Toward a typology of driving under the influence of alcohol and drugs.

Abstract

Most research on driving under the influence (DUI) has relied upon variable-centered methods that examine predictors/correlates of DUI. In the present study, we utilize a person-level approach-latent class analysis (LCA)-to model a typology of individuals reporting DUI. This allows us to understand the degree to which individuals drive under the influence of a particular substance or do so across multiple substance types. We use public-use data collected between 2016 and 2019 from the National Survey on Drug Use and Health. The analytic sample was 189,472 participants with a focus on those reporting DUI of psychoactive substances in the past-year (n = 24,619). LCA was conducted using self-reported DUI of past-year alcohol, cannabis, cocaine, heroin, hallucinogens, and methamphetamine as indicator variables. More than 1 in 10 Americans reported a DUI within the past-year. One in five people who reported DUI of one substance also reported DUI of at least one additional substance. Using LCA to model heterogeneity among individuals reporting DUI, four classes emerged: "Alcohol Only" (55%), "Cannabis and Alcohol" (36%), "Polydrug" (5%), and "Methamphetamine" (3%). Rates of risk propensity, drug involvement, illicit drug use disorders, and criminal justice system involvement were highest among members of the "Polydrug" and "Methamphetamine" classes. Drug treatment centers should take care to include discussions of the dangers and decision-making processes related to DUI of the full spectrum of illicit substances. Greater investment in drug treatment across the service continuum, including the justice system, could prevent/reduce future DUI episodes.