Occupant-level injury severity analyses for taxis in Hong Kong: A Bayesian space-time logistic model.

This study aimed to identify the factors affecting the crash-related severity level of injuries in taxis and quantify the associations between these factors and taxi occupant injury severity. Casualties resulting from taxi crashes from 2004 to 2013 in Hong Kong were divided into four categories: taxi drivers, taxi passengers, private car drivers and private car passengers. To avoid any biased interpretation caused by unobserved spatial and temporal effects, a Bayesian hierarchical logistic modeling approach with conditional autoregressive priors was applied, and four different model forms were tested. For taxi drivers and passengers, the model with space-time interaction was proven to most properly address the unobserved heterogeneity effects. The results indicated that time of week, number of vehicles involved, weather, point of impact and driver age were closely associated with taxi drivers' injury severity level in a crash. For taxi passengers' injury severity an additional factor, taxi service area, was influential. To investigate the differences between taxis and other traffic, similar models were established for private car drivers and passengers. The results revealed that although location in the network and driver gender significantly influenced private car drivers' injury severity, they did not influence taxi drivers' injury severity. Compared with taxi passengers, the injury severity of private car passengers was more sensitive to average speed and whether seat belts were worn. Older drivers, urban taxis and fatigued driving were identified as factors that increased taxi occupant injury severity in Hong Kong.