Constructing and testing a framework for dynamic risk assessment.

Abstract

This paper describes the construction and testing of a framework for dynamic risk assessment. A review of previous studies identified 4 domains into which dynamic risk factors for sexual offending seem to fall. These were sexual interests, distorted attitudes, socioaffective functioning, and self-management. Psychometric indicators for 3 of the domains were identified, and 2 studies are reported using these indicators to test the framework. Study 1 divided men serving a prison sentence for a sexual offense against a child into 2 groups--those with a previous conviction of this kind (Repeaters) and those for whom this was the only time they had been sentenced for such an offense (Current Only). The Repeaters were found to show more distorted attitudes, worse socioaffective functioning, and poorer self-management than did the Current Only group. Study 2 used a simple algorithm to combine these psychometric indicators into an overall "Deviance" classification. Reconviction data was obtained for offenders classified as high, moderate, or low on Deviance. Sexual reconviction was found to be monotonically associated with the Deviance classification. Logistic regression analysis showed that both static variables (Static-99) and the Deviance classification made independent contributions to prediction. It is suggested that risk assessment procedures should combine these 2 approaches.