Discrimination and calibration properties of violence risk scale scores as a function of indigenous Canadian heritage in a multisite forensic-correctional sample

Abstract

The present study examined the discrimination and calibration properties of Violence Risk Scale (VRS; Wong & Gordon, 1999-2023) risk and change scores in a predominantly adult male, combined sample of Indigenous (n = 439) and non-Indigenous, White majority (n = 597) persons with conviction histories for violent offenses; approximately two thirds of whom completed risk-need-responsivity based violence reduction treatment services. Indigenous men tended to score higher on VRS static, dynamic, and total scores and to be classified as higher risk; however, there were no differences between the groups in treatment change. In the aggregate sample, VRS total scores demonstrated broadly medium to large effects in the prediction of violent and general recidivism (median AUCs = .72 [Indigenous] and .71 [non-Indigenous]) across ethnocultural groups. Conversely, VRS change scores (controlling for pretreatment score) were significantly associated with decreased violent and general recidivism for Indigenous persons (median AUC = .62) but considerably less so, with small or lower effects, for non-Indigenous persons (median AUC = .48). These results were upheld when effect sizes were aggregated across the samples through meta-analysis. Calibration analyses demonstrated that integrating risk and change information via logistic regression modelling decreased disparities between ethnoracial groups in rates of recidivism associated with VRS scores. Implications for violence risk assessment, treatment, and management using the VRS with Indigenous persons who have a history of criminal violence are discussed.