0925 Accuracy of Text Message-Based Sleep Diary for Assessment of Sleep Timing and Duration in Youth

Abstract Introduction Accurate assessment of sleep timing/duration is essential for research and clinical practice. Sleep diaries are an inexpensive and convenient method for assessing sleep, but their accuracy in youth remains unclear. Our research group has used text-messages to administer electronic sleep diaries, which have shown high response rates and rapid response times. We hypothesize that these diaries yield accurate self-reported sleep timing and duration compared to sleep/wake estimation from concomitantly worn wrist actigraphy. Methods Youth 12-21y wore wrist actigraphy (ActiGraph GT3XP-BTLE) and completed twice-daily text-message sleep diaries for 2 weeks. Sleep timing and duration were derived using Cole-Kripke and Tudor-Locke algorithms. Mean sleep onset, offset, duration, and wake after sleep onset (WASO) were compared between actigraphy and sleep diary methods using Wilcoxon rank sum and Bland-Altman analyses. Demographic factors associated with improved diary-actigraphy agreement were assessed using linear regression. Results Participants (n=71,52.1%F) with median (IQR) age 19.1y(17.5,21.0) completed actigraphy and sleep diaries. Average sleep onset, offset, and duration were similar between the two assessment methods by Wilcoxon rank sum (p=0.95 (onset), p=0.15 (offset), p=0.56 (duration)). The absolute magnitude of difference [median(IQR)] was 24.3min (8.4,40.2) for sleep onset, 29.4min (12.5,46.1) for sleep offset, and 36.8min (17.1,68.6) for sleep duration. Per Bland-Altman analysis, the mean differences between diary and actigraphy (positive=diary overestimated) were 1.9min (onset), -5.8min (offset), and -5.3min (duration). 95% limits of agreement were -79.16 to 66.9 minutes (onset), -185.8 to 91.2 minutes (offset), and -94.1 to 58.4 minutes (duration). Proportional bias was 6.6min difference per delayed hour sleep offset, but none was present in onset and duration. There was no heteroscedasticity for any parameter. WASO was significantly different between diary data and actigraphy (p< 0.001). Female sex was associated with improved accuracy of reported sleep duration (p=0.028) but not sleep timing, and age and socioeconomic status were not associated with accuracy of any metric. Conclusion Text-message sleep diaries may provide accurate assessment of sleep onset, offset, and duration in youth when compared to actigraphy, but do not accurately capture WASO. Proportional bias in sleep offset suggests delayed awakening is associated with decreased accuracy in reported offset. Most demographic factors were not associated with reporting accuracy. Support (if any) NIH R01DK115648