and reliability towards inducing consumption behaviour should not be neglected (Hu et al., 2019). All the more so when it comes to impulsive purchase (Mahmood et al., 2019). Particularly in periods of ambiguity, the need for announcements that promote stability (e.g., continuous availability of daily consumables) should not be neglected to prevent over-stockpiling that topples the balanced distribution of vital commodities. Dulam et al.’s (2020) proposition is, therefore, recommended, which is to implement a periodical quota policy to combat crisis situations. This policy grants benefits in terms of equalised product distributions and the effective satisfaction of market demands among sellers and retailers. Limitations and Future Research Directions While this study has highlighted the multifaceted nature of the current societal phenomenon in the COVID-19 pandemic, we must acknowledge its limitations. Primarily, the paper remains a cross-sectional study on the short-term implications of panic buying and emotional gratification. The results obtained seemingly neglect the potential long-term influence of present consumption to focus on immediate cognitive judgment across the span of several months. Habitual changes over temporary behavioural adaptations carry greater practical value; as such, the long-term impact of the pandemic on consumers’ perceptions should be assessed in terms emotions and lifestyle related to panic consumption. This research also narrowed its scope to several social and personal factors, leaving other potential variables unexplored. Apart from the controllable factors of consumers’ purchase decisions studied in this paper, uncontrollable factors (e.g., financial and geographical characteristics) may likely exert an influence on the subject matter. Future research can, thus, extend upon sociocultural impacts on situational consumerism from the latter perspective. Additionally, this study adopted a non-probability sampling approach in the selection of its participants, which may have alienated possible consumer segments. Since individuals exercise crisis-situation behaviours in accordance with their demographic attributes (e.g., income level and lifestyle), probability sampling can present more representative results on the total population. Last but not least, the significance of impulsivity as a moderator between actual consumption behaviour and subsequent emotional responses leaves room for the investigation of other potential moderators associating both the variables. Times of crisis understandably drive the stockpiling of commodities (e.g., food and drinks, sanitary appliances) over luxuries; as such, there remains the possibility of impulsivity playing a peripheral role in other factors like the usefulness and relevance of purchased items. Amidst the pandemic, further studies can be conducted in this regard to producing results that guide mindful consumption decisions. Conclusion The present study is one of the few to prevail as a founding groundwork in uncovering a new measurement that provide more conclusive detail for panic buying. Such discovery has, therefore, contributed an alternative measurement to the scholastic evaluation of panic buying as a research variable. Expanding the works of Arafat et al., (2020a, b) and Yuen et al. (2020), the current study has also clarified the importance of both personal and social aspects of consumers’ endeavours for panic consumption during the COVID-19 outbreak. Earlier research have primarily placed emphasis on the rationale behind panic buying, overlooking the emotional repercussions of such impulsivity. Motivational factors aside, findings of the present study acknowledged the significance of post-purchase regret following pandemic-triggered preparatory consumption. We argue the relevance of pre-evaluated purchases, seeing that (1) the outbreak was, to some extent, unforeseeable and disruptive to otherwise habitual routines, and 2) there is a shortage of available resources for the comprehensive evaluation of available alternatives prior to crisis-related consumption. In addition, panic buying, unlike rationality and impulsivity, has a varying degree of influence on regret based on the degree of impulsivity within each purchase. Situational aspects have been inevitably taken into consideration, particularly in linking perceptions (i.e., worry, uncertainty, and misguided information) to overly frantic preparatory consumption. The phenomenon hereby highlights the notable complexity in balancing cases of pre-purchasing distress to well-thought consumption, without the anticipation of post-purchase regret. Thus, in minimising their anticipated regret, consumers have undergone adjustments to their short-term monetary investments from leisure to sustainability (Hall et al., 2020). Buyers would probably continue to endure impulsivity-generated regret when the situation is vague, though their emotional distress depends on cognitive perceptions of each penny spent. Acknowledgements We acknowledge our respondents or participants who have taken part in this research. Data Availability The datasets generated during and/or analysed during the current study are available from the corresponding author on a reasonable request. Declarations Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed Consent Informed consent was obtained from all individual participants included in this study. Conflict of Interest The authors declare that they have no conflict of interest. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. ==== Refs References Ahmad AR Murad HR The impact of social media on panic during the COVID-19 pandemic in Iraqi Kurdistan: Online questionnaire study Journal of Medical Internet Research 2020 22 5 e19556 10.2196/19556 32369026 Addo PC Jiaming F Kulbo NB Liangqiang L COVID-19: Fear appeal favoring purchase behavior towards personal protective equipment The Service Industries Journal 2020 40 7–8 471 490 10.1080/02642069.2020.1751823 Ahorsu, D. K., Lin, C. Y., Imani, V., Saffari, M., Griffiths, M. D., & Pakpour, A. H. (2020). The fear of COVID-19 scale: Development and initial validation. International Journal of Mental Health and Addiction, 1–9,. 10.1007/s11469-020-00270-8 Andrei AG Zait A Vătămănescu E-M Pînzaru F Word-of-mouth generation and brand communication strategy Industrial Management & Data Systems 2017 117 3 478 495 10.1108/IMDS-11-2015-0487 Andrew, S. (2020). The psychology behind why toilet paper, of all things, is the latest coronavirus panic buy. Available at: https://edition.cnn.com/2020/03/09/health/toilet-paper-shortages-novel-coronavirus-trnd/index.html. Accessed 10 Mar 2020. Arafat SMY Kar SK Marthoenis M Sharma P Apu EH Kabir R Psychological underpinning of panic buying during pandemic (COVID-19) Psychiatry Research 2020 289 113061 10.1016/j.psychres.2020.113061 33242817 Arafat SMY Kar SK Menon V Kaliamoorthy C Mukherjee S Alradie-Mohamed A Sharma P Marthoenis M Kabir R Panic buying: An insight from the content analysis of media reports during COVID-19 pandemic Neurology Psychiatry and Brain Research 2020 37 100 103 10.1016/j.npbr.2020.07.002 32834528 Arafat SM Kar SK Kabir R Possible controlling measures of panic buying during COVID-19 International Journal of Mental Health and Addiction 2021 19 6 2289 2291 10.1007/s11469-020-00320-1 32837417 Aragoncillo L Orus C Impulse buying behaviour: An online-offline comparative and the impact of social media Spanish Journal of Marketing - ESIC 2018 22 1 42 62 10.1108/SJME-03-2018-007 Armfield JM Cognitive vulnerability: a model of the etiology of fear Clinical Psychology Review 2006 26 6 746 768 10.1016/j.cpr.2006.03.007 16806621 Ataguba OA Ataguba JE Social determinants of health: The role of effective communication in the COVID-19 pandemic in developing countries Global Health Action 2020 13 1 1788263 10.1080/16549716.2020.1788263 32657669 Avery EJ Park S Perceived knowledge as [Protective] power: Parents' protective efficacy, information-seeking, and scrutiny during COVID-19 Health Communication 2021 36 1 81 88 10.1080/10410236.2020.1847438 33249853 Aydınlıoğlu Ö Gencer ZT Let me buy before I die! A study on consumers’ panic buying behaviours during the Covid-19 pandemic Electronic Turkish Studies 2020 15 6 139 154 Barnes SJ Diaz M Arnaboldi M Understanding panic buying during COVID-19: A text analytics approach Expert Systems with Applications 2021 169 114360 10.1016/j.eswa.2020.114360 Beatty SE Ferrell ME Impulse buying: Modeling its precursors Journal of Retailing 1998 74 2 169 191 10.1016/S0022-4359(99)80092-X Billieux J Rochat L Rebetez MML Van der Linden M Are all facets of impulsivity related to self-reported compulsive buying behavior? Personality and Individual Differences 2008 44 6 1432 1442 10.1016/j.paid.2007.12.011 Bryan BA Ye Y Zhang Je Connor JD Land-use change impacts on ecosystem services value: Incorporating the scarcity effects of supply and demand dynamics Ecosystem Services 2018 32 144 157 10.1016/j.ecoser.2018.07.002 Campbell N Sinclair G Browne S Preparing for a world without markets: Legitimising strategies of preppers Journal of Marketing Management 2019 35 9–10 798 817 10.1080/0267257X.2019.1631875 Camargo LR Pereira SCF Scarpin MRS Fast and ultra-fast fashion supply chain management: An exploratory research International Journal of Retail & Distribution Management 2020 48 6 537 553 10.1108/IJRDM-04-2019-0133 Chakravarty, S. (2020). Covid-19 leads to bicycle boom around the world, triggers problem of shortage. Available at: https://auto.hindustantimes.com/auto/news/covid-19-leads-to-bicycle-boom-around-the-world-triggers-problem-of-shortage-41592189831842.html. Accessed 15 Oct 2020. Cham TH Cheng BL Low MP Cheok JBC Brand image as the competitive edge for Hospitals in Medical Tourism European Business Review 2020 31 1 31 59 Cham, T.-H., Cheah, J.-H., Cheng, B.-L., & Lim, X.-J. (2021a). I Am too old for this! Barriers contributing to the non-adoption of mobile payment. International Journal of Bank Marketing, Forthcoming. Cham TH Cheng BL Ng CKY Cruising down millennials’ fashion runway: A cross-functional study beyond Pacific borders Young Consumers: Insight and Ideas for Responsible Marketers 2021 22 1 28 67 10.1108/YC-05-2020-1140 Cham Tat‐Huei Lim Yet‐Mee Sigala Marianna <>after‐service International Journal of Tourism Research 2022 24 1 140 157 10.1002/jtr.2489 Cham, Tat-Huei., Cheah, Hun-Hwa., Ting, Hiram, & Memon, Mumtaz Ali. (2022b). Will destination image drive the intention to revisit and recommend? Empirical evidence from golf tourism. International Journal of Sports Marketing and Sponsorship , 23(2), 385–409. Cheah J-H Ting H Cham TH Memon MA The effect of selfie promotion and celebrity endorsed advertisement on decision-making processes: A model comparison Internet Research 2019 29 3 552 577 10.1108/IntR-12-2017-0530 Cheah JH Memon MA Richard JE Ting H Cham TH CB-SEM latent interaction: Unconstrained and orthogonalized approaches Australasian Marketing Journal (AMJ) 2020 28 4 218 234 10.1016/j.ausmj.2020.04.005 Chen H Qian W Wen Q The impact of the COVID-19 pandemic on consumption: Learning from high frequency transaction data In AEA Papers and Proceedings 2020 111 307 311 10.1257/pandp.20211003 Cheung CMK Thadani DR The impact of electronic word-of-mouth communication: A literature analysis and integrative model Decision Support Systems 2012 54 1 461 470 10.1016/j.dss.2012.06.008 Churchill GA Jr A paradigm for developing better measures of marketing constructs Journal of Marketing Research 1979 16 1 64 73 10.1177/002224377901600110 Clemens KS Matkovic J Faasse K Geers AL Determinants of safety-focused product purchasing in the United States at the beginning of the global COVID-19 pandemic Safety Science 2020 130 104894 10.1016/j.ssci.2020.104894 32834513 Cohen J Statistical power analysis for the behavioral sciences 1988 2 Erlbaum Cornish LS Why did I buy this? Consumers' post-impulse-consumption experience and its impact on the propensity for future impulse buying behaviour Journal of Consumer Behaviour 2020 19 1 36 46 10.1002/cb.1792 Crockford S Thank God for the greatest country on earth: White supremacy, vigilantes, and survivalists in the struggle to define the American nation Religion, State and Society 2018 46 3 224 242 10.1080/09637494.2018.1483995 Davvetas V Diamantopoulos A “Should have I Bought the other One?” Experiencing regret in global versus local brand purchase decisions Journal of International Marketing 2018 26 2 1 21 10.1509/jim.17.0040 Dellarocas C The digitization of word of mouth: Promise and challenges of online feedback mechanisms Management Science 2003 49 10 1407 1424 10.1287/mnsc.49.10.1407.17308 Depoux, A., Martin, S., Karafillakis, E., Preet, R., Wilder-Smith, A. and Larson, H. (2020). The pandemic of social media panic travels faster than the COVID-19 outbreak. Journal of Travel Medicine 27(3), 1–2 DeVellis RF Scale development: Theory and applications 2003 2 Sage Dulam R Furuta K Kanno T Development of an agent-based model for the analysis of the effect of consumer panic buying on supply chain disruption due to a disaster Journal of Advanced Simulation in Science and Engineering 2020 7 1 102 116 10.15748/jasse.7.102 Eckert S Sopory P Day A Wilkins L Padgett D Novak J Noyes J Allen T Alexander N Vanderford M Gamhewage G Health-related disaster communication and social media: Mixed-method systematic review Health Communication 2018 33 12 1389 1400 10.1080/10410236.2017.1351278 28825501 Eger L Komárková L Egerová D Mičík M The effect of COVID-19 on consumer shopping behaviour: Generational cohort perspective Journal of Retailing and Consumer Services 2021 61 102542 10.1016/j.jretconser.2021.102542 Eriksson M Lessons for crisis communication on social media: A systematic review of what research tells the practice International Journal of Strategic Communication 2018 12 5 526 551 10.1080/1553118X.2018.1510405 Eysenck MW Derakshan N Santos R Calvo MG Anxiety and cognitive performance: Attentional control theory Emotion 2007 7 2 336 353 10.1037/1528-3542.7.2.336 17516812 Faul F Erdfelder E Buchner A Lang AG Statistical power analyses using G* power 3.1: tests for correlation and regression analyses Behavior Research Methods 2009 41 4 1149 1160 10.3758/BRM.41.4.1149 19897823 Fenton-O'Creevy M Dibb S Furnham A Antecedents and consequences of chronic impulsive buying: Can impulsive buying be understood as dysfunctional self-regulation? Psychology & Marketing 2018 35 3 175 188 10.1002/mar.21078 Finau G Tarai J Varea R Titifanue J Kant R Cox J Social media and disaster communication: A case study of cyclone Winston Pacific Journalism Review 2018 24 1 123 137 10.24135/pjr.v24i1.400 Forani, J. (2020). No more panic buying? StatCan says grocery spending is slipping back to normal levels. Available at: https://www.ctvnews.ca/health/coronavirus/no-more-panic-buying-statcan-says-grocery-spending-is-slipping-back-to-normal-levels-1.4934091. Accessed Aug 2020. Forbes, K. (2012). The “Big C”: Identifying and mitigating contagion. National Bureau of Economic Research (NBER), 1–42. Fornell C Larcker DF Structural equation models with unobservable variables and measurement error: Algebra and statistics Journal of Marketing Research 1981 18 3 382 388 10.1177/002224378101800313 Gallagher CE Watt MC Weaver AD Murphy KA “I fear, therefore, I shop!” exploring anxiety sensitivity in relation to compulsive buying Personality and Individual Differences 2017 104 37 42 10.1016/j.paid.2016.07.023 Gazali HM The COVID-19 pandemic: Factors triggering panic buying behaviour among consumers in Malaysia Labuan Bulletin of International Business and Finance 2020 18 1 84 95 Goel A Gupta L Social media in the times of COVID-19 Journal of Clinical Rheumatology 2020 26 6 220 223 10.1097/RHU.0000000000001508 32852927 Grigsby JL Jewell RD Campbell C Have your cake and eat it too: How invoking post-purchase hyperopia mitigates impulse purchase regret Marketing Letters 2020 32 1 75 89 10.1007/s11002-020-09536-6 Guo J Xin L Wu Y Nah FH Tan CH Arousal or not? The effects of scarcity messages on online impulsive purchase International Conference on HCI in Business, Government, and Organizations 2017 Springer 29 40 Gupta S Gentry JW Construction of gender roles in perceived scarce environments–Maintaining masculinity when shopping for fast fashion apparel Journal of Consumer Behaviour 2016 15 3 251 260 10.1002/cb.1565 Gupta S Gentry JW ‘Should I Buy, Hoard, or Hide?’