Assessing day-to-day regularity of sleep-wake patterns: theoretical and practical implications of available metrics

Kurzfassung

Background. Day-to-day changes in sleep-wake patterns are important to quantify because they can result in circadian disruption, a risk factor for health outcomes. Traditionally, sleep regularity has been assessed by comparing each day to the average sleep-wake pattern, using metrics such as standard deviation (StDev) and Interdaily Stability (IS). Recently, metrics have been proposed to instead capture variability between consecutive days: the Sleep Regularity Index (SRI) and the Composite Phase Deviation (CPD). Here, we systematically compared these metrics across a range of sources of day-to-day variability, including naps, awakenings, and missing data. Methods. Sleep-wake patterns were synthetically generated over 2-28 days with a weekdayweekend structure. Daily sleep variability was introduced by randomly drawing daily midsleeps and/or sleep durations from a normal distribution with standard deviation ranging from 0- 120min. Average estimates and 95% confidence intervals (CIs) were calculated for each metric under the following scenarios: (1) ‘scrambling’ the order of days, (2) fragmented sleep (i.e. naps, wake after sleep onset (WASO), and all-nighters), (3) varying number of days, and (4) randomly vs. non-randomly (i.e. very early/late sleep more likely to be missing) missing data. Results. (1) Scrambling did not affect IS and StDev values but did affect SRI and CPD values, showing that the metrics measure sleep regularity on different time scales: global vs. circadian. (2) SRI and IS behaved similarly for naps and WASO but differed for all-nighters: SRI values increased (more regular) when all-nighters exceeded 50% of nights, whereas IS yielded monotonically lower (less regular) scores. (3) When based on £ 7 days, StDev and IS overestimated how regular patterns were by up to 40% whereas SRI and CPD were more stable, yet with wider CIs requiring up to 40% larger samples. (4) All metrics were highly sensitive to non-randomly missing data but remarkably stable for up to 50% randomly missing data. Conclusions. All examined metrics have been used for quantifying sleep regularity, yet they measure different aspects and should be seen as complementary rather than redundant. Studies should consider including more than one metric and examining mechanistic links between circadian disruption and sleep regularity.