Digital Phenotyping of Rest-Activity Rhythms and Biological Aging from Longitudinal Monitoring with Commercial Wearable Devices in All of Us

Poor circadian health is increasingly recognized as a determinant of aging and chronic diseases, yet longitudinal evidence in free-living populations remains limited. Most prior studies have been restricted to cross-sectional designs or short 7-day monitoring, precluding insight into long-term aging dynamics. To address this gap, we analyzed multi-year consumer wearable data linked with electronic health records from the All of Us Research Program to evaluate circadian rest-activity rhythms as longitudinal predictors of biological aging. Among 2,222 participants (median age 60.6 years, 68.5% female) contributing 8,447 person-years of Fitbit activity data with annual biological age estimates (PhenoAge), we performed high-dimensional digital phenotyping integrating functional data analysis with conventional rhythm metrics. Higher rhythm intensity reduced the odds of accelerated aging by 26-46%, greater regularity lowered the odds by 9-13%, whereas delayed acrophase increased the odds by 22%. Sex-stratified analyses revealed universal protection from rhythm intensity in both sexes, but stronger timing- and regularity-related vulnerabilities to accelerated aging in females (12-18% higher odds). In contrast, males exhibited a biphasic instability phenotype, characterized by early-morning surges and late-evening rebounds, uniquely linked to accelerated aging. This study provides the first large-scale longitudinal evidence establishing circadian rest-activity rhythms derived from consumer wearables as digital biomarkers of aging trajectories. With the growing scalability and ubiquity of consumer devices, our findings pave the way toward scalable aging risk assessment, targeted interventions, and advancing digital precision medicine to promote healthy longevity at the population level.