Systematic Review Evaluates ML Models Predicting Future Falls
According to the JMIR preprint by Gao et al. (2026), the authors screened 6,865 records and included 27 longitudinal studies that developed or validated machine learning (ML) or deep learning (DL) models to predict future falls in community-dwelling adults aged >=60. Per the paper, 22 studies used ML and 5 used DL; prediction horizons ranged from 3 months to 4 years and reported fall incidence varied from 1.6% to 46.6%. Gao et al. searched major databases through August 31, 2025, excluded real-time detection, simulated/no-fall and inpatient studies, assessed risk of bias with PROBAST, and meta-analyzed logit-transformed AUCs using the Hartung-Knapp-Sidik-Jonkman (HKSJ) random-effects method with 95% prediction intervals estimated by four approaches. The authors also ran leave-one-out robustness checks and examined small-study effects with funnel plots and Egger-type tests.
