Multivariable stepwise regressions (Table 2) show that younger individuals were significantly more likely to be non-abstinent, and movement to the next oldest age category reduced the odds of non-abstinence by an average of 27%. Importantly, the confidence intervals were narrow and extremely similar across models, implying that the effect of age was robust to model specification. In regard to help-seeking and problem severity, having attended at least one 12-step meeting and the number of DSM-IV dependence symptoms were both significantly related to non-abstinence. In the fully saturated models, any twelve-step attendance decreased odds of non-abstinence by 57–76% (Model 4), while each additional DSM symptom decreased odds of non-abstinence by 73–89% (Model 4).
