Examining the effects of interactions between driver-vehicle units on driver injury severities of work zone crashes

Most of the crash severity analysis at the level of driver-vehicle units merely considers the effects of driver and vehicle factors of the own unit, while neglects the interactive effects from the other unit. This study aims to examine the interactions effects in work zone crashes, using a two-vehicle crash dataset from the Crash Report Sampling System. To simultaneously accommodate the order nature, within-crash correlation, and temporal correlation, a Bayesian hierarchical generalized ordered Logit model with random walk is proposed, to establish a relationship between the driver injury severity and the observed factors related to drivers and vehicles of both units, together with those pertaining to roadway, work zone, environment, and crash configurations. The results of model estimation indicate that driver injury severity is significantly affected by the driver gender and age, vehicle type and age of both units, which explicitly revealed the interactive effects between driver-vehicle units within the same work zone crash. Additionally, the driver injury severity is also associated with their speeding and seat belt use behaviors, type of work zones, speed limit, number of lanes, season, day of week, time of day, collision manner, and the involvement of alcohol and rollover. Their marginal effects are also estimated. Some factors, such as speeding and rollover, are expected to increase the probability of serious and fatal injury up to 6% or more. Based on the findings, some strategies for preventing severe traffic injury in work zones are recommended. Besides, significant within-crash correlation and temporal correlation are uncovered, and the values of model assessment indicators justify the appropriateness and strength of the proposed model.