School-based driver education for the prevention of traffic crashes

3 randomised controlled trials, n = 17,965 (Stock 1983 DeKalb County n = 16,338; Wynne-Jones 1984 New Zealand n = 848; Strang 1982 Australia n = 779)

Young people aged 15-24 who had not yet obtained a driving licence, in the USA, New Zealand and Australia; trials conducted 1982-1984. Review content assessed as up-to-date 17 May 2006.

A Cochrane review restricted to randomised trials of school-based driver education versus no driver education — an unusually clean evidence base for a school subject, and the only place in this archive where a life-skills course has been tested at this scale under randomisation. The DeKalb County trial (Stock 1983) alone randomised 16,338 high-school students by central computer with well-concealed allocation to one of two driver-education programmes or a control group given no formal instruction, and followed official crash and violation records for two to four years. The review's decisive design choice is stated explicitly and is the right one: driving SKILL was excluded as an outcome, because treating improved skill as a surrogate for reduced crashes assumes the very relationship under test. The limitation is age — all three trials are from 1982-84, and contemporary programmes (the review discusses the UK DSA schools programme) are far less intensive, typically a 50-minute presentation with no behind-the-wheel component, which the reviewers argue is the worst combination: it could raise licensing without touching skill.

Driver education produces earlier licensing and no reduction in crashes. In DeKalb, 87% (9,510/10,894) of the driver-education group had been licensed since course completion versus 84.3% (4,588/5,444) of controls (RR 1.04, 95% CI 1.02-1.05); crash involvement was 27.5% versus 26.7% (RR 1.03, 95% CI 0.98-1.09). In New Zealand the median time from enrolment to licensure was 111 days with driver education versus 300 days without for males (t = 7.19, p < 0.001) and 105 versus 415 days for females (t = 9.88, p < 0.001). Crash relative risks were 1.01 (CI 0.83-1.23), 1.03 (0.98-1.09) and 1.10 (0.76-1.59) across the three trials — every point estimate above one. The reviewers' conclusion is that driver education leads to early licensing, provides no evidence of reduced crash involvement, and may produce a modest but potentially important INCREASE in the proportion of teenagers involved in traffic crashes. Their worked example: if a schools programme raised the proportion of licensed teenagers by just 2%, an additional 27 British teenagers might be killed or seriously injured each year as a result.

None material. Individual randomisation within schools in all three trials, with central computer allocation and good concealment in the largest. Outcomes are official police and licensing records rather than self-report in the two largest trials.

Three independent randomised trials on three continents agree on both findings — earlier licensure, no crash reduction. Vernick et al. (1999) reach the same conclusion from a separate systematic review of nine studies using a different inclusion rule, so the result replicates across review teams as well as across trials.