A correlated random parameters ordered probit approach to analyze the injury severity of bicycle-motor vehicle collisions at intersections

Abstract

Bicycle-motor vehicle (BMV) accidents hold paramount importance due to their substantial impact on public safety. Specifically, road intersections, being critical conflict points, demand focused attention to reduce BMV crashes effectively and mitigate their severity. The existing research on the severity analysis of these crashes appears to have certain gaps that warrant further contribution. To address the mentioned limitations, this study first integrates multiple pre-collision features of the bicycles and vehicles to classify crash types based on the mechanism of the crashes. Then, the correlated random parameters ordered probit (CRPOP) model is employed to examine the factors influencing injury severity among bicyclists involved in intersection BMV crashes in Pennsylvania from 2013 to 2018. To gain deeper insights, this study conducts a separate analysis of crash data from 3-leg intersections, 4-leg intersections, and their combined scenarios, followed by a comparative examination of the results. The findings revealed that the presented crash typing approach yields new insights regarding injury severity outcomes. Moreover, in addition to exhibiting a comparable statistical performance contrasting to the more restricted models, the CRPOP model identified hidden correlations between three random parameters. Furthermore, the study demonstrated that analyzing combined crash data from the two intersection types obscured certain factors that were found significantly influential in the injury outcomes through analyzing sub-grouped data. Consequently, it is recommended to implement tailored countermeasures for each type of intersection.