- Consumers’ responses to perceived scarcity The International Review of Retail, Distribution and Consumer Research 2019 29 2 178 197 10.1080/09593969.2018.1562955 Guzmán-González JI Sánchez-García FG Ramírez-de los Santos S Gutiérrez-Rodríguez F Palomino-Esparza D Telles-Martínez AL Worry and perceived risk of contagion during the COVID-19 quarantine in the Jalisco population: Preliminary study Salud Mental 2020 43 6 253 261 10.17711/SM.0185-3325.2020.035 Hair JF Black WC Babin BJ Anderson RE Multivariate data analysis 2019 8 Cengage Learning Hall, R. (2020). Coronavirus: Why people are panic buying toilet paper according to panic experts. Available at: https://www.independent.co.uk/news/world/americas/coronavirus-toilet-paper-panic-buying-covid-19-uk-australia-a9403351.html. Accessed Aug 2020. Hall MC Prayag G Fieger P Dyason D Beyond panic buying: Consumption displacement and COVID-19 Journal of Service Management 2020 32 1 113 128 10.1108/JOSM-05-2020-0151 Hamilton R Thompson D Bone S Chaplin LN Griskevicius V Goldsmith K Hill R John DR Mittal C O’Guinn T Piff P Roux C Shah A Zhu M The effects of scarcity on consumer decision journeys Journal of the Academy of Marketing Science 2019 47 3 532 550 10.1007/s11747-018-0604-7 Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis. Available at: http://www.personal.psu.edu/jxb14/M554/specreg/templates.pdf. Assessed 13 Aug 2020. He H Harris L The impact of Covid-19 pandemic on corporate social responsibility and marketing philosophy Journal of Business Research 2020 116 176 182 10.1016/j.jbusres.2020.05.030 32457556 Herman B Fauzi R Pongpanich S 48 hours public response to Corona epidemic status in Indonesia. Perceived risk and panic buying European Journal of Public Health 2020 30 5 996 10.1093/eurpub/ckaa067 32388567 Hidayanto AN Ovirza M Anggia P Ayuning Budi NF Phusavat K The roles of electronic word of mouth and information searching in the promotion of a new E-Commerce strategy: A case of online group buying in Indonesia Journal of Theoretical and Applied Electronic Commerce Research 2017 12 3 69 85 10.4067/S0718-18762017000300006 Hockey GRJ Compensatory control in the regulation of human performance under stress and high workload: A cognitive-energetical framework Biological Psychology 1997 45 1–4 73 93 10.1016/S0301-0511(96)05223-4 9083645 Hoque, M. Z., & Alam, M. N. (2018). What determines the purchase intention of liquid milk during a food security crisis? The role of perceived trust, knowledge, and risk. Sustainability, 10(10), 1–22. Hosen Mosharrof Ogbeibu Samuel Giridharan Beena Cham Tat-Huei Lim Weng Marc Paul Justin Individual motivation and social media influence on student knowledge sharing and learning performance: Evidence from an emerging economy Computers & Education 2021 172 104262 10.1016/j.compedu.2021.104262 Hu J Zhen S Yu C Zhang Q Zhang W Sensation seeking and online gaming addiction in adolescents: A moderated mediation model of positive affective associations and impulsivity Frontiers in Psychology 2017 8 699 10.3389/fpsyg.2017.00699 28529494 Hu X Chen X Davison RM Social support, source credibility, social influence, and impulsive purchase behavior in social commerce International Journal of Electronic Commerce 2019 23 3 297 327 10.1080/10864415.2019.1619905 Huang Y-F Kuo F-Y How impulsivity affects consumer decision-making in e-commerce Electronic Commerce Research and Applications 2012 11 6 582 590 10.1016/j.elerap.2012.09.004 Imai T Takahashi K Hoshuyama T Hasegawa N Lim MK Koh D SARS risk perceptions in healthcare workers, Japan Emerging Infectious Diseases 2005 11 3 404 15757555 Irshad M Ahmad MS Malik OF Understanding consumers’ trust in social media marketing environment International Journal of Retail & Distribution Management 2020 48 11 1195 1212 10.1108/IJRDM-07-2019-0225 Islam T Pitafi AH Arya V Wang Y Akhtar N Mubarik S Xiaobei L Panic buying in the COVID-19 pandemic: A multi-country examination Journal of Retailing and Consumer Services 2021 59 102357 10.1016/j.jretconser.2020.102357 Jezewska-Zychowicz M Plichta M Krolak M Consumers' fears regarding food availability and purchasing behaviors during the COVID-19 pandemic: The importance of trust and perceived stress Nutrients 2020 12 9 2852 10.3390/nu12092852 Jin X Li J Song W Zhao T The impact of COVID-19 and public health emergencies on consumer purchase of scarce products in China Front Public Health 2020 8 617166 10.3389/fpubh.2020.617166 33344410 Kaltenegger HC Laftman SB Wennberg P Impulsivity, risk gambling, and heavy episodic drinking among adolescents: A moderator analysis of psychological health Addictive Behaviors Reports 2019 10 100211 10.1016/j.abrep.2019.100211 31463359 Kaplan AM Haenlein M Users of the world, unite! The challenges and opportunities of Social Media Business Horizons 2010 53 1 59 68 10.1016/j.bushor.2009.09.003 Karapinar I Eru O Cop R The effects of consumers’ FoMo tendencies on impulse buying and the effects of impulse buying on post- purchase regret: An investigation on retail stores Broad Research in Artificial Intelligence and Neuroscience 2019 10 3 124 138 Kaur, A. and Malik, G. (2020). Understanding the psychology behind panic buying: A grounded theory approach. Global Business Review, forthcoming. Kay AC Eibach RP Compensatory control and its implications for ideological extremism Journal of Social Issues 2013 69 3 564 585 10.1111/josi.12029 Khorrami MS Esfidani MR Delavari S The effect of situational factors on impulse buying and compulsive buying: Clothing International Journal of Management, Accounting and Economics 2015 2 8 823 837 Kim JH Ritchie JB McCormick B Development of a scale to measure memorable tourism experiences Journal of Travel Research 2012 51 1 12 25 10.1177/0047287510385467 Koh D Takahashi K Lim MK Imai T Chia SE Qian F Fones C SARS risk perception and preventive measures, Singapore and Japan Emerging Infectious Diseases 2005 11 4 641 10.3201/eid1104.040765 15834989 Kotler P Marketing management 2006 12 Prentice Hall Kunimatsu MM Marsee MA Examining the presence of anxiety in aggressive individuals: The illuminating role of fight-or-flight mechanisms Child & Youth Care Forum 2012 41 3 247 258 10.1007/s10566-012-9178-6 Laato S Islam AKMN Farooq A Dhir A Unusual purchasing behavior during the early stages of the COVID-19 pandemic: The stimulus-organism-response approach Journal of Retailing and Consumer Services 2020 57 102224 10.1016/j.jretconser.2020.102224 Laros FJM Steenkamp J-BEM Emotions in consumer behavior: A hierarchical approach Journal of Business Research 2005 58 10 1437 1445 10.1016/j.jbusres.2003.09.013 Lazarus RS Folkman S Stress, appraisal, and coping 1984 Springer publishing company Lee GY Yi Y The effect of shopping emotions and perceived risk on impulsive buying: The moderating role of buying impulsiveness trait Seoul Journal of Business 2008 14 2 67 92 10.35152/snusjb.2008.14.2.004 Lehmann TA Krug J Falaster CD Consumer purchase decision: Factors that influence impulsive purchasing Revista Brasileira de Marketing 2019 18 4 196 219 10.5585/remark.v18i4.13345 Lester D Measuring Maslow's hierarchy of needs Psychological Reports 2013 113 1 15 17 10.2466/02.20.PR0.113x16z1 Li M Zhao T Huang E Li J How does a public health emergency motivate people's impulsive consumption? An empirical study during the COVID-19 outbreak in China International Journal of Environmental Research and Public Health 2020 17 14 5019 10.3390/ijerph17145019 Li Q Chen T Yang J Cong G Based on computational communication paradigm: Simulation of public opinion communication process of panic buying during the COVID-19 pandemic Psychology Research and Behavior Management 2020 13 1027 1045 10.2147/PRBM.S280825 33390730 Liang S Ye D Liu Y The effect of perceived scarcity: Experiencing scarcity increases risk taking The Journal of Psychology 2020 155 1 59 89 10.1080/00223980.2020.1822770 33048657 Lim XJ Cheah JH Cham TH Ting H Memon MA Compulsive buying of branded apparel, its antecedents, and the mediating role of brand attachment Asia Pacific Journal of Marketing and Logistics 2020 32 7 1539 1563 10.1108/APJML-03-2019-0126 Lins S Aquino S Development and initial psychometric properties of a panic buying scale during COVID-19 pandemic Heliyon 2020 6 9 e04746 10.1016/j.heliyon.2020.e04746 32895636 Lin S-P Shih H-C Huang Y-C Emotional states before and after impulsivity Social Behavior and Personality: An International Journal 2009 37 6 819 824 10.2224/sbp.2009.37.6.819 Liu H Liu W Yoganathan V Osburg VS COVID-19 information overload and generation Z's social media discontinuance intention during the pandemic lockdown Technological Forecasting and Social Change 2021 166 120600 10.1016/j.techfore.2021.120600 34876758 Liu, Q., Xie, Y., & Chang, S. (2018). Y-Generation Digital Natives' Impulsive Buying Behavior, 2018 3rd Technology Innovation Management and Engineering Science International Conference (TIMES-iCON). IEEE, 1–5. Liu W Yue XG Tchounwou PB Response to the COVID-19 epidemic: The Chinese experience and implications for other countries International Journal of Environmental Research and Public Health 2020 17 7 2304 10.3390/ijerph17072304 Lu, L., Lee, L., Wu, L., & Li, X. (2022). Healing the pain: does COVID-19 isolation drive intentions to seek travel and hospitality experiences?. Journal of Hospitality Marketing & Management, forthcoming Loxton M Truskett R Scarf B Sindone L Baldry G Zhao Y Consumer Behaviour during Crises: Preliminary Research on How Coronavirus Has Manifested Consumer Panic Buying, Herd Mentality, Changing Discretionary Spending and the Role of the Media in Influencing Behaviour Journal of Risk and Financial Management 2020 13 8 166 10.3390/jrfm13080166 Ma A Kay AC Compensatory control and ambiguity intolerance Organizational Behavior and Human Decision Processes 2017 140 46 61 10.1016/j.obhdp.2017.04.001 M'Barek, M., & Gharbi, A. (2011). The moderators of post purchase regret. Journal of Marketing Research and Case Studies, 2021, 1–16. MacKenzie SB Podsakoff PM Common method bias in marketing: Causes, mechanisms, and procedural remedies Journal of Retailing 2012 88 4 542 555 10.1016/j.jretai.2012.08.001 Maddux JE Rogers RW Protection motivation and self-efficacy: A revised theory of fear appeals and attitude change Journal of Experimental Social Psychology 1983 19 5 469 479 10.1016/0022-1031(83)90023-9 Mahmood K Rashid MA Hussain G Personality and post-purchase consumer regret experienced after impulse buying: Cross-theoretical approach with individual differences moderator International Transaction Journal of Engineering, Management, & Applied Sciences & Technologies 2019 10 12 1 17 Malhotra NK Kim SS Patil A Common method variance in IS research: A comparison of alternative approaches and a reanalysis of past research Management Science 2006 52 12 1865 1883 10.1287/mnsc.1060.0597 Mannelli C Whose life to save? Scarce resources allocation in the COVID-19 outbreak Journal of Medical Ethics 2020 46 6 364 366 10.1136/medethics-2020-106227 32277018 Marchand A Hennig-Thurau T Wiertz C Not all digital word of mouth is created equal: Understanding the respective impact of consumer reviews and microblogs on new product success International Journal of Research in Marketing 2017 34 2 336 354 10.1016/j.ijresmar.2016.09.003 Martin-Neuninger R Ruby MB What does food retail research tell us about the implications of coronavirus (COVID-19) for grocery purchasing habits? Frontiers in Psychology 2020 11 1448 10.3389/fpsyg.2020.01448 32581987 Mertens G Gerritsen L Duijndam S Salemink E Engelhard IM Fear of the coronavirus (COVID-19): Predictors in an online study conducted in March 2020 J Anxiety Disord 2020 74 102258 10.1016/j.janxdis.2020.102258 32569905 Mills MF Preparing for the unknown… unknowns: ‘doomsday’ prepping and disaster risk anxiety in the United States Journal of Risk Research 2018 22 10 1267 1279 10.1080/13669877.2018.1466825 Mitchell, V. W. (1992). Understanding consumers’ behaviour: Can perceived risk theory help?. Management Decision, 30(3), 26–31. Mou J Shin D Effects of social popularity and time scarcity on online consumer behaviour regarding smart healthcare products: An eye-tracking approach Computers in Human Behavior 2018 78 74 89 10.1016/j.chb.2017.08.049 Mukerjee K The impact of brand experience, service quality and perceived value on word of mouth of retail bank customers: Investigating the mediating effect of loyalty Journal of Financial Services Marketing 2018 23 1 12 24 10.1057/s41264-018-0039-8 Muralidharan S Rasmussen L Patterson D Shin J-H Hope for Haiti: An analysis of Facebook and Twitter usage during the earthquake relief efforts Public Relations Review 2011 37 2 175 177 10.1016/j.pubrev.2011.01.010 Naeem, M. (2020a). The role of social media to generate social proof as engaged society for stockpiling behaviour of customers during Covid-19 pandemic. Qualitative Market Research: An International Journal 24(3), 281–301. Naeem M Understanding the customer psychology of impulse buying during COVID-19 pandemic: Implications for retailers International Journal of Retail & Distribution Management 2020 49 3 377 393 10.1108/IJRDM-08-2020-0317 Nazione S Perrault E Pace K Impact of information exposure on perceived risk, efficacy, and preventative behaviors at the beginning of the COVID-19 pandemic in the United States Health Communication 2021 36 1 23 31 10.1080/10410236.2020.1847446 33183090 Nazri MA Omar NA Ramly SM Ab Hamid SN Mohd Hashim AJC Analysing the influence of perceived scarcity, negative feelings, and status consumption on food waste among consumers Journal of Environmental Treatment Technique 2021 9 1 33 36 New Straits Times. (2020). WHO COVID envoy fears third wave, calls Europe response 'incomplete'. Available at: https://www.nst.com.my/world/world/2020/11/641947/global-covid-19-cases-surpass-55-million-1328048-dead. Accessed Sept 2020. Nichols BS The development, validation, and implications of a measure of consumer competitive arousal (CCAr) Journal of Economic Psychology 2012 33 1 192 205 10.1016/j.joep.2011.10.002 Ozer L Gultekin B Pre- and post-purchase stage in impulse buying: The role of mood and satisfaction Journal of Retailing and Consumer Services 2015 22 71 76 10.1016/j.jretconser.2014.10.004 Parsad C Prashar S Vijay TS Sahay V Role of in-store atmospherics and impulse buying tendency on post purchase regret Journal of Business and Management 2019 25 1 1 24 Podsakoff PM Organ DW Self-reports in organizational research: Problems and prospects Journal of Management 1986 12 4 531 544 10.1177/014920638601200408 Podsakoff PM MacKenzie SB Podsakoff NP Sources of method bias in social science research and recommendations on how to control it Annual Review of Psychology 2012 63 539 569 10.1146/annurev-psych-120710-100452 21838546 Porter L Golan GJ From subservient chickens to Brawny Men: A comparison of viral advertising to Television advertising Journal of Interactive Advertising 2006 6 2 26 33 Prentice, C., Quach, S., & Thaichon, P. (2020). Antecedents and consequences of panic buying: The case of COVID‐19. International Journal of Consumer Studies 46(1), 132–146. Rook DW The buying impulse Journal of Consumer Research 1987 14 2 189 199 10.1086/209105 Saleh, M. A. E.