title stringlengths 1 1.19k | keywords stringlengths 0 668 | concept stringlengths 0 909 | paragraph stringlengths 0 61.8k | PMID stringlengths 10 11 |
|---|---|---|---|---|
Researchers’ stance | Interviews were carried out and transcribed by two of the authors (JS and RT). Both are female, trained clinicians (nurse and psychologist), well-acquainted with the practices of MET, but neither was a therapist at the clinic where treatments were conducted. Authors JS and SIH (both female) were both active as research coordinators in the RCT, and acquainted with the individuals who participated in the study. Authors SIH (female) and AH (male) are both trained MI therapists, but were not active as therapists in the current project. The authors’ pre-understanding of MI/MET as clinicians at the treatment site could potentially have had an impact on the analysis. The two-legged position as both therapists and researchers was key in critically evaluating and interpreting data. Lastly, author CN (female) is a senior researcher with long experience of qualitative research. CN is affiliated with a different university and was therefore not involved in the data collection process, or active as a therapist at the clinic, which allowed for a more distanced view of the data. | PMC10357895 | ||
Methodological approach | Thematic analysis in accordance with the six phases defined by Braun and Clarke [ | PMC10357895 | ||
Results | PMC10357895 | |||
MITI-scoring results | The analysis of MITI scores among the involved therapists resulted in the following mean scores (including cutoffs for the accepted level according to the MITI in square brackets); (1) Technical score = 3.2 (SD = 0.79), [> = 3.0]; (2) Relational score = 3.9 (SD = 0.53), [> = 3.5]; (3) Percentage of complex reflections = 0.6 (SD = 0.14), [> = 0.4]; (4) Reflection to question ratio = 3.3 (SD = 3.3), [> = 1.0]. All results of MITI scoring were above the threshold of acceptable levels of adherence to MITI-protocol [ | PMC10357895 | ||
Analysis | Five themes were identified (example of themes, see Table Examples of the analysis process | PMC10357895 | ||
Discussion | EVENTS, DISORDERS | The aim of the current study was to investigate how patients who were treated with MET perceived their treatment and if it was sufficient to achieve the desired change. To our knowledge, this is the first qualitative interview study that explores patient experiences of receiving MET for AUD. Participants reported the therapist relationship as important for feeling secure and engaging in treatment. However, a lack of guidance on how to reach personal goals, as well as what would be an appropriate goal, was perceived as less helpful to some. Participants also expressed a wish for a longer treatment period, to increase their self-efficacy to change. Lastly, participants were hesitant or even negative towards bringing a SO to the treatment sessions.Participants described a positive, supportive, and non-judgmental relationship with the therapist as a particularly important aspect of the treatment. These are specific qualities suggested to represent MI spirit [In the current study, participants reported that therapists did not interfere with their drinking goals or strategies to approach those goals. Some of the participants described that this approach made them feel like the owner of their problem. This resembles the results of Jones and colleagues, where therapists were perceived to emphasize patients’ autonomy which in turn was important for the patients’ motivation to change [Some participants described that there was no room for talking about what they experienced as a failure or negative events, subjects which in their opinion, could have been beneficial to reflect on. A recent study of a brief MI-session addressing risky use of alcohol in emergency units found that discussing negative consequences related to alcohol consumption led to an increase in patients’ perceived readiness to change [Participants reported that their decision to seek treatment was one step in an ongoing motivational process and that they were already motivated to change at treatment entry. They hoped and expected that treatment would consolidate their wish to change, rather than expecting that the treatment per se would make them change. Seeking treatment may thus have been the most important step in the process from contemplation to readiness for change in these individuals. This is somewhat unexpected, as MI is primarily described as a method which would suit individuals in a pre-contemplation or contemplation stage of change [Participants in this study expressed that the treatment had been helpful regarding both increased awareness and increased control over drinking behavior. However, some had expected to have been more successful in reducing alcohol consumption than they actually were. The general opinion among participants was that they wanted treatment to continue for a longer period than the given 12 weeks to establish and maintain change. A more flexible approach to MET treatment length in clinical settings may thus be beneficial to patients’ sense of self-efficacy.In the current study, participants were encouraged to bring a SO to the treatment sessions but, interestingly, none of them did. They were, however, in favor of the idea of involving a SO, and some stated that they already had a supporting dialogue with their SO. In practice, most were reluctant to their SO participating in treatment and preferred to interact with their therapist in private. The effect of involving a SO in the treatment of substance use disorders has been examined in relation to several treatment methods [ | PMC10357895 | |
Strengths and limitations | The current study was conducted in an addiction specialist setting, and treatment was performed by highly trained and experienced therapists. By the measures of treatment integrity derived from MITI scores, we could assume that our results were based on MET performed by therapists who were adherent to MITI protocol. Further, this kind of study design gives a good understanding of perceptions in a group of individuals. It does not aim to generalize the results in a quantitative manner. Instead, descriptions of context, process of analysis, and appropriate quotations can inform and enhance readers’ understandings of how the findings can be transferred to other settings or groups.A limitation to the study is that MET sessions were primarily provided through video meetings instead of face-to-face, due to the COVID pandemic. Some participants expressed a desire to meet their therapist in person, and some also considered the video format to have affected their decision not to bring a SO. However, the video format was not mentioned as the main reason for not bringing a SO. Lastly, one limitation was that participants who were offered to participate in interviews were the ones who retained in the study at the 26-week follow-up. Hence, we did not receive information from participants who dropped out from the study prematurely (16.4%). | PMC10357895 | ||
