Challenges in implementing and assessing outcomes of school start time change in the UK: experience of the Oxford Teensleep study

2 schools recruited against a target of 100 state secondary schools; the trial never delivered outcome data

Intended population was UK Year 10/11 students (ages 14-16) in state secondary schools

This is a report of a failed recruitment, not a results paper, and is graded D accordingly. It is recorded because it answers the most important structural question about this topic: why does no randomized trial of school start times exist? Teensleep was designed as a properly randomized trial of a 10:00 a.m. start and/or a sleep-education programme. Despite extensive media coverage, only two of a target 100 schools volunteered.

The delayed start time itself, and the prospect of being randomized to it, were the overwhelming reasons schools gave for refusing to participate. The authors conclude that evaluating a change in school start times through an RCT is unfeasible in the UK. This is the direct explanation for why the evidence base consists of district-level natural experiments and quasi-experiments rather than trials, and why the Cochrane review rated the body of evidence very low: schools will not cede scheduling control. It also implies that any school which DID volunteer would be a self-selected outlier, so even a successful trial would carry a selection problem of its own.

Not applicable — no outcome data were collected. Relevant indirectly: because randomization is unobtainable, the field is structurally dependent on designs where allocation is made by districts, which is why the district-wide-imposition logic carries so much of the causal weight here.