Abstract 4369416: Impact of Long COVID on Physical Activity and Sleep: a Matched Cohort Study using Wearable and EHR data From The All of Us Research Program

Backround: Long COVID (LC) is associated with negative health effects, but its impact on health behaviors such as activity and sleep remains unclear. Using data from the All of Us Research Program, we examined whether individuals with LC experience long-term changes in these behaviors. We hypothesized that LC would be associated with decreased activity and disrupted sleep patterns. Methods: We identified participants in AOU with a documented positive COVID-19 test result in their EHR, linked Fitbit data, and ≥180 days of valid step or sleep data before and after diagnosis (N=800). We then applied a validated machine learning algorithm using 200 clinical predictors to calculate a Long COVID probability score (0-1). Participants were stratified into high (HLCP, >0.5) and low (LLCP, ≤0.5) LC probability groups. Separate control cohorts with no record of COVID-19 were matched 5:1 to the HLCP (1620 controls) and LLCP (2535 controls) groups by age (±5 years), sex, BMI, and baseline activity (average daily steps six months prior to diagnosis/match date). Sleep irregularity was defined as the standard deviation of sleep duration. The effect of LC was measured by comparing average pre-COVID baseline step count and sleep SD values with the average values 330–360 days post-diagnosis. Results: The HLCP group (N=324; 72% female; median age 60 [IQR 47;70]; BMI 30 [25;36]) and LLCP group (N=507; 74% female; age 57 [44;68]; BMI 28 [25;33]) were demographically comparable to their respective matched controls. At baseline, median step counts were 6,446[4566;8644] for HLCP and 7,919[5775;10,657] for LLCP. Both groups showed changes in steps in the days preceding and following date of diagnosis, reaching a median delta of -1,994[-3,351;-972] steps for HLCP and -1,971[-3,582;-816] steps for LLCP (Fig. 1). At 330–360 days post-diagnosis, LLCP participants had a median step count change of -144[-1,601;+974] from baseline. HLCP participants showed a larger sustained reduction, with a median change of -345[-1,801;+812] (Fig. 2). After 52 weeks, sleep irregularity increased in HLCP (p=0.02) but not in LLCP (p=0.11) compared to controls (Fig. 3). Conclusion: In this cohort, individuals with a high probability of developing LC exhibited prolonged reductions in physical activity and persistent sleep disruptions for at least one year following diagnosis. These findings suggest that LC may contribute to sustained changes in health behaviors with potential implications for long-term health.