title stringlengths 1 1.19k | keywords stringlengths 0 668 | concept stringlengths 0 909 | paragraph stringlengths 0 61.8k | PMID stringlengths 10 11 |
|---|---|---|---|---|
Data availability | The datasets generated during and/or analyzed during the current study are available from the corresponding authors on reasonable request. | PMC9870900 | ||
Competing interests | The authors declare no competing interests. | PMC9870900 | ||
References | PMC9870900 | |||
Background | RECURRENCE, ATRIAL FIBRILLATION (AF) | A number of models have been reported for predicting atrial fibrillation (AF) recurrence after catheter ablation. Although many machine learning (ML) models were developed among them, black-box effect existed widely. It was always difficult to explain how variables affect model output. We sought to implement an explainable ML model and then reveal its decision-making process in identifying patients with paroxysmal AF at high risk for recurrence after catheter ablation. | PMC9936738 | |
Methods | Between January 2018 and December 2020, 471 consecutive patients with paroxysmal AF who had their first catheter ablation procedure were retrospectively enrolled. Patients were randomly assigned into training cohort (70%) and testing cohort (30%). The explainable ML model based on Random Forest (RF) algorithm was developed and modified on training cohort, and tested on testing cohort. In order to gain insight into the association between observed values and model output, Shapley additive explanations (SHAP) analysis was used to visualize the ML model. | PMC9936738 | ||
Results | RECURRENCES, RECURRENCE | In this cohort, 135 patients experienced tachycardias recurrences. With hyperparameters adjusted, the ML model predicted AF recurrence with an area under the curve of 66.7% in the testing cohort. Summary plots listed the top 15 features in descending order and preliminary showed the association between features and outcome prediction. Early recurrence of AF showed the most positive impact on model output. Dependence plots combined with force plots showed the impact of single feature on model output, and helped determine high risk cut-off points. The thresholds of CHA | PMC9936738 | |
Conclusion | RECURRENCE, PAROXYSMAL ATRIAL FIBRILLATION | An explainable ML model effectively revealed its decision-making process in identifying patients with paroxysmal atrial fibrillation at high risk for recurrence after catheter ablation by listing important features, showing the impact of every feature on model output, determining appropriate thresholds and identifying significant outliers. Physicians can combine model output, visualization of model and clinical experience to make better decision. | PMC9936738 | |
Supplementary Information | The online version contains supplementary material available at 10.1186/s12872-023-03087-0. | PMC9936738 | ||
Keywords | PMC9936738 | |||
Introduction | drug-refractory | ATRIAL FIBRILLATION (AF) | One of the primary treatment options for atrial fibrillation (AF) is rhythm control. Catheter ablation is the first-line treatment for drug-refractory paroxysmal AF, and pulmonary vein isolation, the cornerstone of AF ablation, eradicates approximately 90% of AF triggers in principle [Numerous studies have proposed a variety of prediction models, such as the HATCH, APPLE, or CAAP-AF score; however, the results differ greatly [The use of machine learning (ML) algorithms in medicine has been gaining popularity and is helping physicians in clinical decision-making. ML algorithms can learn the association between multiple patient variables and clinical outcomes automatically. A large number of ML algorithms are non-parametric and are not limited by variable collinearity. Moreover, ML algorithms are appropriate for processing high-dimensionality data, given reasonable optimization [ | PMC9936738 |
Methods | PMC9936738 | |||
Study design and population | AF, mitral stenosis, drug-refractory | MITRAL STENOSIS, PRIMARY CARDIOMYOPATHY, HYPERTHYROIDISM, HYPERTROPHIC CARDIOMYOPATHY, SICK SINUS SYNDROME | A consecutive cohort of patients was included in the study from the First Affiliated Hospital of Air Force Medical University between January 2018 and December 2020. Patients over 18 years old with drug-refractory paroxysmal AF, who received their first catheter ablation procedure were eligible. Exclusion criteria included: (i) persistent, long-standing persistent or permanent AF; (ii) valvular AF, which was defined as AF occurring with moderate or severe mitral stenosis or surgical valve replacement; (iii) AF with primary cardiomyopathy (e.g., Hypertrophic Cardiomyopathy); (iv) reversible AF (e.g., AF associated with hyperthyroidism); (v) suspected compensated AF because of Sick Sinus Syndrome; (vi) patients who failed to follow-up. The electronic medical record system of the Information Department of the hospital showed 471 patients who met the criteria in the above period and were thus enrolled in the study. This study conformed to Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement and was approved by the ethics committee of the hospital, according to the principles of the Declaration of Helsinki. Written informed consent was acquired by all 471 patients before radiofrequency or cryoballoon ablation. | PMC9936738 |
Data collection | ATRIAL FIBRILLATION | With "Atrial Fibrillation" as the search term, all patients hospitalized in the department of cardiovascular medicine from January 2018 to December 2020 were retrieved, and basic information of patients, including age, sex and ID number, were collected. Retrieval logic written in R language was used to automatically retrieve other relevant information such as medical history, physical examination, echocardiogram indexes, laboratory examination, procedural recordings, discharge medication and follow-up records, from the medical record management system. Basic information and data extracted by the retrieval logic were reviewed by two expert physicians to ensure accuracy. Patients were excluded based on the joint consensus of the two physicians if their information did not meet the criteria. The data collection flowchart is shown in Fig. Study flowchart. LAAC left atrial appendage closure, PVI, pulmonary vein isolation | PMC9936738 | |