Conclusions | The part of treatment that participants most appreciated was the opportunity to verbalize thoughts and feelings about their problems in a non-judgmental and supportive environment. Notably, they had started their change process before seeking treatment. A less helpful aspect was the non-directive approach to goal-setting, which increased autonomy for some, but was not helpful to others. Furthermore, all participants experienced a need for longer treatment. Lastly, SOs can play an important role as supporters, even though it may not be necessary or desirable to include them in treatment sessions. | PMC10357895 | ||
Future directions | Future research on patient experiences from various MI applications in AUD treatments may inform further development of these methods. Research on mechanisms of behavioral change (MOBC) in MI has been suggested to be prematurely narrow, focusing on technical components primarily involving therapist behavior [ | PMC10357895 | ||
Acknowledgements | The authors would like to thank the participants in the study, the clinical staff, and the therapists at the study sites. | PMC10357895 | ||
Author contributions | Conceptualization: SIH, JS, RT, AH, CN. Methodology: SIH, JS, RT, AH, CN. Investigation: SIH, JS, RT, AH, CN. Formal analysis: SIH, JS, RT, AH, CN. Writing original draft: SIH, JS, CN. Review and editing: SIH, JS, RT, AH, CN. Project administration: SIH, JS, AH. Supervision: AH, CN. Funding acquisition: AH, CN. Each author certifies that their contribution to this work meets the standards of the International Committee of Medical Journal Editors. All authors read and approved the final manuscript. | PMC10357895 | ||
Funding | Open access funding provided by Karolinska Institute. The study was funded by the Swedish Research Council for Health, Working life and Welfare (FORTE) (2018–00716), the Research Council of the Swedish Alcohol Retailing Monopoly (SRA) (FO-2016-0042; 2017-0045; FO-2018-0080), and the Stockholm County Council (20170688). | PMC10357895 | ||
Availability of data and materials | The qualitative data analyzed in the current study is not publicly available. Translated pseudonymized data are available from the corresponding author on request. | PMC10357895 | ||
Declarations | PMC10357895 | |||
Ethics approval and consent to participate | This trial was approved by the Regional Ethics Review Board in Stockholm (DNR: 2016/634-31/2). | PMC10357895 | ||
Consent for publication | Participants have approved that data are used in a scientific publication. | PMC10357895 | ||
Competing interests | Author AH is a co-author of the translated Swedish MET-manual used in the current trial. Remaining authors have no competing interests. | PMC10357895 | ||
References | PMC10357895 | |||
Background: | death, PVD | EVENTS, PERIPHERAL VASCULAR DISEASE (PVD), PVD | Patients with peripheral vascular disease (PVD) are often underdiagnosed and undertreated. Nocturnal nondipping blood pressure (BP) pattern, as diagnosed by ambulatory BP monitoring (ABPM), is associated with increased cardiovascular risk, but has not been studied in patients with PVD. We aimed to investigate if a nondipping BP pattern predicts cardiovascular events or all-cause death in outpatients with PVD. | PMC10408241 |
Methods: | PVD | Consecutive outpatients with carotid or lower-extremity PVD were examined with 24-hour ABPM ( | PMC10408241 | |
Results: | EVENTS | In the cohort (mean age 70; 40% women), 137 events occurred during a 5.1-year median follow-up; incident rate of 7.35 events per 100 person-years. Nondipping was significantly associated with outcome (hazard ratio 1.55, 95% CI 1.07–2.26, | PMC10408241 | |
Conclusion: | carotid artery disease, lower-extremity PVD, PVD | CAROTID ARTERY DISEASE, EVENTS, PVD | In a cohort of outpatients with PVD, nondipping was an independent risk factor for future cardiovascular events or mortality and seemed to be a strong predictor in patients with carotid artery disease but not in lower-extremity PVD. Additional studies are needed to evaluate the clinical utility of ABPM for improved prevention in these high-risk patients. | PMC10408241 |
Background | Atherosclerotic peripheral vascular disease, atherosclerosis, cardiovascular disease, PVD | CARDIOVASCULAR DISEASE, DISEASE, PVD, EVENTS, ATHEROSCLEROSIS | Atherosclerotic peripheral vascular disease (PVD) is a common disease defined as systemic arterial atherosclerosis outside the aorta and coronary and intracranial arteries.BP is a leading risk factor for cardiovascular disease in the general populationWe hypothesized that nondipping is a relevant prognostic marker for an increased risk for cardiovascular events also in patients with PVD. We aimed to investigate the association between nondipping and the incidence of cardiovascular events or all-cause mortality during a long-term follow-up of outpatients with confirmed carotid or lower-extremity PVD. | PMC10408241 |
Methods | PMC10408241 | |||
Study population | atherosclerotic, Peripheral Arterial Disease | PERIPHERAL ARTERIAL DISEASE | Analyses were based on patients included in the Peripheral Arterial Disease in Västmanland study (PADVa), based on patients with atherosclerotic PVD.In total, 452 patients (73.6%) accepted the invitation to join the study. Everyone in the study was offered ABPM, of whom 35 individuals declined. We excluded patients with < 10 daytime or < five night-time ABPM measurements (The study was approved by the Ethics Committee of Uppsala University, Sweden (Dnr 2005:382). All participants gave their written informed consent. The study is identified as ClinicalTrials.gov Identifier NCT01452165. | PMC10408241 |
Examination protocol | diabetes mellitus, Hypertension, cardiovascular disease | DIABETES MELLITUS, HYPERTENSION, CARDIOVASCULAR DISEASE | All patients were invited to attend the Department of Clinical Physiology and were examined according to a standard examination protocol, including an extensive self-administered questionnaire including smoking status (current smoking defined as regular smoking within the past year), medical history, and current medication. Self-reported diagnoses of cardiovascular disease and diabetes mellitus were confirmed from the medical records. Hypertension was defined as present if diagnosed by a physician and treated with antihypertensive medication. | PMC10408241 |
Blood samples | chronic kidney disease | Participants fasted overnight, and venous blood samples were taken by trained staff and immediately sent to the accredited Laboratory of Clinical Chemistry, Västmanland County Hospital, Västerås. The estimated glomerular filtration rate (eGFR) was calculated from serum creatinine levels standardized by isotope dilution mass spectrometry (Synchron LX or UniCel DxC instruments; Beckman Coulter Inc., Brea, CA, USA) using the chronic kidney disease epidemiology (CKD-EPI) formula. | PMC10408241 | |