-H. (2012). An investigation of the relationship between unplanned buying and post-purchase regret. International Journal of Marketing Studies , 4(4), 106–120. Santini FDO Ladeira WJ Vieira VA Araujo CF Sampaio CH Antecedents and consequences of impulse buying: A meta-analytic study RAUSP Management Journal 2019 54 2 178 204 10.1108/RAUSP-07-2018-0037 Sarwar, M. A., Awang, Z., Habib, M. D., Nasir, J., & Hussain, M. (2020). Why did I buy this? Purchase regret and repeat purchase intentions: A model and empirical application. Journal of Public Affairs, forthcoming. Saunders M Lewis P Thornhill A Research methods for business students 2012 6 Pearson Shimp TA Bearden WO Warranty and other extrinsic cue effects on consumers' risk perceptions Journal of Consumer Research 1982 9 1 38 46 10.1086/208894 Shorey S Ang E Yamina A Tam C Perceptions of public on the COVID-19 outbreak in Singapore: A qualitative content analysis Journal of Public Health (oxford, England) 2020 42 4 665 671 10.1093/pubmed/fdaa105 Simonson I Carmon Z Dhar R Drolet A Nowlis SM Consumer research: In search of identity Annual Review of Psychology 2001 52 249 275 10.1146/annurev.psych.52.1.249 11148306 Singh H Chakrabarti S Defining the relationship between consumers and retailers through user-generated content: Insights from the research literature International Journal of Retail & Distribution Management 2020 49 1 41 60 10.1108/IJRDM-03-2020-0080 Slack N Singh G Sharma S Impact of perceived value on the satisfaction of supermarket customers: Developing country perspective International Journal of Retail & Distribution Management 2020 48 11 1235 1254 10.1108/IJRDM-03-2019-0099 Sokić K Horvat Đ Martinčić SG How impulsivity influences the post-purchase consumer regret? Business Systems Research Journal 2020 11 3 14 29 10.2478/bsrj-2020-0024 Song W Jin X Gao J Zhao T Will buying follow others ease their threat of death? An analysis of consumer data during the period of COVID-19 in China International Journal of Environmental Research and Public Health 2020 17 9 3215 10.3390/ijerph17093215 Sorokowski P Groyecka A Kowal M Sorokowska A Białek M Lebuda I Dobrowolska M Zdybek P Karwowski M Can information about pandemics increase negative attitudes toward Foreign groups? A case of COVID-19 outbreak Sustainability 2020 12 12 4912 10.3390/su12124912 Suciu, P. (2020). Misinformation spreading faster than the actual Coronavirus. Available at: https://www.forbes.com/sites/petersuciu/2020/02/03/misinformation-spreading-faster-than-the-actual-coronavirus/#2e12cf7ecabd. Accessed August 2020. Sullivan D Landau MJ Kay AC Toward a comprehensive understanding of existential threat: Insights from Paul Tillich Social Cognition 2012 30 6 734 757 10.1521/soco.2012.30.6.734 Tai Z Sun T Media dependencies in a changing media environment: The case of the 2003 SARS epidemic in China New Media & Society 2007 9 6 987 1009 10.1177/1461444807082691 Tan K-L Sia JK-M Tang DKH To verify or not to verify: using partial least squares to predict effect of online news on panic buying during pandemic Asia Pacific Journal of Marketing and Logistics 2022 34 4 647 668 10.1108/APJML-02-2021-0125 Taylor S Landry CA Paluszek MM Fergus TA McKay D Asmundson GJG COVID stress syndrome: Concept, structure, and correlates Depression and Anxiety 2020 37 8 706 714 10.1002/da.23071 32627255 Teachman BA Stefanucci JK Clerkin EM Cody MW Proffitt DR A new mode of fear expression: Perceptual bias in height fear Emotion 2008 8 2 296 301 10.1037/1528-3542.8.2.296 18410203 Tucker T Online word of mouth: Characteristics of Yelp.com reviews The Elon Journal of Undergraduate Research in Communications 2011 2 1 37 42 Twedt DW Does the" 9 Fixation" in retail pricing really promote sales? Journal of Marketing 1965 29 54 UN International Strategy for Disaster Reduction (UNISDR) UNISDR terminology on disaster risk reduction 2009 UNISDR Wang C Pan R Wan X Tan Y Xu L McIntyre RS Choo FN Tran B Ho R Sharma VK Ho C A longitudinal study on the mental health of general population during the COVID-19 epidemic in China Brain, Behavior, and Immunity 2020 87 40 48 10.1016/j.bbi.2020.04.028 32298802 Wang HH Hao N Panic buying? Food hoarding during the pandemic period with city lockdown Journal of Integrative Agriculture 2020 19 12 2916 2925 10.1016/S2095-3119(20)63448-7 Wang L Yan Q Chen W Drivers of purchase behavior and post-purchase evaluation in the Singles’ Day promotion Journal of Consumer Marketing 2019 36 6 835 845 10.1108/JCM-08-2017-2335 Williams AD Grisham JR Impulsivity, emotion regulation, and mindful attentional focus in compulsive buying Cognitive Therapy and Research 2012 36 5 451 457 10.1007/s10608-011-9384-9 Wu Y Xin L Li D Yu J Guo J How does scarcity promotion lead to impulse purchase in the online market? A field experiment Information & Management 2021 58 1 103283 10.1016/j.im.2020.103283 Yangui W Hajtaïeb El Aoud N Consumer behavior and the anticipation of a total stockout for a food product: Proposing and validating a theoretical model The International Review of Retail, Distribution and Consumer Research 2015 25 2 181 203 10.1080/09593969.2014.951675 Yasir A Hu X Ahmad M Rauf A Shi J Ali Nasir S Modeling impact of word of mouth and E-Government on online social presence during COVID-19 outbreak: A multi-mediation approach International Journal of Environmental Research and Public Health 2020 17 8 2954 10.3390/ijerph17082954 Yoshizaki HTY de Brito Junior I Hino CM Aguiar LL Pinheiro MCR Relationship between panic buying and Per Capita income during COVID-19 Sustainability 2020 12 23 9968 10.3390/su12239968 Yuen KF Wang X Ma F Li KX The psychological causes of panic buying following a health crisis International Journal of Environmental Research and Public Health 2020 17 10 3513 10.3390/ijerph17103513 Zeelenberg M van Dijk WW Manstead ASR Reconsidering the relation between regret and responsibility Organizational Behavior and Human Decision Processes 1998 74 3 254 272 10.1006/obhd.1998.2780 9719654 Zheng Y Yang X Liu Q Chu X Huang Q Zhou Z Perceived stress and online compulsive buying among women: A moderated mediation model Computers in Human Behavior 2020 103 13 20 10.1016/j.chb.2019.09.012
PMC009xxxxxx/PMC9006085.txt	==== Front Interv Neuroradiol Interv Neuroradiol INE spine Interventional Neuroradiology 1591-0199 2385-2011 SAGE Publications Sage UK: London, England 35404161 10.1177/15910199221093896 10.1177_15910199221093896 Original Research Article Neutrophil–Lymphocyte ratio is associated with poor clinical outcome after mechanical thrombectomy in stroke in patients with COVID-19 https://orcid.org/0000-0003-4461-7005 Al-Mufti Fawaz 1 Khandelwal Priyank 2 Sursal Tolga 1 https://orcid.org/0000-0003-4737-3026 Cooper Jared B. 1 https://orcid.org/0000-0002-1952-4555 Feldstein Eric 1 https://orcid.org/0000-0002-8859-8574 Amuluru Krishna 3 Moré Jayaji M. 1 Tiwari Ambooj 4 Singla Amit 2 https://orcid.org/0000-0003-0131-5699 Dmytriw Adam A 5 Piano Mariangela 6 Quilici Luca 6 Pero Guglielmo 6 Renieri Leonardo 7 https://orcid.org/0000-0002-0432-5414 Limbucci Nicola 7 https://orcid.org/0000-0002-8024-4712 Martínez-Galdámez Mario 8 https://orcid.org/0000-0003-3351-668X Schüller-Arteaga Miguel 8 Galván Jorge 8 Arenillas-Lara Juan Francisco 8 Hashim Zafar 9 https://orcid.org/0000-0002-9616-1188 Nayak Sanjeev 9 Desousa Keith 10 Sun Hai 11 Agarwalla Pankaj K. 2 Sudipta Roychowdhury J 12 Nourollahzadeh Emad 12 Prakash Tannavi 2 Xavier Andrew R 13 Diego Lozano J 14 Gupta Gaurav 11 Yavagal Dileep R 15 Elghanem Mohammad 16 Gandhi Chirag D. 1 Mayer Stephan A. 1 1 Department of Neurosurgery, New York Medical College, 8138 Westchester Medical Center , Valhalla, New York, USA 2 Department of Neurological Surgery, University Hospital Newark, 12286 New Jersey Medical School , Rutgers, New Jersey, USA 3 Department of Neurointerventional Radiology, 178242 Goodman Campbell Brain and Spine , Indianapolis, Indiana, USA 4 Department of Neurology, Brookdale and Jamaica Hospital Center, 12297 NYU School of Medicine , Brooklyn, New York, USA 5 Neuroradiology and Neurointervention Service, 1861 Brigham & Women's Hospital , Harvard Medical School, Boston, Massachusetts, USA 6 Department of Neuroradiology, 9338 ASST Grande Ospedale Metropolitano Niguarda , Milan, Italy 7 Department of Radiology, Neurovascular Unit, Careggi University Hospital, Florence, Italy 8 Department of Interventional Neuroradiology, 16238 Hospital Clínico Universitario de Valladolid , Valladolid, Spain 9 Department of Radiology, University Hospital of North Midlands, Stoke-on-Trent, UK 10 Department of Neurology, 5799 Northwell Health , Long Island, New York, New York, USA 11 Department of Neurological Surgery, Rutgers Robert Wood Johnson Medical School, New Jersey Medical School, New Brunswick, New Jersey, USA 12 Department of Neurology & Radiology, 25044 Robert Wood Johnson University Hospital , Rutgers, New Jersey, USA 13 Department of Neurology, Saint Joseph Health, 2956 Detroit Medical Center , Detroit, Michigan, USA 14 Department of Radiology, 8790 University of California Riverside , Riverside, California, USA 15 Department of Neurology, Miller School of Medicine, Miami, Florida, USA 16 Department of Neurology, 12216 University of Arizona-Tucson , Tucson, Arizona, USA Fawaz Al-Mufti, MD, Westchester Medical Center, New York Medical College School of Medicine, 100 Woods Road, Macy Pavilion 1331, Valhalla, NY 10595, USA. Email: Fawaz.Al-Mufti@wmchealth.org 11 4 2022 11 4 2022 1591019922109389618 1 2022 15 3 2022 24 3 2022 © The Author(s) 2022 2022 SAGE Publications This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. Background The neutrophil–lymphocyte ratio (NLR) is emerging as an important biomarker of acute physiologic stress in a myriad of medical conditions, and is a confirmed poor prognostic indicator in COVID-19. Objective We sought to describe the role of NLR in predicting poor outcome in COVID-19 patients undergoing mechanical thrombectomy for acute ischemic stroke. Methods We analyzed NLR in COVID-19 patients with large vessel occlusion (LVO) strokes enrolled into an international 12-center retrospective study of laboratory-confirmed COVID-19, consecutively admitted between March 1, 2020 and May 1, 2020. Increased NLR was defined as ≥7.2. Logistic regression models were generated. Results Incidence of LVO stroke was 38/6698 (.57%). Mean age of patients was 62 years (range 27–87), and mortality rate was 30%. Age, sex, and ethnicity were not predictive of mortality. Elevated NLR and poor vessel recanalization (Thrombolysis in Cerebral Infarction (TICI) score of 1 or 2a) synergistically predicted poor outcome (likelihood ratio 11.65, p = .003). Patients with NLR > 7.2 were 6.8 times more likely to die (OR 6.8, CI95% 1.2–38.6, p = .03) and almost 8 times more likely to require prolonged invasive mechanical ventilation (OR 7.8, CI95% 1.2–52.4, p = .03). In a multivariate analysis, NLR > 7.2 predicted poor outcome even when controlling for the effect of low TICI score on poor outcome (NLR p = .043, TICI p = .070). Conclusions We show elevated NLR in LVO patients with COVID-19 portends significantly worse outcomes and increased mortality regardless of recanalization status. Severe neuro-inflammatory stress response related to COVID-19 may negate the potential benefits of successful thrombectomy. Neutrophil lymphocyte ratio COVID acute ischemic stroke large vessel occlusion edited-statecorrected-proof typesetterts19 ==== Body pmcIntroduction The current SARS-CoV-2 virus pandemic has affected 21 million individuals worldwide and has accounted for 760,000 deaths at the time of the writing of this paper. Coronavirus entry into the host cells is reportedly mediated via the angiotensin-converting enzyme ACE 2 receptor. The primary complication is acute lung injury resulting in type-1 respiratory failure, with a significant proportion of patients requiring intensive care. However, in addition to these respiratory features, the disease can affect multiple organs including the cardiovascular and gastrointestinal systems. Emerging evidence suggests neurotropism of the novel SARS-CoV-2 virus, which causes a wide range of neurologic complications.1–3 The neutrophil–lymphocyte ratio (NLR), a marker of acute physiological stress, is an established prognostic marker in patients with cancer, cardiac disease, sepsis, and acute neurological disorders such as subarachnoid hemorrhage and acute ischemic stroke.4–8 NLR has been demonstrated to have superior predictive value when compared to leukocyte count alone in predicting poor clinical outcomes in the aforementioned conditions.4–8 More recently, elevated NLR has been confirmed as a negative short-term prognostic indicator for patients with COVID-19.9,10 We sought to describe the role of NLR in predicting poor outcome in COVID-19 patients with acute ischemic stroke (AIS) due to large vessel occlusion (LVO) undergoing mechanical thrombectomy. We hypothesize that systemic inflammation and physiological stress reflected by an elevated NLR may outweigh any potential benefit of mechanical thrombectomy. We herein examine the relationship between NLR and vessel recanalization in COVID-19 patients treated with mechanical thrombectomy for LVO stroke. Methods Data availability statement The data that support the findings of this study are available from the corresponding author, upon reasonable request Patient selection We conducted an international multi-center retrospective study of laboratory-confirmed COVID-19 patients with acute LVOs consecutively admitted between March 1, 2020 and May 1, 2020 in 12 stroke centers located in the US, UK, Spain, and Italy. The ethical review boards of these institutions approved the study. As this was a retrospective study, waiver of informed consent was obtained. Data collection & analysis We tabulated the total number of hospitalized COVID-19 patients and obtained detailed information in all consecutive LVO cases within this cohort. Per the ASA/AHA guidelines, patients met criteria for large vessel occlusion strokes if they had a common carotid occlusion, internal carotid artery occlusion, middle/ anterior cerebral artery occlusion, basilar artery occlusion, or occlusions in multiple vascular territories. All patients who had LVO in our cohort underwent mechanical thrombectomy. Collected data included demographics, past medical history, baseline clinical status, imaging results, and details of stroke treatment and complications. Details related to the diagnosis of COVID-19 included clinical presentation, laboratory findings, and pulmonary CT findings on admission, treatment regimens, and clinical outcomes. The diagnosis criteria and classification of COVID-19 severity was defined by the American Thoracic Society criteria.11 Neutrophil (PMN) and lymphocyte counts were analyzed as percentages of the total WBC population. NLR was calculated as the ratio of the percentage of neutrophils over the percentage of lymphocytes. Outcome assessment The primary outcome measure was poor outcome, defined as death or discharge to a skilled nursing facility. Good outcome was defined as discharge to home or an acute rehabilitation facility. These definitions have been previously used in the literature, both as clinically significant definitions of outcome as well as helpful divisions for statistical analysis.12–14 Our secondary outcomes were relationships between NLR, TICI score, and mechanical ventilation. Statistical analysis Data analyses were performed with Stata statistical software (Stata Version 16.1, Statacorp LLC). P values of ≤ .05 were considered significant. Individual variables were analyzed using chi-square or t-test, followed by logistic regression. Univariate associations with a P < .05 were then tested further to assess in multivariable models to assess interactions and combined predictive capabilities. Initial analysis of outcome was performed using individual variables NLR, TICI, and mechanical ventilation, using an ordered logistic regression. Each individual variable was predictive of outcome. While other inflammatory biomarkers were ran on our multivariate model, such as white blood cell count, d-dimer, serum ferritin, C-reactive protein, fibrinogen, procalcitonin, and platelet count, only elevated NLR remained an individual predictor of outcome and mortality. Receiver operating characteristic (ROC) analysis was undertaken to determine if a particular cutoff for NLR, TICI, or ventilator days may be more helpful in predicting outcomes. compare the predictive value of NLR and the TICI score on outcome (Table 1, Figure 1). The NLR cutoff was determined to be > = 7.272 with 78.57% of patients correctly classified with this cutoff (AUC .7427). TICI score (TICI 0, 1, and 2a in one group and TICI 2b, 2c, and 3 in the other) correlated with outcome in 73.68% of