Clustering drivers by demographic and aberrant driving behavior: implications for traffic injury risk and safety interventions

• Three distinct driver clusters were identified with different road traffic injury severities. • The high-risk cluster included young males with poor driving behavior and limited experience. • The low-risk cluster showed fewer violations, shorter drives, and higher education. • Cluster membership was significantly associated with injury severity levels. • Findings support behavior-based stratification in traffic injury risk prediction. Road traffic injuries remain a leading cause of death and disability globally. This study aimed to identify distinct clusters of drivers based on demographic characteristics and multidimensional aberrant driving behaviors, and to examine their association with the severity of traffic-related injuries. A secondary analysis was conducted on a cross-sectional dataset of 800 licensed drivers from Iran. Data included demographics, crash history, traffic fines, and responses to the Driver Behavior Questionnaire. Principal Component Analysis and Hierarchical Clustering on Principal Components were used to derive latent driver clusters. Cluster profiles were compared using statistical tests. An Extreme Gradient Boosting classifier was developed to assess the discrimination of the identified clusters. Three driver clusters were identified. Cluster 3, comprising mostly young, less-educated males with limited driving experience, showed significantly higher mean scores in all aberrant behavior dimensions and the highest proportion of severe injuries (40.7%; p < 0.001). Cluster 2, characterized by younger, more educated females with lower driving exposure and fewer behavioral violations, had the lowest severe injury rate (15.6%). Cluster 1 demonstrated intermediate characteristics and injury risk, positioned between Clusters 2 and 3 in terms of demographic, behavioral, and outcome variables. Classification performance was high (accuracy = 92.6%, AUC = 0.987), supporting the clustering scheme. Multidimensional clustering of demographic and behavioral data can effectively identify high-risk driver subgroups. These findings highlighted the value of behaviorally informed risk stratification in traffic safety frameworks, providing a foundation for precision-targeted interventions and injury prevention efforts.