functions differed between these two groups. In the third model we added demographic, treatment and extra treatment variables in that order, to determine a final model with a best fit in predicting relapse. After finding our best fit model, we ran additional models with interactions between status at 1 year and the other covariates to test for moderation by non-problem use. Effects of the initial treatment (day hospital versus outpatient) and randomization (randomized versus not) were also tested in each model. These two variables did not significantly (p < .05) predict relapse and their inclusion did not change the pattern of results, so they were not included in the final models.
