1255 Longitudinal Modeling of Sleep Timing, Duration, and Social Jetlag Across the Adult Lifespan Using High-Resolution Wearable Data

Abstract Introduction Reliable longitudinal sleep measurements are critical for chrono-epidemiology, as misalignment and variability are strongly associated with chronic diseases. Self-reporting methods or short-term monitoring can fail to capture longer term patterns required for risk stratification. Therefore, we used device-measured sleep data from the All of Us (AoU) research program to characterize age, sex, and environmental contributors to sleep patterns. Methods We analyzed 22,428,610 nights of filtered Fitbit sleep data from 42,290 AoU participants. We calculated nightly onset, offset, and duration, excluding non-primary sleep and episodes >18h. Linear Mixed-Effects Models (LMMs) with person as random effects were used to control for employment, weekend status, sex, seasonality, and age incorporating quadratic age terms and employment weekend interactions. Results The cohort (71% White, 67% Female, 52 ± 17y) contributed a median of 109 nights (mean: 530) of recorded sleep (mean duration 7.45h, median 7.53h). All reported associations were significant (p< 0.001). Age showed a quadratic trend in sleep timing across the lifespan. Sleep midpoint showed a maximum advance of ~42 minutes between young adulthood (age 18, 2:37 AM) and older age (age 57, 1:54 AM). Sleep duration increased with age gaining over 34 minutes between ages 18 (7.45h) and 85 (8.02h). Furthermore, the employment-weekend interaction showed a delay in midpoint, onset, and offset and increased duration across all groups. Participants "Employed for Wages" showed the greatest social jetlag (average midpoint delay) of ~42 minutes on weekends (4:08 AM) compared to weekdays (3:26 AM). This shift was accompanied by a duration extension of ~28 minutes on weekends (7.95h) vs. weekdays (7.48h). Finally, seasonality analysis revealed an 11-minute decrease in sleep duration in June compared to the maximum observed in January. Conclusion This large-scale analysis confirms that sleep timing and duration are dynamically determined by non-linear age effects, strong social constraints, temporal effects, and measurable environmental factors. Our findings provide population-level characterization essential for future genetic and clinical studies targeting personalized interventions for circadian misalignment. Support (if any) All of Us Research Program, Office of the Director, NIH