Preventing Pedestrian Injury Severity at Signalized Intersections: A Bayesian Multilevel Parametric Survival Model

This study intended to investigate the influencing factors of pedestrian injury severity at signalized intersections, considering the heterogeneity issue of unobserved factors at different signalized intersections and the spatial attributes with survival models. To achieve the objectives, a Bayesian multilevel parametric survival model was developed, in which the survival model addressed the pedestrian severity levels varying with time, while the panel data model accommodated the heterogeneity attributed to unobserved factors and spatial features within the Bayesian framework. The pedestrian-related crash data of Hong Kong metropolitan area from 2008 to 2012 were integrated, involving 376 signalized intersections with 2,090 pedestrian severity samples. By comparing the proportional hazard (PH) and accelerated failure time (AFT) models, the PH model with exponential distribution showed priority to the other models. Results revealed that pedestrian age, injury location, pedestrian special circumstance, pedestrian contributory, number of pedestrian stream, obstruction, road type, presence of tram/LRT stops and bus stops were potentially significant factors of increasing the pedestrian injury severity probability. The findings provide useful insights for practitioners and policy makers to improve pedestrian safety at signalized intersections.