Ankle–brachial index and carotid ultrasound | BLOOD | Blood pressure in both arms and ankles was measured in all participants in a supine position after at least 5 minutes of rest. The ankle BP was measured in the bilateral dorsalis pedis and posterior tibial arteries using an inflatable leg-cuff, an aneroid sphygmomanometer, and a handheld Doppler instrument with a 5-MHz probe. The ankle–brachial index (ABI) was calculated by dividing the highest ankle pressure by the highest BP of both arms. An ABI of ⩽ 0.90 or ⩾ 1.40 in either leg was defined as abnormal. Ultrasound examinations of the carotid and lower-limb arteries have been described in detail. | PMC10408241 | |
Clinical and ambulatory blood pressure | Office BP was measured manually by trained technicians and obtained from the nondominant arm or the other arm if the systolic BP was > 10 mmHg higher. The BP was measured in the supine position after a minimum of 5 minutes of rest and was rounded up to the nearest 2 mmHg. Using the arm from which the office BP was obtained, the ABPM-04 instrument (Meditech Ltd, Budapest, Hungary) was applied for 24-hour ABPM, with readings taken every 20 minutes.Nondipping was defined as a reduction in night-time systolic BP ⩽ 10% of the daytime average systolic BP and dipping as a nightly fall by more than 10% of the daytime average systolic BP value. | PMC10408241 | ||
Outcomes | stroke, death, ICD-10, heart failure | DISEASES, MYOCARDIAL INFARCTION, STROKE, EVENT, HEART FAILURE | The participants were followed through the Swedish population and in-patient registries until a cardiovascular endpoint, all-cause death, or December 31, 2013, at which time the remaining participants were censored. The cardiovascular endpoints were defined as hospitalization or death caused by myocardial infarction (International Classification of Diseases 10th Revision [ICD-10] code I21), stroke (ICD-10 I61 or I63), and heart failure event (ICD-10 I11.0, I25.5, or I50). As endpoints, we also included hospitalization because of percutaneous coronary intervention or coronary artery bypass grafting. | PMC10408241 |
Statistics | lower extremity PVD | REGRESSION, CAROTID ARTERY DISEASE | Baseline data are presented as mean ± SD or frequency and percentage. Highly skewed variables (hs-CRP) are presented as median (25In sensitivity analyses, we performed weighted Cox regression based on the propensity score to balance the baseline characteristics of the dippers and nondippers. We performed these analyses in the entire study cohort and in subgroups with lower extremity PVD and carotid artery disease, respectively. The propensity score was obtained by multiple logistic regression, including all variables from Baseline characteristics of the study cohort.Values are mean ± SD, median (25th, 75th percentiles) or Kruskal–Wallis test Medication as defined in the COPART risk score (i.e., aspirin, ACE-I, ARB, or statins).Amb, ambulatory; ABI, ankle–brachial index; ACE-I, angiotensin converting enzyme inhibitor; amb, ambulatory; ARB, angiotensin receptor blocker; BMI, body mass index; BP, blood pressure; eGFR, estimated glomerular filtration rate; hs-CRP, high-sensitivity C-reactive protein; ICA, internal carotid artery.Statistical analyses were performed using SPSS statistical software for Windows, version 26 (IBM, Armonk, NY, USA) and R 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria). Two-sided | PMC10408241 |
Results | PMC10408241 | |||
Baseline characteristics | The baseline characteristics of the patients, in total and divided into dippers and nondippers, are presented in | PMC10408241 | ||
Cardiovascular events | EVENTS, EVENT | The number of incident cardiovascular events during follow-up with numbers at risk and incidence rates among the participants are presented in Risk and incidence of events in the whole cohort and by dipping status.PYAR, person-years at risk.Kaplan–Meier curve displaying survival free from cardiovascular event in dippers and nondippers. | PMC10408241 | |
Cox regression analyses | stroke, diabetes | MYOCARDIAL INFARCTION, STROKE, CAROTID STENOSIS, EVENTS, HEART FAILURE, HYPERTENSION, DIABETES | Nondipping was significantly associated with adverse outcomes in all multivariable models, with a hazard ratio (HR) of 1.55 (95% CI 1.07–2.26, Risk of cardiovascular events and all-cause mortality in nondippers versus dippers (multivariable Cox regression).Based on the full study cohort (Model A adjusted for age and sex. Model B adjusted for age, sex, office and ambulatory 24-hour systolic and diastolic blood pressures. Model C adjusted for age, sex, ambulatory 24-hour systolic and diastolic blood pressures, total cholesterol, BMI, smoking, diabetes, hypertension, previous myocardial infarction, previous stroke, heart failure, eGFR, ABI, and internal carotid stenosis. Model D adjusted for variables included in the COPART risk score, i.e., age, previous myocardial infarction, hs-CRP, ABI, eGFR, and medication with statins, aspirin, and angiotensin receptor blocker.ABI, ankle–brachial index; BMI, body mass index; eGFR, estimated glomerular filtration rate; hs-CRP, high-sensitivity C-reactive protein. | PMC10408241 |
Sensitivity analyses | REGRESSION | In sensitivity analyses, we reached a satisfactory balance of baseline characteristics between dippers and nondippers in the weighting based on propensity score in the entire study cohort (In the weighted Cox regression, nondipping was significantly associated with outcome in the whole cohort (HR 1.45, 95% CI 1.04–2.03, | PMC10408241 | |
Discussion | PMC10408241 | |||
Principal findings | death, PVD | ADVERSE EVENTS, EVENTS, PVD | In the present sample of outpatients with PVD, nondipping was an independent predictor of cardiovascular events or all-cause death. Nondipping remained significantly associated with prognosis when adjusting for age, sex, 24-hour BP levels, and established cardiovascular risk factors. Dipping status also improved risk prediction when added to the validated COPART risk score. Our findings show that the previously reported association between nondipping and adverse events in the general population also holds for patients with PVD. | PMC10408241 |