cases (AUC .6462). There was no statistical significance between the AUC of the NLR and TICI score (AUC: .7427 vs. .6462; p = .51). Figure 1. ROC curves for NLR and TICI: comparative predictive value of NLR and TICI score on outcome compared to reference curve. Table 1. ROC curves for NLR and TICI. Obs ROC Area Std. Error [95% Confidence Interval] NLR 28 0.74 0.11 0.54–0.95 -TICI 28 0.65 0.12 0.42–0.89 P = 0.51. After cutoffs were identified on ROC curve, outcome was converted to binary where variables were still predictive in a logistic regression. Dichotomous outcomes helped to achieve sufficient statistical significance for analysis. Results Demographic and clinical features Out of a total of 6698 patients admitted with COVID-19 to 12 stroke centers during the study period, the incidence of LVO stroke was 38/6698 (.57%). Mean age of the patients with LVO stroke was 62 years (range 27–87), 50% were female, and the hospital mortality rate was 29% (11/38). Of these 38 patients, 28 had NLR measured on admission and were included in the subsequent analysis. There were no major differences between LVO patients with or without an admission NLR with regard to demographics, stroke characteristics, or mortality (Table 2). Eight of 38 patients (21.1%) had internal carotid artery-middle cerebral artery tandem occlusions. Three out of nine patients with NLR >7.2 had tandem occlusions (p = .11). Table 2. comparison of patients with and COVID-19 With and without admission NLR data. Variables Without NLR Data (N = 10) With NLR Data (N = 28) P-value Age (years) 60.8 ± 5.7 62.9 ± 2.9 0.36 Female 5 (50) 14 (50) 1.0 Ethnicity 0.125 White 5 (50) 17 (61) Black 2 (20) 7 (26) Hispanic 1 (10) 4 (5) Other 2 (20) 0 (0) NIHSS score 20.1 ± 2.3 18.8 ± 1.4 0.31 TICI Score 0.73 TICI 1 0 (0) 1 (4) TICI 2a 2 (20) 4 (14) TICI 2b 5 (50) 10 (36) TICI 3 3 (30) 13 (46) Mechanical Ventilation 1 (33) 7 (30) 0.92 Discharge to SNF 2 (20) 1 (4) 0.11 Died in hospital 3 (30) 7 (26) 0.80 NLR, TICI score, and outcome Seven patients (7/28, 25%) had a normal NLR of ≤3, whereas nine patients (9/28, 32.1%) had an admission NLR ≥ 7.2, indicative of moderate-to-severe physiologic stress (Table 3). Among the patients who had NLR data, seven patients (7/28, 25%) died. Mean NLR was similarly higher in those who died compared to those who lived (9.2 vs. 4.7, P = .015) (Table 4). Mean NLR was higher in the population that had poor outcome (discharge to SNF or death) compared to those with good outcome (discharge to home or ARF, 8.7 vs. 4.8, P = .019) (Table 3). Among the patients with poor recanalization score (TICI 0, 1, or 2a) three patients died [3/7 (43%) versus 7/31 (23%), p = .27]. Five patients with TICI 0, 1, or 2a scores had poor outcome (nursing home, death) [5/7 (71%) versus 8/31 (26%), p = .02] (Table 5). Logistic regression suggests that patients with poor outcome (nursing home or death) are 7.2 times more likely to be in the cohort with poor TICI scores (TICI 0, 1, or 2a), (OR 7.2, CI95% 1.2–44.7, p = .034). Table 3. NLR and outcome. Group Obs Mean Std. Error Std. Dev. [95% Confidence Interval] Good Outcome 19 4.76 0.82 3.56 3.05–6.48 Poor Outcome 9 8.66 1.94 5.83 4.17–13.14 Combined 28 6.01 0.89 4.68 4.20–7.83 Table 4. NLR and mortality. Group Obs Mean Std. Error Std. Dev. [95% Confidence Interval] Survival 20 4.73 0.77 3.46 3.11–6.35 Mortality 7 9.15 2.44 6.45 3.19–15.11 Combined 27 5.88 0.91 4.72 4.01–7.75 Table 5. TICI grade and outcome. Group TICI 0, 1, 2a TICI 2b, 2c, 3 Total Good Outcome 2 23 25 Poor Outcome 5 8 13 Total 7 31 38 Patients with a NLR > 7.2 were 6.8 times more likely to die (OR 6.8, CI95% 1.2–38.6, p = .03) and approximately 6 times more likely to have a poor outcome (OR 5.9, CI95% 1.3–27.3, p = .02). Ordered logistic regression of outcome and TICI score is statistically significant and suggests that patients with better TICI grade tend to have better outcomes, and it also predicts that patients with worse TICI scores (TICI 0, 1, or 2a) are more likely to have worse outcomes than those with better TICI scores (TICI 2b, 2c, or 3), (OR 5.6, CI95% 1.2–26.0, p = .029). When added to our model as a covariate, both NLR (OR 1.24, 95% CI 1.02–1.50, P = .03) and a TICI score of 0, 1 or 2a (OR 10.78, 95% CI 1.30–89.63, P = .03) independently predicted poor outcome. In a multivariate analysis, the predictive significance of TICI grade on poor outcome was lost when an NLR > 7.2 cutoff was included (NLR > 7.2 p = .043, TICI p = .070). Mechanical ventilation, ventilator days, and outcome In our COVID-19 LVO cohort, invasive mechanical ventilation (>24 h) was a highly significant predictor of mortality [7/8 (88%) ventilated patients versus 3 of 30 (10%) non-ventilated patients, P = .001] and poor outcome [7/8 (88%) ventilated patients versus 6/30 (20%) non-ventilated patients, P = .001]. These were all patients ventilated in the setting of severe COVID-19 pneumonia. Note that there were 2 of 30 non-ventilated patients that were intubated for <24 h. Average number of ventilator days was significantly higher in patients who died compared with patients who survived (5.3 vs. 1.1, days, p = .02), as well as those with poor outcomes compared to better outcomes (4.7 vs. 1.2, p = .03). ROC analysis was undertaken to assess the predictive value of a longer intubation period on outcomes. The ventilator days optimal cutoff for both mortality and poor outcome was determined to be > = 3 days, and with this cutoff 92% (AUC .88) of patients correctly classified with mortality and 88.5% (AUC .83) of patients correctly classified with poor outcome. While number of ventilator days alone was not found in our cohort to be predictive of poor outcome or mortality, invasive mechanical ventilation (>24 h) is a highly significant predictor of mortality (OR 119, p = .001) and also of poor outcome (OR 56, p = .002). Also, patients with NLR > 7.2 were almost 8 times more likely to be in the cohort of patients who underwent prolonged (>24hr) invasive mechanical ventilation (OR 7.8, CI95% 1.2–52.4, p = .03). When we applied multivariate logistic regression analysis to the 3 variables of NLR, TICI, and mechanical ventilation, mechanical-ventilation(>24hr) had such a strong correlation with mortality [mortality: mechanical-ventilation(>24hr) p = 0.002, NLR(>7.2) p = .14, TICI(0,1,2a) p = .72] and poor outcome [mortality: mechanical-ventilation(>24 h) p = .01, NLR(>7.2) p = .10, TICI(0,1,2a) p = .12] in our cohort that it overpowered the effects of NLR and TICI on outcome, both of which lost their significance. Discussion Immune dysregulation has long been purported to be involved in pathophysiology of acute ischemic stroke, and more recently COVID-19. Severe inflammatory responses induced by COVID-19 have been shown to contribute to weak immune adaption, and consequently imbalanced immune responses. Evidence has shown that patients with stroke and COVID-19 have a higher mortality rate, as well as a prolonged hospital length of stay.15 Circulating biomarkers may be surrogate markers for inflammatory and immune status in COVID-19, and have furthered the understanding of the rapidly evolving pandemic. An elevated NLR represents an immune dysregulation state, manifested as an elevated innate immune response (represented by higher neutrophil counts) and a decreased adaptive immune response (represented by lower lymphocyte counts), causing immunosuppression by lowering the cytolytic activities of lymphocytes.1–3,16 As a marker of systemic subclinical inflammation, the NLR has been demonstrated to predict worse clinical outcomes in patients with cancer, cardiac disease and sepsis. The NLR has also been implicated in a number of neurological conditions including subarachnoid hemorrhage and stroke. Studies examining NLR in hemorrhagic stroke have shown that NLR is an independent predictor of worse functional outcome after acute intraparenchymal hemorrhage.17 Prior studies have also defined a relationship between NLR and the development of delayed cerebral ischemia in subarachnoid hemorrhage.18 The NLR, easily calculated by routine blood counts, can be a potentially powerful tool in the prognostication of patients with COVID-19 suffering from AIS due to LVO. In addition to predicting a higher risk of hemorrhage and mortality, elevated NLR values may also identify patients who are at risk for other complications especially other post-stroke infection. Given these recent data, the NLR may present itself as an appealing candidate to assess the neurotropic inflammatory component of COVID-19.4–8,19–22 Though elevated NLR was linked with increased mortality and poor outcome in our cohort, this was a result ultimately dwarfed by the relationship of mechanical ventilation on mortality. When controlling for mechanical ventilation in our model, all effect of NLR on outcome disappeared. Mechanical ventilation may thus fill the role of a mediating variable between elevated NLR and outcome, where elevated NLR might predict outcome through its prediction of mechanical ventilation status. Recently, the NLR has been shown to be a potential predictor of prognosis in COVID-19.23 Specifically, an increased neutrophil count has been shown to be associated with heightened risk of ARDS in COVID-19 patients, while lymphopenia has been associated with worse outcomes.24 An elevated NLR was found to be an independent predictor of poor clinical outcome with an AUC of 0.841, specificity 63.6% and sensitivity of 88%.23 Liu et al. demonstrate that patients with COVID-19, aged ≥50 and with an NLR ≥3.13 are predicted to develop critical illness, and should thus have rapid access to an intensive care if necessary.25 In a study of 74 hospitalized patients with confirmed COVID-19, an NLR of >4 (P = .046) predicted admission to the ICU, reinforcing the theory of a close association between a hyper-inflammatory state and COVID-19 pathogenesis.26 Our results confirm these prior data and lend increasing evidence towards using the NLR as a consistent prognosticator in COVID-19. In our multivariate analysis, patients with a NLR >7.2 were 6.8 times more likely to die, and approximately 6 times more likely to have a poor outcome. Patients with NLR > 7.2 were almost 8 times more likely to be in the cohort of patients who underwent prolonged (>24hr) invasive mechanical ventilation, which was heavily predictive of mortality and poor outcome. This is not the first paper to attempt to identify risk factors for mortality in LVO patients with COVID-19, and, indeed, it has been previously shown that there is an association of elevated NLR values on development of poor outcome in this patient population: Goyal et. al demonstrated that higher admission NLR values are independent predictors of symptomatic intracranial hemorrhage and 3-month mortality in LVO patients treated with mechanical thrombectomy.27–29 In an observational study of 60 patients with acute ischemic stroke, Lin et. al showed that patients with the SARS-Cov-2 virus had a higher NLR compared to those without infection.30 In a retrospective analysis of 116 patients with AIS secondary to LVO, a NLR≥5.9 predicted poor outcome and death at 90 days that remained significant when controlling for age, treatment with IV tPA, and recanalization.31 In our study we found that patients with an elevated NLR >7.2 were more likely to die and have a poor outcome, even when controlling for the effect of poor recanalization. Our results confirm the well-known association between poor angiographic recanalization and outcome. Patients in our cohort with TICI 0–2a were approximately 7 times more likely to have poor outcome. Furthermore, our finding of the synergistically predictive ability of an elevated NLR and low TICI score on worse outcome improves prognostication. Patients with elevated NLR and TICI 0–2a are nearly 12 times more likely to do poorly. While recanalization determined by TICI scores are a simple and immediate post-angiographic metric to guide treatment and provide some prognostication, when examined in isolation, they are not always reliable. This is evidenced by the fact that several patients fail to do well despite TICI 3 recanalization (i.e “futile recanalization”), and identifying these patients is the subject of ongoing research.32 In our multivariate analysis, patients with elevated NLR went on to have poor clinical outcome, despite favorable recanalization. This finding is significant as it suggests that in certain patients, the neuro-inflammatory process in COVID-19 may outweigh the potential benefit of a successful thrombectomy. This idea is corroborated by a recent study introducing the Poor Outcome of Endovascular Treatment With Successful Recanalization (PREDICT) scale. Notably, one of the individual components of PREDICT scale is admission NLR.33 PREDICT has been shown to have good discrimination and satisfactory calibration in a multi-center cohort of 332 LVO patients treated with thrombectomy. Identification of patients with perhaps both elevated PREDICT scores and NLR, may lead to improved prognostication, both for patient and family counseling, and for allocation of resources. As the allocation of healthcare resources continues to be an ever-present dilemma during this pandemic, NLR can be incorporated as a scoring variable, to predict response after such a procedure. Accurate identification of patients with COVID-19 who would reliably do poorly despite successful thrombectomy may also lead to more efficient ICU utilization. Several limitations of this study deserve mention. This is a retrospective evaluation of prospectively databases maintained during the height of the pandemic. We were unable to compare baseline demographic and disease-related features between the patients with LVO and the main cohort of COVID-19 patients. Similarly, we also were unable to provide an estimated frequency of stroke among the overall cohort of COVID-19 patients who did not undergo brain imaging due to being too ill to transfer or under sedation. Due to the fact that all of our patients were sick enough to be admitted to the hospital, multiple confounding factors could have contributed to the observed complications and outcomes. This includes the inability to attribute NLR as a marker for neuro-specific inflammatory stress versus a systemic inflammatory response in the setting of COVID-19. Furthermore, while we know that the data for this study was collected in pre-Omicron and pre-Delta phases of the pandemic, we do not have granular data on the specific subtype of COVID-19 these patient's had. Finally, the small number of included patients may limit the generalizability of our findings to other regions throughout the world. Conclusion An elevated NLR in patients with COVID-19 and AIS due to LVO is a significant indicator of mechanical ventilation which portents significantly worse outcomes and increased mortality regardless of the TICI score. This suggests that clinical scenarios may exist where the inflammatory response in COVID-19 outweighs any potential benefit of a successful thrombectomy. Further study is necessary to characterize the details of this relationship. Declaration of conflicting interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding: The author(s) received no financial support for the research, authorship, and/or publication of this article. ORCID iDs: Fawaz Al-Mufti https://orcid.org/0000-0003-4461-7005 Jared B. Cooper https://orcid.org/0000-0003-4737-3026 Eric Feldstein https://orcid.org/0000-0002-1952-4555 Krishna Amuluru https://orcid.org/0000-0002-8859-8574 Adam A Dmytriw https://orcid.org/0000-0003-0131-5699 Nicola Limbucci https://orcid.org/0000-0002-0432-5414 Mario Martínez-Galdámez https://orcid.org/0000-0002-8024-4712 Miguel Schüller-Arteaga https://orcid.org/0000-0003-3351-668X Sanjeev Nayak https://orcid.org/0000-0002-9616-1188 ==== Refs References 1 Petrie HT Klassen LW Kay HD . Inhibition of human cytotoxic t lymphocyte activity in vitro by autologous peripheral blood granulocytes. J. Immunol 1985; 134 : 230–234.3871101 2 el-Hag A Clark RA . Immunosuppression by activated human neutrophils. Dependence on the myeloperoxidase system. J. Immunol 1987; 139 : 2406–2413.2821114 3 Provencio JJ . Inflammation in subarachnoid hemorrhage and delayed deterioration associated with vasospasm: a review. Acta Neurochir 2013; 115 : 233–238. 