The validity and reliability of the violence risk scale-youth version (VRS-YV)
The present study examines the validity, reliability, and psychometric properties of a newly developed violence risk assessment and treatment planning measure for youth – the Violence Risk Scale-Youth Version (VRS-YV; Lewis, Wong, & Gordon, 2004). Composed of 4 static and 19 dynamic items, the VRS-YV is designed to assess violence risk, identify targets for treatment, and evaluate changes in risk as a function of treatment. Change is evaluated through a modified application of Prochaska et al.’s (1992) Transtheoretical Model of Change. Stages of change ratings made pre- and post-treatment are summed across dynamic items to yield change scores. The VRS-YV, Youth Level of Services/Case Management Inventory (YLS/CMI; Hoge & Andrews, 2003), and Psychopathy Checklist-Youth Version (PCL-YV; Forth, Kosson, & Hare, 2003) were each rated from file information on a sample of 133 young offenders (68 males and 65 females) who had received assessment and/or treatment services from a community mental health facility in Saskatoon, Saskatchewan. All youths had been charged or convicted of a violent offense. This tended to be a rather high risk sample with a large proportion of Aboriginal youths. The VRS-YV demonstrated good internal consistency (VRS-YV total α = .91) and interrater reliability (VRS-YV total ICC = .90), while most of the individual items had acceptable inter-item (mean r = .32) and item total correlations (range r = .30 to .70). Male and female youths displayed few differences on the three risk measures or their respective scale components; however, Aboriginal youths scored significantly higher on these measures than non-Aboriginal youths, with the trend being particularly strong among males. The VRS-YV showed good convergence with the YLS/CMI and PCL-YV. The three measures significantly postdicted violent offending, that is, youth who were repeat violent offenders tended to score significantly higher on each of the measures, than first time violent youth (i.e., those who had no previous history of violence). Similar postdiction was observed for general criminal offending. Recidivism data were available for roughly half of the total sample (n = 62) over a mean follow-up time of approximately 2 years. Preliminary evidence was obtained for the predictive accuracy of the VRS-YV with respect to violent and general recidivism. Predictive accuracy statistics were comparable to those obtained for the YLS/CMI and PCL-YV, with correlations generally being in the .40 to .50 range and Areas Under the Curve (AUCs) in the mid .70s to low .80s. There was also a significant relationship between VRS-YV risk level and both violent and general recidivism. Survival analyses further confirmed that the VRS-YV was able to differentiate those who were more likely to recidivate and more likely to do so more quickly. Change ratings were available for a small subsample of youth (n = 39), which were used to compute post-treatment dynamic ratings. Youths appeared to demonstrate some degree of change, indicating possible therapeutic progress after receiving treatment services. While the trends for the change results tended to be in the expected direction across several of the analyses, the small nature of the sample precluded meaningful interpretation of these findings. In sum, these data provide preliminary evidence for the ability of the VRS-YV to evaluate risk and predict violent and general recidivism with comparable accuracy to that of two well-known and psychometrically robust instruments in the field. The results further demonstrate that the VRS-YV, YLS/CMI, and PCL-YV can have predictive validity for future violent and general recidivism among a diverse sample of youth that includes both male and female, Aboriginal, and community-based youth, living in the province of Saskatchewan. Limitations and future directions are discussed.
treatment, violence risk assessment, young offenders
Doctor of Philosophy (Ph.D.)