Comparison with the literature | diabetes and renal disease, stroke, hypertensive, carotid artery disease, PVD | ADVERSE EVENTS, STROKE, CARDIAC EVENTS, PVD, CAROTID ARTERY DISEASE, EVENTS | The results that nondipping is a risk factor for adverse prognosis in patients with PVD is in line with studies in hypertensive patients and in the general population, as well as in patients in hemodialysis.The prevalence of nondipping is higher in high-risk cohorts; for example, in patients with diabetes and renal disease. As expected, the prevalence of nondipping was markedly higher (43%) in our cohort compared to a general hypertensive population (25%).Smoking is a particularly strong risk factor for PVD, and a higher proportion of the participants in this study were smokers than the general population in Sweden.Interestingly, nondipping was associated considerably stronger with incident adverse events in our subgroup with carotid artery disease than in the subgroup with abnormal ABI. Previous data suggest that the ABPM nondipping pattern is more strongly associated with incident stroke events than with cardiac events. | PMC10408241 |
Potential mechanisms | cardiovascular disease | OBSTRUCTIVE SLEEP APNEA, CARDIOVASCULAR DISEASE | The pathogenetic mechanism behind nondipping and cardiovascular disease is incompletely understood. However, it is believed to have multiple causes, such as activity during the day, the depth and quality of sleep, and activity of the sympathetic nervous system, among others.Obstructive sleep apnea is associated with a significantly increased risk of nondipping. | PMC10408241 |
Clinical relevance | PVD | PVD | A previous study suggests that bedtime antihypertensive drug administration can partially restore a normal dipping pattern.The COPART risk score is a validated risk score for patients with PVD that is evaluated for long-term prediction of all cause and CV mortality. | PMC10408241 |
Strengths and limitations | carotid artery disease | CAROTID ARTERY DISEASE, PVD | Strengths of the study include the well-characterized study participants, the long-term follow-up, and the clinically generalizable study population of heterogeneous outpatients with carotid and/or lower-extremity PVD, both symptomatic and asymptomatic.Limitations include that the study population was limited to outpatients of European origin who were found to have lower-extremity and/or carotid artery disease at a visit to a vascular ultrasound laboratory. The invited patients who declined to join the study ( | PMC10408241 |
Conclusions | lower-extremity PVD, ICA stenosis, PVD | EVENTS, PVD | A nondipping 24-hour BP pattern was associated with an increased risk of cardiovascular events and mortality in outpatients with PVD. Nondipping is an independent risk factor for adverse outcomes, and this association might be more potent in patients with ICA stenosis than in patients with lower-extremity PVD. Further studies are needed to evaluate if a nocturnal BP profile can be used clinically to improve risk prediction and to investigate if treatment of nocturnal BP can diminish nondipping and improve prognosis. | PMC10408241 |
Supplemental Material | PMC10408241 | |||
References | PMC10408241 | |||
Background | critically ill | CRITICALLY ILL | In critically ill patients, measured creatinine clearance (CrCl) is the most reliable method to evaluate glomerular filtration rate in routine clinical practice and may vary subsequently on a day-to-day basis. We developed and externally validated models to predict CrCl one day ahead and compared them with a reference reflecting current clinical practice. | PMC10327364 |
Methods | A gradient boosting method (GBM) machine-learning algorithm was used to develop the models on data from 2825 patients from the EPaNIC multicenter randomized controlled trial database. We externally validated the models on 9576 patients from the University Hospitals Leuven, included in the M@tric database. Three models were developed: a “Core” model based on demographic, admission diagnosis, and daily laboratory results; a “Core + BGA” model adding blood gas analysis results; and a “Core + BGA + Monitoring” model also including high-resolution monitoring data. Model performance was evaluated against the actual CrCl by mean absolute error (MAE) and root-mean-square error (RMSE). | PMC10327364 | ||
Results | All three developed models showed smaller prediction errors than the reference. Assuming the same CrCl of the day of prediction showed 20.6 (95% CI 20.3–20.9) ml/min MAE and 40.1 (95% CI 37.9–42.3) ml/min RMSE in the external validation cohort, while the developed model having the smallest RMSE (the Core + BGA + Monitoring model) had 18.1 (95% CI 17.9–18.3) ml/min MAE and 28.9 (95% CI 28–29.7) ml/min RMSE. | PMC10327364 | ||
Supplementary Information | The online version contains supplementary material available at 10.1186/s13054-023-04553-z. | PMC10327364 | ||
Keywords | PMC10327364 | |||
Background | AKI, Critical illness, critically ill, acute kidney injury | CRITICAL ILLNESS, CRITICALLY ILL | Critical illness often affects kidney function. Epidemiologic studies have shown that 40–60% of intensive care unit (ICU) patients have episodes of acute kidney injury (AKI) [Existing machine-learning predictions for kidney function have focused on predicting the onset of AKI [Despite the importance and need for continuous kidney function prediction, to the best of our knowledge, no prediction models for daily prediction of CrCl in critically ill patients exist. Hence, this study aims to develop and validate prediction models that apply machine learning algorithms to routinely collected patient data to predict CrCl one day ahead. | PMC10327364 |
Methods | PMC10327364 | |||
Prediction tasks and CrCl definition | This study aims to predict the CrCl of the next patient day. CrCl was calculated by daily 24-h urine output (UO), urinary creatinine (UCr), and serum creatinine (SCr) without correction for an average body surface area: CrCl (ml/min) = UCr(mg/dL) × 24-h UO(ml/day) /SCr(mg/dL)/1440 (min/day). In an additional analysis, the same methodology was applied to develop models to predict the average CrCl over the next two days ahead (Additional file | PMC10327364 | ||
Study databases with inclusion and exclusion criteria | The large multicenter EPaNIC randomized controlled trial (RCT) database [External validation was performed on a dataset of 20,930 patients of the University Hospitals Leuven included in the large multicenter M@tric database between 2013 and 2018 [ | PMC10327364 | ||
Feature engineering | USA).The | REGRESSION | Only data up to the day of predicting CrCl were used as input to the models. The considered data included: 1) admission data: demographics, diagnosis, and comorbidities, 2) time-series data such as minute-by-minute monitoring data and daily or hourly laboratory results, 3) medication-related data, 4) time-related data: day of the week, and day from ICU admission. Data were retrieved from both the EPaNIC study database (Filemaker Pro®; FileMaker Inc, FileMaker International) and the patient data management system (PDMS) database (Microsoft SQL Server®; Microsoft®, Redmond, Washington, USA).The minimum, maximum, mean, standard deviation, linear regression slope, fast Fourier transform (FFT), cepstrum analysis, autoregressive analyses, and first-order derivative were applied to derive characteristics from the timestamped data. All the features with more than 10% missing values were excluded. For the remaining features, missing values were imputed with the mean and the mode from the development cohort for continuous data and categorical data, respectively. Finally, continuous data were standardized to zero mean and unit variance, and categorical data without order relation were converted into a form with binary data for each category. | PMC10327364 |
