Study 1
Carr, Baker, and Cassidy (2016) conducted a quasi-experimental study to examine the effectiveness of the Community Reporting Engagement Support and Training (CREST) enhanced day reporting center (DRC) in New Haven, Conn., on recidivism outcomes of probationers with mental illness. The treatment group consisted of all CREST clients on probation who were admitted and discharged between July 1, 2007, and July 1, 2012 (n = 188). The control group consisted of all individuals on probation in the same geographical district as the treatment group between Jan. 1, 2006, and Jan. 1, 2009. To be eligible for inclusion in the control group, probationers had to have never been referred to CREST and have a mental illness (measured by a score of 15 or above on the Adult Substance Use Survey–Revised [ASUS–R]) Mood Disruptions scale. This resulted in a control group of 425 individuals.
The average age of the treatment group was 35.7 years old, and the participants were mostly male (82 percent). More than half of the treatment group was Black (59 percent), 23 percent were White, 15 percent were Hispanic, and 3 percent had race/ethnicity that was not disclosed. The primary diagnosis was of a psychotic disorder such as schizophrenia (60 percent), followed by bipolar disorder (18 percent). Most treatment group participants (78 percent) were diagnosed with a comorbid substance use disorder in addition to their psychiatric diagnoses and were previously convicted of a felony offense (87 percent). The average age of the control group was 33.6 years, and the participants were mostly male (65 percent). Of the control group, 44 percent were Black, 31 percent were White, and 25 percent were Hispanic. Most participants in the control group were previously convicted of a felony (81 percent). Because the control group data were pulled from state probation records, clinical data (such as mental illness diagnosis) were not available.
Propensity score matching was conducted to control for differences between the treatment and control groups in gender, age, race, ASUS–R Mood score, and previous convictions. Because the original control group was larger than the treatment group, a 1:2 match ratio was set (i.e., two control group participants were matched to each treatment group participant). After accounting for missing data, the propensity score matching resulted in a final sample of 81 participants in the treatment group and 146 participants in the control group (n = 227). There were no statistically significant differences between the treatment and control groups on baseline characteristics.
Outcomes of interest included any new conviction and new felony conviction following admission to CREST (for the treatment group) or probation supervision (for the control group). Data were collected from the state probation department records. Owing to the varying length of follow-up periods, a time-to-event Cox proportional hazard regression was performed using time to any conviction and time to a felony conviction as dependent measures, and CREST status as the independent variable. Subgroup analyses were conducted to examine the impact completion of the program made on treatment group participants.