Longitudinal Digital Phenotyping of Circadian Rest-Activity Rhythms via Wearables as Biomarkers for Late-Life Function, Cognition, and Neuropsychiatric Health

Background: Circadian rest–activity rhythmicity, a manifestation of circadian rhythms, characterizes 24–hour activity patterns. Growing evidence links disruption of circadian rhythms in late life to adverse outcomes, including functional and cognitive declines. Yet, most studies have been cross–sectional or restricted to a single monitoring period, lacking longitudinal assessment. Consequently, it remains unknown whether deterioration or improvement in circadian rest–activity rhythmicity reflects late–life vulnerability in aging populations. Methods: We analyzed data from the National Health and Aging Trends Study (NHATS), a nationally representative U.S. cohort of adults aged ≥70 years. In Cycles 11 and 12, participants completed 7–day wearable monitoring and assessments of functional, cognitive, and neuropsychiatric outcomes. Circadian biomarkers were derived using cosinor and non–parametric methods (amplitude, MESOR, acrophase, pseudo–F, RA, IS, IV, M10, L5). Participants were classified into four transition groups (Optimal, Improved, Deteriorated, Adverse). Survey–weighted regression and Cox models adjusted for demographics and comorbidities estimated associations, with multiple testing correction. Results: At baseline, weaker rhythm intensity (low amplitude, MESOR, M10, RA) and poor stability (high fragmentation, low regularity) were associated with greater ADL disability, lower SPPB, weaker grip strength, and poorer recall. Longitudinal analyses revealed a graded hierarchy of risk. Participants with deteriorating rhythms represented the most dynamic risk state: declining amplitude was linked to reduced SPPB (β= −0.71, 95% CI: −0.98 to −0.44) and diminished immediate recall (β= −0.60, 95% CI: −0.91 to −0.29). Declining regularity was also associated with impaired function (SPPB: β= −0.52, 95% CI: −0.82 to −0.21) and cognition (immediate recall: β= −0.43, 95% CI: −0.68 to −0.18, delayed recall: β= −0.45, 95% CI: −0.70 to −0.20). By contrast, rhythmicity improvement aligned with stabilization. For neuropsychiatric outcomes, high fragmentation was linked to probable dementia, while poor regularity was tied to anxiety/depression. Conclusion: Circadian rest–activity rhythmicity is a robust and dynamic determinant of late–life outcomes. Persistent weakness or deterioration in intensity and regularity was linked to accelerated decline in function and cognition, whereas improvement aligned with stabilization. These findings position circadian rhythmicity as both an early biomarker of vulnerability and a potential modifiable target. Digital phenotyping via wearables offers a scalable, noninvasive framework for early risk detection, personalized intervention, and resilience promotion in aging populations.