Supplementary Material for: Monitoring gait and physical activity of elderly frail individuals in free-living environment: a feasibility study

Introduction: Frailty is conventionally diagnosed using clinical tests and self-reported assessments. However, digital health technologies (DHTs), such as wearable accelerometers, can capture physical activity and gait during daily life, enabling more objective assessments. In this study, we assess the feasibility of deploying DHTs in community-dwelling older individuals, and investigate the relationship between digital measurements of physical activity and gait in naturalistic environments and participants’ frailty status, as measured by conventional assessments. Methods: Fried Frailty Score (FFS) was used to classify fifty healthy individuals as non-frail (FFS=0, n/female=21/11, mean±sd age:71.10±3.59 years), pre-frail (FFS=1-2, n/female=23/9, age:73.74±5.52 years), or frail (FFS=3+, n/female=6/6, age:70.70±6.53 years). Participants wore wrist-worn and lumbar-worn GENEActiv accelerometers (Activinsights Ltd., Kimbolton, UK) during three in-lab visits, and at-home for 2 weeks, to measure physical activity and gait. After this period, they completed a comfort and usability questionnaire. Compliant days at-home were defined as follows: those with ≥18 hours of wear time, for the wrist-worn accelerometer, and those with ≥1 detected walking bout, for the lumbar-worn accelerometer. For each at-home measurement, a group analysis was performed using a linear regression model followed by ANOVA, to investigate the effect of frailty on physical activity and gait. Correlation between at-home digital measurements and conventional in-lab assessments was also investigated. Results: Participants were highly compliant in wearing the accelerometers, as 94% indicated willingness to wear the wrist device, and 66% the lumbar device, for at least 1 week. Time spent in sedentary activity and time spent in moderate activity as measured from the wrist device, as well as average gait speed and its 95th percentile from the lumbar device were significantly different between frailty groups. Moderate correlations between digital measurements and self-reported physical activity were found. Conclusions: This work highlights the feasibility of deploying DHTs in studies involving older individuals. The potential of digital measurements in distinguishing frailty phenotypes, while unobtrusively collecting unbiased data, thus minimizing participants’ travels to sites, will be further assessed in a follow-up study.