Testing the robustness of daily associations of affect with alcohol and cannabis use

Etiological models of alcohol and cannabis use disorders hypothesize that people are more likely to use and consume more of these substances when they experience heightened negative affect, yet recent EMA studies found no evidence for the daily association between negative affect and alcohol/cannabis use (but positive affect consistently predicted subsequent alcohol/cannabis use). However, the theory underlying affect regulation is vague and has been translated into many different statistical tests across the literature, with inconsistent results. For that reason, here we tested the affect regulation hypothesis in a high-risk community sample of young adults (N = 287) across hundreds of statistical models to provide a robust understanding of whether and when affect regulation is supported in everyday life. We systematically varied statistical models along several specifications that are relevant to the interpretation of the affect regulation hypothesis. We varied the operationalization of negative and positive affect (e.g., NA: anger, sadness, anxiety, distress), the time scale at which affect could be associated with alcohol/cannabis use (daily average prior to use, daily maximum prior to use, daily variability prior to use, most recent affect report prior to use), and the population for whom the association should be present (e.g., full sample, AUD criterion count of 2 and 6, AUDIT score of 3+ and 8+, above-average coping and enhancement motives). We then fitted all of these models, which resulted in a specification curve of ranging effect sizes and confidence intervals. Results indicated that across all models negative affect was associated with a decreased likelihood to use alcohol (OR = 0.93, p = .015) and cannabis (OR = 0.87, p < .001) later that day and positive affect was associated with an increased likelihood to use alcohol (OR = 1.17, p < .001) and cannabis (OR = 1.12, p = .004) later that day. Individual model results differed based on the operationalization of affect (e.g., anxiety was associated with a reduced likelihood to drink but anger was not). This indicates that associations between affect and substance use might not be universal and depend on the specific emotions assessed and used to compute indicators of general negative and positive affect.

Jun 24, 2024