Examining injury severity in left turning crashes at intersections

Abstract

In response to the increasing concerns and challenges for most frequently left-turn crashes at intersections, partial proportional odds models, for which some of the beta coefficients vary across variables, are proposed to examine and understand the influence of contributory factors (i.e. human attributes, traffic flow features, roadway geometrics, and environmental factors, etc.) on injury severity involved in left-turn crashes, using the selected 317 crash data over latest 6 years from Xian city. The results show that partial proportional odds model has better performance than general ordered logit or probit probability approach. Specifically, the aged and younger drivers are more prone to cause left-turn crashes, and the increasing effect of trucks involvement, impact points of both vehicles, environmental factors, safety belt usage, alcohol and/or drugs are also significantly associated with higher injury severities, which was underestimated or underreported in previous researches.