DRIVE is a prospective cohort study of a large (≥20 000) nonrepresentative sample of newly licensed young drivers in New South Wales, Australia.1 Its objective is to identify risk and protective factors for motor vehicle crashes (MVCs) (or crashes) by observing subgroup differences over time. A previously observed protective factor was participation in the New South Wales Health Reduce Risk Increase Student Knowledge (RRISK) resilience program.2 RRISK brings together grade 11 students from diverse schools for a 1-day, whole-of-community seminar that is preceded and followed by multiple peer-led teacher and parent support activities. It addresses common youth risks, including alcohol and drug use, partying, and driver and passenger risks. The focus is on empowering students to personalize learning into awareness of their own and their friends’ risk propensities, contributing factors, and how to minimize these, including forward planning and back-up strategies. The current research aims were to explore whether crash differences by program participation persisted over 13 years and to explore any differences by crash type, including severity.Detailed DRIVE methods are reported elsewhere.1,3 Briefly, all drivers aged 17 to 24 in New South Wales, Australia, on their first independent (postlearner) license during 2003–2004 were invited to complete a detailed online questionnaire, which included RRISK participation, and gave permission to access their police-recorded MVC records in future years. Survey and crash data to 2016 were linked by an independent body and were provided deidentified (ethics approvals: Aboriginal Health & Medical Research Council, University of New South Wales, and New South Wales Population & Health Services; reference: HREC/16/CIPHS/9).Univariate and multivariate negative binomial regression models were analyzed, with the number of MVCs of legally permitted vehicles as the primary outcome and MVC police codes for wet road surface, dark lighting, and severity (no injury versus minor injury to fatal) as secondary outcomes (other MVC types were too infrequent to include). We imputed missing values in the survey data using chained equations with 30 imputation cycles. Table 1 includes baseline survey variables considered as potential confounders, with time in study (from survey completion to either December 31, 2016, or [all cause] date of death) included as an offset. Analyses were performed by using Stata 15 (Stata Corp, College Station, TX).The cohort comprised 20 806 participants; 54.1% of the cohort was female (Table 1). Other demographics differed by group, with more program participants aged 17 and living in a regional and disadvantaged area, reflecting program delivery locations. Overall, 20.4% of the cohort had at least 1 crash as a driver, including 4.7% crashes on wet roads, 6.1% crashes in dark lighting, and 9.5% crashes resulting in injury.Table 2 summarizes resulting rate ratios. Unadjusted models revealed lower risk for program participants than nonparticipants for all crashes, crashes in dark lighting, and injury crashes but not crashes on wet roads. When adjusted for confounders, injury crashes were no longer statistically different; however, program participants were 24% less likely to have had any crash and 42% less likely to crash in darkness than nonparticipants.In this prospective study, rich baseline information on known confounders and access to statewide records were available for a large cohort of drivers for up to 13 years. Even when known confounders were accounted for, a lower risk of crash persisted for resilience program participants, including crashes in darkness. Injury crashes indicated similar positive findings but were not statistically different after adjustment for confounders.Limitations of the observational study design, including potential selection bias (more by schools than individuals) and complex changes in baseline measures over time, such as crash risk exposures and lifestyle changes (including cessation of driving), albeit unlikely influenced by program participation, have been reported previously.2,3 At-fault status was unknown; however, driver errors contribute to the majority of young novice crashes.4 Regardless, positive outcomes persisted, strongly indicating that setting up youth to be safer, responsible drivers early has real potential to reduce their lifetime risk of a crash that is significant enough to warrant police attendance. Research increasingly identifies that road risks cluster with other youth risks and that addressing these collectively and early (including at predriving age) is likely to have better outcomes than addressing them in isolation.5–7High-quality evaluations of school-based road safety initiatives are limited or treat any driver education program as equal.8,9 This contributes to overriding beliefs that driver education does not work rather than that good programs need to be identified from the proliferation of traditional programs. Positive outcomes are evident for more nuanced initiatives.10With road crashes persisting as a leading cause of preventable deaths and serious injuries, the current research warrants increased efforts to improve the evidence base and thereby the quality of youth road safety initiatives to set them up for a lifetime of safer driving.