Digital Healthcare Approaches for Fall Detection and Prediction in Older Adults: A Systematic Review of Evidence from Hospital and Long-Term Care Settings

Background and Objectives: Falls are a leading cause of morbidity and mortality among older adults in hospitals and long-term care facilities (LTCFs). Digital healthcare approaches are increasingly being applied to fall detection and prevention; however, their effectiveness remains uncertain. This review evaluated the effectiveness, usability, and clinical applicability of detection- and prediction-based systems in institutional settings. Materials and Methods: We systematically searched major international and Korean databases—PubMed, Embase, Ovid-MEDLINE, CINAHL, the Cochrane Library, IEEE, KMbase, KISS, KoreaMed, and RISS—for studies published up to December 2024. The eligible studies included randomized controlled trials, quasi-experimental, and observational studies involving older adults in hospitals or LTCFs. Two reviewers independently screened the studies, extracted data, and assessed their quality using standardized tools. Results: Thirty-three studies comprising 20 fall detection systems and 13 fall prediction models were included. Detection systems using inertial, pressure, radar, or multimodal sensors have improved monitoring and achieved high usability (>80% acceptance); however, they did not consistently reduce fall incidence or the occurrence of injurious falls. For instance, one trial reported a nonsignificant reduction in injurious falls (aRR 0.56, 95% CI 0.17–1.79), whereas another trial observed a nonsignificant increase (aIRR 1.60, 95% CI 0.83–3.08). Frequent false alarms contribute to alarm fatigue. The prediction models showed moderate-to-strong discrimination. Gradient boosting and neural networks performed best for continuous gait features, while regression and boosting approaches were effective for categorical EHR data. Most models lacked external validation and were not linked to clinical care pathways. Conclusions: Digital approaches show potential for fall prevention in hospitals and LTCFs; however, current evidence remains inconsistent and limited. Detection systems improve surveillance but offer limited preventive effects, whereas prediction models demonstrate technical promise without establishing clinical benefits. Future research should refine the technology, validate models externally, and integrate them into patient-centered workflows.