The Level of Service Inventory-Ontario Revision: A recidivism follow-up study within a sample of male young offenders

This study represents the first attempt to evaluate the predictive validity of the Level of Service Inventory-Ontario Revision (LSI-OR; Andrews, Bonta, & Wormith, 1995) on young offenders. The LSI-OR is a risk/need assessment measure designed to assess risk of recidivism and other offender problems, and to classify offenders for supervision and treatment (Andrews & Bonta, 1998). Although the scale has been used with young offenders, it has not yet been adequately validated with this population. The sample consisted of 154 16- to 18-year old, Caucasian, male, young offenders from Northern Ontario, who had been sentenced to custody or community (probation) terms. Following a 3.8-year follow-up period, 50.6% of the sample had been reconvicted, with 20.1% for a violent offence. Recidivists scored significantly higher than nonrecidivists on the LSI-OR total score and on six of the eight risk and need subscales. Correlations between the LSI-OR total score and general and violent recidivism were .46 and .31, respectively. ROC analyses indicated significant effects for LSI-OR total and combined scales with general offending (AUCs of .77 to .80) and with violent offending (AUCs of .72 to .74). Controlling for location (custody-community), hierarchical multiple regression analyses revealed the LSI-OR's incremental predictive validity for general, violent, and nonviolent recidivism. The General Risk/Need and Responsivity Sections made unique contributions to the final model of binary nonviolent recidivism. General Risk/Need contributed to the prediction of binary violent recidivism. Among the General Risk/Need subsections, Criminal history and Leisure/recreation as well as Criminal history and Education/employment uniquely contributed to the model of nonviolent and violent charges, respectively. The results support the use of the LSI-OR instrument with 'older' young offenders, especially those in the community, but should be cross validated on another (larger) sample of 16- to 18-year old offenders. Future research exploring the role of young offenders' responsivity concerns, in particular their motivational problems and use of denial/minimization, is suggested because these items also discriminated between recidivists and nonrecidivists in the current study