This research explored the use of count models to determine design and weather factors correlated with median crossover crashes on Washington State highways. The study 1) developed a roadside data system that can be consistently and systematically used in all six regions of Washington State; 2) developed a decision matrix comprising geometric, environmental, and traffic factors for estimating crossover probability ranges; and 3) examined the impacts of barriering. Longitudinal data for the period 1990 to 1994 containing crash information on vehicle crossovers on non-barriered medians on Washington State highways were used as the dataset for this study. Two types of statistical models were examined: 1) a model that forecasts the mean number of yearly median crossovers, and 2) a model that examines the contribution of roadway geometrics, median widths, weather, traffic volumes and roadside characteristics to the annual societal cost of median crossovers.
Results of the study suggest these design policies: Barrier all medians less than or equal to 50 feet wide; do not install barriers for medians wider than 60 feet; consider case-by-case barriering medians in the 50-foot to 60-foot range.