presents baseline demographic and clinical characteristics of the sample divided by psychiatric severity. The following variables differed significantly by severity groups: DSM-IV alcohol dependence, DSM-IV cocaine dependence, treatment study, treatment site, ethnicity, and the following ASI composite scores: medical, alcohol, drug, and family, p < .05.
Baseline variables by psychiatric severity level.
As presented in , evidence for the criterion validity of the psychiatric severity groupings is supported by significant group differences on history of psychiatric medication prescription, past year inpatient psychiatric hospitalization, and lifetime history of suicide attempt, p < .05. Also, current psychological distress (BSI-GSI scores) differed significantly between the groups, p < .001. Tukey’s post-hoc testing revealed all three groups differed significantly from each other, with participants in the low severity group reporting the least distress and the high severity participants reporting the greatest distress, p < .001.
An ANCOVA was performed with weeks retained in treatment as the dependent variable. As shown in , treatment condition was significantly associated with retention, p < .001. The adjusted weeks retained in treatment (± standard error [SE]) was 7.4 (0.3) for CM and 5.5 (0.4) for standard treatment. A significant interaction of treatment condition by psychiatric severity was also found for weeks retained in treatment, F(2,378) = 4.04, p < .05. As seen in , as psychiatric severity increases the number of weeks retained in standard treatment decreases; conversely, as psychiatric severity increases, the number of weeks retained rose slightly for patients assigned to CM. No other independent variables were significantly associated with retention, p > .05. Although not shown, we also tested a model in which the continuous variable, ASI baseline psychiatric composite score, was substituted for the ordinal psychiatric severity variable. The results of this analysis were consistent with the previous analysis with a significant interaction between psychiatric severity and treatment condition, F(1,380) = 8.01, p < .01.
Analysis of Covariance of weeks retained in treatment (n = 392).
Weeks retained in treatment (± standard error) by psychiatric severity group and treatment condition.
Consistent with group differences in retention, the total number of urine/breath samples provided differed between standard and CM treatment conditions, F(1,380) = 16.8, p < .001, and was on average (+ SE) 9.3 (0.5) and 11.9 (0.4), respectively. However, the number of samples provided did not differ significantly by psychiatric severity groups, F(2,380) = 0.7, p = .46. The mean number of samples provided (± SE) was 11.0 (0.6), 10.1 (0.6), and 10.7 (0.6) for participants in the low, moderate, and high psychiatric severity groups, respectively.
Sequential binary logistic regression investigated the relationship between psychiatric severity and treatment condition with LDA (0–7 weeks abstinence vs. 8–12 weeks). Study, gender, age, ethnicity, annual income, and DSM-IV alcohol and cocaine dependence were entered in the first step, and then psychiatric severity, treatment condition, and the treatment condition by psychiatric severity interaction were entered in the second step. The assumption of predictor variable linearity was tested via Box-Tidwell approach with age and annual income, and neither were found to violate this assumption. The Hosmer-Lemeshow goodness-of-fit statistic indicated that the model fit was adequate, χ2(8, n = 392) = 1.53, p = .992.
provides a summary of the final model for LDA. Both steps of variables entered were significant (Step 1: χ2
(5) = 18.93, p
= .002; Step 2: χ2
(5) = 30.95, p
= .001) and the model correctly predicted classification group (i.e., 0–7 weeks LDA and 8–12 weeks LDA) in 72.2% of the sample. Study and treatment condition were the only variables significantly associated with LDA, p
< .05. Participants in the Petry et al. (2005b)
study were more likely to achieve 8 or more week of continuous abstinence than those in Petry et al. (2004
studies. Participants in the CM conditions were more likely to achieve 8 or more weeks of abstinence than those in the standard treatment condition (see ). Although treatment condition had a substantial impact on achieving longer durations of abstinence, there was no effect of psychiatric severity on LDA. These analyses were re-run twice, first omitting the CM group that was not reinforced for abstinence (from Petry et al., 2006
), and second using a median split of LDA (0–4 weeks abstinent vs. 5–12 weeks abstinent). In both cases, results were similar with study and treatment condition as the only significant variables associated with LDA, p
< .05 (data not shown).
Logistic regression: Eight or more consecutive weeks of objectively confirmed abstinence during treatment (n =392).
Figure 2 Percentage of participants achieving eight or more consecutive weeks of abstinence by psychiatric severity group and treatment condition. Note: The n provided is number of participants within the psychiatric severity by treatment condition group associated (more ...)
An ANCOVA was performed with percent of abstinent samples provided out of the total possible as the dependent variable. Independent variables significantly associated with percent of abstinent samples provided were study, F
(2,378) = 16.6, p
< .001, and treatment condition, F
(1,378) =12.6, p
< 001. The adjusted percent of samples abstinent (± SE) was 49.2% (0.02) for CM and 37.8% (0.03) for standard treatment. For study, the adjusted percent of samples abstinent (± SE) was 33.3% (0.03) for Petry et al. (2004)
, 54.3% (0.03) for Petry et al. (2005b)
, and 42.8% for Petry et al. (2006)
. Psychiatric severity was not associated with percent of abstinent samples provided, F
(2,378) = 1.89, p
= .152, nor was the interaction between psychiatric severity and treatment condition, F
(2,378) = 1.47, p
Within the CM treatment conditions that reinforced goal-related activity completion (N = 5 of 6 CM conditions), we evaluated the relation of psychiatric severity on number of activities completed. The same independent variables were included in the analyses. Treatment study was related to number of activities completed, F
(2,222) = 11.60, p
< .001, with Tukey’s post-hoc testing finding that participants in Petry et al. (2005b)
completed more treatment-related activities than participants in Petry et al. (2004)
and Petry et al. (2006)
. Psychiatric severity was not significantly related to number of treatment-related activities completed, F
(2,222) = 0.57, p
= .557, which was on average (± standard error) 14.9 (1.63), 15.1 (1.81), and 17.0 (1.73) for the low, moderate, and high severity groups, respectively.
Finally, the total amount of CM reinforcement earned for completion of target behaviors (Log 10 transformation) was investigated across the psychiatric severity groups. Amount of CM reinforcement did not differ by psychiatric severity, F(2, 269) = 2.31, p = .101, with untransformed mean dollar amount (standard deviation) of $171.34 ($231.35), $208.98 ($277.67), and $187.66 ($267.72) for the low, moderate, and high severity groups, respectively.