Can a Body-Fixed Sensor Reduce Heisenberg's Uncertainty When It Comes to the Evaluation of Mobility? Effects of Aging and Fall Risk on Transitions in Daily Living

Background: Functional performance-based tests like the Timed Up and Go test (TUG) and its subtasks have been associated with fall risk, future disability, nursing home admission, and other poor outcomes in older adults. However, a single measurement in the laboratory may not fully reflect the subject's condition and everyday performance. To begin to validate an approach based on long-term, continuous monitoring, we investigated the sit-to-walk and walk-to-sit transitions performed spontaneously and naturally during daily living. Methods: Thirty young adults, 38 older adults, and 33 elderly (idiopathic) fallers were studied. After evaluating mobility and functional performance in the laboratory, participants wore an accelerometer on their lower back for 3 days. We analyzed the sit-to-walk and walk-to-sit transitions using temporal and distribution-related features. Machine learning algorithms assessed the feature set's ability to discriminate between the different cohorts. Results: 5,027 transitions were analyzed. Significant differences were observed between the young and older adults (p <. 044) and between the fallers and older adults (p <. 032). Machine learning algorithms classified the young and older adult with an accuracy of about 98% and the fallers and the older adults at 88%, which was better than the results achieved using traditional laboratory assessments (~72%). Conclusions: Features extracted from the multiple transitions recorded during daily living apparently reflect changes associated with aging and fall risk. Long-term monitoring of temporal features and their distribution may be helpful to provide a more complete and accurate assessment of the effects of aging and fall risk on daily function and mobility.