O031 How many nights? Examining the reliability of sleep intraindividual variability

Abstract Introduction Individual sleep–wake patterns are characterized by the intraindividual mean and variability, which captures night-to-night fluctuations. Although research has focused on mean sleep parameters, higher sleep duration and timing variability are associated with negative physical and mental health outcomes. Most accelerometry-based sleep studies assess 5-14 nights, which may be insufficient for reliable variability estimates. This study quantifies how many nights are needed for reliable sleep intraindividual mean and variability estimates. Methods Participants (n = 10 412, M-age = 37.75 ± 10.65, 50% women) wore a validated biometric device for 1-year (3 700 492 person-nights). Total sleep time (TST), sleep onset, mid-sleep, sleep offset, wake after sleep onset (WASO), and sleep percentage were derived from heart rate and accelerometer data. Reference intraindividual means (iM) and standard deviations (iSD) were calculated from all (363) nights. Reliability correlation coefficients compared reference values with 2-362 consecutive nights randomly selected per participant, with r ≥ .80 considered reliable. Results Reliable estimates of sleep iM required 2-8 nights (TST = 8 nights; sleep timing = 2 nights; WASO and sleep percentage = 4 nights). Reliable iSD estimates required substantially longer durations, 41-67 nights (TST = 43 nights; sleep timing = 41 nights; WASO and sleep percentage = 67 nights). iSDs from 7- or 14-nights showed poor reliability (r7-Nights = .48–.57; r14-Nights = .59–.68). Discussion Eight nights of accelerometry data are sufficient to estimate intraindividual mean sleep, but 41-67 nights are needed to reliably estimate intraindividual sleep variability. Findings highlight the importance of tailoring study periods to the sleep metric of interest. Longer monitoring periods may improve diagnostic accuracy and inform recommendations for sleep assessments in clinical and research settings.