AbstractRoadway improvements to reduce the frequency of crashes are of the utmost priority to transportation agencies. To a great extent, implementation of improvement programs depends on the reliable identification of roadway segments with high crash risk. Among all crash types, wrong-way driving (WWD) crashes are considered random in nature and are a major safety concern. The Federal Highway Administration defines WWD specifically for high-speed divided highways and access ramps. This definition excludes all other roadway classifications when a crash occurs in the opposing direction to the legal flow of traffic. Screening 5 years of crash data in Minnesota revealed that WWD resulted in crashes on other types of roadway functional classes. This work aimed to (1) introduce a new term/acronym to the literature for driving in the wrong direction (DWD) on all roadway functional classes, (2) apply a set of count data models to estimate the occurrence of DWD crashes, (3) identify roadway geometric features of high-risk segments for DWD crashes, (4) investigate random effects of covariates due to unobserved factors, and (5) calculate elasticity effects of variables. Final models’ specifications indicate that the negative binomial (NB) mixed effect model was found to be the best-fit model. Focusing on DWD crashes, we uncovered the factors contributing to higher DWD crash-risk segments: log of average annual daily traffic (AADT), number of lanes, sidewalk, and shoulder type. The change in frequency of crashes is also investigated using marginal effects, and safety interventions for preventing DWD crashes are also discussed. Transportation agencies can use the findings of this research, in terms of contributing factors and their relative effects on DWD crashes, to deploy appropriate countermeasures at high-risk locations.