Machine-learning algorithm, feature selection methods, and clinical prediction models | The prediction models were trained with the gradient-boosting regressor method [For each prediction task, three models with progressively more features were developed which are meant to be utilized sequentially, based on the data availability at the bedside.A “A “A “ | PMC10327364 | ||
Internal and external validation | Models were developed and internally validated on the EPaNIC database with tenfold cross-validation. At the external validation stage, models trained on the entire EPaNIC database were applied to the previously unseen external validation cohort to assess generalizability. To examine the model usefulness, model performance was further compared against a reference reflecting the current clinical practice: using as prediction for one-day ahead the same CrCl value of the day of prediction. This reference CrCl was henceforth referred to as | PMC10327364 | ||
Evaluation metrics for predictive performance | Mean absolute error (MAE) and root-mean-square error (RMSE) were computed for all available patient-days, stable days, and unstable days for each model in both cohorts. Both MAE and RMSE measured the errors between the model predictions and the target CrCl values, with RMSE more sensitive to large errors. Predictive performance was also evaluated visually with scatter plots and plots of daily MAE and RMSE for all available patient-days, stable days, and unstable days during the first week of ICU stay. As multiple patient-days were available in many patients, no overall p-value can be calculated as this may be biased by repeated measures, but we compared the MAE on a day-by-day basis with the Diebold–Mariano test [ | PMC10327364 | ||
Descriptive analyses and software used | Python 3.7.4 (Python Software Foundation, | PMC10327364 | ||
Results | PMC10327364 | |||
Study cohorts | PMC10327364 | |||
External validation cohort | For the external validation of the developed models, data from 53,943 patient-days from 9576 patients were used, corresponding to 45.8% of the University Hospitals Leuven patients included in the M@tric database. Compared to the development cohort, the age was younger (median (IQR) 65.6 (54.6–75) years, p < 0.01), emergency admission was less frequently (n = 3231, 33.7%, p < 0.01), APACHE II score was lower (median (IQR) 17 (13–21), p < 0.01), cardiac surgery was still the major admission diagnosis but occurred less often (n = 3229, 33.7%, p < 0.01), and average CrCl over the entire ICU stay was similar (median (IQR) 93.1 (56.2–133.6) ml/min, p = 0.7) (Table | PMC10327364 | ||
Features selected for CrCl prediction | Among the ten most predictive variables of the three models, seven were related to CrCl, one to urea level, and the remaining two were the baseline characteristics age and body mass index (BMI) (Fig. Top ten most important features of different models. The red, green, and blue bar plots are the results for the Core, Core + BGA, and Core + BGA + Monitoring models, respectively | PMC10327364 | ||
Discussion | toxicity, critically ill | CRITICALLY ILL | In this study, we presented three models to predict daily CrCl in critically ill adults, based on information derived from routinely collected clinical data. The predictive performance remained similar when adding high-resolution data. The developed models were externally validated on previously unseen patients with good performance. Finally, the models demonstrated smaller prediction errors than using the CrCl of the day of prediction (reflecting the current clinical practice. This is mainly explained by the better performance of the prediction model as compared to the reference of the previous day on the days with a bigger change in the CrCl (i.e., ‘unstable days’), whereas the model performed similarly as the reference of the previous days on the days that the CrCl remained stable (‘stable days’). To the best of our knowledge, this study presents the first machine-learning algorithm for daily CrCl prediction in the ICU.There are many reasons why there is a need for such daily prediction of CrCl during the entire ICU stay. First, measured urinary CrCl is currently considered the most suitable method to estimate the GFR in clinical practice [Having a reference to compare against helps to understand whether the models could have clinical usefulness. Compared to the current clinical practice of assuming the same CrCl as the day of prediction, our developed models reduced the RMSE from 40.1 to 28.9 ml/min. Importantly, in the subgroup of patient-days with stable kidney function, the developed models demonstrated a clinically insignificant larger RMSE, around 4 ml/min on average, than the reference. This difference, however, is too small to be of clinical relevance and unlikely to cause treatment failure or drug toxicity due to altered renal clearance. Noticeably, in the subgroup of patient-days with unstable kidney function (comprising 30–40% of all patient-days), the developed models had clinically relevant smaller RMSEs, around 23 ml/min on average. This subgroup analysis of days with high CrCl instability clearly exhibited our models’ capability of better capturing the dynamics of kidney function. Nevertheless, despite the large reduction in prediction errors during unstable days, whether or not the models help improve drug dosing and patient outcomes still needs to be investigated prospectively.Our study has many strengths. First, the use of a general ICU population instead of a specific subset of patients made it more generally applicable, and the daily prediction truthfully reflected the fluctuating kidney function on each patient day, allowing for risk stratification and drug dose adjustment. Second, the reporting of this study was performed following the Transparent Reporting of a Multivariate Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) guidelines [There are several limitations in our study. First, the development cohort was based on a RCT database in Belgium dating back to 2010, which might limit its generalizability in other settings. However, model performance remained unchanged when externally validated on a very large database with patient data collected up to 2018. Second, the use of high-resolution data might be difficult to implement in hospitals with limited resources, and some settings