Feature selection | embolism | EMBOLISM, INFLAMMATION | To ensure the robustness of the ML classification model, features were selected if they: (i) characterized the cardiac function, risk of systemic embolism or the level of inflammation, or (ii) promoted the occurrence and development of AF. Laboratory and echocardiographic characteristics were considered as missing values if they were not collected four weeks prior to the ablation procedure. Characteristics with more than 15% of missing values were excluded. Finally, 30 features were selected. Details of the features are presented in Additional file | PMC9936738 |
End point and follow-up | atrial flutter, atrial tachycardia | RECURRENCE, RECURRENCE, ATRIAL TACHYCARDIA, ATRIAL FLUTTER, INFLAMMATORY RESPONSE | Recurrence of AF post-procedure was defined as the first episode of any type of atrial tachycardia, including but not limited to atrial flutter and AF, sustained for more than 30 s after the blanking period (BP). BP refers to the first three months post-procedure. In this period, any occurrence of atrial tachycardia was identified as early recurrence of AF (ERAF), which may relate to inflammatory response induced by electrical burn or freezing injury, rather than a true relapse [After the catheter ablation procedure, the outpatient follow-up was scheduled at one, three, six and every six months thereafter. At each visit, atrial tachycardia episodes were captured and confirmed by 12-lead electrocardiogram, Holter or ambulatory monitors. Communication with patients was established through online or telephone follow-up to ensure timely feedback on patients or inquire about the control of AF. | PMC9936738 |
Explainable ML model | The Random Forest (RF) algorithm was used in the building of an explainable ML model. | PMC9936738 | ||
Classifier illustration | DTS | RF is an ensemble algorithm based on the Decision Tree (DT). The basic idea of RF is bagging: features selected into the RF model are simultaneously voted by all independent DTs in the forest, and this process conforms to the principle of majority subordination. RF integrates the results of all DTs, obtaining higher model performance than any individual DT, while avoiding overfitting. At the same time, this algorithm can effectively solve the problem of feature collinearity because of its “if…else” calculation logic [ | PMC9936738 | |
Model development, validation and testing | overfitted | DTS | The median of each feature was used to fill in the missing values, preserving the distribution of data to the greatest extent. However, underestimation of the weight of features using this method is possible. All patients and their specific characteristics in this cohort were randomly divided into the training and testing cohort (70% and 30% of the dataset, respectively). Because most of the hyperparameters are overfitted by default, a validation set, as an internal testing cohort, was needed to optimize the hyperparameters to ensure the better generalization of the ML model. This method enabled the loss reduction, model modification and determination of the weight of every selected feature. Therefore, the training cohort was randomly split into a training set and validation set in the ratio of 9:1. A tenfold cross validation method was used to evaluate the accuracy of the model in the training cohort to optimize the hyperparameters. The split of the datasets is shown in Fig. The optimal hyperparameters based on accuracy were determined step by step: first, learning curves were plotted to select the appropriate number of DTs, which may effectively reduce calculation time and avoid introducing too many systematic errors; second, repeated grid search was employed to determine the partial optimal hyperparameters, which were overfitted by default; then, adjusted feature numbers were included in the ML model according to the fitting situation of model; finally, according to the above results, grid search was executed to find out the global optimal hyperparameters. | PMC9936738 |
Model explanation | Shapley additive explanations (SHAP) analysis is derived from a game theory concept and has the advantage of gaining insight into the association between observations and clinical outcome [ | PMC9936738 | ||
Statistical analysis | SD | REGRESSION, RECURRENCE | Because the missing values for most laboratory examination and echocardiographic indexes were replaced with their medians, they were described as the median (IQR) and compared using Mann–Whitney U test. Continuous variables with normal distribution were expressed as the mean ± SD and compared by one-way ANOVA. Categorical data were described as numbers and percentages, and the Chi square test or Fisher’s exact test were used where appropriate. Receiver Operating Characteristic (ROC) curve, Precision-Recall curve and Decision Curve Analysis were used on the testing cohort to show the ability of ML model to classify correctly. Youden index was used to determine the threshold of each continuous covariate in the training cohort. Kaplan–Meier curve was used to describe time to AF recurrence. Univariable and multivariable Cox proportional hazard regression analyses were used to explore the potential risk factors. Risk ratio (RR) with 95% confidence interval (CI) was used to describe the potential risk factors. A two-sided | PMC9936738 |
Results | PMC9936738 | |||
Patients and baseline characteristics | tachycardias | RECURRENCES | A total of 1496 patients were screened in this study. Finally, 471 patients were enrolled. The median follow-up time was 25 months (IQR: 13–36 months). Patient follow-up totaled an aggregate of 954.3 patient-years with 135 patients (14.2/100 patient-years) experiencing tachycardias recurrences (Additional file The entire cohort was divided based on the clinical outcome and 336 patients were assigned into the sinus rhythm maintenance group. Baseline characteristics are shown in Table Patient characteristics at baselineData are mean ± SD, median (IQR) or n (%). Missing values were filled in with the medianCHA | PMC9936738 |
Model development | DBP diastolic blood pressure | ATRIAL FIBRILLATION, RECURRENCE | Based on the 7:3 ratio, 329 and 142 patients were randomly assigned into the training and testing cohorts, respectively. Baseline characteristics are summarized in Table Baseline characteristics between training cohort and testing cohortData are mean ± SD, median (IQR) or n (%)AF duration indicates time since first AF diagnosis. BMI body mass index, DBP diastolic blood pressure, SBP systolic blood pressure, AF paroxysmal atrial fibrillation, RAAS renin–angiotensin–aldosterone system, ERAF early recurrence of atrial fibrillation. *Model performance in testing cohort | PMC9936738 |