4 Kumar R Geuna E Michalarea V , et al. The neutrophil-lymphocyte ratio and its utilisation for the management of cancer patients in early clinical trials. Br J Cancer 2015; 112 : 1157–1165.25719834 5 Nunez J Nunez E Bodi V , et al. Usefulness of the neutrophil to lymphocyte ratio in predicting long-term mortality in st segment elevation myocardial infarction. Am J Cardiol 2008; 101 : 747–752.18328833 6 Akpek M Kaya MG Lam YY , et al. Relation of neutrophil/lymphocyte ratio to coronary flow to in-hospital major adverse cardiac events in patients with st-elevated myocardial infarction undergoing primary coronary intervention. Am J Cardiol 2012; 110 : 621–627.22608360 7 Cho KH Jeong MH Ahmed K , et al. Value of early risk stratification using hemoglobin level and neutrophil-to-lymphocyte ratio in patients with st-elevation myocardial infarction undergoing primary percutaneous coronary intervention. Am J Cardiol 2011; 107 : 849–856.21247535 8 Yalcinkaya E Yuksel UC Celik M , et al. Relationship between neutrophil-to-lymphocyte ratio and electrocardiographic ischemia grade in stemi. Arq Bras Cardiol 2015; 104 : 112–119.25424159 9 Kalemci S Sarıhan A Zeybek A . Association between NLR and COVID-19. Int Immunopharmacol 2020; 88 : 106917.32889243 10 Haghjooy Javanmard S Vaseghi G Manteghinejad A , et al. Neutrophil-to-lymphocyte ratio as a potential biomarker for disease severity in COVID-19 patients. J Glob Antimicrob Resist 2020; 22 : 862–863.32810639 11 Jamil S Mark N Carlos G , et al. Diagnosis and management of COVID-19 disease. Am J Respir Crit Care Med 2020; 201 : 19–27. 12 Al-Mufti F Amuluru K Sahni R , et al. Cerebral venous thrombosis in COVID-19: a New York Metropolitan Cohort Study. American Journal of Neuroradiology 2021; 42 : 1196–1200. doi:10.3174/ajnr.A7134 33888450 13 Richardson S Hirsch JS Narasimhan M , et al. Presenting characteristics, comorbidities, and outcomes among 5700 patients hospitalized with COVID-19 in the New York city area. JAMA 2020; 323 : 2052–2059. doi:10.1001/jama.2020.6775 32320003 14 Stein LK Mocco J Fifi J , et al. Correlations between physician and hospital stroke thrombectomy volumes and outcomes: a nationwide analysis. Stroke 2021; 52 : 2858–2865. doi:10.1161/STROKEAHA.120.033312 34092122 15 Yaghi S Ishida K Torres J , et al. SARS-CoV-2 and stroke in a New York healthcare system. Stroke 2020; 51 : 2002–2011.32432996 16 Bianchi ME . Damps, pamps and alarmins: all we need to know about danger. J Leukoc Biol 2007; 81 : 1–5. 17 Liu S Liu X Chen S , et al. Neutrophil-lymphocyte ratio predicts the outcome of intracerebral hemorrhage: a meta-analysis. Medicine (Baltimore) 2019; 98 : e16211.31261573 18 Al-Mufti F Amuluru K Damodara N , et al. Admission neutrophil-to-lymphocyte ratio predicts delayed cerebral ischemia following aneurysmal subarachnoid hemorrhage. J Neurointerv Surg 2019; 11 : 1135–1140.30979846 19 Halazun HJ Mergeche JL Mallon KA , et al. Neutrophil-lymphocyte ratio as a predictor of cognitive dysfunction in carotid endarterectomy patients. J Vasc Surg 2014; 59 : 768–773.24571940 20 Demirci S Demirci S Kutluhan S , et al. The clinical significance of the neutrophil-to-lymphocyte ratio in multiple sclerosis. Int J Neurosci 2016; 126: 700–706. 21 Tokgoz S Kayrak M Akpinar Z , et al. Neutrophil lymphocyte ratio as a predictor of stroke. J Stroke Cerebrovasc Dis 2013; 22 : 1169–1174.23498372 22 Tokgoz S Keskin S Kayrak M , et al. Is neutrophil/lymphocyte ratio predict to short-term mortality in acute cerebral infarct independently from infarct volume? J Stroke Cerebrovasc Dis 2014; 23 : 2163–2168.25106834 23 Yang A-P Liu J-P Tao W-Q , et al. The diagnostic and predictive role of NLR, d-NLR, and PLR in COVID-19 patients. Int Immunopharmacol 2020; 84 : 106504.32304994 24 Terpos E Engelhardt M Cook G , et al. Management of patients with multiple myeloma in the era of COVID-19 pandemic: a consensus paper from the European myeloma network (EMN). Leukemia 2020; 34 : 2000–2011.32444866 25 Liu J Liu Y Xiang P , et al. Neutrophil-to-lymphocyte ratio predicts critical illness patients with 2019 novel coronavirus in the early stage. J Transl Med 2020; 18 : 206.32434518 26 Ciccullo A Borghetti A Lombardi F , et al. Neutrophil-to-lymphocyte ratio and clinical outcome in COVID-19: a report from the Italian front line. Int J Antimicrob Agents 2020; 56 : 106017.32437920 27 Altschul DJ Esenwa C Haranhalli N , et al. Predictors of mortality for patients with COVID-19 and large vessel occlusion. Interv Neuroradiol 2020; 26 : 623–628. doi:10.1177/1591019920954603 32862753 28 Esenwa C Cheng NT Luna J , et al. Biomarkers of coagulation and inflammation in COVID-19–associated ischemic stroke. Stroke 2021; 52 : e706–e709. doi:10.1161/STROKEAHA.121.035045 34428931 29 Goyal N Tsivgoulis G Chang JJ , et al. Admission neutrophil-to-lymphocyte ratio as a prognostic biomarker of outcomes in large vessel occlusion strokes. Stroke 2018; 49 : 1985–1987.30002151 30 Lin C Arevalo YA Nanavati HD , et al. Racial differences and an increased systemic inflammatory response are seen in patients with COVID-19 and ischemic stroke. Brain Behav Immun Health 2020; 8 : 100137.32904928 31 Brooks SD Spears C Cummings C , et al. Admission neutrophil-lymphocyte ratio predicts 90 day outcome after endovascular stroke therapy. J Neurotinterv Surg 2014; 6 : 578–583. 32 van Horn N Kniep H Leischner H , et al. Predictors of poor clinical outcome despite complete reperfusion in acute ischemic stroke patients. J Neurointerv Surg 2021; 13 : 14–18.32414889 33 Wang H Zhang M Hao Y , et al. Early prediction of poor outcome despite successful recanalization after endovascular treatment for anterior large vessel occlusion stroke. World Neurosurg 2018; 115 : e312–e321.29673825
PMC009xxxxxx/PMC9006087.txt	==== Front J Infect Prev J Infect Prev spbji BJI Journal of Infection Prevention 1757-1774 1757-1782 SAGE Publications Sage UK: London, England 10.1177_17571774211046989 10.1177/17571774211046989 Letter to the Editor Feasibility of hygienic clinical attire for doctors during COVID-19: A university hospital experience https://orcid.org/0000-0002-0535-3062 James Hannah Phillips Andrew Trauma & Orthopaedic Surgery, 61138 Coventry and Warwickshire Hospital , Coventry, Ireland Hannah James, Trauma & Orthopaedic Surgery, Coventry and Warwickshire Hospital, Clifford Bridge Road, Coventry CV2 2DX, Ireland. Email: hsmith22@doctors.org.uk 11 4 2022 7 2022 11 4 2022 23 4 190191 © The Author(s) 2022 2022 Infection Prevention Society This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. Infection control nosocomial infection quality improvement typesetterts10 ==== Body pmcPersonal protective equipment (PPE) for healthcare staff has been central to reducing nosocomial transmission of COVID-19 (Public Health England, 2020: p6–36). The role of doctors’ clinical attire in transmitting the virus in hospital settings has not been reported in the literature. The survival of SARS-CoV-2 on clothing is currently unknown, but the virus has been shown to remain viable and infectious on surfaces for several days, depending on inoculum shed (Doremalen et al., 2020). Fomite transmission was associated with nosocomial spread and super-spreading events with SARS-CoV-1, and the stability of the two viruses has been demonstrated to be similar (Chen et al., 2004). In contrast to nursing and allied healthcare staff who have an established uniform policy with evidence-based hygiene measures (NHS England and NHS Improvement, 2020:p2–6), doctors habitually commute to the hospital in work clothing and visit public places en route (Oxtoby, 2015). This is poor infection control practice in ordinary times, but in the context of a pandemic may represent an underappreciated route of viral transmission. We report the experience of piloting hygienic ‘ward wear’ scrubs for doctors at a large 1000-bed National Health Service University teaching hospital. The hospital employs 1022 doctors, all of whom were offered use of the new scrubs on a loan basis during the first wave of the UK COVID-19 pandemic in March 2020. The provision consisted of three sets of a unisex navy blue poly-cotton scrub suit of standard design, with the hospital logo and ‘doctor’ embroidered on the chest. Doctors were responsible for laundering their own scrubs, as evidence shows that domestic laundering at 60°C is as effective as commercial washing for decontaminating healthcare clothing (NHS England and NHS Improvement, 2020:p2–6). Doctors using the scrubs were instructed to change into and out of them on hospital premises using existing changing room facilities and store their clothing with their personal belongings. Data was obtained by a web survey with 10 questions in October 2020. Questions included multiple choice, Likert scale questions and free-text. All doctors employed at the hospital (n = 1022) were invited by email to take part. The objectives were to map the demographics of uptake of the ward wear initiative; to obtain feedback on users’ perceptions of hygiene, comfort and professionalism; and to understand the barriers to use among doctors who chose not to take part in the pilot. Of 1022 doctors, 504 (49%) opted to use the new scrubs. 169 doctors completed the survey (14%). 135 (80%) of respondents were using the new scrubs at the time of the survey, 6 months after the start of the initiative. There was an equal gender split amongst respondents who had tried the scrubs (the ‘uptake’ group) and those who chose not to (the ‘decline’ group). The grade of respondents was comparable between the uptake and decline groups and broadly representative of the proportions of each grade in current employment at the hospital. Of respondents; 50% were consultants, 25% speciality trainees, 15% foundation doctors and 10% staff and associate speciality (SAS) grades. Within the uptake group (n = 135), 75% said the current pandemic had increased their awareness of the importance of hygienic clinical attire for doctors and 60% intend to continue wearing the scrubs beyond the pandemic. 77% stated willingness to launder the scrubs at home, whereas 23% felt this should be the hospitals’ responsibility. Less than half (40%) felt that doctors should have a mandatory uniform. The uptake group reported the scrubs were comfortable and professional looking, with mean scores of 7.2 and 8.4, respectively, on a Likert scale of 1–10, where 1 represents ‘not atall’ and 10, ‘extremely’. The trust logo and ‘doctor’ designation was perceived positively, with mean Likert scores of 8.7 and 8.2. Regarding fit, 67% of female respondents would like a female-specific cut top to be offered, and 30% of respondents of any gender wanted a wider size range beyond S–XL. 96% reported that the scrubs met their personal modesty requirements. Of the decliners group (n = 34), the main reasons for not adopting new scrubs were already wearing theatre blues as part of job role (25%); agreement in principle with the initiative but not liking the style, colour or fit of the provision (25%); already bought own scrubs (15%) and preferring to wear own clothes for work (15%). When asked what would encourage the decliners to start wearing the ward wear attire, 38% would like their grade and 31% would like their speciality embroidered on as they felt ‘doctor’ was descriptively inadequate. 35% would use them if the hospital was responsible for laundry, 25% would use them if there was a wider size range and 15% would if there was a choice of style. In summary, doctors clothing may be an unappreciated route of transmission of SARS-CoV-2 virus and introduction of hygienic, hospital-provided ward wear may serve in part of mitigate this risk. In this pilot study, we have demonstrated that it is acceptable to doctors to wear and self-launder hygienic ward wear in a university hospital setting and it is feasible for them to comply with on-site changing requirements without provision of any extra facilities. We have identified the main barriers to uptake, which other hospitals may wish to consider when implementing a similar initiative. ORCID iD Hannah James https://orcid.org/0000-0002-0535-3062 The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding: The author(s) received no financial support for the research, authorship, and/or publication of this article. ==== Refs References Chen YC Huang LM Chan CC , et al. (2004) SARS in hospital emergency room. Emerging Infectious Diseases 10 : 782–788.15200809 van Doremalen N. Bushmaker T. Morris DH , et al. (2020) Aerosol and surface stability of SARS-CoV-2 as compared with SARS-CoV-1. The New England Journal of Medicine 382 : 1564–1567.32182409 NHS England and NHS Improvement (April 2020). Uniforms and Workwear: Guidance for NHS Employers. London: NHSE. Available at: www.england.nhs.uk/about/equality/equality-hub/uniforms-and-workwear ((accessed on 18 December 2020)) Oxtoby K (2015) Scrubs, suit, or jeans-what should doctors wear to work? Bmj: British Medical Journal 351 : h3611. Public Health England . COVID-19: Guidance for the Remobilisation of Services within Heath and Care Settings. Infection Prevention and Control Recommendations. August 2020 Version 1.0. Available at: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachments_data/file/910885/COVID-19_Infection_prevention_and_control_guidance_FINAL_PDF_20082020.pdf (accessed on 18 December 2020)
PMC009xxxxxx/PMC9006088.txt	==== Front Int J Health Serv Int J Health Serv JOH spjoh International Journal of Health Services 0020-7314 1541-4469 SAGE Publications Sage CA: Los Angeles, CA 35404167 10.1177/00207314221092354 10.1177_00207314221092354 II. Social Determinants of the COVID-19 pandemic Discontinuation of Health Interventions Among Brazilian Older Adults During the Covid-19 Pandemic: REMOBILIZE Study https://orcid.org/0000-0003-3218-1769 Coelho de Amorim Juleimar Soares 1 Ornellas Giulianna 1 Lloyd-Sherlock Peter 2 Pereira Daniele Sirineu 3 da Silva Alexandre 4 Duim Etienne 5 Lima Camila Astolphi 6 Perracini Monica Rodrigues 67 1 133629 Instituto Federal de Educação, Ciência e Tecnologia, Rio de Janeiro (RJ) , Brazil 2 School of International Development, 6106 University of East Anglia , Norwich, UK 3 Escola de Educação Física, Fisioterapia e Terapia Ocupacional da UFMG 4 Department of Collective Health, 146840 Faculdade de Medicina de Jundiaí , Jundiaí, Brazil 5 37896 Hospital Israelita Albert Einstein , São Paulo, SP, Brazil 6 Master's and Doctoral Program in Physical Therapy, 149944 Universidade Cidade de São Paulo , São Paulo, São Paulo, Brazil 7 Master's and Doctoral Programs in Gerontology, Faculty of Medical Sciences, Universidade Estadual de Campinas, Campinas, São Paulo, Brazil Juleimar Soares Coelho de Amorim, Instituto Federal do Rio de Janeiro, Rua Prof. Carlos Wenceslau, 343, Realengo, CEP 21710-240, Rio de Janeiro, RJ, Brasil. Email: juleimar@yahoo.com.br 7 2022 7 2022 7 2022 52 3 330340 08 9 2021 22 12 2021 21 2 2022 © The Author(s) 2022 2022 SAGE Publications This article is made available via the PMC Open Access Subset for unrestricted re-use and analyses in any form or by any means with acknowledgement of the original source. These permissions are granted for the duration of the COVID-19 pandemic or until permissions are revoked in writing. Upon expiration of these permissions, PMC is granted a perpetual license to make this article available via PMC and Europe PMC, consistent with existing copyright protections. The objective of this study was to analyze changes in access to health interventions during the pandemic among Brazilian older adults and to investigate the factors associated with social and health inequalities. We conducted an online survey with Brazilian adults aged 60 + years between May and June 2020. A multidimensional questionnaire was used to investigate access to health interventions during the pandemic and associated factors. Of 1482 participants, 56.5% reported health care before the pandemic, and 36.4% discontinued it during the pandemic. The discontinuation rate