Deer-vehicle accidents (DVAs) are a growing concern in the United States, due to rising White-tailed deer (Odocoileus virginianus) populations, and increasing traffic volume and road density. Currently, managers implement changes at the roadside; however, it is not sufficient to mitigate at this scale when additional factors are having an effect on the system. This study examines how variables measured at differing scales predict DVA frequency in Onondaga County, New York. Models that incorporated only roadside characteristics, the traditional focus for management, performed poorly. Moreover, the inclusion of deer abundance did not improve model fit. At the roadside scale, models that incorporated both deer crossing motivation (grazing, cover) and ability (barriers, visibility) factors showed higher goodness-of-fit. The Ability and Motivation themes incorporating roadside and spatial context factors received the highest weight, based on Akaike's Information Criterion, (wi = 0.67), suggesting that incorporating variables from differing scales is important for mitigating DVAs.