Validity of the LSI-R:SV, LS/RNR and VRS risk assessment instruments in a sample of male serious violent offenders in Australia

Abstract

This study explored the predictive, discriminative, and incremental validity of the LSI-R:SV, LS/RNR, and VRS in a sample of 420 Australian males who were sentenced to prison for a serious violent offence. Scores on the LSI-R:SV and LS/RNR were predictive of all recidivism outcomes (any, non-sexual violent, and non-violent), while scores on the VRS predicted any and non-sexual violent recidivism only. When analysed across time, the overall discriminative validity for all instruments improved. In general, both the LSI-R:SV and VRS performed better for the recidivism outcome for which they were originally developed (i.e. general and non-sexual violent recidivism, respectively). When examining incremental validity, the VRS was combined with each of the LS instruments in separate models but did not significantly contribute to the models. Of the LS instruments, only the LS/RNR incrementally added to the prediction of non-sexual violent recidivism when combined with the VRS. Correctional and forensic mental health services that use the VRS in combination with one of the LS suite of instruments may benefit from delineating prediction from treatment to clarify the purpose of each instrument and reduce staff confusion.