All HBV vaccine completion analyses were intent to treat. Preliminary analyses, including chi-square and t tests and analysis of variance, were used to assess comparability of the three programs at baseline and to examine unadjusted correlates of vaccine completion. Multiple logistic regression modeling was used to assess program effects on treatment completion, controlling for potential confounders. Stepwise backward multiple logistic regression analysis was used to create a model of treatment completion; predictors included variables that were associated with vaccine completion at the .15 level in the preliminary analyses.