Assessment of Commercial Truck Driver Injury Severity as a Result of Driving Actions

Abstract

The disaggregate modeling approach is a new trend in the literature to analyze the injury severity of truck-involved crashes. The assessment of truck driver injury severity based on driver action is still missing in the literature. This paper presents an extensive exploratory analysis that highlights significant variability in the severity of truck drivers’ injuries based on various action types (i.e., aggressive driving, failure to keep proper lane, driving too fast, and no improper driving). Binary logistic regression with the Bayesian random intercept approach was developed to examine the factors contributing to fatal or any injuries of truck drivers using 10 years (2007–2016) of historical crash data in Wyoming. Log-likelihood ratio tests were performed to justify that separate models by various driving action types are warranted. The results demonstrated the effects of various vehicle, driver, crash, and roadway characteristics, combined with truck driver-specific action, on the corresponding severity of driver injury. The gross vehicle weight, age and gender of the driver, time of day, lighting condition, and the presence of junctions were found to have significantly different impacts on the severity of truck driver injury in various driving action-related crashes. With the incorporation of the random intercept in the modeling procedure, the analysis found a strong presence (27%–33%) of intra-crash correlation in driver injury severity within the same crash. Finally, based on the findings of this study, several recommendations are made.

Investigating the Temporal Instability in Injury Severity Outcomes of Clear and Adverse Weather Crashes on Rural Mountainous Highways

“…For this reason, the injury severity levels were combined into two levels: no injury and fatality/injury. The severity levels are often combined to form a binary response with 0 = no injury crashes, and 1 = fatality/injury crashes ( 13 , 26 , 44 ). Similarly, the injury severity levels in this study are represented by probability π ij of a fatality/injury severity outcome, and 1 − π ij is the corresponding probability of a no injury severity outcome.…”

“…Injury severity levels are often combined to form just two categories and as a result the response variable then becomes dichotomous, with 0 representing no injury, minor injury, or non-severe injury, and 1 representing fatal injury, major injury, or severe injury. Binary logistic models have been more often utilized in the literature to model the binary response crash injury severity ( 13 , 25 , 26 ).…”

“…Crash injury severity analysis using a Bayesian approach has also gained popularity ( 13, 20, 25–28 ). In a Bayesian modeling framework, prior distributions on parameters can be assigned which, when combined with the likelihood of the observed data, provide posterior probability distributions of the parameters.…”

“…Various MCMC sampling techniques are used in the literature, such as the Gibbs sampler, Metropolis Hastings, and the combination of both ( 13 , 25 , 27 ), implemented using the freeware Winbugs ( 30 ). A new MCMC method, known as the No-U-Turn Hamiltonian Monte Carlo (NUTS HMC), has gained more attention simply because of its superior efficiency and robustness over the other sampling techniques ( 20 , 26 , 31 , 32 ). The adoption of the NUTS HMC has been slow in the literature, mainly because of its underlying sophisticated mathematics.…”

“…The adoption of the NUTS HMC has been slow in the literature, mainly because of its underlying sophisticated mathematics. However, recent developments in the Stan programming language in R® interface have allowed the implementation of the NUTS HMC and the interpretation of the model results much more easily than was previously possible ( 20 , 26 , 33 ).…”