SHAP analysis | SHAP dependence | The top 15 features with descending importance ranked by mean absolute SHAP values are shown in the summary plot (Fig. Summary plot for ML model. Summary plot Decision plot and force plot for ML model. SHAP dependence plots show the impact of a single feature on outcome prediction (Fig. The impact of single feature on the outcome prediction. Dependence plots and force plots showed how single features influenced model output. Red dashed lines represented SHAP = 0. Thresholds were indicated by orange arrows. AF atrial fibrillationFor continuous variables, the influence of outliers on model output is visualized in a decision plot (Fig. | PMC9936738 | |
Statistical significance of thresholds | SBP systolic blood pressure | REGRESSION, ATRIAL FIBRILLATION | Statistical thresholds of continuous variables in the training cohort evaluated by Youden index were similar to those determined by SHAP analysis (Table Thresholds of continuous variables in training cohort based on Youden indexBMI body mass index, AF atrial fibrillationUnivariable Cox proportional hazard regression based on high-risk thresholds. AF atrial fibrillation, SBP systolic blood pressure, LAD left atrial diameter | PMC9936738 |
Discussion | RECURRENCE | In this study, an explainable ML model was developed and tested to reveal its decision-making process in identifying patients with paroxysmal AF at high risk for recurrence after catheter ablation. The top 15 features and their specific impact on outcome prediction were revealed through SHAP analysis. ERAF showed the most positive impact on model output, and female patients had a higher risk of AF recurrence compared to male patients. CHA | PMC9936738 | |
Risk factors identification | RECURRENCE | Risk assessment for AF recurrence after catheter ablation remains a topic worth exploring. According to current guidelines, despite the fact that a series of prediction scores have been evaluated, their prediction ability is moderate; the most powerful predictor is ERAF [ | PMC9936738 | |
Relationship between single feature and model output | SHAP analysis could illuminate the effect of every feature on the outcome from an objective perspective. CHA | PMC9936738 | ||
Threshold determination | A threshold determines the score of each continuous variable in the clinical model, for example, age in CHAThe thresholds of laboratory examination indexes evaluated by SHAP analysis were inaccurate. For example, the threshold of NT-proBNP was 52.28 pg/ml, which was less than the upper limit of regular medical reference. This was clearly an incorrect outcome. Stenwig et al. also reported that the decision-making process of LR did not correspond to common medical theory: a linear relationship was detected between temperature and mortality, with higher temperatures being associated with better prognosis [ | PMC9936738 | ||
Model performance in high-dimensional data | REGRESSION | Multivariable Cox regression results showed that ERAF, AF duration and left ventricular ejection fraction were statistically significant, with an AUROC value of 0.513 (Additional file Currently, there are conflicting reports as to whether ML models perform better than traditional statistical methods. Moncada et al. [Recently, three-dimensional data, from computer tomography or magnetic resonance images for example, have already been used in automatic feature selection by DL and risk stratification models have been built based on the selected features, to identify patients at high risk [ | PMC9936738 | |
Explainable ML model in clinical practice | Explainable ML model can make clinical practice more accurate. When the characteristics of each patient are entered into the model, the model first predicts whether he or she is at high risk. After that, we can understand the decision-making process of the model through SHAP analysis. According to the weight of different features and thresholds of features, physicians are also able to make rational clinical decisions. If there is any error in decision-making process, physicians can clearly see the unreasonable judgements in model prediction, to combine model outputs and clinical experience to further confirm whether patient is at high risk. Besides, with the increase of sample size in the training cohort of model, the accuracy of the model will also be improved. Perhaps human–computer interaction will become a trend in clinical practice in the future. | PMC9936738 | ||
Limitations | chronic obstructive pulmonary disease | CHRONIC OBSTRUCTIVE PULMONARY DISEASE | There were several limitations in this study. The sample size in our study was small. Median values were used to fill in the missing data for laboratory examinations and echocardiographic indicators, which may reduce power of test and representativeness. Potential risk factors proposed by other studies such as abnormal estimated glomerular filtration rate, and chronic obstructive pulmonary disease were not considered because of insufficient confirmed cases [ | PMC9936738 |
Conclusions | RECURRENCE, PAROXYSMAL ATRIAL FIBRILLATION | This study showed that the explainable machine learning model can effectively reveal its decision-making process in identifying patients with paroxysmal atrial fibrillation at high risk for recurrence after catheter ablation by listing important features, showing the impact of every feature on model output, determining appropriate thresholds and identifying significant outliers. Physicians could clearly see the unreasonable judgements in model prediction, to combine model outputs and clinical experience and assist in decision-making. Studies based on high-dimensional databases with large sample size are necessary to further confirm the universality of this finding. | PMC9936738 | |
Acknowledgements | Not applicable. | PMC9936738 | ||
Author contributions | MH | FY conceived and supervised the study. YM and DZ participated in the design of the study, performed statistical analysis of the data and drafted the manuscript. JX and HP participated in clinical follow-up. JL, LG and SZ participated in data collection and verification. MH revised this paper. All of the authors read and approved the final manuscript. | PMC9936738 | |
