Exploring the need to model severity of single- and multi-occupant vehicles crashes separately: A case of crashes at highway-rail grade crossings

Abstract

In recent years, traffic safety researchers have attempted to separate single-vehicle and multi-vehicle crashes when analyzing crash severity, considering the significant differences in the mechanism of occurrence of the two crash types. However, regardless of the number of vehicles involved in a crash, the severity of a crash is defined by the most severe injury outcome sustained by the occupants, not vehicles. Thus, this study evaluated a need for conducting a separate severity analysis for crashes involving a single occupant (SO) and multiple occupants (MO). Ten-year data (2009–2018) of crashes that involved a collision between a single vehicle and a train at the highway-rail grade crossings (HRGCs) across the United States was used as a case study. Crashes were grouped based on occupancy level; that is, crashes involving SO were separated from the ones involving MO. As expected, MO crashes had higher injury and fatality rates than SO crashes. Three Multinomial Logit (MNL) models were developed to analyze the crash severity of SO crashes, MO crashes, and total crashes. The study found several differences in associated factors when SO crashes and MO crashes were modeled separately. Overall, combining SO and MO crashes tend to either underestimate or overestimate the actual impact of the predictor variable on a specific crash type. Among the variables, train speed and vehicle speed during crash showed a great difference. The findings provide evidence that the severity analysis of the SO and MO crashes should be performed separately as they have different characteristics.

Keywords

• Injury severity
• Single-occupant crashes
• Multi-occupant crashes
• Highway-rail grade crossings