Large truck fatal crash severity segmentation and analysis incorporating all parties involved: A Bayesian network approach

Studies on fatal crashes involving large trucks are typically focused on the examination of injury severity at the personal level rather than the crash level. In this study, instead of adopting the predefined, commonly used rule of transformation of individual injury severity to crash level severity, we develop a data specific transformation approach. Crash-level severity for fatal large-truck involved crashes is first identified applying a two-step clustering method. A Bayesian network is then constructed to investigate the impacts of influential factors on the crash severity. Data obtained from the Fatality Analysis Reporting System in 2019, are used to calibrate the model. The results indicate that the number of vehicles involved in the crash, collision manner, truck-airbag deployment, overturn of vehicles, driver's speeding offense, and gross vehicle weight are the key determinants of crash injury severity. Uni- and two-dimensional inference analyses were conducted to determine the relations among these factors and their association with crash severity.

Automated summary

Studies on fatal crashes involving large trucks primarily focus on individual injury severity rather than crash severity. This study proposes a data-specific transformation approach, identifying crash-level severity using a two-step clustering method and investigating influential factors using a Bayesian network. The results highlight the determinants of crash injury severity, including the number of vehicles involved, collision manner, truck-airbag deployment, overturn of vehicles, driver's speeding offense, and gross vehicle weight.