Prediction of Alcoholics Anonymous affiliation from pretreatment factors : methods and theory to inform provider referrals

Introduction: Treatment providers commonly emphasize and teach Alcoholics Anonymous (AA) philosophies and beliefs in treating alcohol disorders. The high rate of post-treatment relapse into harmful drinking suggests a need for extended care beyond initial treatment, and practitioners commonly refer clients to AA for aftercare. Although An abundance of correlational research suggests that greater AA affiliation is moderately and consistently associated with improved drinking outcomes, the vast majority of those who try AA drop out or substantially curtail affiliation ("disaffiliation") within a year.Despite this, AA generally is believed to represent an affordable and effective adjunct and extension of initial treatment. Objective: The purpose of this study is to predict AA affiliation using information available at treatment and to advance knowledge surrounding affiliative processes. Literature from previous research and theory is used to develop hypotheses and a model of affiliation involving domains of motivation, coping skills and cultural fit. Methods: Using the Project MATCH data, two classification methods (binary logistic regression and artificial neural networks) are used to test hypotheses and predict AA affiliation at one-year post-treatment. Affiliation was operationalized to include dimensions of (1) meeting attendance or exposure, and (2) level of involvement or participation in AA activities. The affiliation construct was measured using the Alcoholics Anonymous Involvement scale (AAI) administered at 12 months post-treatment (N = 1,506). The study attempts to optimize classification accuracy on a dichotomous response variable that includes "disaffiliates" or "moderate to high affiliates" class membership. Results: Significance tests of 58 predictor variables suggested that the pretreatment AAI, divorced and separated marital statuses, age.treatment assignment, treatment site (representing inpatient or outpatient subjects),guilt/worry surrounding drinking and religiosity are significant predictors of affiliation.Prior affiliation (the pretreatment AAI) is a good predictor of affiliation. Despiteconsiderable prior theory and empirical evidence, motivation, severity, self-efficacy and external help-seeking measures were not significant predictors. Results are explained using theories borrowed from the organizational culture and climate literature. Attempts To predict affiliation were moderately successful (kappa = .42, sensitivity = 74%,specificity = 68%).