NCJ Number
215973
Journal
Criminal Behaviour and Mental Health Volume: 16 Issue: 3 Dated: 2006 Pages: 155-166
Date Published
2006
Length
12 pages
Annotation
This study examined the level of predictive accuracy for general and violent postrelease reoffending when the following four risk-prediction instruments had low agreement on their risk predictions: the Psychopathy Checklist-Revised (PCL-R), Level of Service Inventory-Revised (LSI-R), Violence Risk Appraisal Guide (VRAG), and the General Statistical Information on Recidivism (GSIR).
Abstract
The findings show that predictive accuracy for reoffending is poor when there is low agreement among risk estimates. Disagreement among instruments reduced predictive accuracy for all the instruments except the GSIR. Disagreement had a moderating effect on both the PCL-R and LSI-R, but not on the VRAG and GSIR. Since the GSIR was the least susceptible to a lack of consensus among instruments, the authors pose a working hypothesis that instruments which rely on clinical judgment (PCL-R, LSI-R, and VRAG) are more susceptible to error when compared with instruments that rely on static, historical variables (GSIR). This hypothesis requires further testing. Study participants were 209 incarcerated adult males sentenced to 2 or more years in prison. The PCL-R, LSI-R, and VRAG were scored by trained masters and doctorate-level clinicians following an interview and file review. The GSIR was scored primarily by parole officers at the time the offender was admitted to custody. In a few cases, the GSIR was scored retrospectively by the authors. Instrument scores were determined in advance of release, so the raters were unaware of the reoffending outcome. Reoffending was recorded from official police records. 4 tables and 45 references