Evaluating the Predictive Validity of the COMPAS Across Psychiatric Diagnoses
Abstract
The Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) is a widely used actuarial risk assessment instrument in correctional settings across the United States. While several studies have explored the validity of the COMPAS, there is a dearth of research evaluating the performance of the COMPAS among persons with mental health disorders. In fact, only one published report has examined the predictive validity of the COMPAS using a sample of mental health court participants with primarily SMI diagnoses. This is especially concerning considering evidence of the over-representation of mental illness among correctional populations. The current study sought to address this gap in the literature by examining associations between COMPAS risk scores and recidivism among a sample of recently released inmates with mental health disorders (N = 123). Specifically, this study evaluated potential differences in the predictive validity of the COMPAS across individuals with differing mental health and substance use disorders. This study drew data from a sample of post-incarcerated offenders referred or mandated by NY State for reentry services. Outcome data consisted of any new arrest during the 12 months following admission to the program. Statistical analyses included independent samples t-tests, ANOVAs, and chi-square tests. To test whether mental health and substance use diagnoses moderated the relationship between COMPAS risk scores and recidivism, this study included a series of multiple logistic regressions with diagnosis and COMPAS scores as predictors. Finally, the implications of this research were discussed as well as suggestions for future research.
Subject Area
Clinical psychology
Recommended Citation
Collins, Aidan, "Evaluating the Predictive Validity of the COMPAS Across Psychiatric Diagnoses" (2022). ETD Collection for Fordham University. AAI29210931.
https://research.library.fordham.edu/dissertations/AAI29210931
