Unsupervised Assessment of Frailty Status Using Wearable Sensors: A Feasibility Study among Community-Dwelling Older Adults.

Objectives: This study examined whether community-dwelling older adults can independently capture wearable sensor data that can be used to classify frailty status.

Methods: Fifty-one older adults (age 77.5 ± 8.4 years, height 163.6 77.5 ± 8.4, weight 72.0 ± 13.5 kg, female 76%) took part in this investigation. Participants independently captured physical activity and physical function data at home using a smartwatch and a research-grade inertial sensor system for 48-hours. Machine learning classifiers were used to determine whether the data obtained can discriminate between frailty levels.

Results: Models incorporating variables from both the smartwatch and inertial sensor system were successful in the prediction of frailty status.

Discussion: This study has demonstrated the ability of older adults to collect data which can be used to indicate their frailty risk. This may enable earlier intervention and lessen the impact of frailty on the individual and society as a whole.