Ordinal Discrete-Choice Analyses of Wisconsin Cross-Median Crashes

Cross-median crashes (CMCs), in which a vehicle crosses the highway median, are one of the most severe crashes due to high speeds and risk of collision with an opposing vehicle. This paper describes the ordinal discrete-choice modeling efforts for investigating the nexus between the severity propensity and miscellaneous variables pertinent to roadway safety for singleand multi-vehicle CMCs which occurred between 2001 and 2007 in Wisconsin. Ordinal Logit (ORL) and Probit (ORP) regression models were employed for severity analyses. For multi-vehicle CMCs, both models revealed road surface condition has a significant effect on the severity. Adverse road surfaces enhance the likelihood of being involved in a more severe multi-vehicle CMC if one occurs. Winter snow or ice impacts the CMC severity, and logically Wisconsin’s geographical location plays a significant role. Although both models found the speed limit significant, they revealed different severity propensities, implying this factor should be treated cautiously and the necessity of applying different discrete-choice models to severity analyses if more comprehensive understanding is pursued. Final ORP model for single-vehicle CMCs shows alcohol/drug use, lane curvature, and unfriendly roadway visibility exacerbate the severity if a single-vehicle CMC occurs. Interestingly, dry road surface is found to significantly incur more severe consequences, which implies more severe single-vehicle CMCs are closely related to maintaining high speeds. ORL modeling results were found statistically invalid for single-vehicle CMC severity analyses. The median width and average daily traffic were found insignificant factors for both multi-vehicle and single-vehicle CMCs.