A Hierarchical Modeling Approach to Predict Pedestrian Crash Severity

The crashes involving pedestrians are increasing at an alarming rate over the years in urban Bangalore. The past year witnessed 1592 crashes in urban Bangalore out of which 338 pedestrians were killed and 1254 were injured. The gravity of the problem requires immediate attention of road safety experts, engineers, and other stakeholders who are directly or indirectly involved in traffic safety. Hence, an in-depth analysis of this problem is required to draw out sound and reliable engineering countermeasures that will address the safety issues of pedestrians and improve the safety performance of roadways. For this, data was collected from Bangalore City Traffic Police, and it was found that the maximum number of pedestrian vehicle collisions occurs in the Bangalore-Chennai Highway, which is our study location. This study mainly focuses on identifying the factors contributing to the severity of pedestrian crashes in urban mid-blocks. From the literature review, it was identified that pedestrian crashes are associated with characteristics of pedestrian, location, land use, environment, and crash. For modeling vehicle-pedestrian crashes, the review of the literature shows that different regression techniques such as logit and probit-models are widely used. Crash patterns may vary across locations, and this variation is not accounted in the traditional models. The traditional models cannot accommodate the array of variables at multiple levels like regional, site, crash, and driver-vehicle unit level. To increase the accuracy of model prediction, a hierarchical modeling approach is considered in the present study. This study considers traffic, geometric, and environmental variables and identifies their association with vehicle-pedestrian crashes.