Uncovering contextual risk patterns in cannabis-involved fatal crashes: A data-driven approach to public...

As cannabis legalization expands across the United States, concerns about its impact on road safety and public health continue to grow. This study examined fatal crashes involving cannabis-involved drivers using national data from the Fatality Analysis Reporting System (FARS) between 2018 and 2022, focusing on cases where cannabis was toxicologically confirmed in the driver's bloodstream.

To uncover underlying crash typologies, we applied Cluster Correspondence Analysis (CCA), a two-way dimension reduction method optimized for categorical data, to reveal patterns across roadway environments, driver demographics, crash dynamics, and environmental conditions.

The analysis revealed six distinct crash clusters: rural straight-roadway single-vehicle collisions, high-speed multi-vehicle crashes with lane conflicts, single-vehicle crashes on curves with loss of control, turning and yielding errors at intersections, unusual user and road conditions with pedestrian involvement, and nighttime urban crashes involving vulnerable road users. These findings highlighted the intersection of tetrahydrocannabinol (THC)-positive toxicology and systemic infrastructure vulnerabilities that contribute to fatal outcomes in cannabis-involved crashes.

By using a method designed for complex categorical datasets, this research provided novel insights into the multifaceted risks associated with drug-impaired driving. The results could inform targeted countermeasures, such as improved roadway lighting, intersection design, and behavioral interventions, offering a data-driven foundation for public health-oriented traffic safety strategies.