Sociodemographic, Behavioral, and Cognitive Predictors of Alcohol-Impaired Driving in a Sample of U.S. College Students

Abstract

Alcohol-impaired driving continues to be a major public health concern, particularly among college students. The current study examined whether sociodemographic, behavioral, and cognitive variables predicted alcohol-impaired driving in a sample of college students. Data were collected via telephone interviews from a random sample of undergraduates, ages 18–25 years old, stratified by sex and class in school. Using hierarchical logistic regression analyses (n = 330), results revealed that higher levels of weekly alcohol use, being age 21 or older, and perceived difficulty in obtaining alternative transportation were associated with a greater likelihood of drinking and driving. In addition, perceived likelihood of drinking and driving-related consequences was associated with a lower likelihood of drinking and driving. Knowledge of the .08% per se and zero tolerance laws did not predict alcohol-impaired driving. Findings are discussed in terms of their implications for college media campaigns designed to reduce alcohol-impaired driving.

We are grateful to our many colleagues on the Common Ground project who helped conduct the student surveys. This study was supported by grant U01-AA-014749 from the National Institute on Alcohol Abuse and Alcoholism (NIAAA) awarded to Mark D. Wood. The views expressed in this article are those of the authors and do not necessarily reflect the official position of NIAAA.

Notes

Note. Correlations were computed as follows: Pearson product-moment correlation coefficients between two continuous variables, point-biserial correlations between one continuous and one dichotomous variable, and phi coefficients between two dichotomous variables.

^a For fraternity/sorority status, “0 = nonmember” and “1 = current member/pledge of a fraternity or sorority.” For sex, “0 = female” and “1 = male.” Weekly alcohol use was defined as the typical number of drinks consumed per week. Age was coded as “0 = under 21” and “1 = 21 or older.” Perceived likelihood of negative consequences was coded on a 5-point response scale ranging from “1 = not at all likely” to “5 = extremely likely.” Knowledge of deterrence laws was scored from “0 = did not correctly answer the BAC limit for either the .08% per se law or the zero tolerance law” to “2 = correctly answered the BAC limits for both laws.” Perceived availability of alternative transportation was scored on a 5-point response scale, ranging from “1 = very easy” to “5 = very difficult.” The sex-specific alcohol-impaired driving measure was dichotomized as “0 = did not drink and drive in the past 30 days” and “1 = did drink and drive at least once in the past 30 days.”

∗p < .05;

∗∗p < .01;

∗∗∗p < .001.

¹Analyses were conducted also after inputting continuous predictors using mean substitution, yielding a sample of n = 370. The significant predictors obtained after mean substitution were the same as those presented here.

²Parallel ordinary least squares regression analysis using a square-root transformation of the sex-specific drinking and driving outcome measure closely approximated the results from the logistic regression analyses presented here, with the significant predictors for Steps 1 and 2 being consistent across the two types of analysis.

Note. Odds ratio estimates are adjusted for other predictors and are given for significant predictors only.

∗p < .05;

∗∗p < .01;

∗∗∗p < .001.

³Zero-inflated Poisson regression models also were conducted in Mplus Version 4.2 since the dependent variable was a count variable with an overabundance of zeros. The count portion of the model did not add to the interpretation of the results. Therefore, only the logistic regression analyses are presented. Specifically, only age was a significant predictor in the count portion of the model in Step 1, while none of the predictors was significant in the count portion of the model in Step 2. Results for the binary portion of the model approximated the results reported here using logistic regression, such that the significant predictors for Steps 1 and 2 were consistent across the two types of analysis, with the exception of sex not being significant at Step 1.

⁴We ran logistic regression models separately for those who were under 21 years of age (n = 202) and those who were 21 or older (n = 128) to determine whether knowledge of the .08% per se law and zero tolerance law were predictive of alcohol-impaired driving after accounting for age. For both age groups, results showed that knowledge of neither law predicted alcohol-impaired driving.