MnDOT’s Rail Administration Section monitors the safety performance of more than 4,000 public rail grade crossings and develops the Department’s Railroad-Highway Grade Crossing Safety Improvement Program. A total of 445 vehicle-train crashes occurred at the public crossings during the 10-year study period, extending from 2004 through 2013. Of these crashes, 52 involved fatalities and another 129 involved injuries. Analysis of the crash data reveals two key points; the density of fatal plus injury crashes is very low - 0.004 per grade crossing per year and almost 91 percent of the crossings had NO crashes of any kind during the study period. This data combined with the fact that Rail Administration staff used a pair of crash prediction models that included consideration of crash history (along with crossing geometry and train/vehicle exposure factors) to identify candidates for safety investment raised concern among MnDOT staff regarding whether or not the models placed too high a priority on prior crash history. This set of facts resulted in undertaking a study to determine how well the prediction models fit Minnesota’s data and to attempt to identify a set of grade crossing risk factors that would support a statewide systemic evaluation. The study determined that both the FRA Accident Prediction Model and the Texas Hazard Index were poor fits with Minnesota’s crash data. The study also identified a set of volume, speed, design and surrounding area features that were over-represented at crossings with crashes and that the crossings with multiples of these features present had the highest crash densities and could be considered high priority candidates for safety investment.