In order to deeply understand the relationship between crash occurrence and highway features on suburban highways, 161 segments of 8 suburban highways in Shanghai were chosen and a model was established to analyze the relationship between safety and macro highway characteristics, geometry designs, access features, traffic characteristics and environmental settings to deepen the understanding of crash occurrence on suburban highways. The road segments from the same highway were correlated because of the same design criteria and similar traffic characteristics, which violated the traditional models assumption that all samples should be independent with each other. Therefore, hierarchical negative binomial model was developed to analyze risk factors. The results show that more slight accidents occur in inner suburb than those in outer suburb; the number of signalized intersections per kilometer has a significant influence on slight, severe accidents and total accidents and long and equiliium interval is best for safety; the probability of accident occurrence is higher for the segments with more access points. In addition, there are more severe accidents happening on curve segments.