The analyses were conducted in three stages. First, the socio-demographic characteristics, alcohol use patterns, psychiatric and substance use comorbidity, family history and treatment for mood symptoms were compared across participants who did or did not develop alcohol dependence, as well as for those who did or did not have persistence of alcohol dependence. Second, these same characteristics were compared across participants by report of alcohol self-medication for mood symptoms at baseline separately for the study samples used to assess each outcome. These analyses were conducted to identify potential confounders and variables to be included in computation of propensity score weights described below. In the third stage of the analyses, the association of self-medication with each alcohol dependence outcome was assessed in the total sample and in strata based on socio-demographic characteristics, mood disorder diagnosis, and treatment history to examine whether the effect was more pronounced in some population subgroups than others. These analyses were conducted using both bivariate and multivariate models. The multivariate logistic regression models adjusted for socio-demographic, psychopathology, substance use, and treatment variables and whether or not the
