Alcohol-related motor vehicle crashes are a major preventable cause of death in the United States. One potential factor that may modulate the influence of alcohol intoxication on driving-related decision-making is posttraumatic stress. The current study evaluated the influence of induction of acute trauma-related stress (via script-driven imagery) during alcohol intoxication (.06% BrAC) on driving-related risky decision-making – willingness to drive, driving-related decision-making (i.e., attempted red light runs), and driving-related reaction time (i.e., braking latency) – among 56 trauma-exposed (currently symptomatic) adult drinkers from the community (M = 25.32; 46.4% female). Results indicated that trauma-related stress may exacerbate willingness to drive during a state of acute alcohol intoxication, but, alternatively, may have only a minimal-to-moderate effect on performance-based, driving-related decision-making (i.e., red light runs), and a potentially mitigating impact on driving-related reaction time (i.e., braking latency) under the influence of alcohol. Generally, results suggest that trauma-related stress may differentially impact varying aspects of driving-related risky decision-making, above and beyond the influence of alcohol. Implications for theoretical modeling for driving-related decision-making during acute intoxication and for the advancement of education and intervention efforts, as well as suggestions for future directions, including methodological and procedural improvements, are discussed.

The purpose of this research was two-fold: (1) further validate the Virtual Reality Stroop Task HMMWV [VRST; Stroop stimuli embedded within a virtual high mobility multipurpose wheeled vehicle] via a comparison of the 3-dimensional VRST factor structure to that of a 2-dimensional computerized version of the Stroop task; and (2) model the performance of machine learning [ML] classifiers and hyper-parameters for an adaptive version of the VRST. Both the 3-D VRST and 2-D computerized Stroop tasks produced two-factor solutions: an accuracy factor and a reaction time factor. The factors had low correlations suggesting participants may be focusing on either responding to stimuli accurately or swiftly. In future studies researchers may want to consider throughput, a measure of correct responses per unit of time. The assessment of naive Bayes (NB), k-nearest neighbors (kNN), and support vector machines (SVM) machine learning classifiers found that SVM classifiers tended to have the highest accuracies and greatest areas under the curve when classifying users as high or low performers. NB also performed well but kNN algorithms did not. As such, SVM and NB may be the best candidates for creation of an adaptive version of the VRST.