The first study by Martin, Butzin, and Inciardi (1995) used a quasi-experimental design to examine the differences among 457 subjects, recruited between the summer of 1990 and the spring of 1994, who participated in the following conditions:
- Primary treatment in prison at KEY program with no subsequent treatment (KEY group, n=43)
- Primary treatment in prison at KEY program and secondary (transitional) treatment at CREST Outreach Center (KEY-CREST group, n=32)
- Primary treatment at CREST with no prior prison-based treatment (CREST, n=176)
- No treatment other than HIV/AIDS prevention education (comparison, n=206)
Subjects in the KEY-CREST group were not assigned randomly, but by prison personnel who determined eligibility based on institutional classification screening, criminal history, and correctional counselor interviews.
The CrimeSolutions.gov review of this study focused on the comparisons between the KEY-CREST group and the no- treatment (other than HIV/AIDS prevention education) comparison group.
The sample was mostly male (81 percent), African American (72 percent), with the remaining participants mostly white. There were a few statistically significant differences between the groups at baseline. The KEY-CREST group included more African American participants who, on average, had longer criminal histories. Also, all of the KEY-CREST subjects were male, as there was no therapeutic community available for female inmates at the time. There were no other statistically significant differences between the treatment and comparison group on baseline characteristics.
Subjects were interviewed just prior to leaving prison and again about 6 months after release. Self-reported measures were collected on sexual activities, criminal history, drug abuse treatment history, psychosocial and mental health status, and sociodemographics. The two outcome measures of interest were intended as a rough index of risk of HIV infection. One measure was the use of any injection drugs since release, and the other measure was report of condom use. The primary mode of statistical analysis was logistic regression, with additional chi-square analyses.
Two other outcomes of interest, drug-free and arrest-free, were measured at a later timepoint (Martin et al. 1999). The study authors did not conduct subgroup analyses.Study 2
A follow-up study by Martin and colleagues (1999) looked at the outcome results 1 year and 3 years after participants left work release. The study examined the same four groups as the previous study by Martin, Butzin, and Inciardi (1995). The study authors also looked at results from participants who received no in-prison treatment, were assigned to regular work release, were CREST dropouts, were CREST graduates without aftercare, and were CREST graduates with aftercare. However, the CrimeSolutions.gov review of this study focused on the comparisons between the initial KEY-CREST group (n
=68) and the initial comparison group (n
The comparison group was predominately male (82 percent), 68 percent African American, 27 percent white, and 4 percent Hispanic. Their average age was 29 years. The KEY-CREST group was 74 percent male, 85 percent African American, and 13 percent white. Their average age was 31.
The outcome measures of interest were relapse (drug-free) and recidivism (arrest-free). Each outcome variable combined information from repeated self-report and official data. To be considered drug-free, the study participant must have reported no illegal drug use and tested negative for drugs on the urine screen at every follow-up point. Similarly, arrest-free was measured as no self-reports of arrest and no official arrest record for new offenses since release from prison (excluding parole violations). The study authors relied on bivariate and multivariate logistic regression to analyze the data and did not conduct subgroup analyses of KEY-CREST participants.