1060 — Randomized Controlled Trial of a Relational Agent to Reduce Risky Alcohol Use

Lead/Presenter: Steven Simon,  COIN - Bedford/Boston
All Authors: Simon SR ((Center for Healthcare Organization and Implementation Research, Boston; Harvard Medical School)), Livingston N (Mental Health Service, VA Boston Healthcare System), Sawdy M (Center for Healthcare Organization and Implementation Research, Boston) Yeksigian C (Center for Healthcare Organization and Implementation Research, Boston) Bickmore T (College of Computer and Information Science, Northeastern University) Zhou S (College of Computer and Information Science, Northeastern University) Saitz R (Boston University School of Public Health) Kressin N (VA Boston Healthcare System and Boston University School of Medicine) Rubin A (Center for Healthcare Organization and Implementation Research, Boston)

Objectives:
Risky drinking remains a significant issue among Veterans. Although the efficacy of brief intervention and referral to specialty care for risky drinking is well established, implementation of these approaches often falls short. "Relational Agent" (RA) intervention technologies can mimic live human interactions and may offload PCP burden while providing intervention support to Veterans drinking above guidelines. In this study, we conducted a randomized controlled trial to evaluate a Relational Agent intervention programmed to provide brief intervention and referral to treatment for Veterans identified in primary care whose alcohol consumption exceeded guideline-recommended levels.

Methods:
We randomized 179 Veterans who screened positive on the Alcohol Use Disorders Identification Test (AUDIT-C) to Treatment as Usual (TAU) vs. TAU+RA intervention conditions. The RA consisted of two 15-minute interactive and motivationally-based sessions, with the first session occurring during their primary care visit and the second session one month later. Outcome variables included change in drinking from baseline to three months, and rates of referral to specialty care for alcohol treatment.

Results:
Veterans were similarly distributed with respect to age, sex, race/ethnicity, employment, and presence of VA-recognized disability. Using generalized multilevel mixed modeling, we found that drinking trajectories were similar between conditions; however, negative consequences associated with drinking decreased more rapidly among TAU+RA Veterans, b = -0.24 , p = 0.02. Additionally, 26 subjects in RAI (29%) were referred for specialty treatment, compared with 1 in the TAU (1%; p < 0.05).

Implications:
In this primary care-based RCT, a tailored RA intervention engaged Veterans and resulted in decreased drinking consequences and increased referrals for specialty treatment of alcohol-use disorder. RA interventions, shown to be effective in various health care settings, may be a useful tool to incorporate in primary care; this approach may be tested through other channels, such as web-based platforms and mobile devices, for broader impact on risky drinking and other sensitive health conditions.

Impacts:
Relational Agent interventions can expand intervention reach to Veterans drinking above guidelines, who might not otherwise receive alcohol intervention. Relational Agents can also offset PCP workload while facilitating Veteran referrals to specialty care for alcohol use-disorder treatment.