ANALYSIS OF THE MAGNITUDE AND PREDICTABILITY OF MEDIAN CROSSOVER CRASHES UTILIZING LOGISTIC REGRESSION

A median crossover crash (MCC) is defined as an accident in which a vehicle traverses the median area and penetrates the opposing travel lane. Crashes vary from vehicles coming to rest in the opposing lane, to vehicles passing through the opposing lane without hitting an opposing vehicle, to head-on or sideswipe impacts with opposing vehicles. Until recently, the magnitude, characteristics, and causes of MCCs were not widely investigated. The objective of this research was to determine the magnitude, severity, and predictability of MCCs in Wisconsin state. A total of 15,194 crash reports from Wisconsin’s median divided freeways and expressways were analyzed for the period of 2001-2003. The results of this analysis identified 631 reported MCCs over this three-year period. The magnitude of MCCs indicated that this crash type is a considerable issue in Wisconsin and required additional investigation in order to determine the causes of these crashes and to develop appropriate countermeasures. To identify the significant attributes of MCCs, ordinal logistic regression models were developed to predict MCC severity based on a number of predictors, including: roadway and driver characteristics, traffic operations, incident management, temporal elements, and environmental factors. Crash severity was selected as the response variable so that the significant variables identified and associated countermeasures developed were focused initially on improving safety, although reducing the frequency was also of critical interest. Statistically, the initial analysis found four predictors to have significant effects on the severity of the 615 MCCs: Weather, Road Condition, Median Crossover Type, and Total ADTs. Additionally, further analysis showed that Number of Vehicles Involved and Crossover Type affected crash severity on typical Wisconsin highways. Moreover, Crossover Type was found to be responsible for aggravating the MCC severity when traffic volume is either higher or lower. The results indicate that modeling MCC severity as an ordinal response is statistically appropriate, and resultant findings could be used by traffic authorities to determine the probability of crash severity based on a set of predictors and facilitate the decision-making process regarding roadway safety enhancement measures such as median barriers.