Association of Delaying School Start Time With Sleep Duration, Timing, and Quality Among Adolescents (START study)

455 students across 5 Minneapolis-St Paul public high schools, followed grade 9 (2016) to grade 11 (2018)

US public high school students, mean age 15.2 at baseline; 2 schools delayed start by 50 and 65 min, 3 comparison schools held at 7:30 am

Difference-in-differences on district-initiated bell-time changes with wrist actigraphy as the outcome — the treatment is assigned by the district to whole schools, not chosen by students or families, and the outcome is objectively measured rather than self-reported. Two annual follow-ups establish persistence. Residual threats: only 5 schools (clustering is coarse), districts chose to change rather than being randomized, and there is scope for differential attrition and family relocation over 2 years.

Delayed-start students gained a mean 41 min (95% CI 25-57) of objectively measured school-night sleep at follow-up 1 and 43 min (95% CI 25-61) at follow-up 2, relative to the change in comparison schools. Bedtimes did not drift later — sleep onset differences were minimal — so the extra time is a net gain, not a reallocation. Weekend catch-up sleep fell by 24 min (95% CI -51 to 2) and 34 min (95% CI -65 to -3), consistent with reduced sleep debt rather than a shifted schedule. Sleep efficiency and fragmentation were essentially unchanged.

Very low for the sleep-duration estimate. Bell time is imposed on an entire school population regardless of genotype, chronotype, or family background, and the comparison schools absorb secular trends. Chronotype and sleep-duration heritability cannot generate this contrast because they do not differ systematically between treated and comparison schools at baseline. Residual risk lives in the district's non-random decision to change and in 2-year attrition.