“How we roll:” The effects of human, vehicle, and environmental contributors to the occurrence and severity of rollover crashes

Introduction: Although rollovers account for 3% of vehicle crashes, they result in approximately one-third of all occupant deaths. This study investigates the effects of human, vehicle, and environmental factors on the occurrence and severity of rollovers in single-vehicle crashes. We emphasize the importance of a safe system approach that incorporates safer speeds, safer vehicles, and equitable design to prevent and mitigate rollover crashes while addressing the diverse needs and challenges of all road users. Methods: Using data from NHTSA’s Crash Report Sampling System and the New Car Assessment Program’s Safety Ratings, we applied logit regression and XGBoost models to identify the significant predictors of rollover likelihood and injury severity. We estimated the Levels of Automation in vehicles and their impact on the occurrence and severity of rollovers. We also used SHAP analysis to interpret the XGBoost model predictions. Results: Our findings reveal that younger drivers, impairments, and device-related distractions on high-speed limit roadways increase rollover risks, while higher vehicle automation levels and seatbelt usage reduce them. We also find that environmental factors, such as road alignments and surface conditions, have complex impacts on rollover occurrence and severity. Notably, NHTSA’s rollover possibility value, with a significant positive coefficient of 1.54, indicates that rollovers are more likely to occur as the values rise. Practical Applications: We highlight the potential of emerging vehicle technologies to reduce rollover vulnerability. Additionally, we emphasize the need for inclusive road safety measures that cater to the needs of all road users. These insights provide valuable guidance for future transportation safety strategies.