AI Framework Personalizes Fall Risk Prevention for Older Adults.

Falls among older adults pose a significant public health challenge, stemming from complex and changing interactions of various risk factors. Traditional methods for identifying these risks often fall short because they are static and fail to adapt to an individual's evolving health profile over time, especially when dealing with sparse and delayed fall events in heterogeneous longitudinal data. To address this, researchers developed PAFIR (Personalized and Adaptive Feature selection framework for fall risk Identification and pRevention). PAFIR frames adaptive feature selection as a reinforcement learning problem, processing longitudinal multimodal health data. It simultaneously models the structural relationships between assessment variables and the temporal dynamics observed in wearable device data. The framework learns adaptive selection policies across repeated study visits, guided by reward signals derived from actual fall incidents. Applied to data from the PEER clinical trial, PAFIR demonstrated superior capability in capturing both longitudinal and structural patterns of feature relevance compared to existing state-of-the-art baselines. By dynamically adapting the selected features over time, PAFIR enables the creation of more personalized and timely fall prevention strategies, promising a significant improvement in elder care.