was 64.4% (95% CI 61.1-67.6). Participants with higher educational level (nine or more years of education: OR 0.34; 95% CI 0.17-0.70) and higher income (eight or more times the minimum wage: OR 0.54; 95% CI 0.36-0.81) were associated with less probability of discontinuation. Presenting multimorbidity (OR: 1.42; 95% CI 1.06-1.90) and polypharmacy (OR: 0.61; 95% CI 0.46-0.81) were associated with discontinuity in health interventions. Our study showed that structural health inequities in access to health care shaped the rates of discontinuation in health care interventions during the COVID-19 pandemic. Strategic actions should be set up to actively monitor socially vulnerable older adults and strengthen community-based services to mitigate the discontinuation of health care interventions. older adults geriatric care discontinuation of health care COVID-19 pandemic Instituto Federal do Rio de Janeiro Edital Integrado n. 1 typesetterts19 ==== Body pmcThe SARS-CoV-2 pandemic, the novel coronavirus, has reached worldwide and grown into a public health crisis, requiring interventions to reduce the disease, such as distancing and social isolation.1,2 However, there was no consensus on a national policy for protective measures in Brazil, generating an alarming scenario of more than 17 million confirmed cases, with almost 500,000 deaths (June 2021).3 Not surprisingly, health services and health care workers were fighting on the frontlines to combat the pandemic, with severe consequences for the system and those in need of care.4,5 The coronavirus disease 2019 (COVID-19) has a significant impact among older adults, who represent, globally, 31% of cases of infection, 45% of hospitalizations, 53% of admissions to intensive care units, and 80% of deaths.6,7 In Brazil, hospital mortality increased with age, corresponding to 42%, 55%, and 66% of hospitalized patients aged 60 to 69 years, 70 to 79 years, and 80 years or more, respectively.8 Furthermore, older adults with low education, low income, and black and brown skin color were more affected by COVID-19 and had more severe disease cases, revealing structural social and health inequities. These vulnerable groups have a greater number of comorbidities and disabilities, increasing the risk of aggravation and death as a result of COVID-19.8–10 Necessary measures for COVID-19 care and containment of the pandemic can negatively affect older adults’ physical and mental health. Recommendations to stay at home and to practice social distancing increased sedentary behavior and decreased physical activity levels, resulting in harmful consequences for mobility. In addition, restricted mobility outside the home increases loneliness and social isolation, while also reducing access to health care.1,11–14 Since the beginning of the COVID-19 pandemic, there has been a record of health disruption in 90% of countries,15 including outpatient care, elective surgical procedures, and non-essential services.16 Interruptions in services during the COVID-19 pandemic may impact ongoing health care and cause harmful consequences, such as insufficient monitoring and treatment of noncommunicable diseases, delayed diagnoses, and surgical interventions.17 Health inequities greatly impact the course of the COVID-19 pandemic, with a disproportionate burden on people living with high socioeconomic vulnerability. It ultimately might hurt interruptions in health care.18 Societies with higher levels of income inequality are associated with poorer health outcomes, including a lower life expectancy, higher rates of obesity, and disabilities since the pandemic period. The evidence before the COVID-19 pandemic pointed to differences and inequalities in the access to and continuity of health treatments in developing countries. In this case, the concept of access was related to maintening or improving health by receiving health care. The determinants of access and continuity of health care services are related to contextual and individual factors. According to the classical model of Andersen and Newman, individual factors encompass predisposing characteristics (such as age and sex), enabling characteristics (such as education level and income), and health needs.19 Adherence to long-term therapeutic programs, such as rehabilitation, is an additional challenge due to the lack of transport and social support, sociodemographic factors, and health needs. A previous study observed a variation in the adherence according to the socioeconomic characteristics of the users.16 Disability and pain, for example, are greater predictors for the use of more health services.20,21 The search for health facilities is greater among women, older adults, and those living with a spouse. Having fewer comorbidities and low income is associated with not searching for care.5,21,22 Access and utilization of health services is also related to health systems and their organization, as the Brazilian Unified System is responsible for free, universal provision of health services and programs.22–24 However, in Brazil, the main problems have included hours of operation of health care facilities20 and delays in receiving care in the public health system, including scheduling a doctor visit, waiting to get the visit, accessing the health unit, and receiving home visits.20,22,23 Social determinants can influence the interruption of health services and care during the COVID-19 pandemic, especially income, educational level, and clinical conditions (multimorbidity and polypharmacy). However, few studies have assessed treatment interruption during the COVID-19 pandemic among Brazilian older adults. Understanding the relationships between health system preparedness, health system capacity, and disruption is particularly important in a country marked by vast inequalities in socioeconomic characteristics (eg, housing and employment status) and other health risks (age structure, burden of chronic disease, and disability). Inequalities and lack of continuity in health care result in fragmented treatments, higher hospital admissions, and avoidable visits to the emergency room.25 Monitoring the discontinuation of health care among older adults can lead to collaboration and development of strategies for resuming care approaches after the period of social restriction. In the COVID-19 pandemic context, these surveys are necessary to prepare health professionals and policymakers to meet urgent demands and needs, as well as to expand on the existing literature that examines the lack of social and health equality and discontinuation of the use of and access to health care. We hypothesized that as health services were discontinued in Brazil during the pandemic, this did not occur homogenously within the older population, but instead that those experiencing more social vulnerability were more affected. The present study aimed to analyze changes in access to health interventions during the pandemic among Brazilian older adults and to investigate the factors associated with social and health inequalities. Methods This research was a cross-sectional study, as part of the Study Network on Mobility in Aging (Remobilize Study), following STROBE guidelines for observational studies. The Remobilize Study has been conducting a longitudinal study with adults who are 60 years old and older and are residents of all five macro-regions of Brazil, aiming to assess the impact of the pandemic on mobility among Brazilian older adults. Baseline data were collected between May and June 2020, with a follow-up schedule of three, six, and 18 months. In this study, we analyze the baseline data, which refer to the first quarter of 2020, when the COVID-19 pandemic was declared a public health emergency in Brazil. Data Collection Procedures Community-dwelling older adults (age 60 years or more) of both genders were included. The study excluded participants who were bedridden and/or living at long-term care facilities. Participants who reported dementia or cognitive decline, severe vision or hearing problems, or severe communication limitations could participate in the survey through interviews answered by proxy or by telephone contact with a trained interviewer. According to the snowball sampling methodology, participant recruitment was carried out by convenience, in the local community, and through social media. Snowball sampling methodology is one in which a small number of individuals with the same characteristics indicate other people in their social network or community to participate in the study.26 In this particular study, participants were recruited virtually by phone, social media groups (Facebook, Instagram, and WhatsApp), researchers’ friends, family, patients, and ex-patients, who recruited more people and so on. We contacted community leaders and allied health professionals working in vulnerable regions to include participants with varied educational and income levels, ethnicities, and genders. More details can be found on the research homepage (https://www.remobilize.com.br/us) and in another publication (Perracini et al, 2021). The Research Ethics Committee approved the University Cidade de São Paulo (CAAE: 31592220.6.0000.0064). Participants agreed with the terms of the survey by clicking on the button “I agree to participate” (for the online survey) or verbally (for telephone interviews). Outcome We considered the discontinuation of health care, and a range of self-care and health promotion behaviors before and during the COVID-19 pandemic, as the outcome variable. Health interventions were defined as services aiming to prevent and cure illnesses and to maintain health and well-being.27,28 These health interventions included medical appointment and follow-up (diagnosis and treatment), use of medications, rehabilitation (conventional physiotherapy, psychology, occupational therapy, speech therapy, hydrotherapy, global postural re-education), dentistry, and integrative therapies (acupuncture, yoga). Health behaviors included physical activities and health promotion, such as a gym, Pilates, hydrogymnastics, gymnastics, dance, seniors’ gym, walking, and sports. Therefore, we categorized the variable into “continuity” (carrying out some care before and during the COVID-19 pandemic) and “discontinued” (carrying out some maintenance before but not during the COVID-19 pandemic). Independent Variables The independent variables included were sex, age group (60-69, 70-79, and 80 years old and over), race (white, black, brown, and yellow/indigenous, per the official classifications of the National Institute of Geography and Statistics in Brazil), income in minimum wages (up to 1, 2-3, 4-7, 8-10, and above 10), educational level in years (illiterate, 1-4, 5-8, and 9 or more), number of comorbidities, polypharmacy (report of routine use of four or more medications 29), functional limitation, and presence of pain. We selected these variables according to the classic model by Andersen and Newman (1973),19 which considers individual factors such as predisposing characteristics (age and sex), facilitators (such as educational level and income), and self-reported health needs (comorbidity, polypharmacy, pain, and functional limitation), in addition to the social determinants of illness and health care in Brazil.30 Based on participants’ self-reports of medical diagnoses, the selected comorbidities were arthritis, osteoporosis, asthma, congestive heart failure, myocardial infarction, neurological disease, stroke or transient ischemic attack, peripheral vascular diseases, diabetes mellitus type I or II, upper gastrointestinal diseases, depression, anxiety or panic attacks, visual impairment, hearing impairment, degenerative disc diseases, obesity, high blood pressure, urinary incontinence, fecal incontinence, and dizziness/vertigo. We considered older people reporting two or more conditions as having multimorbidity.31 We assessed activity limitations using the Brazilian Multidimensional Functional Assessment Questionnaire (BOMFAQ). The participants reported their difficulty (yes/no) to perform 15 daily activities, including eight basic activities of daily living (BADLs) (eg, eating, bathing, and dressing) and seven instrumental activities of daily living (IADLs) (eg, shopping, taking medication at the correct time, and walking nearby home). The number of activities performed with difficulty was summed (0-15), and the participants were categorized into two groups: those with mild (0-3 activities) and those with moderate to severe functional limitation (≥ 4 activities).32 Data Analysis We compared all variables in the study among older adults who continued and discontinued some health care, using the Pearson's chi-square test to select the variables to be included in the multiple models. The discontinuation rate refers to the difference in the proportion of people who performed some care during and before the COVID-19 pandemic divided by the proportion of those who performed it before the COVID-19 pandemic. We used odds ratio (OR) and respective 95% confidence intervals by logistic regression to estimate associations between independent variables and the outcome. Factors that showed statistically significant associations (p < 0.05) with the outcome in the final model were selected as independent variables for adjustment of a multiple logistic regression model, which aims to estimate the predicted discontinuation probabilities. Predicted probabilities were presented with respective confidence intervals (95%). All analyses were performed using Stata software version 14.0 (StataCorp LLC, College Station, TX), and a statistical significance of 5% was considered. Results Demographics In total, 1482 older adults were included in this study: 69.5% (n = 1030) and 30.5% (n = 452) answered the questionnaire by weblink and phone, respectively. As illustrated in Figure 1, more than half of the participants (56.5%) reported performing some health care before the COVID-19 pandemic. Of the total, 36.4% discontinued this care at the beginning of the COVID-19 pandemic. The discontinuation rate was 64.4% (95% CI: 61.1-67.6). Figure 2 shows the main health care discontinued during the pandemic, highlighting medical follow-up (47.5%), physical activity (30.9%), and rehabilitation (14.6%). Figure 1. Flowchart of entry of participants in the study, together with prevalence and rates of health care before and discontinuation during COVID-19 pandemic among Brazilian older adults. Remobilize Study, 2020. Figure 2. Discontinuity of health care before and during the COVID-19 pandemic among Brazilian older adults, according to self-report. Remobilize Study, 2020. Associations Between Social and Health Conditions and Discontinuity Table 1 shows the sociodemographic characteristics, region of residence, health conditions, and functionality of the sample. The sample was mainly composed of female older adults (74.0%), between 60 and 69 years old (56.1%), of race white (61.8%), with lower income (62.4% up to three minimum wages), with a high educational level (60.9% aged nine years or more), and from the Southeast and Northeast regions (together, 85.75%). Table 1. Sociodemographic Characteristics, Region of Residence, Health Conditions, and Functionality of Brazilian Older Adults, According to Reports of Health Care Before and During the COVID-19 Pandemic. REMOBILIZE Study, 2020. Variables Total 1482 (100%) Health care p-value Discontinuation 1062 (75.5%) Continuation 346 (24.5%) Sex 0.297 Female 1096 (74.0) 798 (75.1) 250 (72.2) Male 386 (26.0) 265 (24.9) 96 (27.8) Age group (in years) 0.774 60–69 831 (56.1) 593 (55.8) 190 (54.9) 70–79 420 (28.4) 308 (29.0) 98 (28.3) 80 + 229 (15.5) 161 (15.2) 58 (16.8) Skin color/ethnicity 0.012 White 914 (61.7) 624 (58.8%) 233 (67.6%) Black 100 (6.8) 80 (7.5%) 17 (4.9%) Yellow, brown, and indigenous 466 (31.5) 357 (33.7%) 95 (27.5%) Income (minimum wagea) <0.001 Up to 1 512 (34.5) 416 (39.1) 82 (23.7) 2–3 413 (27.9) 299 (28.1) 96 (27.8) 4–7 267 (18.0) 171 (16.1) 74 (21.4) 8 + 290 (19.6) 177 (16.7) 94 (27.2) Educational level (years of study) <0.001 Illiterate 117 (7.9) 105 (9.9) 11 (3.2) 1–4 282 (19.0) 221 (20.8) 54 (15.6) 5–8 181 (12.2) 145 (13.6) 31 (9.0) 9 + 902 (60.9) 592 (55.7) 250 (72.2) Region 0.744 South 56 (3.8) 39 (3.6) 14 (4.1) Southeast 638 (43.1) 467 (44.0) 150 (43.4) Midwest 54 (3.6) 40 (3.8) 8 (2.3) Northeast 630 (42.6) 444 (41.8) 151 (43.6) North 102 (6.9) 72 (6.8) 23 (6.6) Comorbidityb (two or more) 639 (43.2) 478 (45.0) 121 (35.0) 0.001 Polypharmacyc 458 (30.9) 293 (27.6) 150 (43.4) <0.001 Presence of pain 406 (27.4) 295 (28.8) 92 (26.6) 0.674 Functional limitation moderate to severed 314 (21.2) 232 (21.8) 75 (21.7) 0.954 Missing: 74 participants (4.9%), referring to the older adults who did not undergo treatment before the COVID-19 pandemic. a Minimum wage = R$1.045,00. b Includes arthritis, osteoporosis, asthma, congestive heart failure, myocardial infarction, neurological disease, stroke or transient ischemic attack, peripheral vascular disease, diabetes mellitus type I or II, upper gastrointestinal disease, depression, anxiety or panic attacks, visual changes, hearing impairment, degenerative disc diseases, obesity, high blood pressure, urinary incontinence, fecal incontinence, and dizziness/vertigo. c Routine use of four or more medications. d Score ≥ 4 on the BOMFAQ, which assesses the following daily activities (ADLs and IADLs): lying down or getting out of bed, eating, brush the hair, walking on a plane, taking a shower, dressing, going to the bathroom on time, climbing stairs, taking medication, walking the neighborhood, shopping, preparing meals, cutting toenails, driving, and cleaning the house. We found that sociodemographic social disparities (income, education, and race) and health condition (comorbidities and polypharmacy) were associated with discontinuing health care. Income of up to one minimum wage (39.1% of participants) was associated with the discontinuation of health care. Compared to those who continued health care, those with race black (7.5%) and brown, yellow, or indigenous (33.7%) had higher discontinuation rates. For older adults with up to eight years of education, the discontinuation rate was higher than the continuity rate. On the other hand, a higher continuity rate for older adults with more than nine years of education was observed (72.2%). In addition, among those who discontinued health care, 45.0% reported two or more comorbidities, and 27.6% were using polypharmacy. Table 2 presents the odds ratios (95% CI) of the adjusted multiple model. Income more than eight times the minimum wage (OR = 0.54, 95% CI = 0.36-0.81), educational level higher than nine years (OR = 0.34, 95% CI = 0.17-0.70), and polypharmacy (OR = 0.61, 95% CI = 0.46-0.81) were the associated variables that were less likely to be associated with discontinued health care. Those who presented two or more comorbidities had a higher chance of discontinued health care (OR = 1.42, 95% CI = 1.06-1.90). Table 2. Factors Associated with Treatment Discontinuation During the COVID-19 Pandemic. REMOBILIZE Study, 2020. Variables Odds Ratio 95% IC Female sex (ref.: male) 1.18 0.88–1.57 Age group (in years) 60–69 1.00 70–79 0.94 0.70–1.27 80 + 0.84 0.56–1.27 Skin color/ethnicity White 1.00 Black 1.39 0.79–2.45 Yellow, brown, or indigenous 1.06 0.79–1.41 Income (minimum wagea) Up to 1 1.00 2–3 0.76 0.53–1.09 4–7 0.65 0.43–0.98 8 + 0.54 0.36–0.81 Educational level (years of study) Illiterate 1.00 1–4 0.50 0.25–1.00 5–8 0.56 0.27–1.20 9 + 0.34 0.17–0.70 Region South 1.00 Southwest 1.01 0.52–1.93 Midwest 1.34 0.49–3.67 Northeast 0.94 0.49–1.81 North 0.92 0.42–2.03 Comorbidityb (two or more) 1.42 1.06–1.90 Polypharmacyc 0.61 0.46–0.81 Presence of pain 1.17 0.87–1.57 Functional limitation moderate to severed 0.95 0.66–1.37 a Minimum wage = R$1.045,00. b Includes arthritis, osteoporosis, asthma, congestive heart failure, myocardial infarction, neurological disease, stroke or transient ischemic attack, peripheral vascular disease, diabetes mellitus type I or II, upper gastrointestinal disease, depression, anxiety or panic attacks, visual changes, hearing impairment, degenerative disc diseases, obesity, high blood pressure, urinary incontinence, fecal incontinence, and dizziness/vertigo. c Routine use of four or more medications. d Score ≥ 4 on the BOMFAQ, which assesses the following daily activities (ADLs and IADLs): lying down or getting out of bed, eating, brush the hair, walking on a plane, taking a shower, dressing, going to the bathroom on time, climbing stairs, taking medication, walking the neighborhood, shopping, preparing meals, cutting toenails, driving, and cleaning the house. We found that older adults with an income of more than eight times the minimum wage (63.8%, 95% CI: 54.8-62.9), those who use polypharmacy (69.3%, 95% CI: 64.9-73.8), and those with more than nine years of education (72.1%, 95% CI: 68.8-75.5) were the ones with the lowest probability of discontinued health care. In contrast, illiterate older adults, those with an income of up to one times the minimum wage (80.2%, 95% CI: 65.9-84.4), and those with two or more comorbidities (78.9%, 95% CI: 75.4-82.4) were the most likely to interrupt health care. Figure 3 shows how the average discontinuation rate varied according to federation states represented by the participants in this study. Rio Grande do Sul, Santa Catarina, São Paulo, Acre, Goiás, Pará, Maranhão, Piauí, Paraíba, Alagoas, and Sergipe were the states with the highest discontinuation rates, ranging from 91.6% to 100%. Amazonas, Paraná, Rio de Janeiro, Espírito Santo, Ceará, and Rio Grande do Norte were the states with the lowest rates of discontinuation, from 56.2% to 85.6%. Figure 3. Variation in the discontinuation rate of health care for older adults representatives of each Brazilian state. Remobilize Study, 2020. Discussion In this study, we found that greater levels of social inequality (as measured by income and educational status) and the presence of multimorbidity and polypharmacy in community-dwelling older adults were related to an increased likelihood of discontinuation of health interventions, even after controlling for potential confounders factors. This study indicates an overall health care discontinuation rate of 64.4% among Brazilian older adults during the first months of the COVID-19 pandemic. Among all types of health care services, medical interventions were the type most often discontinued during the pandemic. Older adults with a higher number of comorbidities, low educational levels, and low income were more likely to experience discontinuation in their health care. On the other hand, wealthy older adults with high educational levels and polypharmacy were less likely to interrupt health care interventions during the COVID-19 pandemic. Furthermore, there were multiple and varied discontinuation rates of health care in different regions of Brazil. Although continuity of health care is influenced by a variety of factors, including a person's cultural background, context environment, personal beliefs, and system organization,25 our findings arguably suggest the existence of an association between social and health inequality and poor continuation of health care generally among older adults during the COVID-19 pandemic. Several studies have reported some type of discontinuation rate. Almeida and colleagues (2021) described a worsening in the health status of 29.4% of the participants and a need for continuity care of treatment in 25.5%. Cancellations of medical appointments were reported by 68.8% of Americans over 64 years old, and nearly half of them reported that surgeries or medical procedures were canceled due to the COVID-19 pandemic.1 In the same study, 1.4% of the participants reported not having the medical care they needed.1 Macinko and colleagues (2020) also reported that 17.2% of Brazilian adults aged 50 and over had their medical appointments and elective surgical canceled during pandemic. Although our results pointed out a higher discontinuation rate when compared with the Brazilian study5 and a similar rate when compared with the American study,1 our results encompass a wider range of health care activities, including rehabilitation services, medical appointments, surgeries, diagnostic procedures, and health promotion activities, while they only analyzed medical interventions. Furthermore, our study included participants aged 60 years and older, which is different from the age group selected in these studies (50 + years and 64 + years, respectively). Our results showed that health care discontinuation during the COVID-19 pandemic reflected Brazil's long-lasting structural social inequalities. The variables that explained higher rates of discontinuation in health interventions were income and educational inequality. In the literature, some studies have pointed to the widening of the social distance between the “haves” and “have nots,” together with literacy divides, as possible explanations for poorer population health outcomes; both have been described as directly proportional to seeking health care, besides being associated with difficulty in accessing the health system.33,34 This is incredibly daunting in Brazil, where older people are among the poorest half of the population, with an average monthly income of R$850 (US$166.25) in 2019.35 In this population, illiteracy rates reach 18.0% and are even higher among older adults who are black and brown (30.7%).36 It is worth highlighting that, in homes where older adults are not the only source of income, they are still responsible for 70.6% of the total household income.37 The economic and political crises of the 2010s had already reduced the relative advantage of low-income older adults in spending directly on health care.37 However, older adults with lower incomes have a 30% higher chance of having catastrophic expenditures (ratio between disbursement for health and total household income) than those with higher income.38 In 2020, with the COVID-19 pandemic, there was an 18% loss in the income generated by work in the Brazilian population.37 Although most Brazilian older adults count on income from pensions and retirement (73.6%), researchers found that an even more significant decline in income for this group (22.0%)37 could be explained by the reduction in non-essential work activities during the pandemic period.38 It was possible to observe a gradient effect between higher income and lower chance of discontinued health care, highlighting that income inequality influences health care adherence. Contradictory to our results, Macinko and colleagues (2020)5 showed a higher prevalence of medical appointments cancellation (PR = 1.77; 95% CI: 1.32-2.38) among older adults with high educational levels (more than nine years of schooling). However, our results were not restricted to doctor appointments. We also observed a lower likelihood of discontinued health care among older adults with more than eight times the minimum wage. Older people with a high income and high educational level have a wider range of alternatives to continue their health care, such as using digital technologies (eg, telemedicine, telerehabilitation),39 private cars for transportation to health care appointments, and exams that are frequently covered by their insurance packages. Curiously, multimorbidity and polypharmacy markedly influenced discontinuation of interventions. Previous studies show that people with multimorbidities present 30% more probability of seeking health care.20,40 Patients with multimorbidity and polypharmacy are recognized as high utilizers of health care resources. The COVID-19 pandemic led to unique daily life experiences for members of the older population because they were recommended to stay at home and restrict social contact. A substantial increase in doctor appointments cancellations of 1.5 and 1.9 times higher among older adults with two and three comorbidities was also observed.5 Fear of being infected and the widespread information that older people with multimorbidity were at a high risk of death and intensive care unit hospitalization possibly explain this population's adherence to social distancing.4 Older adults who used polypharmacy were less likely (OR = 0.61 95% CI: 0.46-0.81) to discontinue their health treatments. This can be explained by the fact that many medications need a prescription to be purchased. Older adults who attend regular medical appointments have a 1.9 times greater chance of using polypharmacy.41 Although the use of multiple medications is not ideal for older adults due to drug interactions, the ideology of medicalization and the health care model centered on exposure to drugs still prevail as a concept for treating diseases.41 The probability of health care discontinuation was lower in all age groups among those who consumed four or more medications simultaneously (0.70%, 0.69%, and 0.67%, respectively, for 60-69 years old, 70-79 years old, and 80 + years old) compared to those who did not use polypharmacy. During the COVID-19 pandemic, essential initiatives to guide patients and physicians were necessary for keeping compliance with drug treatment. As a strategy to minimize long-term negative results, the European Society of Cardiology issued an information note for physicians and patients not to discontinue antihypertensive drug treatment, as it could lead to a greater risk of infection and severity of COVID-19.42 The Brazilian government recommended actions to avoid interruptions in health care,43 but with a fragmented and uncoordinated effort. Other institutions, such as Fiocruz, also developed documents to support actions for the continuity of health care among specific groups, such as individuals under mental health care. Our results also revealed that about a third of the older adults reduced their practice of physical exercise during the COVID-19 pandemic. Increased sedentary behavior was reported by 14.9% of the Brazilian population during the COVID-19 pandemic compared to the pre-pandemic period. Motivation to perform exercise changed for the Brazilian general population before and after the pandemic, decreasing 4.4% for performance-related motivation and increasing 6.7% for health-related motivation.13 Our results showed higher interruption values, which can be explained because they are related to an older sample. Foreseen consequences of an older population that discontinued crucial health care actions such as physical activity and disease management are alarming. If this reality remains in the medium and long terms, a higher rate of falls, cardiorespiratory problems, obesity, and difficulty in engaging with medication and worse psychological conditions could be seen shortly.44 Additionally, Brazil has huge social and health inequities,23 which were deeply worsened during the COVID-19 pandemic. There is an expected increase in health demand in post-COVID-19 treatment, which will add to the consequences of the discontinued health care and overload the Unified Health System, which is already fragile in terms of human and financial resources and poorly prepared for long-term care. The COVID-19 pandemic amplified substantial health care inequities. Although we cannot infer causality, our study provides some insights into changes to access to health interventions and ways to customize continuity based on socioeconomic and health conditions. Indeed, our findings of an association between more significant social and health inequality levels among older adults and an increasing likelihood that respondents would discontinue their health care may support this theory. Intervention support could be strengthened among older adults with a low socioeconomic level, including emphasis on self-health management and routine care, creation of an alternative for health interventions, and deepening of public understanding of access and utilization.45 Restriction of social interaction and mobility during the pandemic among older adults confined at home needs to be actively monitored by health services. The rapid pace of change during the pandemic revealed the limitations of the delivery modes and raises questions about whether our current health care system, and its financing, can support these changes and ensure