Comparison of Sleep Regularity Index scores calculated by open-source packages and implications for outcomes research: rationale and design of the RIRI statement (Reporting Items for Regularity Indices)

The day-to-day regularity of sleep-wake timing refers to time-varying patterns of behavioral cycles, which co-occur with temporally associated environmental exposures and circadian rhythms. Introduced in 2017, the Sleep Regularity Index (SRI) has enabled rigorous study of the health and performance implications of the day-to-day regularity of sleep-wake timing. Since its introduction, multiple calculators have been published to facilitate SRI scoring from timestamped sleep-wake data; however, the comparability of these calculators had not previously been evaluated. Here, we sought to (1) estimate SRI usage and method of calculation in peer-reviewed studies published since its establishment; (2) compare SRI scores calculated by two widely used SRI calculators, sleepreg and GGIR; (3) compare results from prospective assessments of the relationship between sleepreg SRI scores versus GGIR SRI scores and previously examined health outcomes. We found that amidst increasing use of the SRI, non-disclosure, and heterogeneity in the method of SRI calculation are common. Additionally, among more than 70 000 adults with accelerometer-derived sleep-wake data, SRI scores calculated by two widely used open-source packages differed markedly, both in absolute and relative values. Applied to prospective clinical outcome models for all-cause mortality, incident type 2 diabetes, and incident atrial fibrillation or atrial flutter, the method of calculation alone meaningfully changed results and interpretations. In light of these findings, we developed and introduced a 14-item RIRI statement (Reporting Items for Regularity Indices) to standardize reporting and promote reproducibility in research involving the SRI or complementary regularity indices. Statement of Significance Health and performance implications of sleep variability have been unveiled in part due to improvements in operationalizing the day-to-day regularity of sleep-wake cycles using the Sleep Regularity Index (SRI). However, in a critical review of publications that have used the SRI, we identified marked non-disclosure and heterogeneity in the method of SRI calculation. Moreover, using the same accelerometer-based sleep-wake estimates among 73 794 adults, we found that SRI scores calculated by different calculators differ considerably and consequentially, including altered findings in prospective time-to-event models. We therefore developed and propose a 14-item RIRI statement (Reporting Items for Regularity Indices) to standardize reporting and promote reproducibility of research involving the SRI, with generalizability to all research that compares temporally matched data.