Detection of impaired drivers in sobriety checkpoints is typically low. This study determines if a set of observable driver, vehicle and situational variables has potential value in identifying impaired drivers passing through sobriety checkpoints. The study predicts the blood alcohol concentrations (BAC) of a randomly selected sample of nighttime drivers and classifies them into BAC groups. The data were obtained from a 1998 roadside survey in three British Columbia communities. Multinomial logit models and discriminant models were estimated to describe, explain, and classify the BAC of nighttime drivers. Several factors increase the odds of being an impaired driver: age group, 26 to 45; education, high school or less; trip origin, bar; passengers, group same sex as driver; and advancing hour of night. The classification function correctly classified more than 65% of all drivers and 62% of impaired drivers. Information in the model is not sufficient to make a reliable determination of BAC but could be used to increase the likelihood of identifying impaired drivers in sobriety checkpoints, thereby improving the efficiency of enforcement operation. (Author/publisher).
Abstract