might even struggle to have the necessary data for the lower-resolution Core model. Third, there might be a selection bias resulting from the exclusion from the analyses of patient-days with KRT on the day of prediction and in previous week, or of patient-days when less than 2 consecutive CrCls were available, or patient-days after the first 90 days in ICU. These exclusion criteria were necessary to ensure reliable CrCl prediction models could be developed. Fourth, this study was based on retrospective data, and the developed models still need prospective validation in independent cohorts. Fifth, the model performance was not compared against novel biomarkers such as cystatin C and proenkephalin that may be less biased, and urea clearance that may add valuable information especially in those with low CrCl. However, measured CrCl is a fast and cheap test, which are important characteristics as the measurements were taken on a daily basis. Sixth, the measurement of creatinine changed from the Jaffe method in the development cohort to the enzymatic method in the validation cohort, and it was found that the Jaffe method yielded higher creatinine values [ | PMC10327364 |
Conclusions | We have shown that CrCl can be accurately predicted one day in advance during ICU stay. We have also demonstrated the robustness of the developed models on previously unseen patients in external validation. The developed models’ usefulness has been shown in comparison with a reference reflecting current clinical practice, mainly on the patient-days with high kidney function instability. Despite the promising performance, these findings should be prospectively validated in independent patient populations, before these prediction models can be further used for risk stratification or incorporated into a pharmacokinetic model to support a more optimized dose regimen. | PMC10327364 | ||
Acknowledgements | The authors would like to thank Prof. Dr. Miet Schetz for her contribution to the EPaNIC database, and the members of the M@tric research group for designing the M@tric database, which were used to develop and validate the CrCl prediction models, respectively. | PMC10327364 | ||
Take-home message | Daily creatinine clearance can be accurately predicted one day ahead, by machine-learning models using routinely collected clinical data. In the future, these models could be useful for hydrophilic drug dosage adjustment or stratification of patients at risk | PMC10327364 | ||
Author contributions | CYH, FG, GDV, GM did conception and design; CYH, FG, PW, LM, PM, MC, GM done collection and assembly of data; CYH, FG, GDV, GM were involved in data analysis and interpretation; CYH, FG, GDV, GM wrote the manuscript; all authors contributed to critical revision of the manuscript: and the final approval of manuscript. All authors read and approved the final manuscript. | PMC10327364 | ||
Funding | CYH receives funding from the Taiwan-KU Leuven scholarship. GM is funded by the research foundation, Flanders (FWO) as senior clinical investigator (1843123N). JG is granted a postdoctoral research fellowship by the Clinical Research and Education Council of the University Hospitals Leuven. MC receives funding from the Research Foundation Flanders (FWO) (Grant No. 1832817N) and Onderzoeksraad, KU Leuven (Grant No. C24/17/070) and from the Private Charity Organization “Help Brandwonden Kids.” | PMC10327364 | ||
Availability of data and materials | The datasets generated and/or analyzed during the current study are not publicly available due to no prior agreement with the ethical committee but are available from the corresponding author on reasonable request. | PMC10327364 | ||
Declarations | PMC10327364 | |||
Ethical approval and consent to participate | This study was conducted on the basis of prior informed consent by all patients or their legal representatives, and the consent forms included the permission to use the data for additional research (S50404). Approval for the use of these patient data in the present study was obtained from the Ethics Committee of University Hospitals Leuven (S61364). The study was conducted in compliance with the principles of the Declaration of Helsinki and its later revisions. | PMC10327364 | ||
Consent for publication | Not applicable. | PMC10327364 | ||
Competing interests | On behalf of all authors, the corresponding author states that there is no conflict of interest. | PMC10327364 | ||
References | PMC10327364 | |||
Background: | Parkinson’s disease, PD | These authors contributed equally to this work.The sequence effect is the progressive deterioration in speech, limb movement, and gait that leads to an inability to communicate, manipulate objects, or walk without freezing of gait. Many studies have demonstrated a lack of improvement of the sequence effect from dopaminergic medication, however few studies have studied the metric over time or investigated the effect of open-loop deep brain stimulation in people with Parkinson’s disease (PD). | PMC10357155 | |
Objective: | To investigate whether the sequence effect worsens over time and/or is improved on clinical (open-loop) deep brain stimulation (DBS). | PMC10357155 | ||
Methods: | PD, repetitive wrist flexion extension | Twenty-one people with PD with bilateral subthalamic nucleus (STN) DBS performed thirty seconds of instrumented repetitive wrist flexion extension and the MDS-UPDRS III off therapy, prior to activation of DBS and every six months for up to three years. A sub-cohort of ten people performed the task during randomized presentations of different intensities of STN DBS. | PMC10357155 | |
Results: | The sequence effect was highly correlated with the overall MDS-UPDRS III score and the bradykinesia sub-score and worsened over three years. Increasing intensities of STN open-loop DBS improved the sequence effect and one subject demonstrated improvement on both open-loop and closed-loop DBS. | PMC10357155 | ||
Conclusion: | DISEASE PROGRESSION | Sequence effect in limb bradykinesia worsened over time off therapy due to disease progression but improved on open-loop DBS. These results demonstrate that DBS is a useful treatment of the debilitating effects of the sequence effect in limb bradykinesia and upon further investigation closed-loop DBS may offer added improvement. | PMC10357155 | |
INTRODUCTION | PD, peripheral muscle fatigue, bradykinesia | MOVEMENT DISORDERS, DISEASE | The sequence effect is the progressive deterioration in ongoing movement that is not related to peripheral muscle fatigue [Although different components of bradykinesia can be treated with levodopa and/or deep brain stimulation (DBS), a critical unmet need for improving the lives of people with PD is that the sequence effect does not respond to dopaminergic medication [Our goal in this study was to investigate whether the sequence effect improved during subthalamic nucleus (STN) DBS in people with PD and whether there was a ‘dose” or DBS intensity dependence. We measured the sequence effect during repetitive wrist flexion-extension, correlated it with the Movement Disorders Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS III) and with its sub-score of lateralized bradykinesia and demonstrated that it became worse over time in the off therapy state, but improved with STN DBS. | PMC10357155 |
