New Tool Improves Teen Recidivism Prediction: Once Validated, Instrument Could Be a Valuable Aid in the Allocation of Limited Treatment Resources

Article excerpt

DENVER -- The assessment of recidivism risk among teenage sexual offenders is poised for a major step forward in terms of precision and credibility as a result of a new actuarial tool.

The Juvenile Sexual Offense Recidivism Risk Assessment Tool (J-SORRAT) is an empirically developed instrument that has performed very well in its initial large-scale test. Its value will be as a screening tool aiding in differential allocation of limited treatment resources, and indirectly for secondary prevention, Douglas L. Epperson, Ph.D., asserted at a meeting of the National Adolescent Perpetration Network.

"If you have an identified juvenile who has offended sexually and you can divert that person to a nonoffending path, you've saved, what, 5, 10, 30 victims? It's hard to know for sure for any one perpetrator [who is] redirected. But to redirect people, we need to know who the high-risk kids are so we can give them appropriate-intensity treatment. Even more importantly, we can keep the low-risk kids segregated from the high-risk kids so they're not preyed upon and their risk level does not increase," explained Dr. Epperson, professor of psychology and associate dean at Iowa State University, Ames.

Validated actuarial risk-assessment tools have made major contributions to the evaluation of adult sex offenders in recent years, but such tools do not yet exist for juvenile offenders. Dr. Epperson and his coinvestigators at Iowa State and Utah's Juvenile Justice Services decided to remedy the situation by developing the J-SORRAT through sophisticated statistical analyses of data contained in the archived files of 636 youths aged 12-17 when found guilty of a sex offense in Utah from 1990 to 1992.

This population comprised virtually all the juvenile sex offenders adjudicated in the state during that period. They were followed through 2003. Arrest for a new sex offense prior to age 18 occurred in 13.2%. The arrest rate for a new sex offense prior to 2004, when subjects were in their mid-20s to 30, was 19.8%.

The initial version of J-SORRAT used 12 variables that identified recidivism risk. The test had an 81.3% positive predictive value for sexual offense recidivism prior to age 18, a 95.9% negative predictive value, 72.6% sensitivity, and 97.5% specificity, for an overall accuracy of 94.2%. The area under the receiver operating characteristics curve (AUC), a favored statistical measure of accuracy, was 0.955. For comparison, an AUC of 0.5 is no better than chance, while tests with a value of 0.8-0.9 are considered excellent.

There was a problem with the original, though. Some of the variables were continuous and nonlinear, making the test too complicated for easy clinical use. So the investigators developed the J-SORRAT Research Version 2. It relies upon categorical scoring, with points awarded for 10 variables. With an AUC of 0.925, it is only slightly less accurate than the original cumbersome version.

Here's an example of how J-SORRAT-2 works. Past sexual offense history is the most powerful predictor of future risk, so J-SORRAT-2 awards 0 points for first offenders, 1 point for a single prior adjudication, 2 points for two, and a maximum 3 points for three or morem prior sex offense adjudications.

The other nine independent predictive variables included duration of sex offending history; commission of any felony sex offense in a school or public place; the use of deception, grooming, or enticement in connection with any sex offense; the number of prior adjudications for nonsexual offenses; prior sex offense treatment status; and diagnosis of a paraphilia.

Paraphilia cannot technically be diagnosed in childhood using DSM-IV criteria. …