Funding | The present work was financially supported by General program of National Natural Science Funds of China (Grant No. 81970274). | PMC9936738 | ||
Availability of data and materials | The datasets used and/or analyzed during the current study can be available from the corresponding author on reasonable request. | PMC9936738 | ||
Declarations | PMC9936738 | |||
Ethics approval and consent to participate | This study confirmed to Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement and was approved by the ethics committee of the First Affiliated Hospital of Air Force Medical University according to the principles of the Declaration of Helsinki. Written informed consent were provided by all patients before procedure. | PMC9936738 | ||
Consent for publication | Not applicable. | PMC9936738 | ||
Competing interests | The authors declare no competing interests. | PMC9936738 | ||
References | PMC9936738 | |||
Background | type 2 diabetes mellitus, diabetes | DISEASE, TYPE 2 DIABETES MELLITUS, COMPLICATIONS, DIABETES | Medically underserved people with type 2 diabetes mellitus face limited access to group-based diabetes care, placing them at risk for poor disease control and complications. Immersive technology and telemedicine solutions could bridge this gap. | PMC10209787 |
Objective | diabetes | DIABETES | The purpose of this study was to compare the effectiveness of diabetes medical group visits (DMGVs) delivered in an immersive telemedicine platform versus an in-person (IP) setting and establish the noninferiority of the technology-enabled approach for changes in hemoglobin A | PMC10209787 |
Methods | TYPE 2 DIABETES MELLITUS | This study is a noninferiority randomized controlled trial conducted from February 2017 to December 2019 at an urban safety net health system and community health center. We enrolled adult women (aged ≥18 years) who self-reported African American or Black race or Hispanic or Latina ethnicity and had type 2 diabetes mellitus and HbA | PMC10209787 | |
Results | Of 309 female participants (mean age 55, SD 10.6 years; n=195, 63% African American or Black; n=105, 34% Hispanic or Latina; n=151 IP; and n=158 in VW), 207 (67%) met per-protocol criteria. In the intention-to-treat analysis, we confirmed noninferiority for primary outcomes. We found similar improvements in mean HbA | PMC10209787 | ||
Conclusions | diabetes | DIABETES | In this noninferiority randomized controlled trial, immersive telemedicine was a noninferior platform for delivering diabetes care, eliciting comparable glycemic control improvement, and enhancing patient engagement, compared to IP DMGVs. | PMC10209787 |
Trial Registration | ClinicalTrials.gov NCT02726425; https://clinicaltrials.gov/ct2/show/NCT02726425 | PMC10209787 | ||
Introduction | T2DM, diabetes | TYPE 2 DIABETES MELLITUS, DIABETES | Minority and low-income women with type 2 diabetes mellitus (T2DM) face widening disparities in diabetes care and clinical outcomes, highlighting the pressing need to improve diabetes care for underserved communities [Telehealth solutions have gained unprecedented traction with the onset of the COVID-19 pandemic. Early evidence has shown that virtual worlds (VWs) and virtual reality platforms are feasible and potentially more effective alternatives to IP programming [To our knowledge, the possibilities of avatar-based VW DSME have not been rigorously tested. We developed an immersive telemedicine platform, linking an interactive VW learning environment with videoconferencing software, to overcome the common barriers to diabetes group-based care while maintaining clinical effectiveness at scale. We implemented Women in Control 2.0 (WIC2) in 2015 to study the comparative effectiveness of delivering DMGVs in a VW versus the traditional IP classroom for women from Black, African American, Hispanic, or Latina backgrounds with uncontrolled T2DM (trial protocol in | PMC10209787 |
Methods | PMC10209787 | |||
Trial Design | T2DM | From February 2017 to October 2019, we recruited 17 cohorts of African American or Black or Hispanic or Latina women with uncontrolled T2DM. A total of 309 participants were enrolled and randomly assigned to the VW or IP DMGV conditions. Participants attended 8 weekly DMGVs and were followed for 6 months. | PMC10209787 | |
Participants | T2DM | Eligible participants were adult women (≥18 years) who self-identified as African American, Black, Hispanic, or Latina with uncontrolled T2DM, defined by a hemoglobin A | PMC10209787 | |
Recruitment | PMC10209787 | |||
Overview | We identified participants from Boston Medical Center and a local community health center using a weekly electronic medical record query. We contacted eligible participants with an introductory letter and follow-up call [ | PMC10209787 | ||
Randomization and Masking | After stratification by language, we used one-to-one block randomization (alternating blocks of 6 and 8) to assign participants to the VW or IP DMGV conditions. A biostatistician generated the randomization sequence, and randomization occurred after informed consent. We randomized participants prior to obtaining baseline data. Investigators were blinded to the randomization process, but assignments were revealed to participants and investigators post consent. | PMC10209787 | ||
Intervention | diabetes | DIABETES | Assigned to cohorts of 6-12 participants based on study arm, participants convened in clinical or virtual settings for 8 weekly DMGVs. Each session lasted approximately 120 minutes and started with the completion of an intake form to document acute or chronic symptoms, health system usage, and self-management activities, followed by the measurement of vital signs and the delivery of DSME. Sessions included a one-on-one clinical consult. Study clinicians were 4 board-certified physicians and 2 nurse practitioners. Nonclinical group facilitators received training on core DSME topics and facilitation skills from lead faculty (SEM, PG). All participants received a paper curriculum booklet.Prior to the first virtual DMGV, staff provided laptops and wireless internet to VW participants and conducted IP computer training. All participants then met weekly for 8 weeks, according to a session schedule. During DMGVs, all participants received the same WIC2 curriculum, which was adapted from During each session, a clinician met individually with participants (in a separate physical space or via secure telehealth platform or telephone depending on study group) to review blood glucose readings and hyper or hypo glycemic data, conduct diabetes medication reconciliation, and address concerns. Recommendations for medication adjustments were based on an algorithm [To ensure fidelity of DMGV protocols and standard operating procedures, we used checklists, audits of session recordings, and participant observation field notes.Following the 8-week DMGV sessions, participants entered a 16-week maintenance period. They were encouraged to self-monitor (tracking blood glucose, blood pressure, diet, and exercise) using a paper booklet or mobile app. No formal DMGVs occurred.Illustration of avatars in the virtual world. | PMC10209787 |
