The first hypothesis questions whether prison-based treatments (collectively) were associated with positive changes over time in methamphetamine-using inmates. To address this, we tested whether the means calculated for the total sample on each of the motivation, psychosocial functioning, and criminal thinking scales (from the CJ CEST and CTS) had changed significantly by the end of primary treatment (see the right-hand column of ). All tests were significant as evaluated in the multi-level analysis. However, because of the large sample size (N = 2,026), effect sizes also were examined. Based on Cohen's D index (J. Cohen, 1988
), the changes in Desire For Help (D = .19) and Treatment Readiness (D =.12) would be considered to be “small” (as they had a D of .20 or less), whereas changes in Risk Taking (D = .36), Hostility (D = .47), and Depression (D = .49) fell into the “small to medium” range (with D sizes greater than .20 but less than .50). The Anxiety (D = .50), Decision Making (D = .66), and Self Esteem (D = .70) changes were in the “medium to large” effect size range.
Intake to Treatment LS Means on Client Evaluation of Self and Treatment Scales for Indiana Male Methamphetamine User Inmates (N = 2026)
Overall improvements in criminal thinking orientation from intake to the end of treatment phase also were significant, but their effect sizes were smaller than those found for the psychosocial scales above. The effect size for Cold Heartedness was small (D=.13), while the remaining CTS scales were between “small” and “medium”. These include Entitlement (D = .28), Justification (D = .35), Personal Irresponsibility (D = .31), Rationalization (D = .23), and Power Orientation (D = .26). Thus, decreases on the motivation scales were “small,” whereas positive changes for the set of psychosocial functioning were generally in the “medium to large” range and criminal thinking scales were “small” to “medium.”
It should be noted that changes in motivation scales were actually “decreases” from their very high intake score values. These measures are designed to assess initial desire, pressures, and readiness to enter a treatment program. After client engagement in treatment, however, the meaning of these scales shift by virtue of how motivation constructs are expressed therapeutically. While they serve as significant pretreatment predictors of subsequent treatment progress (e.g., Simpson, 2004
), they have limited and only specialized value as during-treatment measures and therefore are not included below in further discussions about change. Reporting of these measures here, however, demonstrates that the overall trend toward “positive” changes in other scales is not a global or indiscriminate response to all measures used.
Another aspect of multi-level analyses is that it provides information on the uniformity of results across the sites studied. That is, it provides a test statistic for whether knowledge of the site identity is needed for prediction of the outcome. The between-site variance for the intercept was significant for two of the six CJ CEST scales [Anxiety (Z = 2.32, p < .01), Hostility (Z = 2.24, p < .02)] and three of the six CTS scales [Entitlement (Z = 1.86, .04), Personal Irresponsibility (Z = 1.69, p < .05), and Cold Heartedness (Z = 1.85, p < .04)]. In other words, differences in the magnitude of change over time among program sites for these measurements cannot be attributed simply to random sampling variations for these five scales. This was not unexpected because inmates were not randomly assigned to treatments; however, the number of scales that was significant is relatively small compared to the number of outcomes analyzed, and only two of the Z-statistics exceeded 1.96. This suggests a degree of uniformity in the changes occurring across multiple sites.
3.1. Changes within treatment programs
Having evidence for “overall improvements” across all treatments is useful, but it is the effectiveness of each of the three treatment approaches that is of greater interest. Results summarized in show indeed that changes on the CJ CEST and CTS scales from intake to end of treatment were significant for all three program types. Changes on the psychological measures were substantial, falling in the “moderate” (Depression and Anxiety) and “large” (Self Esteem and Decision Making) ranges for each treatment. This also was true for social functioning where the changes for Hostility and Risk Taking scores were in the “moderate to large” range.
Effect sizes for the CTS changes within the three treatment programs were smaller. For OTP, all of the changes were closer to “small” than to “moderate,” while for Modified-TC most of the changes were in the “moderate” range. The changes for CLIFF-TC lay in between, with only three effect sizes larger than .30.
3.2. Treatment comparisons after covariance adjustments
Although treatment comparisons reported above might be suggestive about how treatments are differentially effective, they are limited by the fact that their inmates were not equated or randomly assigned. To adjust for these program-level differences, a multi-level covariate analysis model was applied to test for “adjusted” treatment group differences on CJ CEST and CTS means measured at the end of treatment (). Age, race, the Client Problem Profile Index (CPPI) composite measure, and the corresponding CJ CEST intake measure were used as covariates in the analysis of each outcome.
