Risk assessment for domestic violence offenders: predicting probation outcomes

Risk assessment instruments play a significant role in correctional intervention and guide decisions about supervision and treatment. Although advances have been made in risk assessment over the past 50 years, limited attention has been given to risk assessment for domestic violence offenders. This study investigates the use of the Domestic Violence Screening Inventory (DVSI) and the Offender Screening Tool (OST) with a sample of 573 offenders convicted of domestic violence offenses and sentenced to supervised probation in Maricopa County, Arizona. The study has two purposes. The first is to assess the predictive validity of the existing assessment tools with a sample of domestic violence offenders, using a number of probation outcomes. The second is to identify the most significant predictors of probation outcomes. Predictive validity is assessed using crosstabulations, bivariate correlations, and the Receiver Operating Characteristic (ROC) curve. Logistic regression is used to identify the most significant predictors of probation outcomes. The DVSI and the OST were found to be predictive of probation outcomes and were most predictive of the outcomes petition to revoke filed, petition to revoke filed for a violation of specialized domestic violence conditions, and unsuccessful probation status. Significant predictors include demographics, criminal history, current offense, victim characteristics, static factors, supervision variables and dynamic variables. The most consistent predictors were supervision variables and dynamic risk factors. The supervision variables include being supervised on a specialized domestic violence caseload and changes in supervision, either an increase or decrease, during the probation grant. The dynamic variables include employment and substance abuse. The overall findings provide support for the continued use of the DVSI and the OST and are consistent with the literature on evidence-based practices for correctional interventions. However, the predictive validity of the assessments varied across sub-groups and the instruments were less predictive for females and offenders with non-intimate partner victims. In addition, study variables only explained a small portion of the variation in the probation outcomes. Additional research is needed, expanding beyond the psychology of criminal conduct, to continue to improve existing risk assessment tools and identify more salient predictors of probation outcomes for domestic violence offenders.