Investigating Contributing Factors to Fatal Non-Intersection Crashes Involving Elderly Pedestrians Using Association Rule Mining

Fatal pedestrian crashes involving elderly individuals (aged 65+) are a growing public safety concern in the United States. Notably, approximately 80% of fatalities occurring away from intersections underscores the critical need to examine non-intersection crash dynamics. This study investigates contributing factors to such non-intersection fatalities involving elderly pedestrians using Association Rule Mining (ARM) applied to the Fatality Analysis Reporting System (FARS) data from 2019 to 2023. A two-stage analytical framework was employed: first, random forest-based feature selection identified the most influential variables; second, the Apriori algorithm uncovered frequent co-occurring crash patterns. The analysis yielded 26 high-lift rules (lift > 2.6) associated with elderly pedestrians being pronounced dead at the scene. A striking and novel finding was the consistent presence of solo drivers (DRIVER_ALONE = TRUE) in all high-risk scenarios, suggesting that unaccompanied drivers may face heightened risks of missing or failing to respond to elderly pedestrians, particularly in challenging environments. Other common antecedents included roadside pedestrian locations, poor lighting conditions (e.g., “Dark – Not Lighted”), and overnight hours (12 A.M.–6 A.M.). Alcohol involvement, lack of traffic controls, and disabling vehicle damage further contributed to fatal outcomes. These findings highlight the compounded dangers of isolated, low-visibility, and uncontrolled settings, especially when drivers are alone. The study underscores the need for targeted safety interventions such as enhanced roadside lighting, improved infrastructure in non-intersection areas, and time-specific driver awareness efforts. By identifying these high-risk patterns, this research provides actionable insights to inform policy and infrastructure strategies aimed at protecting vulnerable elderly pedestrians.