Predicting falls in community-dwelling older adults using a composite wearable sensor score: Preliminary results

Background: Falls represent a leading cause of disability and mortality in the elderly population, yet traditional screening methods, such as the Timed Up and Go (TUG) test, often lack the sensitivity to detect subtle motor deficits. This study aims to evaluate whether gait parameters derived from wearable inertial sensors, improve fall risk prediction in community-dwelling older adults compared to conventional methods.Methods: This prospective longitudinal study enrolled participants aged ≥65 years in Rome, Italy. At baseline, participants underwent the Short Functional Geriatric Evaluation (SFGE) and performed the TUG and 10-Meter Walk Test (10MWT) wearing a tri-axial accelerometer. A 6-month follow-up was conducted to record falls incidence. A multivariate logistic regression model was developed to identify predictors of future falls, and Receiver Operating Characteristic (ROC) analysis was used to determine optimal cut-off values.Results: A total of 234 participants aged ≥65 years were enrolled. At the 6-month follow-up, 26.6% of the 94 participants who completed follow-up experienced a fall. The strongest predictors of future falls were history of previous falls (OR 6.8) and antero-posterior asymmetry during TUG (OR 5.8).  A composite predictive score incorporating fall history, gait asymmetry, step count, and test duration achieved an Area Under the Curve (AUC) of 0.85, significantly outperforming single-parameter assessments. Notably, the standard CDC-recommended TUG cut-off (≥12s) demonstrated high sensitivity (92%) but poor specificity (15.2%) in this cohort. Conversely, a data-driven cut-off of ≥16s offered a better balance of predictive accuracy (Youden Index 0.261 vs. 0.072).Discussion: The integration of wearable sensor data into a multiparametric assessment reveals a distinct "motor signature" in at-risk elderly individuals characterized by gait asymmetry and variability. The multiparametric approach shows a stronger association with falls than is observed for the TUG, supporting the adoption of instrumented gait analysis for precision fall prevention strategies in community settings.