A multiverse analysis of the affect-alcohol use association

PROJECT SUMMARY The notion that people consume alcohol in order to regulate both positive (drink-to-enhance) and negative (drink-to-cope) emotions is central to theories of the development and treatment of alcohol use disorder (AUD). Evidence for affect regulation is robust from both experiments and people’s retrospective, global self-reports of their drinking motives. However, evidence from ecological momentary assessment (EMA) studies have failed to produce compelling evidence of affective regulation of alcohol use. A central problem is that affect regulation theories are vague as to when, where, how, and for whom they are relevant, leading to potentially hundreds of ways that these theories can be operationalized. The lack of empirical consensus means that studies designed to test a single affect regulation hypothesis are ultimately less informative because they cannot rule out or rule in plausible alternative hypotheses. We will test hypotheses across timescales, measures, persons, and situations using an adaptive, multi-burst EMA design. Our design will allow us to weigh the evidence for the most theoretically informed and empirically supported affect regulation hypotheses in the same dataset. Using specification curve analysis in combination with cross-validation in a hold-out sample, we will be able to test the robustness and generalizability of each hypothesis across hundreds of specifications of timescale, affect, and alcohol outcomes. We will simultaneously test multiple specific and contextualized predictions of the affect–alcohol use association that can inform etiological and intervention research. We aim to do this in a large (n = 500) adult regular and hazardous drinkers collected across three sites who will complete three high intensity bursts of EMAs (10 per day over 10 days. By using these data to compare all reasonable ways that affect regulation may be specified, the proposed project will identify sets of hypotheses about affect regulation of alcohol use that are either well supported by data, or which should no longer be investigated.