that they improve quality and equity.46 The most immediate changes were the scaling up of telehealth and in-home care, which may be particularly useful in managing both care of older adults patients and routine visits. For example, as an alternative to face-to-face health care, the Federal Councils of health professionals recognized teleassistance during the COVID-19 pandemic as an option.47 Although promising, this resource does not reach a large part of the Brazilian population, especially older people with low educational levels and low income, who are much more likely to have digital illiteracy and poor access to the Internet. Our study suggests that public policies in Brazil designed to improve access to and use of health services for the population may work best when supported by policies to promote greater economic equality among Brazilian older adults, even considering the pandemic context. Limitations Although the present study is one of the few carried out on this topic during the pandemic and including participants living in different regions of Brazil, it is essential to highlight some limitations. First, we included a convenience sample using a digital interface, the interviews were conducted online and by phone calls, and there was unequal distribution between geographic macro-regions. Using a digital interface may have excluded some groups of older people and therefore does not guarantee our sample represents Brazil's older population in general. It is noteworthy that women, people who define themselves as white, and people with higher educational levels account for a disproportionate share of our sample in comparison to national data. This limits the external validation of our study, but it was the only feasible approach in the context of the COVID-19 pandemic. A second limitation is that the study has a cross-sectional design, precluding causal or risk inferences. For example, some older adults could have stopped using health services due to no longer needing them or completing treatment for acute or temporary conditions. There is a need for future longitudinal studies to explore these issues and to identify specific strategies for continuing treatments under unfavorable circumstances. Third, memory bias may have reduced the reliability of participant responses about their health care before the COVID-19 pandemic. Conclusion There appears to be a strong association between social and health inequality and the likelihood that older adults will self-report discontinuation health care during the COVID-19 pandemic period. High discontinuation rates in health interventions among Brazilian older adults were identified, and individuals with multimorbidity, low income, and low educational levels were more affected during the pandemic. Strategies for maintaining services within social protection measures could partially mitigate the adverse effects of discontinuity in health care during the COVID-19 pandemic. Proactive and coordinated actions to seek out vulnerable older groups that interrupted treatments are urgently needed to mitigate the negative consequences of discontinued health care. While addressing unmet needs remains a priority, studies investigating the impacts of discontinuation among specific groups might help avoid redundancies in health care provision and better allocate health systems resources in the post-pandemic era. Author Biographies Juleimar Soares Coelho de Amorim, PhD, is a teacher at the Federal Institute of Rio de Janeiro (IFRJ). He holds a PhD in public health (aging epidemiology) from the Oswaldo Cruz Foundation (FIOCRUZ) and a master’s degree in rehabilitation sciences from the State University of Londrina. His research and practices are focused on geriatric health and rehabilitation, aging, and health education. He's also a member of IFRJ's Ethics Research Committee. Giulianna Ornellas is a physical therapy graduation student at the Federal Institute of Rio de Janeiro. Peter Lloyd-Sherlock, PhD, is a professor at the University of East Anglia's School of International Development. He holds a PhD in economics from the London School of Economics. His experience includes health politics for older people in low-income countries. He's the leader of a study about health services for vulnerable older Brazilians at the Oswaldo Cruz Foundation (FIOCRUZ). He also has been a consultant for the World Health Organization's Primary Healthcare for Older People Programme. Daniele Sirineu Pereira, PhD, is a teacher at Federal University of Minas Gerais (UFMG). She holds postdoctoral, PhD, and master's degree in rehabilitation sciences from UFMG. Her research is focused on geriatric health and rehabilitation. Alexandre da Silva, PhD, is a teacher at Medical” Jundiaí High School. He holds a PhD in public health from São Paulo University and a master's degree in rehabilitation from Federal University of de São Paulo. His research is focused on health politics, racism, inequalities, and aging. He is also a member of the work group Racism and Health at the Brazilian Public Health Organization. Etienne Duim, PhD, is a teacher at Israelita Albert Einstein High School. She holds a PhD in epidemiology from São Paulo University, with a doctoral stay at Maastricht University. Her research is focused on public health and aging. Camila Astolphi Lima, PhD, is a post-doctoral student in physical therapy at UNICID, with a doctoral stay at the University of Alberta. She has experience in geriatric health care, which is also the focus of her research. Monica Rodrigues Perracini, PhD, is a teacher at the master's and doctoral program at City University of São Paulo and State University of Campinas (UNICAMP). She holds a post-doctoral post at The George Institute for Global Health, as well as a PhD in rehabilitation from Federal University of São Paulo and a master's degree in education from UNICAMP. She is also a board member at the Fragility Fracture Network; a consultant for the Global Network on Long-Term Care, which is tied to the World Health Organization Program “Global Strategy and Action Plan on Ageing and Health”; and the coordinator of the Remobilize Study. Declaration of Conflicting Interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding: This work was orted by the Instituto Federal do Rio de Janeiro (IFRJ). Professor Monica Perracini has received a researcher productivity grant (309838/2017-7) from the Brazilian National Council for Scientific and Technological Development. Author contribution: G.O participated in data collection, conception to manuscript, interpretation, final revision. M.R.P participated in concept design, data collection, data analysis, interpretation, drafting, critical revision, and approval of the article. D.S.P, A.S., and E.D participated in concept design, interpretation, critical revision, and approval of the article. C.A.L.: participated in concept design, conception to manuscript, interpretation, critical revision, and approval of the article. J.S.C.A participated in data collection, data analysis, interpretation, critical revision, and approval of the article. ORCID iD: Juleimar Soares Coelho de Amorim https://orcid.org/0000-0003-3218-1769 ==== Refs References 1 Heid AR Cartwright F Wilson-Genderson M Pruchno R . Challenges experienced by older people during the initial months of the COVID-19 pandemic. Gerontologist. 2021;61 (1 ):48–58. doi:10.1093/geront/gnaa138 32955079 2 Fuller HR Huseth-Zosel AH . Lessons in resilience: initial coping among older adults during the COVID-19 pandemic. Gerontologist. 2021;61 (1 ):114–125. doi:10.1093/geront/gnaa170 33136144 3 Coronavírus Brazil. Ministry of Health. Brazil, 2021. Available from: https://covid.saude.gov.br. [Cited 2021 Jun 30]. 4 Lima-Costa MF Mambrini JVM Andrade FB Peixoto SWV Macinko J . Social distancing, use of face masks, and hand washing among participants in the Brazilian longitudinal study of aging: the ELSI-COVID-19 initiative. Cad Saúde Pública. 2020;36 (3 ):e00193920. doi:10.1590/0102-311X00193920 5 Macinko J Woolley NO Seixas BV Andrade FB Lima-Costa MF . Health care seeking due to COVID-19 related symptoms and health care cancellations among older Brazilian adults: the ELSI-COVID-19 initiative. Cad Saúde Pública. 2020;36 (3 ):3:e00181920. doi:10.1590/0102-311X00181920 6 Esme M Koca M Dikmeer A , et al. Older adults with coronavirus disease 2019: a nationwide study in Turkey. J Gerontol A Biol Sci Med Sci. 2021;76 (3 ):e68–e75. doi:10.1093/gerona/glaa219 32871002 7 Couteur DG Anderson RM Newman AB . COVID-19 through the lens of gerontology. J Gerontol A Biol Sci Med Sci. 2020;75 (9 ):e119–e120. doi:10.1093/gerona/glaa077 32222763 8 Ranzani OT Bastos LS Gelli JGM , et al. Characterisation of the first 250000 hospital admissions for COVID-19 in Brazil: a retrospective analysis of nationwide data. Lancet Respir Med. 2021;9 (4 ):407–418. doi:10.1016/S2213-2600(20)30560-9 33460571 9 Terracciano A Stephan Y Aschwanden D , et al Changes in subjective age during COVID-19. Gerontologist. 2021;61 (1 ):13–22. doi:10.1093/geront/gnaa104 32766780 10 Bui CN Peng C Mutchler JE Burr JA . Race and ethnic group disparities in emotional distress among older adults during the COVID-19 pandemic. Gerontologist. 2021;61 (2 ):262–272. doi:10.1093/geront/gnaa217 33367785 11 Schrack JA Wanigatunga AA Juraschek S . After the COVID-19 pandemic: the next wave of health challenges for older adults. J Gerontol A Biol Sci Med Sci. 2020;75 (9 ):e121–e122. doi:10.1093/gerona/glaa102 32315025 12 Kim HH Jung JH . Social isolation and psychological distress during COVID-19 pandemic: a cross-national analysis. Gerontologist. 2021;61 (1 ):103–113. doi:10.1093/geront/gnaa168 33125065 13 Sonza A Sá-Caputo DC Bachur JA Araújo MGR Trippo KV Gama DRN . Brazil Before and during COVID-19 pandemic: impact on the practice and habits of physical exercise. Acta Biomed. 2020;92 (1 ):e2021027. doi:10.23750/abm.v92i1.10803 33682804 14 OPAS. Organização Pan-americana de Saúde. Considerações sobre a reabilitação durante o surto de COVID-19. 23 p., 2020. Portuguese. 15 WHO. World Health Organization. Pulse survey on continuity of essential health services during the COVID-19 pandemic. 21 p., 2020. 16 Bettger JP Thoumi A Marquevich V , et al. COVID-19: maintaining essential rehabilitation services across the care continuum. BMJ Glob Health. 2020;5 (5 ):1–7. doi:10.1136/bmjgh-2020-002670 17 Moynihan R Sanders S Michaleff ZA , et al. Impact of COVID-19 pandemic on utilization of healthcare services: a systematic review. BMJ Open. BMJ Open. 2021;11(3):e045343. doi: 10.1136/bmjopen-2020-045343 18 Bambra C Riordan R Ford J Matthews F . The COVID-19 pandemic and health inequalities. J Epidemiol Community Health. 2020;74 (11 ):964–968. doi: 10.1136/jech-2020-214401 32535550 19 Andersen R Newman JF . Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc. 2005;83 (4 ):1–28. doi:10.1111/j.1468-0009.2005.00428.x 20 Levorato CD Mello LM Silva AS Nunes AA . Factors associated with the demand for health services from a gender-relational perspective. Cien Saude Colet. 2014;19 (4 ):1263–1274. doi:10.1590/1413-81232014194.01242013 24820609 21 Dellaroza MSG Pimenta CAM Lebrao ML Duarte YA . Associação de dor crônica com uso de serviços de saúde em idosos residentes em São Paulo. Rev Saúde Pública. 2013;47 (5 ):914–922. Portuguese. doi:10.1590/S0034-8910.2013047004427. 24626496 22 Silva AMM Mambrini JVM Peixoto SV Malta DC Lima-Costa MF . Use of health services by Brazilian older adults with and without functional limitation. Rev Saúde Pública. 2017;51 (1 ):1s–10s. doi:10.1590/S1518-8787.2017051000243 23 Osorio RG Servo LMS Piola SF . Unmet health care needs in Brazil: an investigation about the reasons for not seeking health care. Cien Saude Colet. 2011;16 (9 ):3741–3754. doi:10.1590/S1413-81232011001000011 21987318 24 Paim J Travassos C Almeida C Bahia L Macinko J . The Brazilian health system: history, advances, and challenges. Lancet. 2011;377 (9779 ):1778–1797. doi:10.1016/S0140-6736(11)60054-8 21561655 25 WHO. World Health Organization. Continuity and coordination of care. 68 p., 2018. 26 Biernacki P Waldorf D . Snowball sampling: problems and techniques of chain referral sampling. Sociol Methods Res. 1981;10 (2 ):141–163. doi:10.1177/004912418101000205 27 Carrasquillo O . Health care utilization. In: Gellman MD Turner JR , eds. Encyclopedia of behavioral medicine. Springer; 2013: 136 p. 28 WHO. World Health Organization. Health systems: improving performance. 215 p., 2000. 29 WHO. World Health Organization. Medication without harm. 2017. 30 Veras RP Caldas CP Cordeiro HA . Models of health care for the elderly: rethinking the meaning of prevention. Physis. 2013;23 (4 ):1189–1213. doi:10.1590/S0103-73312013000400009 31 WHO. World Health Organization. Multimorbidity: technical series on safer primary care. 24 p., 2016. 32 Lins AES Simon KA Ramos LR . Functional performance in elderly women from an open university for the elderly in a northeast urban area. Geriatr Gerontol Aging. 2013;7 (1 ):20–27. 33 Almeida WS Szwarcwald CL Malta DC , et al. Changes in Brazilians’ socioeconomic and health conditions during the COVID-19 pandemic. Rev Bras Epidemiol. 2021;6(23):e200105. doi:10.1590/1980-549720200105 34 Singu S Acharya A Challagundla K Byrareddy SN . Impact of social determinants of health on the emerging COVID-19 pandemic in the United States. Front Public Health. 2020;8:406. doi:10.3389/fpubh.2020.00406 35 IBGE. Pesquisa nacional por amostra de domicílios contínua: educação 2019. Instituto Brasileiro de Geografia e Estatística; 2020. 36 Faustino CG Levy RB Canella DS Oliveira C Novaes HMD . Income and out-of-pocket health expenditure in living arrangements of families with older adults in Brazil. Cad Saúde Pública. 2020;36 (3 ):e00040619. doi:10.1590/0102-311X00040619 32267373 37 Camarano AA . Depending on the income of older adults and the coronavirus: orphans or newly poor? Cien Saude Colet. 2020;25 (2 ):4169–4176. doi:10.1590/1413-812320202510.2.30042020 33027353 38 Bernardes GM Saulo H Fernandez RN Lima-Costa MF Andrade FB . Catastrophic health expenditure and multimorbidity among older adults in Brazil. Rev Saúde Pública. 2020;54 (125 ):1–11. doi:10.11606/s1518-8787.2020054002285 39 Bardi G Bezerra WC Monzeli GA Pan LC Braga IF Macedo MDC . Pandemic, social inequality and necropolitics in Brazil: reflections from social occupational therapy. Revisbrato. 2020;4 (2 ):496–508. doi:10.47222/2526-3544.rbto34402 40 Souza ASS Faerstein E Werneck GL . Multimorbidity and use of health services by individuals with restrictions on habitual activities: the Pró-Saúde Study. Cad Saúde Pública. 2019;35 (11 ):e00155118. doi:10.1590/0102-311X00155118 31691782 41 Pereira KG Peres MA Iop D , et al. Polypharmacy among the elderly: a population-based study. Rev Bras Epidemiol. 2017;20 (2 ):335–344. doi:10.1590/1980-5497201700020013 28832855 42 Simone G . Position statement of the ESC Council on hypertension on ACE-inhibitors and angiotensin receptor blockers. ESCardio; 2020. 43 Ministério da Saúde. COVID-19 guia orientador para o enfrentamento da pandemia na Rede de Atenção à Saúde. 4th ed. Brasília, 2021. Available https://www.conasems.org.br/wp-content/uploads/2021/04/Covid-19_guia_orientador_4ed.pdf Access 16 Aug 2021. 44 Palmer K Monaco A Kivipelto M , et al. The potential long-term impact of the COVID-19 outbreak on patients with non-communicable diseases in Europe: consequences for healthy ageing. Aging Clin Exp Res. 2020;32(7):1189–1194. doi:10.1007/s40520-020-01601-4 45 Werner RM Glied SA . COVID-induced changes in health care delivery – can they last? N Engl J Med. 2021;385 (10 ):868–870. doi: 10.1056/NEJMp2110679 34449184 46 Werner RM Hoffman AK Coe NB . Long-term care policy after COVID-19 — solving the nursing home crisis. N Engl J Med. 2020;383(10):903–905. doi: 10.1056/NEJMp2014811 47 Caetano R Silva AB Guedes ACCM , et al. Challenges and opportunities for telehealth during the COVID-19 pandemic: ideas on spaces and initiatives in the Brazilian context. Cad Saúde Pública. 2020;36 (5 ):e00088920. doi:10.1590/0102-311x00088920 32490913

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.