Individualized Driver Risk Stratification and Targeted Prevention of Road Traffic Crashes: Development, Validation, and a Controlled Intervention Study
Litsenziya: cc-by-nc-nd
Annotatsiya
Background Driver-risk stratification can support targeted prevention, but its practical value requires transparent scoring, discrimination assessment and appropriately qualified outcome evaluation. Methods A 38-item scored questionnaire (range 0–114), a 20-item knowledge test and a composite index were applied within a three-stage driver-safety program: baseline screening (n=1,258), 365-day validation (n=120; eight crashes), and a nonrandomized 24-month training comparison (n=832; 21 crashes). Crash proportions were compared using Fisher exact tests. Dated paired scores and an exploratory covariate-adjusted logistic model were analyzed. Results The raw questionnaire score discriminated validation crashes with AUC 0.7963 (bootstrap 95% CI 0.5290–0.9558); the composite I_RISK yielded AUC 0.6217 (95% CI 0.4817–0.7560). At raw score >60, sensitivity was 87.5% and specificity 58.0%. In high-risk groups, crashes occurred in 2/210 trained versus 12/204 comparison drivers (p=0.00586); in moderate-risk groups, 0/212 versus 7/206 (p=0.00670). Dated month-24 paired scores were available for 794/832 drivers. The exploratory adjusted training association was OR 0.0899 (95% CI 0.0207–0.3911; p=0.00132). Conclusions The study illustrates a reproducible risk-to-training pathway with favorable observational outcomes. Discrimination was stronger for the raw questionnaire than for the composite score. Nonrandom allocation, few crash events and source-linkage limitations preclude causal inference and warrant external prospective validation.
Hali tarjima qilinmagan