Fall prediction from wearables showed uneven accuracy in older adults
A review found higher specificity than sensitivity for predicting falls in older adults, highlighting a gap between overall prediction accuracy and identifying future fallers.
Based on the published abstract. The full paper may contain additional methods, results and limitations.
The 30-second takeaway
Wearable-based models were better at correctly classifying people who would not fall than at identifying those who would. That distinction matters when judging how reassuring a low-risk result should be. The review assessed prediction accuracy, and its abstract does not demonstrate that using these devices reduces falls or improves the outcomes of prevention programs.
The abstract does not report the prediction horizon or summarize risk-of-bias findings. Performance varied with study characteristics, and the pooled results do not establish accuracy for any particular device or person.
Understanding the accuracy measures
Sensitivity describes how well a model identifies people who later fall; specificity describes how well it identifies those who do not. The pooled results were stronger on specificity, so some future fallers went unidentified.
Differences between prediction models
Studies using machine learning models showed stronger overall discrimination. This subgroup result may reflect differences between studies; it does not establish that a particular machine learning device is the best choice for an individual.
The original publication
Accuracy of wearable devices in predicting falls in older adults: a systematic review and meta-analysis.
Mou C, Yan X, Miao X et al.
Front Public Health · 2026
- PubMed ID
- 41889634
- Record checked
AI-assisted research and writing. This explains one selected publication; it is not a complete review of everything known. Our approach.
One more question, understood.
Keep track of the research you’ve explored.