UC Berkeley

Background

Alcohol continues to adversely affect the lives of Americans, particularly individuals suffering from addiction. The majority of treated alcoholics relapse to alcohol abuse or dependence. Between 66% and 80% of adults relapse in the six months after an episode of community- or hospital-based drug or alcohol treatment and 40% will re-enter treatment. Thus, the aftercare and follow-up plan have potential to affect long-term treatment success. Research that tracks treatment outcomes for alcohol addiction has shown that while a variety of treatment interventions are effective, the progress clients make in treatment is frequently undermined if they are surrounded by or reside in an environment that triggers relapse. Although individual-level risk factors for alcoholism have been well-established they do not fully explain variability in recovery suggesting that environmental and social factors need to be explored.

Neighborhood alcohol outlet density (AOD) and Alcoholics Anonymous (AA) affiliation are two environmental and social factors that show promise for intervention on a moderate time scale. Despite the contributions of research on how one’s living and social environment can affect alcohol use, very little is known about the impact of AOD and AA on recovery over time. To date, there are no studies examining the effect of AOD on relapse among alcoholics in recovery. Similarly, in spite of a vast body of literature on AA, few studies have examined the effectiveness of long-term affiliation with AA on relapse. Moreover, no studies have utilized parameters based on a causal inference framework to examine the potential impacts of these factors on relapse and recovery.

Methods

Using a 7-year prospective cohort study of alcoholics in recovery, the purpose of this dissertation was to estimate the effects of AOD and AA on relapse (past 30-day abstinence), applying improved analysis techniques. The most widespread statistical method in studies of AOD and AA associations with drinking rely on conventional regression. This approach does not appropriately adjust for time-dependent confounding, and the modeling assumptions may not always be met. An alternative approach is to estimate parameters motivated by the causal inference literature, which can be interpreted as estimates of the outcome under hypothetical interventions to the exposure of interest. In this framework, a key step is careful consideration of the assumptions necessary to interpret the parameter as a causal effect. The current work is stronger than past work with respect to some of the assumptions. In the first chapter, I estimate the longitudinal impact of AOD on abstinence using a parameter motivated by the causal inference literature. In the second chapter, I again examine the longitudinal impact of AOD on abstinence with a focus on specific types of alcohol outlets. In the third chapter, I examine the longitudinal impact of AA participation on abstinence. For all study questions, I use data-adaptive estimation (SuperLearning) combined with a recently released R package, Longitudinal Targeted Maximum Likelihood Estimation (ltmle), an estimation method that encourages an explicit process for specifying and estimating target parameters to address causal questions that specifically incorporate time-dependent confounders.

Significance

This work will contribute to epidemiologic research in several ways. First, we hope to begin to fill the gap in the literature on the association between neighborhood AOD and drinking among alcoholics in recovery. Second, we aim to determine whether specific alcohol outlet types confer distinct drinking risks among alcoholics in recovery. Third, we hope to contribute to the limited literature examining the long-term impact of AA on alcohol recovery. Moreover, this work represents the first application of ltmle to the field of alcohol epidemiology. We hope to demonstrate how it can provide a powerful way of estimating parameters with direct public health relevance using observational data. Extensions of this research can help to improve understanding of how environmental and social contexts contribute to alcohol recovery, and identify ways to optimize future interventions in this area. Conceptually, this work will contribute to efforts aimed at promoting recovery by examining to what extent AOD exposure and AA participation are associated with drinking among alcoholics over time, two areas that warrant further research. Understanding the interrelationships between neighborhood context, social network, and subsequent alcohol use is critical to better understand alcohol relapse and recovery.