Cochrane-Style Systematic Review Protocol — AI / ML Prediction Models for IoT-Based Remote Patient Monitoring of Chronic Disease (v1.2, 2026-08-29)

Abstract

Protocol version 1.2 (2026-08-29), superseding v1.1 (2026-05-12). Methodology specification for a Cochrane-style systematic review of AI and machine-learning prediction-model studies in IoT-based remote patient monitoring of chronic disease (chronic kidney disease, heart failure, diabetes, multi-morbidity). Methodological anchors. Cochrane Handbook for Systematic Reviews of Prognosis Research and Prediction Models (Riley, van der Windt, Croft and Moons, OUP 2019), and Cochrane Methodological Expectations conduct standards C1-C75. Appraisal and extraction. PROBAST+AI (Moons et al. 2025, BMJ) as the primary risk-of-bias tool; CHARMS extraction under a 100% dual independent process. Synthesis. REML with the Hartung-Knapp-Sidik-Jonkman correction on the logit-transformed c-statistic, reported with a 95% prediction interval; SWiM 9-item narrative synthesis where pooling is infeasible; GRADE per the Foroutan concept papers (2024, 2021, 2020). A pre-specified validation-design meta-regression tests partition-level leakage as a mechanism for optimistic discrimination estimates. Reporting and equity. PRISMA 2020, with TRIPOD-SRMA, PRISMA-S, PRISMA-Equity 2012 and PRISMA-Abstracts 2020; PROGRESS-Plus equity extraction with SAGER sex and gender disaggregation; AMSTAR-2 and ROBIS pre-submission self-audit. Dating. The specification is dated 2026-05-12 (v1.1) and 2026-08-29 (v1.2). This Zenodo record was first deposited on 2026-08-28, and each version's publication date is the date that version was deposited. The deposit is retrospective and makes no claim that the protocol was publicly available before 2026-08-28. Registration anchors. OSF Registries at https://osf.io/pkq8c/, PROSPERO (CRD pending), and this Zenodo record. AI tool disclosure per RAISE 1 (2025) and ICMJE 2023. This deposit is the protocol companion to the data-only reproducibility package (DOI 10.5281/zenodo.20057329).