Outcomes | The primary outcomes were mean changes in (1) HbA | PMC10209787 | ||
Data Collection and Management | Baseline data collection included sociodemographic characteristics, HbA | PMC10209787 | ||
Sample Size Calculations | We used the average overtime change in HbA | PMC10209787 | ||
Statistical Analysis | SENSITIVITY | Sociodemographic characteristics were compared by arm using chi-square and Fisher exact tests as appropriate for categorical variables and 2 sample Accelerometry data was used to calculate participants’ mean change in physical activity behavior from baseline to 6 months. For each participant, we randomly selected the 2 weekdays with the longest wear-time. We considered missing wear time data in a 24-hour day as sedentary activity. For each weekday, we estimated total MET-hours by a weighted sum of the number of hours in light (1.5 MET), moderate (4 MET), vigorous (6 MET), and very vigorous (8 MET) activity as measured by the accelerometer using the Freedson et al cut points [Sensitivity analyses were performed to evaluate the influence of language preference on our primary outcome results. Participant characteristics with, versus without, baseline HbA | PMC10209787 | |
Ethics Approval | This study was conducted according to the CONSORT (Consolidated Standards of Reporting Trials) guidelines [ | PMC10209787 | ||
Results | PMC10209787 | |||
Fidelity | EVENTS | The 17 study cohorts were conducted with 98% fidelity to the 8-week curriculum. The median number of sessions attended by participants was 6 in the IP arm and 7 in the VW arm. Among participants who attended WIC2 DMGVs, 98.2% (1618/1648 total events) completed the clinician consult and intake forms. In the VW condition, participants completed the clinical consult via telehealth (540/823, 65.6%), telephone (54/823, 6.6%), or either modality (230/823, 27.9%). | PMC10209787 | |
PP Results | Among the 207 participants who attended at least 6 DMGVs, within-group mean HbA | PMC10209787 | ||
Lifestyle Behaviors | diabetes | DIABETES | Self-management behaviors were assessed through a weekly self-report to detect changes in diet, exercise, and diabetes-related medication. Nearly a third of all participants (89/309, 28.8%) reported ≥1 dietary change, with a greater proportion in the VW group compared to the IP group (55/158, 34.8% vs 34/151, 22.5%).Of all participants, 65.7% (203/309) engaged in at least 20 minutes of exercise weekly during the intervention (VW: 106/158, 67.1% vs IP: 97/151, 64.2%). Only 17.2% (53/309) reported any type of change (increase, decrease, or switch) in their diabetes medication regimen (VW: 29/158, 18.4% vs IP: 24/151, 15.9%). | PMC10209787 |
Adverse Events | ADVERSE EVENT | One study-related severe adverse event in the VW group occurred due to emotional distress. | PMC10209787 | |
Discussion | PMC10209787 | |||
Principal Findings | depressive symptoms, depression, T2DM, diabetes distress, diabetes | DIABETES | To our knowledge, this is the first fully powered clinical trial to demonstrate the effectiveness of delivering DMGVs using an immersive 3D telemedicine platform versus IP care. Both approaches were similarly effective in reducing mean HbAOur preliminary study compared IP versus immersive DSME delivery among 89 low-income African American women with uncontrolled T2DM [Nearly half of our participants at baseline had measurable depressive symptoms and diabetes distress. Prior research has revealed a strong correlation between depression, diabetes distress, and uncontrolled diabetes [Given our pilot study showed increased physical activity among study participants, the null finding in physical activity in WIC2 was unexpected [ | PMC10209787 |
Limitations | We acknowledge several study limitations. We had a small imbalance in HbA | PMC10209787 | ||
Conclusions | Shakiyla, Diabetes, Digestive, diabetes | KIDNEY DISEASES, DIABETES, RECRUITMENT, DIABETES | Immersive technologies can reduce disparities by improving effectiveness and access to evidence-based diabetes care. We showed that when given the tools, adults from digitally underserved communities robustly adopt health technology tools with improved health outcomes. More effort is warranted to design technology tailored to the needs, capabilities, and life perspectives of diverse communities to avoid leaving behind those most in need of better health care.We extend our gratitude to the clinicians and research study staff who supported Women in Control 2.0 (WIC2). Clinicians A Mansa Semenya, MD, MPH; Elena Hill, MD; Adi Rattner, MD; Aissatou Gueye, NP; and Cleopatra Ferrao, NP delivered the WIC2 curriculum and performed clinical consults with participants. Katherine Melo, Jenna Bhaloo, Jennifer Albuquerque, Shakiyla Woods, Maria Pompeya Gomez, and Eddie V Gomez supported participant recruitment, data collection, and intervention activities as support staff members. All study personnel received compensation for their work. We additionally thank all the women who participated in WIC2 and the community advocacy groups who helped us achieve our mission.Research reported in this publication was supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under Award Number R01DK106531. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. SEM 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.Authors' Contributions: SEM and PG conceived and designed the study. SEM, PG, AB, IAM, KNP, and JM-H carried out the study intervention and collected participant data. MRW, MJR, AB, IAM, and BAD performed data validation and analyses. SEM, MRW, MJR, AB, and BAD interpreted results. All authors contributed to the preparation of the manuscript, critically reviewed the results, and approved the manuscript for submission. SEM obtained funding. JM-H, KNP, AB, and BAD provided administrative, technical, and material support.Conflicts of Interest: SEM is a consultant on health communication and relationship-centered care and has provided workshops and lectures on this topic funded by pharmaceutical and other industry sponsors. No product endorsement is permitted during these programs. SEM also holds equity in See Yourself Health LLC, a digital health service provider.Women in Control 2.0 (WIC2) trial protocol.Additional statistical analyses.CONSORT eHEALTH checklist (V 1.6.1). | PMC10209787 |