MATERIALS AND METHODS | PMC10357155 | |||
Participants | PD | Twenty-one individuals (5 female) with clinically established PD underwent bilateral implantation of DBS leads (model 3389, Medtronic PLC) in the STN. The two leads were connected to the implanted investigative neurostimulator (Activa | PMC10357155 | |
Experimental protocol | PMC10357155 | |||
Off stimulation longitudinal testing | PD, bradykinesia | Experiments were conducted at initial programming (IP, before the initial activation of the DBS system), which took place 1 month after implantation of the DBS leads, and subsequently at 6 months intervals out to 3 years after the IP visit; the maximum number of total visits was 7 (IP, 6, 12, 18, 24, 30, 36 months). All experimental testing was done in the off-medication state, which entailed the withdrawal of long-acting dopamine agonists for 48 hours, dopamine agonists and controlled release carbidopa/levodopa for 24 h, and short acting medication for 12 h prior to the study visit. At the follow-up visits, stimulation was turned off for 60–75 min. The participant then performed a single trial of repetitive wrist-flexion extension (rWFE) task, which we have previously validated as a measure of bradykinesia in PD [ | PMC10357155 | |
Titrations | tremor, fatigue | A sub-cohort of 10 individuals (3 female) completed a stimulation amplitude titration experiment, off medication. These participants had been implanted with investigative deep brain neurostimulators that recorded STN local field potentials synchronously with kinematic signals during a variety of tasks. Testing was done once optimized on DBS settings and 1.5–3 years after their initial programming visit, and at a time when they tolerated changing of stimulation. Participants in this study did not have tremor that superseded the entire trial of the WFE task. In some subjects, only one wrist was assessed if the subject had excess neural artifact that resulted in that limb/STN being excluded from the experiment. Furthermore, due to protocol time restrictions and fatigue for the subjects, one or both wrists were assessed (randomized in order). Participants performed five trials of the rWFE task, where a single trial was performed unilaterally on one randomized side. Each trial was performed during randomized presentations of STN DBS at 0% (no DBS), 25%, 50%, 75%, and 100% of V | PMC10357155 | |
Closed-loop DBS | In one participant, data were collected during the rWFE task in three stimulation conditions from the same wrist: off stimulation (one trial), clinical open-loop stimulation (olDBS) (one trial), and neural closed-loop stimulation (NclDBS) (five trials, each with a varying stimulation delay period). The NclDBS condition used the beta burst driven adaptive closed-loop algorithm [ | PMC10357155 | ||
Data acquisition and analysis | PMC10357155 | |||
Kinematic data acquisition | Movement was measured using solid-state gyroscopic wearable sensors (sampled at 1 kHz) attached to the dorsum of each hand (Motus Bioengineering, Inc, Benicia, CA) and with synchronized video recordings from a USB web camera (C930e, Logitech, Lausanne, Switzerland). Sampling rates for the gyroscope and video data were 1 kHz and 30 frames per second, respectively. | PMC10357155 | ||
Kinematic data analysis | tremor | DECAY | For each movement epoch, an automated algorithm was used to quantify the sequence effect. The angular velocity data was low-pass filtered in MATLAB using zero-lag 4th order Butterworth filters with a 4 Hz cutoff frequency at the above sampling frequency. Peaks for analysis were chosen based on the maximum angular velocity between each zero crossing for each cycle of flexion-extension. For traces with excess tremor that required extra smoothing to find accurate zero crossings, the angular velocity data was low-pass filtered again using zero-lag 4th order Butterworth filters starting with a 2.5 Hz cutoff frequency, and if subsequent filtering was required the data was re-filtered using a cutoff frequency that decreased by 0.5 Hz until no further filtering was required or the cutoff frequency decreased to 0 Hz, at which case the trace was not usable as zero crossings could not be accurately detected. This filtered trace was only used to identify the zero crossings, which were then applied to the peak detection on the original angular velocity data.Since some trials displayed multiple epochs of sequence effect in which the participant was able to reset following initial decrements in angular velocity, an automated algorithm was used to determine if the trial should be broken up into one or multiple epochs. First, a 3-point moving average was calculated on the angular velocity peaks to better dynamically visualize and model the trend in behavior as well as filter out periodic fluctuations and noise. To further protect the sequence effect models against overfitting from sudden behavioral fluctuations, the percent change was calculated on the moving average of peaks in two ways: between the current peak and the subsequent peak (denoted as PC1) as well as between the current peak and the next 2 peaks (denoted as PC2). A negative percent change represented a smaller angular velocity from previous, typically seen during the sequence effect epoch, and a positive percent change represented a larger angular velocity from previous, which if large enough represented the end of a sequence effect epoch. Upon inspection across the cohort’s rWFE traces, a threshold of 20% was empirically derived, where once the percent change in angular velocity crossed this threshold, a sequence effect epoch had been completed.Following epoching, an exponential curve was fit to the first epoch of decay in a trace using the following criteria: for those where the maximum angular velocity is greater than 100 degrees/s and there are at least 10 full cycles of rWFE present, the initial point of the fitted exponential was chosen as the maximum of the first 10 peaks. For traces where the maximum angular velocity was less than 100 degrees/s and there were at least 5 full cycles of rWFE present, the initial point was chosen as the maximum of the first 5 peaks. For traces that fit neither criterion, the initial point was chosen as the first peak. The algorithm starts a new epoch when either the PC1 or PC2 trace crossed the 20% threshold and the current peak is at least 40% of the first epoch’s maximum peak (representing an accurate pick up in behavior). When this is true, the first epoch will end with the last point with negative percent change prior to crossing the 20% threshold and the second epoch will begin to be fit with an exponential function, and this process repeats until the entire trace’s sequence effect epochs have been fit with exponential curves.The sequence effect behavior in each epoch was modeled using the exponential decay function equation:
Where The slope of the decay was normalized by the initial angular velocity, Finally, to express the sequence effect metric as a percentage where a higher number was indicative of greater (i.e., worse) sequence effect, the inverse of For cases where the epoch is overall increasing and an exponential growth curve was more suitable for modeling the behavior, the epoch was modeled using | PMC10357155 |
Statistical analysis | REGRESSION | Statistics were computed using MATLAB (version 9.9, The MathWorks Inc. Natick, MA, USA). Pearson correlations were used to assess associations between the sequence effect during rWFE and both total MDS-UPDRS III scores and the bradykinesia sub-score. A linear mixed effects regression model was performed to analyze the effect of time (in months) on sequence effect. In this case, the sequence effect metric was included as the dependent variable and visit month was used as a fixed effect with subject as a random intercept and time as a random slope. For analysis of the titrations results, Kolmogorov-Smirnov tests and theoretical-sample quantile plots were used to assess the normality of the distribution of sequence effect at each stimulation condition. Based on the titrations experiment results, a repeated measures ANOVA was used to compare the effect of stimulation level on sequence effect across all STNs at the five stimulation conditions. Paired | PMC10357155 | |
RESULTS | Movement Disorder, Parkinson’s disease, Parkinson’s Disease Rating Scale | MOVEMENT DISORDER | Kinematic data from 42 hands of 21 well-characterized individuals with Parkinson’s disease, off therapy, were included in the analysis. Participant demographicsIP, initial programming; UPDRS, Unified Parkinson’s Disease Rating Scale; MDS, Movement Disorder Society. | PMC10357155 |
Quantification of sequence effect epochs | SE |
Quantification of sequence effect on example WFE traces. (A) WFE trace with minimal sequence effect. (B) WFE with one epoch of substantial sequence effect for the entire trace. (C) WFE split into two epochs based on reemergence of performance around 25 seconds. Fitted exponential line is shown in black. Detected peaks shown by orange open circles. Exponential fit function and subsequent sequence effect (SE) metric shown above each trace.Across the 21 participants throughout the 3 years of repeat visits, there were 212 total trials (Trials over time | PMC10357155 | |
Sequence effect is related to overall motor impairment and bradykinesia |
Scatter plot between sequence effect during rWFE and (A) total MDS-UPDRS III scores and (B) lateralized bradykinesia sub-scores.This relationship was confirmed at each 6-month timepoint for both total MDS-UPDRS III score: (6m: | PMC10357155 | ||
Sequence effect worsens over time |
Sequence effect worsens over time off therapy. Average slope of change over time (thick black line) with individual data overlaid as line plots (light gray) of sequence effect. * indicates significant change over time. Dashed lines represent the 95% confidence interval of the slope estimate.The sequence effect significantly increased over time, off therapy (β= 0.0453 (95% CI: 0.0328 to 0.0578), | PMC10357155 | ||
STN DBS improves the sequence effect |
Boxplots comparing the sequence effect at different intensities of DBS. * indicates
Dose-dependent reduction in mean sequence effect | PMC10357155 | ||
Sequence effect improved during neural closed-loop DBS in one participant |
Example of Wrist Flexion Extension (WFE) (A) OFF therapy, (B) open-loop clinical stimulation, (C). neural closed-loop stimulation. Peaks from each WFE cycle indicated by black dots with the fitted exponential curve overlaid.NclDBS improved the sequence effect in 8 out of the 10 NclDBS conditions across various delay periods between the two hands (Neural closed-loop parameters | PMC10357155 | ||
DISCUSSION | PD, upper extremity bradykinesia | This study found that an objective normalized metric of the sequence effect in upper extremity bradykinesia strongly correlated with the total MDS-UPDRS III score and with the upper and lower extremity lateralized MDS-UPDRS III bradykinesia subscore in people with PD. The sequence effect worsened over time in a longitudinal cohort studied off therapy up to three years after initial programming, demonstrating that it reflects the progression of PD. The sequence effect in limb bradykinesia improved during STN continuous open loop DBS in a dose-dependent manner and in one participant the sequence effect improved further on neural closed loop DBS. | PMC10357155 | |
Measuring the sequence effect in limb movement | PD | One of the difficulties in assessing evidence of the sequence effect and its response to therapies in PD is the paucity of measurement tools that can differentiate the sequence effect from other metrics such as amplitude and frequency. The MDS-UPDRS motor subscale (MDS-UPDRS III) groups together assessments of impairment in amplitude, frequency, and sequence effect into one integer for each item related to limb bradykinesia. Consequently, one cannot discern a specific metric of the sequence effect from the MDS-UPDRS III. Even with quantitative measures, the measure of the sequence effect has varied from a comparison of the first and last multiple of cycles of repetitive movement or gait [ | PMC10357155 | |
The sequence effect is a validated measure of PD motor disability | dystonia, PD motor disability, PD, bradykinesia, Huntington’s disease, PSP | DYSTONIA, MULTIPLE SYSTEM ATROPHY, DISEASE, PROGRESSIVE SUPRANUCLEAR PALSY | In this study the sequence effect was significantly correlated with overall PD motor disability (MDS-UPDRS III), and this was not solely due to lower angular velocities in later stages of disease, as it was normalized by initial peak velocity. The sequence effect was also significantly correlated with the lateralized bradykinesia sub-score on the MDS-UPDRS III; this is expected as bradykinesia is defined as “slowness of movement and decrement in amplitude or speed (sequence effect) as movements are continued”, which is similar to its definition in the first formal diagnostic criteria for PD [Previous literature suggests that the sequence effect appears to be specific to PD and multiple system atrophy but does not appear to be a major feature of progressive supranuclear palsy (PSP), Huntington’s disease, or dystonia [ | PMC10357155 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.