Evaluation of Machine Learning Methods Developed for Prediction of Diabetes Complications: A Systematic Review.

BACKGROUND: With the rising prevalence of diabetes, machine learning (ML) models have been increasingly used for prediction of diabetes and its complications, due to their ability to handle large complex data sets. This study aims to evaluate the quality and performance of ML models developed to predict microvascular and macrovascular diabetes complications in an adult Type 2 diabetes population. METHODS: A systematic review was conducted in MEDLINE®, Embase®, the Cochrane® Library, Web of Science®, and DBLP Computer Science Bibliography databases according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist. Studies that developed or validated ML prediction models for microvascular or macrovascular complications in people with Type 2 diabetes were included. Prediction performance was evaluated using area under the receiver operating characteristic curve (AUC). An AUC >0.75 indicates clearly useful discrimination performance, while a positive mean relative AUC difference indicates better comparative model performance. RESULTS: Of 13 606 articles screened, 32 studies comprising 87 ML models were included. Neural networks (n = 15) were the most frequently utilized. Age, duration of diabetes, and body mass index were common predictors in ML models. Across predicted outcomes, 36% of the models demonstrated clearly useful discrimination. Most ML models reported positive mean relative AUC compared with non-ML methods, with random forest showing the best overall performance for microvascular and macrovascular outcomes. Majority (n = 31) of studies had high risk of bias. CONCLUSIONS: Random forest was found to have the overall best prediction performance. Current ML prediction models remain largely exploratory, and external validation studies are required before their clinical implementation. PROTOCOL REGISTRATION: Open Science Framework (registration number: 10.17605/OSF.IO/UP49X).

Measurement Properties of Patient-Reported Outcome Measures for Diabetes: Systematic Review.

BACKGROUND: The management of diabetes is complex. There is growing recognition of the use of patient-reported outcome measures (PROMs) as a standardized method of obtaining an outlook on patients' functional status and well-being. However, no systematic reviews have summarized the studies that investigate the measurement properties of diabetes PROMs. OBJECTIVE: Our aims were to conduct a systematic review of studies investigating the measurement properties of diabetes PROMs by evaluating the methodological quality and overall level of evidence of these PROMs and to categorize them based on the outcome measures assessed. METHODS: This study was guided by the PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) guidelines. Relevant articles were retrieved from the Embase, PubMed, and PsychINFO databases. The PROMs were evaluated with the COSMIN (COnsensus-based Standards for the selection of health Measurement Instruments) guidelines. RESULTS: A total of 363 articles evaluating the measurement properties of PROMs for diabetes in the adult population were identified, of which 238 unique PROMs from 248 studies reported in 209 articles were validated in the type 2 diabetes population. PROMs with at least a moderate level of evidence for ≥5 of 9 measurement properties include the Chinese version of the Personal Diabetes Questionnaire (C-PDQ), Diabetes Self-Management Instrument Short Form (DSMI-20), and Insulin Treatment Appraisal Scale in Hong Kong primary care patients (C-ITAS-HK), of which the C-PDQ has a "sufficient (+)" rating for >4 measurement properties. A total of 43 PROMs meet the COSMIN guidelines for recommendation for use. CONCLUSIONS: This study identified and synthesized evidence for the measurement properties of 238 unique PROMs for patients with type 2 diabetes and categorized the PROMs according to their outcome measures. These findings may assist clinicians and researchers in selecting appropriate high-quality PROMs for clinical practice and research. TRIAL REGISTRATION: PROSPERO International Prospective Register of Systematic Reviews CRD42020180978; https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020180978.

An Adaptable Framework for Factors Contributing to Medication Adherence: Results from a Systematic Review of 102 Conceptual Frameworks.

