JMIR Publications

ABSTRACT

Background:

Social frailty poses a potential risk even for relatively healthy older adults, necessitating development of early detection and prevention strategies. Recently, consumer-grade wearable devices have gained attention for their ability to provide accurate sensor data, and digital biomarkers for social frailty screening could be calculated from these data.

Objective:

The objective of this study was to explore digital biomarkers associated with social frailty using sensor data recorded by Fitbit devices and to evaluate their relationship with health outcomes in older adults.

Methods:

This cross-sectional study was conducted in 102 community-dwelling older adults. Participants attending frailty prevention programs wore devices of the Fitbit Inspire series on their non-dominant wrist for at least seven consecutive days, during which step count and heart rate data were collected. Standardized questionnaires were used to assess the physical functions, cognitive functions, and social frailty, and based on the scores, the participants were categorized into three groups: robust, social pre-frailty, and social frailty. The sensor data were analyzed to calculate nonparametric and extended cosinor rhythm metrics, along with heart rate-related metrics.

Results:

The final sample included 86 participants who were categorized as robust (n = 28), social pre-frailty (n = 39), and social frailty (n = 19). The mean age of the participants was 77.14 years (SD 5.70), and 90.6% were women (n = 78). Multinomial logistic regression analysis revealed that the step-based rhythm metric, Intradaily Coefficient of Variation (ICV.st), was significantly associated with social pre-frailty. The heart rate metrics, including the delta resting heart rate (dRHR) and UpMesor.hr, showed significant associations with both social frailty and social pre-frailty. Furthermore, the standard deviation of the heart rate (HR.sd) and alpha.hr were significant predictors of social pre-frailty. Specifically, dRHR, defined as the difference between the overall average heart rate and resting heart rate (RHR), exhibited significant negative associations with social pre-frailty (odds ratio [OR] = 0.82, 95% confidence interval [CI] 0.68-0.97, p = 0.024) and social frailty (OR = 0.74, 95% CI 0.58-0.94, p = 0.015). Furthermore, analysis using a linear regression model revealed a significant association of the ICV.st with the Word List Memory (WM) score, a measure of cognitive decline (β = -0.04, p = 0.024).

Conclusions:

This study identified associations of novel rhythm and heart rate metrics calculated from the step count and heart rate recorded by Fitbit devices with social frailty. These findings suggest that consumer-grade wearable devices, which are low-cost and accessible, hold promise as tools for evaluating social frailty and its risk factors through enabling calculation of digital biomarkers. Future research should include larger sample sizes and focus on the clinical applications of these findings. Clinical Trial: UMIN-CTR