Stress-Related Drinking Predicts Alcohol Problems Among University Students: Using Individualized Multilevel Modeling Slopes As Predictors

Introduction: For many students, the college years may be characterized by both experimentation with heavy alcohol use and with the increased stress associated with establishing financial and social independence. For some of these students, alcohol use may become a way of coping with the day-to-day challenges that accompany college life. But what is the consequence of such stress-related drinking?

Methods: We analyzed data from the University Life Study, a longitudinal daily diary study of 744 university students followed from the fall semester of their first year (M age = 18.4 y, SD = .43 y) to the fall semester of their fourth year in college.

Results: Analyses proceeded in two steps. First, we used generalized three-level multilevel modeling (MLM) with random slopes to estimate the day-level association between stressor exposure (using a count of stressful events on each day) and alcohol use, including whether or not any drinking occurred on that day as well as the number of drinks consumed on drinking days. Second, we used empirical Bayes estimates from our multilevel models to generate individualized stress-drinking slopes for each student, with steeper (versus flatter) slopes suggesting students who tended to increase their drinking more strongly on high-stress versus low-stress days. Results showed that, on days when students experienced more versus fewer stressors, they were significantly more likely to engage in drinking (OR = 1.08, 95% CI: 1.03, 1.13) and consumed more drinks once they started drinking (IRR = 1.04, 95% CI: 1.02, 1.06). Next, using individualized MLM slopes as predictors of alcohol problem severity (as measured by the AUDIT), we found that students who increased their odds of drinking more sharply on high- versus low-stressor days (steeper or more positive slopes) showed significantly greater alcohol problem severity in the fourth year of college than students whose odds increased less sharply or did not increase (flatter or more negative slopes; IRR = 1.19, 95% CI: 1.07, 1.32). This relationship held even after controlling for drinking levels on non-stressor days, average levels of stressor exposure, and alcohol problem history.

Conclusions: Our results are innovative because they demonstrate how individualized MLM slopes, which characterized students based on their unique relationship between stressor exposure and alcohol use, can be useful in predicting outcomes relevant to public health and prevention science. As such, our findings provide important information relevant to the prevention of alcohol problems in university students, suggesting that intervention focusing on improving stress management skills could play a part in reducing the prevalence of problem drinking in this population.