Abbreviations | diabetes | DIABETES | Consolidated Standards of Reporting Trialsdiabetes distressdiabetes medical group visitdiabetes self-management educationhemoglobin Ain-personintention-to-treatmetabolic equivalent of taskPatient Health Questionnaire-8per-protocoltype 2 diabetes mellitusvirtual worldWomen in Control 2.0 | PMC10209787 |
Data Availability | SECONDARY | The WIC2 team agrees to share deidentified individual participant data that underlie the results reported in this study, the study protocol, and the statistical analysis plan. Data will be available for 6 months following publication for 5 years. Data will only be shared with academic researchers who provide a methodologically sound proposal to achieve aims related to primary or secondary outcomes and upon completion of a data use agreement. Requests should be directed to Suzanne.Mitchell2@umassmed.edu. | PMC10209787 | |
Methods | Cancer, NRS, pain | ADVANCED CANCER, CANCER | Adults with advanced cancer and scored worst pain ≥ 2/10 on a numeric rating scale (NRS) were recruited from 6 Australian oncology/palliative care outpatient services to the Stop Cancer PAIN trial (08/15-06/19). Out-of-hospital, publicly funded services, prescriptions and costs were estimated for the three months before pain screening. Descriptive statistics summarize the clinico-demographic variables, health services and costs, treatments and pain scores. Relationships with costs were explored using Spearman correlations, Mann-Whitney U and Kruskal-Wallis tests, and a gamma log-link generalized linear model. | PMC9974128 |
Results | cancer type, reflux disease, GORD, pain | Overall, 212 participants had median worst pain scores of five (inter-quartile range 4). The most frequently prescribed medications were opioids (60.1%) and peptic ulcer/gastro-oesophageal reflux disease (GORD) drugs (51.6%). The total average healthcare cost in the three months before the census date was A$6,742 (95% CI $5,637, $7,847), approximately $27,000 annually. Men had higher mean healthcare costs than women, adjusting for age, cancer type and pain levels (men $7,872, women $4,493, p<0.01) and higher expenditure on prescriptions (men $5,559, women $2,034, p<0.01). | PMC9974128 | |
Conclusions | cancer, peptic, pain | CANCER | In this population with pain and cancer, there was no clear relationship between healthcare costs and pain severity. These treatment patterns requiring further exploration including the prevalence of peptic ulcer/GORD drugs, and lipid lowering agents and the higher healthcare costs for men. | PMC9974128 |
Data Availability | Cancer | CANCER | This dataset contains sensitive patient information and patient consent was not obtained to share data. To remain compliant with university, state government, and data custodian ethical requirements, access to the dataset is restricted. Further, data cannot be shared publicly by the authors because the analysis draws on third party data not owned or collected by the authors (MBS and PBS data) and the authors do not have the rights to share these data. The Australian Government place restrictions on access to data to protect the participants' confidentiality and privacy. Access requires processes due to the General Data Protection Regulations (or Australian Privacy Principles) for data distribution. Data were created by linkage of the Stop Cancer PAIN Trial data to Australian Government data sources and permission from the Department of Human Services External Request Evaluation Committee under specific ethics approval. The Stop Cancer PAIN, and MBS and PBS de-identified data are available to researchers from the data custodians (Professor Melanie Lovell ( | PMC9974128 |
Introduction | Pain, cancer pain, pain, cancer-related pain, Cancer | DISEASE, ADVANCED CANCER, CANCER | Approximately 70% of patients living with advanced cancer experience pain [Pain is associated with poorer quality of life [Cancer is the leading cause of social and economic burden in Australia [However, no studies have investigated the relationship between healthcare usage and pain intensity in people with advanced cancer in Australia. Such estimates facilitate modelling of the effect of successfully reducing pain levels on subsequent healthcare utilization through improved pain management. For example, these estimates would usefully inform economic evaluations, i.e. modelled cost-effectiveness, cost-utility and cost-benefit analyses, which systematically compare the costs and benefits of competing strategies and provide information about how best to improve outcomes within funding constraints [Additionally, there is a paucity of research on factors associated with healthcare utilisation and costs in patients living with advanced cancer pain in Australia such as patient, disease or pain-related characteristics. This knowledge could help identify potential opportunities for improvement in cancer pain management. For example, data from the United States (US) suggests that younger age, lower income level and greater pain intensity are associated with higher healthcare costs in outpatients experiencing cancer-related pain [ | PMC9974128 |
Aim | pain | ADVANCED CANCER | The aims of this study were to:identify treatment patterns and corresponding costs of healthcare resource use (government funded) for outpatients living with advanced cancer and pain;explore factors associated with healthcare costs in this population; andexamine the relationship between healthcare costs and pain intensity. | PMC9974128 |
Methods | PMC9974128 | |||
