736 REAL-TIME MOBILITY DATA FROM WEARABLE DEVICES PREDICTING FALLS RISK IN OLDER ADULTS

Abstract Introduction Falls are a common cause of injuries in people over 65 years, usually resulting life-threatening or life-altering sequalae for affected individuals. risk assessment typically requires face-to-face evaluation delivered by trained professional clinical setting, thus limiting their application many older with falls. Therefore, there is need remote monitoring system (RMS) that captures mobility data to assess comprehensively an individual’s true falls undertaking everyday activities usual environment. Methods A pilot study was conducted 16 participants years classified as high (n = 9) low 7) based on fall’s history. RMS developed, which combines inertial measurement unit (IMU) from wearable device and online real-time Machine Learning (ML) generate two-stage fall assessment. In stage one, ML activity motion classifier identified raw IMU either, walking, resting transitional movements (turning sitting standing). two, previously segments were analysed fall-risk classifier, providing objective gait characteristics during motion. incidence recorded prospectively at one year follow-up. Results The achieved 94.8% accuracy differentiating between resting, walking movements. initially retrospective history 87.5% 88.9% sensitivity specificity, comparable other accurate classifiers published literature. retrained prospective follow-up 73% superior the 67% Timed Up Go test prediction. Conclusion Our shows potential machine-learning using information classify person’s falls’ risk.