OBJECTIVE: To summarize the available conceptual models for factors contributing to medication adherence based on the World Health Organization (WHO)'s five dimensions of medication adherence via a systematic review, identify the patient groups described in available conceptual models, and present an adaptable conceptual model that describes the factors contributing to medication adherence in the identified patient groups. METHODS: We searched PubMed®, Embase®, CINAHL®, and PsycINFO® for English language articles published from inception until 31 March 2020. Full-text original publications in English that presented theoretical or conceptual models for factors contributing to medication adherence were included. Studies that presented statistical models were excluded. Two authors independently extracted the data. RESULTS: We identified 102 conceptual models, and classified the factors contributing to medication adherence using the WHO's five dimensions of medication adherence, namely patient-related, medication-related, condition-related, healthcare system/healthcare provider-related, and socioeconomic factors. Eight patient groups were identified based on age and disease condition. The most universally addressed factors were patient-related factors. Medication-related, condition-related, healthcare system-related, and socioeconomic factors were represented to various extents depending on the patient group. By systematically examining how the WHO's five dimensions of medication adherence were applied differently across the eight different patient groups, we present a conceptual model that can be adapted to summarize the common factors contributing to medication adherence in different patient groups. CONCLUSION: Our conceptual models can be utilized as a guide for clinicians and researchers in identifying the facilitators and barriers to medication adherence and developing future interventions to improve medication adherence. PROTOCOL REGISTRATION: PROSPERO Identifier: CRD42020181316.

Development of an Item Bank to Measure Medication Adherence: Systematic Review.

BACKGROUND: Medication adherence is important in managing the progression of chronic diseases. A promising approach to reduce cognitive burden when measuring medication adherence lies in the use of computer-adaptive tests (CATs) or in the development of shorter patient-reported outcome measures (PROMs). However, the lack of an item bank currently hampers this progress. OBJECTIVE: We aim to develop an item bank to measure general medication adherence. METHODS: Using the preferred reporting items for systematic review and meta-analysis (PRISMA), articles published before October 2019 were retrieved from PubMed, Embase, CINAHL, the Cochrane Library, and Web of Science. Items from existing PROMs were classified and selected ("binned" and "winnowed") according to standards published by the Patient-Reported Outcomes Measurement Information System (PROMIS) Cooperative Group. RESULTS: A total of 126 unique PROMs were identified from 213 studies in 48 countries. Items from the literature review (47 PROMs with 579 items for which permission has been obtained) underwent binning and winnowing. This resulted in 421 candidate items (77 extent of adherence and 344 reasons for adherence). CONCLUSIONS: We developed an item bank for measuring general medication adherence using items from validated PROMs. This will allow researchers to create new PROMs from selected items and provide the foundation to develop CATs.

Measurement Properties of Existing Patient-Reported Outcome Measures on Medication Adherence: Systematic Review.

BACKGROUND: Medication adherence is essential for improving the health outcomes of patients. Various patient-reported outcome measures (PROMs) have been developed to measure medication adherence in patients. However, no study has summarized the psychometric properties of these PROMs to guide selection for use in clinical practice or research. OBJECTIVE: This study aims to evaluate the quality of the PROMs used to measure medication adherence. METHODS: This study was guided by the PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) guidelines. Relevant articles were retrieved from the EMBASE, PubMed, Cochrane Library, Web of Science, and CINAHL (Cumulative Index to Nursing and Allied Health Literature) databases. The PROMs were then evaluated based on the COnsensus-based Standards for the selection of health Measurement Instruments (COSMIN) guidelines. RESULTS: A total of 121 unique medication adherence PROMs from 214 studies were identified. Hypotheses testing for construct validity and internal consistency were the most frequently assessed measurement properties. PROMs with at least a moderate level of evidence for ≥5 measurement properties include the Adherence Starts with Knowledge 20, Compliance Questionnaire-Rheumatology, General Medication Adherence Scale, Hill-Bone Scale, Immunosuppressant Therapy Barrier Scale, Medication Adherence Reasons Scale (MAR-Scale) revised, 5-item Medication Adherence Rating Scale (MARS-5), 9-item MARS (MARS-9), 4-item Morisky Medication Adherence Scale (MMAS-4), 8-item MMAS (MMAS-8), Self-efficacy for Appropriate Medication Adherence Scale, Satisfaction with Iron Chelation Therapy, Test of Adherence to Inhalers, and questionnaire by Voils. The MAR-Scale revised, MMAS-4, and MMAS-8 have been administered electronically. CONCLUSIONS: This study identified 121 PROMs for medication adherence and provided synthesized evidence for the measurement properties of these PROMs. The findings from this study may assist clinicians and researchers in selecting suitable PROMs to assess medication adherence.