The Stop Cancer PAIN Trial | cancer-related pain, cancer, cancer pain, pain, NRS, Cancer | CANCER, ADVANCED CANCER, SECONDARY, CANCER | This pragmatic, phase III, stepped-wedge, cluster randomised controlled trial investigated the effectiveness of screening and guidelines for pain with, versus without, implementation strategies for improving cancer pain.From August 2015 to June 2019, adults with cancer and pain presenting to six oncology and palliative care outpatient services across Australia were recruited to the Stop Cancer PAIN Trial [To be eligible, patients had to be, outpatients with a diagnosis of advanced cancer, the ability to self-complete the 0–10 numeric rating scale (NRS) for severity of worst and average pain, reporting a worst pain score of ≥5 (primary outcome) or ≥2 (secondary outcome) and sufficient proficiency in spoken English to complete the secondary outcome measures were eligible to participate in the study [De-identified pain screening data from The primary outcome was the proportion of participants initially reporting a worst pain score of ≥5 who experienced a clinically important improvement of ≥30% 1 week later. Secondary outcomes included mean average pain, quality of life, patient empowerment, and fidelity to the intervention, and were measured in all participants initially reporting a worst pain score of ≥2 at weeks one, two, and four. Overall, there was no statistical difference in pain-related outcomes; the implementation strategies were insufficient to improve pain-related outcomes for outpatients with cancer-related pain [ | PMC9974128 |
Treatment patterns and out-of-hospital healthcare resource utilisation | illness, Cancer | CANCER | The Stop Cancer PAIN Trial database was linked to routinely collected out of hospital services (Medicare Benefits Schedule) and medication data (Pharmaceutical Benefits Schedule) to explore treatment patterns and estimate healthcare resource utilisation. Ethics approval was granted by the South Western Sydney Local Health District Human Research Ethics Committee (HREC)–ethics approval number HREC/14/LPOOL/479 and the data custodian, the Australian Department of Human Services External Request Evaluation Committee (MI4457). Study participants provided written informed consent for access to these data.Approval was granted by the HREC for an opt-out procedure to contact patients at week one as well as a procedure to obtain informed verbal (rather than written) consent to participate. Verbal consent was considered to place less burden on patients who, due to their illness, might be less able to return written consent forms by mail. | PMC9974128 |
Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Schedule (PBS) | Medicare is the publicly funded universal healthcare scheme in Australia providing access to subsidised medical services for all residents [ | PMC9974128 | ||
Outcome measures | cancer | CANCER | Participants demographics and cancer diagnoses details were collected [ | PMC9974128 |
Pain Numeric Rating Scale (NRS) | pain | The pain NRS is a widely used, self-completed, validated pain scale [ | PMC9974128 | |
EORTC QLQ C15-PAL | Cancer | CANCER | The European Organisation for Research and Treatment of Cancer Quality of Life-C15-Palliative questionnaire (‘C15-PAL’) is a shortened version of the QLQ-C30, and contains 15 of the 30 original items [ | PMC9974128 |
EORTC QLU-C10D | cancer | CANCER | The C15-PAL does not provide a total score reflecting the health-related quality of life of people living with cancer and cannot be used to inform economic evaluations because this patient-reported outcome measure is not preference based, i.e. does not enable the calculation of utility values. The EORTC QLU-C10D [ | PMC9974128 |
Data linkage | Cancer | CANCER | Medicare Benefit Schedule and PBS data were requested for all consented participants. The Department of Human Services carried out probabilistic linkage to the Stop Cancer PAIN Trial ID with the MBS and PBS database based on key variables such as date of birth and Medicare number and provided anonymised data to the lead investigator. Data were extracted on 27 November 2019. Services provided through public hospitals such as inpatient, outpatient or emergency care were not recorded in the Stop Cancer PAIN Trial and therefore are not included in the analyses. Consistent with previous Australian healthcare resource utilisation studies, services provided to Department of Veterans Affairs beneficiaries were also excluded because of greater range of services accessible to these beneficiaries compared with the average Australian [ | PMC9974128 |
Analysis | pain | Analyses were performed using SPSS for Windows version 25.0 (SPSS, Inc., Chicago, IL) and Stata version 16.0 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC) on available data. Descriptive summary statistics were estimated for clinico-demographic variables, NRS pain and HrQOL scores, QLU-C10D utilities and healthcare resource use and costs.Medical and allied health services were categorised consistent with the Medicare Benefits Schedule (See Table 1 in Healthcare costs were positively skewed. Consequently, Differences between clinico-demographic sub-groups were assessed using the non-parametric Mann Whitney The relationships between clinico-demographic variables and healthcare resource utilisation were further explored using a generalised linear model (GLM) with a gamma distribution and a log link [ | PMC9974128 | |
Results | PMC9974128 | |||
Sample characteristics | Cancer | CANCER | In total, 30.8% (n = 212) of patients who participated in the Stop Cancer PAIN Trial consented to having their MBS and/ or PBS data accessed. Overall, total healthcare costs and Stop Cancer PAIN Trial data were available for 186 participants (MBS costs, n = 186, PBS costs n = 188) for this study (missing trial data, n = 26). All costs are reported as Australian dollars.The participant demographics and clinical characteristics are summarised in | PMC9974128 |
Participant demographics and clinical characteristics. | pain | IQR = inter-quartile range; NRS = pain numerical rating scale; SD = standard deviation | PMC9974128 | |
2. Factors associated with healthcare costs | pain | Spearman’s rank correlations between total healthcare costs and age, pain intensity and HrQOL and QLU-C10D scores were in the anticipated directions but weaker than expected especially for pain intensity ( | PMC9974128 | |
