The NESARC used a complex sampling design. Analyses used the svy routines of STATA 11.0 to take into account survey weights, clustering and stratification of the data. The propensity score weights described earlier were multiplied by the survey weights and the resulting combined weights were used in analyses of the association of self-medication with alcohol dependence during follow-up. To assess the effectiveness of the IPTW in balancing the composition of the self-medicating and non-self-medicating groups, we compared the characteristics of the groups before and after applying the combined weights. Application of inverse probability of treatment weighting in these analyses was quite successful as the groups in analyses for both alcohol outcomes (incidence of alcohol dependence and persistence of alcohol dependence) were similar with respect to the observed characteristics after using the weights.44 Population attributable fraction (PAF) was computed using the punaf program for STATA software.45 The program implements the method for estimating PAFs as recommended by Greenland and Drescher 46 for cohort and cross-sectional studies.
