|Analyses are reported that explored the predictive characteristics of the items that make up the Community Risk Needs Assessment (CRNA) instrument, which has been used by case managers to assist in setting community supervision levels for offenders in British Columbia since 1996.
A variety of analytical techniques were used, including simple correlation, multiple regression, discriminant analysis, principal components analysis, and logistic regression. The intent of these analyses was not to identify a "best" set of predictors, but first, to ascertain how well the set of CRNA items function with respect to predicting the final supervision level ratings completed by case managers, and second, to determine the utility of the CRNA items to predict future re-offence.
With respect to the first issue, the analyses indicate that there are substantial relationships between case managers' ratings of CRNA items and their global recommended supervision level ratings. Depending on how offender groupings were defined, correct classification rates ranging from 71-83% were attained. The CRNA items were also found to be related as might be expected to subsequent re-offence. It was possible to correctly classify offenders with between 70% and 75% accuracy with regard to whether or not they had subsequent re-offences.
(Return to Program Resources)