Examining in-vehicle distraction sources in relation to crashes using a Bayesian Multinomial Logit model

It is well understood that most crashes are the result of human errors. Among human-related errors, distracted driving, particularly related to cellphones, has received significant attention. Conversely, the underlying factors associated with in-vehicle distractions that are non-cellphone use have not been fully explored. Thus, this paper uses data from driver distraction-related crashes to examine various in-vehicle distraction sources. A Bayesian Multinomial Logit (BMNL) model was developed using 5,078 distracted-driving related crashes from Iowa. Four in-vehicle distraction sources - cellphone use, non-cellphone electronic devices, passengers, and reaching in-vehicle fallen objects - were investigated to determine factors that increase their odds of occurrence. The results suggest that drivers under the influence of alcohol are more likely to be involved in crashes associated with the distraction from cellphones. Furthermore, older drivers are less likely to be involved with distracted driving due to passengers. As expected, the more people in the vehicle, the higher the likelihood a driver can be distracted by passengers. Moreover, the association of driver distraction and speed limit, time of the day, vehicle's age, among others, were evaluated. This study provides useful information for developing and implementing strategies that minimize distractions from all in-vehicle sources.