0751 Comparability of Self-Reported & Wearable-Measured Sleep & Predictors of Agreement in a Preconception Cohort

Abstract Introduction Agreement between self-reported and device-estimated sleep duration has not been well studied among pregnancy planners. As data from wearables provide complementary information to self-reported sleep, assessment via wearable devices may improve characterization of sleep in naturalistic settings and better inform research on sleep health and reproductive outcomes. Methods We analyzed data from Pregnancy Study Online, an Internet-based prospective preconception cohort study. From 2021-2023, we invited U.S. residents aged 21-45 years with pregnancy attempt times ≤6 months at enrollment to participate in a study of wearable sleep-tracking devices. Enrolled participants wore a study-issued Fitbit Inspire 2/3 for 24 hours/day for up to two months and completed all other parent study activities. We examined agreement between self-reported sleep duration in the month before enrollment and device-estimated 30-day average sleep duration (hours/day) within two months after enrollment using Bland-Altman plots and Kappa statistics (one-hour intervals). We used log-binomial regression models to estimate prevalence ratios (PR) and 95% confidence intervals (CI) for predictors of agreement. Results Of the 1,523 invited participants (1,178 females and 345 males), 948 (62%) consented to participate (females: 67% vs. males: 47%). Self-reported sleep duration (median: 7.5; interquartile range [IQR]: 7.0-8.0 hours/day) was longer than wearable-measured (median: 7.2, IQR: 6.3-8.0 hours/day). Self-reported and wearable-measured sleep duration showed poor agreement (weighted Kappa: 0.12; 95% CI: 0.07-0.17). Comparing self-reported to wearable-measured sleep duration, the Bland-Altman plot yielded a mean difference of 0.2 hour/day (95% CI: -3.3-3.8). Outliers from the Bland-Altman plot indicate that participants with shorter wearable-estimated mean sleep duration overestimated their self-reported sleep duration, while participants with longer wearable-estimated mean sleep duration underestimated this measure. Among females, predictors of poorer agreement in short sleep duration (< 7 hours/day) included older age (≥35 years, PR: 0.71; 95% CI: 0.50-1.00), rural residence (PR: 0.64; 95% CI: 0.35-1.16), menstrual irregularity (PR: 0.80, 95% CI: 0.54-1.18), and shift work (PR: 0.83; 95% CI: 0.56-1.23). There were no strong predictors in agreement among males. Conclusion Self-reported and wearable-measured sleep duration showed poor agreement in a preconception cohort. Age, rurality, menstrual irregularity, and shift work predicted poor agreement in short sleep duration among females. Support (if any) R01-HD086742, R01-HD105863