Least Square CEST and CTS Means at End of Treatment PhaseAdjusted for Demographics and CPPI
shows that for treatment comparisons on psychological functioning, only Self Esteem and Decision Making were significant, with inmates in Modified-TC scoring significantly higher than those in either CLIFF-TC or OTP. More importantly, inmates in both CLIFF-TC and Modified-TC were significantly higher at the end of the treatment phase than inmates in OTP on Decision Making. CLIFF-TC inmates also were more improved (i.e., significantly lower) than inmates in OTP on Risk Taking.
For the treatment engagement measures of Treatment Satisfaction, Counselor Rapport, and Treatment Participation, the results varied considerably across the treatment programs. OTP inmates were comparatively more “satisfied” with treatment services, but Modified-TC inmates reported having better Counselor Rapport. Both CLIFF-TC and Modified-TC inmates had higher Treatment Participation scores than those in OTP.
For the CTS scales, significant between-treatment differences were found for Justification, Personal Irresponsibility, Rationalization, and Power Orientation, with inmates in OTP having significantly higher means (i.e., showing more criminal thinking problems) than in the two TC groups. Effect sizes corresponding to the overall treatment differences for the CJ CEST and CTS analyses showed these to be in the small range, similar to previous findings. That is, these score differences between the groups, while significant for most of the scales, tended to be small.
3.3. Inmate attributes as predictors of during-treatment outcomes
Inmate attributes were organized into three general areas – background problems (CPPI), demographics, and initial criminal thinking – for analyses of their relationships with end-of-treatment functioning. None of the covariates representing background (CPPI) and demographic (age and race) variables was consistently predictive of every dependent variable. Of these, CPPI was the most prominent in that it was significant for nine of the dependent variables (Self Esteem, Depression, Anxiety, Decision Making, Hostility, Social Support, Justification, Personal Irresponsibility, and Power Orientation). Direction of the relationships was positive for Depression, Anxiety, Hostility, Justification, Personal Irresponsibility, and Power Orientation, and negative for Self Esteem, Decision Making, and Social Support. That is, more background problems were associated with more psychological issues and higher criminal thinking, but also with lower functioning on self esteem, decision making, and social support.
The associated regression weights show older inmates tended to have lower scores on Self Esteem, Anxiety, Hostility, Risk Taking, Social Support, and Power Orientation. Whites were more likely than non-Whites to have higher levels of Anxiety, Risk Taking, and Entitlement, but lower scores for Justification, Personal Irresponsibility, Rationalization, and Cold Heartedness.
To address the relationship of criminal thinking as an inmate pretreatment attribute affecting treatment progress outcomes, a composite measure of initial criminal thinking (computed as the overall mean of the six criminal thinking scales) was used as an additional covariate in the previous multi-level model. This index was significant for almost every end-of-treatment CJ CEST measure of psychosocial functioning and engagement (Self Esteem, Depression, Anxiety, Decision Making, Hostility, Risk Taking, Treatment Satisfaction, Counselor Rapport, Treatment Participation, Peer Support, and Social Support). That is, criminal thinking (i.e., scored in this analyses as a simple composite index) was positively related to the psychosocial issues of depression, anxiety, hostility, and risk taking, but negatively to decision making and treatment process indicators. Perhaps most importantly in regard to complications for treatment, its strongest relationships were with the treatment engagement measures.
3.4. Treatment engagement effects as a “mediator” of inmate improvements
With pretreatment attributes being related to their end-of-treatment phase functioning, and with during-treatment engagement being related to them as well, attention turned to the possible mediating effects of therapeutic engagement. Thus, the Counselor Rapport and Treatment Participation scale scores were averaged together and then added as another covariate in the previous analytic model in order to address its effects on end-of-treatment psychosocial functioning. These results were very telling as shown in , under Model 2. This treatment engagement composite measure was highly related in a positive way to better inmate functioning scores at the end of treatment for all three programs. Moreover, adding treatment engagement to the model caused all of the previously significant relationships of initial criminal thinking with end-of-treatment functioning measures to become non-significant. That is, treatment engagement was found to compensate for the negative influences that stronger initial criminal thinking tended to have on psychosocial functioning progress during treatment. For the other inmate attributes studied, on the other hand, the inclusion of treatment engagement did not alter the relationships of demographics and background problems (as represented by the CPPI) with end-of-treatment measures. Moreover, treatment modality differences were still significant for Decision Making and Risk Taking. This is noteworthy as it reconfirms the treatment modality results noted previously, even when controlling for treatment engagement. So while treatment engagement explained away initial criminal thinking, it did not do so for the other inmate attributes included in the analytic model or for treatment effects.
Effects of Engagement as Covariate on Relationships of Patient Attributes to End of Treatment Outcomes