Correlations between total healthcare, MBS and PBS costs and age, pain intensity and health-related quality of life. | cancer, pain | CANCER | Sample sizes vary due to missing data; HrQOL = health-related quality of life; MBS = Medicare Benefits Schedule; NRS = pain numerical rating scale; PBS = Pharmaceutical Benefits Schedule; Correlations were interpreted according to Cohen’s guidelines, i.e., “strong” (≥0.51), “moderate” (0.31–0.50), “weak” (0.11–0.30), and “none” (0–0.10). Statistically significant correlations are bolded. + indicates positive direction; -, negative directionIn the bivariate analyses, there was a statistically significant difference in mean total healthcare costs for gender but not age, cancer type or baseline pain level. Mean total healthcare costs were higher for men ($7,924, 95% CI $6,267, $9,581) than women ($5,367, 95% CI $3,975, $6,760) (U = 3546, | PMC9974128 |
Unadjusted, mean healthcare costs by clinico-demographic characteristics. | cancer, lung cancers, pain, head and neck cancers | CANCER, LUNG CANCERS, HEAD AND NECK CANCER | KW = Kruskal-Wallis; MBS = Medicare Benefits Schedule; MWU = Mann-Whitney U; NRS = numeric rating scale; PBS = Pharmaceutical Benefits Schedule; SD = standard deviation. Shaded cells indicate statistically significant differences.Mean total MBS and PBS costs varied by cancer type. Mean total MBS costs were highest for participants diagnosed with head and neck cancers ($5,944, 95% CI $3,291, $8,597), whereas mean total PBS costs were highest for people diagnosed with lung cancers ($4,813, 95% CI $1,340, $8,286). However, there were no other differences detected between PBS and MBS costs for age, gender or baseline pain level. | PMC9974128 |
3. Relationship between healthcare costs and pain intensity | cancer, Pain | CANCER | Pain intensity was not associated with healthcare costs after adjusting for age and sex (see Table 6 in Pain intensity was also not associated with mean total MBS costs after adjusting for age, sex and cancer type (see Table 7 in Exploratory, | PMC9974128 |
Discussion | cancer, pain | CANCER, ADVANCED CANCER, PEPTIC ULCERS, GASTROESOPHAGEAL REFLUX DISEASE | The findings suggest government funded, out-of-hospital costs are, on average, $2,247 per month for people living with advanced cancer and pain, i.e. approximately $27,000 per year, higher than recently reported MBS and PBS costs for the first 12-months following cancer diagnoses in Queensland, Australia (approximately 2012 A$7,224 per person per year) [A smaller proportion of the Stop cancer PAIN Trial participants received concessional benefits compared with the Australian general population, suggesting patients with advanced cancer could incur greater out-of-pocket expenses [Three of the ten most commonly prescribed medications in the sample were the same as those for the general Australian population in 2019–20; pantoprazole and esomeprazole which are largely prescribed for peptic ulcers and gastroesophageal reflux disease, and rosuvastatin for lowering high cholesterol levels [ | PMC9974128 |
Top 10 most frequently prescribed medicines. | cancer type, cancer, breathlessness, pain | DISEASE, CANCER, ADVANCED CANCER | More than one in five people with advanced cancer and pain were prescribed a lipid modifying agent, contrary to guidance to reduce the burden of medications in advanced disease, particularly from medications such as statins which are only prescribed for long term population-level risk reduction [The remaining top ten most frequently prescribed medications in the study sample were related to the treatment of pain (or, to a lesser extent, breathlessness [Few factors were associated with total healthcare costs, contrary to findings reported in the US [Differences in total healthcare costs between males and females with advanced cancer and pain is consistent with previous evidence which suggests a gender difference in total healthcare resource utilisation and costs for people living with cancer [Finally, patterns in MBS costs by cancer type are consistent with population-based estimates of health services costs for people receiving cancer care (with or without pain) in Australia which suggest healthcare costs vary by cancer type and time since diagnosis, possibly driven by differences in treatment modalities and frequency, new targeted therapies and immunotherapies, and associated tests and administrative MBS items [ | PMC9974128 |
Strengths and limitations | ADVANCED CANCER | Consent to access MBS and PBS data was granted by just under a third of study participants and findings may not reflect treatment patterns in the entire study cohort. However, unlike healthcare resource utilisation data collected using other means such as surveys, Medicare and PBS data are not prone to recall bias and typically provide greater accuracy than other methods of measuring costs [Despite these caveats, this analysis provides valuable insights into government funded out-of-hospital costs associated with advanced cancer pain to inform priority setting and policy development. | PMC9974128 | |
Implications for research and practice | peptic, pain | ADVANCED CANCER | The findings identify areas of treatment for outpatients with advanced cancer and pain requiring further exploration and practice change, particularly the high use of peptic ulcer/GORD drugs, lipid modifying agents and corticosteroids. Further research is needed to determine why healthcare costs were higher in men than women with advanced cancer and experiencing pain and to explore both sex and gender-based differences and provider related factors.Health economic research which includes costs related to emergency department and hospital admissions is needed. The authors recommend that future research evaluating interventions to improve pain outcomes include a health economic analysis. Also, the cost effectiveness of many evidence-based non-pharmacological interventions needs further research. | PMC9974128 |
Conclusions | cancer, pain | CANCER | This study provides vital information for informing quality of care and quality use of medicines, resource allocation and developing sustainable health policy. There was no clear relationship between pain intensity and healthcare costs demonstrated in this population with pain. Investigation into the underpinning rationale for higher healthcare costs in men is needed to promote equitable access to cancer care. | PMC9974128 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.