Objective: Reviews of research on Alcoholics Anonymous (AA) have speculated how findings may differ when grouped by client and study characteristics. A meta-analytic review by Emrick et al. in 1993 provided empirical support for this concern but did not explore its implications. This review divided results of AA affiliation and outcome research by sample origin and global rating of study quality. The review also examined the statistical power of studies on AA.
Method: Meta-analytic procedures were used to summarize the findings of 74 studies that examined AA affiliation and outcome. Results were divided by whether samples were drawn from outpatient or inpatient settings and a global rating of study quality that jointly considered use of subject selection and assignment, reliability of measurement and corroboration of self-report. Efficacy of dividing study results was examined by changes in magnitude of correlations and unexplained variance.
Results: AA participation and drinking outcomes were more strongly related in outpatient samples, and better designed studies were more likely to report positive psychosocial outcomes related to AA attendance. In general, AA studies lacked sufficient statistical power to detect relationships of interest.
Conclusions: AA experiences and outcomes are heterogeneous, and it makes little sense to seek omnibus profiles of AA affiliates or outcomes. Well-designed studies with large outpatient samples may afford the best opportunity to detect predictors and effects of AA involvement.