Multi-level Mixed-effects Regression Model for Truck Collisions

Abstract

As freight demand and truck traffic steadily increases in the United States, so does the number of truck crashes. Collisions involving trucks has long been a major safety concern for highway operations since these crashes are often more severe than those involving lighter vehicles. This paper describes the rationale and construction of an estimation model that can be used to assess truck safety in a freeway network. Truck crash data and highway geometric data from California were combined with economic data from the US Census in this study for statistical analysis. Poisson and Negative Binomial regression models were first developed. Multi-Level Mixed-Effects regression model was then established. Within the framework of the multi-level modeling approach, it was assumed that economic conditions in different counties would affect freight demand and thus truck traffic flows and the occurrence of truck collisions. The models developed in this study were established to use the number of trucks involved in crashes as the representative outcome variable or measure to assess the safety performance of truck freeway operations. The outcome of models and associated data analysis revealed that the presence of ramp and freeway segment length were the most important factors affecting the safety performance measured by the number of trucks involved in crashes. Furthermore, the results showed that multi-level mixed-effects models were able to accurately estimate the numbers of truck crashes. Using these models, practitioners can identify freeway sites where truck crashes are more likely to occur and plan accordingly to mitigate safety concerns.