Berry · The Journal of adolescent health : official publication of the Society for Adolescent Medicine 2021 · prospective controlled cohort study · n=2414

Association of Delaying School Start Time With Sleep-Wake Behaviors Among Adolescents.

Cited 11 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective non-randomized controlled cohort study comparing policy-change schools to concurrent control schools.

PubMed 34103238 · doi:10.1016/j.jadohealth.2021.04.030 · record verified 2026-08-27

What was done

The START study followed a prospective cohort of 2,414 ninth-grade students across five Minnesota high schools through 11th grade. At baseline, all schools started at 7:30 or 7:45 a.m. By the 10th- and 11th-grade follow-ups, two intervention schools delayed start times by 50 and 65 minutes, while three comparison schools maintained a 7:30 a.m. start. Six self-reported sleep-wake behaviors were evaluated across all three time points using generalized estimating equation models.

What was found

At baseline, sleep-wake problem prevalence ranged from 11% for oversleeping-related tardiness to 48% for requiring multiple morning wake-up prompts. From 9th to 11th grade, students at policy-change schools experienced smaller increases in the prevalence of daily sleepiness and oversleeping-related school lateness relative to comparison schools. Conversely, awakening too early became more common among students in the delayed-start schools compared to controls. The abstract does not report exact effect sizes, odds ratios, or confidence intervals.

Why it matters

This study provides longitudinal, controlled evidence that delaying high school start schedules mitigates the worsening of adolescent daytime sleepiness and morning tardiness as students age through high school.

Limits

Outcomes relied entirely on self-reported surveys rather than objective sleep metrics like actigraphy. The intervention was non-randomized across only five schools in a single state, and exact quantitative effect sizes or variance measures were omitted from the abstract.

Cited by