Brief Online Interventions for Alcohol Use: A Crowdsourced Study

Alcohol use disorder is a persistent public health concern. Approximately 16 million Americans met criteria for alcohol use disorder in 2015, with the annual economic impact of excessive drinking estimated at $250 billion. Alcohol use can also interact with other licit (e.g., cigarettes) and illicit (e.g., cocaine) substances to increase health risks . At least half of all violent crimes involve alcohol consumption and problem drinking plays an important role in domestic abuse and intimate partner violence . These evident economic, health, and social implications of alcohol misuse highlight the importance of understanding determinants of alcohol use behaviors and developing interventions designed to address maladaptive patterns of use.

Cognitive training to address alcohol and other substance use disorders has received a great deal of attention in the interventions development literature. Training may be broadly divided into two classes: 1) cognitive bias modification and response inhibition (i.e., inhibitory control) and 2) working memory interventions. Inhibitory control training attempts to retrain prepotent responses away from drug-related cues by specifically pairing those cues with no-go signals in training tasks. Working memory training uses cognitive-behavioral tasks (e.g., letter/digit strings, visual search, n-back) to improve information maintenance and manipulation. These brief interventions hold particular appeal because they may be easily incorporated into comprehensive approaches to substance use treatment. For example, cognitive improvements due to brief training may improve engagement with and attention to cognitive-behavioral therapy and compliance with homework and other program-related activities. Generally speaking, training has consistently demonstrated improvements on the trained or similar tasks as well as related concepts (e.g., delay discounting is improved after working memory training). Improvements in dissimilar tasks or different domains are not consistently observed, however, and clinical measures, such as drug use, have seen mixed outcomes.

A significant limitation of the extant literature is the relatively small samples typically evaluated (i.e., 20-40 subjects per condition). It is likely that the effects of cognitive training are of a small-to-medium effect size, however, the relatively low costs of training procedures mean that small effect sizes may not preclude clinical utility. It is also likely that individual characteristics will moderate the utility of these interventions for impacting substance use or related negative health behaviors. The use of small samples precludes testing these types of moderators with adequate statistical power as does the recruitment of comparably homogeneous samples (e.g., student samples, individuals from a single addiction clinic).

An emerging method addressing these concerns is crowdsourcing. Crowdsourcing, such as on Amazon.com's Mechanical Turk (mTurk), allows for the effective and efficient sampling of diverse subjects. We have previously demonstrated a close correspondence between in-person and online findings using mTurk in substance use research providing support for the validity of the resource. We have also recently demonstrated feasibility, acceptability, and validity of conducting intensive longitudinal research related to substance use on this platform. Demonstrating the feasibility and acceptability of applying cognitive training methods through mTurk would help establish this setting for future large sample studies testing novel interventions and/or individual characteristic moderating intervention efficacy.

The purpose of this study is to test the feasibility and acceptability of delivering cognitive training over mTurk. Subjects will be randomized to a 1) inhibitory control training condition, 2) working memory training condition, or 3) control training condition. Recent studies have also demonstrated the feasibility and potential efficacy of delivering brief normative feedback to reduce alcohol consumption through mTurk. In these brief interventions, subjects are provided information about their drinking compared to their same age and gendered peers. Approximately half of the subjects in each cognitive training group will receive normative feedback to evaluate effects on alcohol consumption and possible interactions with cognitive training. This study will focus on alcohol use given the ease and clinical acceptance of alcohol use self-report as a primary outcome.