The accuracy of risk prediction for sexual offenders

Research summary
Vol. 11 No. 1
January 2007

Question

How should we assess the recidivism risk of sexual offenders?

Background

Once someone has been detected committing a sexual offence, it is important to evaluate the risk of recidivism. The effective use of correctional resources involves directing the most intensive interventions toward the highest risk offenders.

Previous research has identified a number of risk factors that are reliably related to recidivism among sexual offenders. These factors include static, historical factors (e.g., number and type of prior offences) as well as potentially changeable characteristics (e.g., intimacy deficits, sexual preoccupations). No single factor, however, is strongly related to recidivism. Consequently, evaluators need to consider a range of risk factors in their overall evaluation of risk.

Different methods have been proposed for conducting risk assessments for sexual offenders. Until recently, the most popular form of risk assessment involved unstructured professional judgement. Most of the research, however, has focused on structured, actuarial tools containing predominantly empirically-derived, static risk factors. Although useful for pure prediction, these empirical actuarial scales provide little information about how to manage and reduce risk. Consequently, recent efforts have focused on developing structured risk tools containing potentially changeable (dynamic) items. This study examined the relative predictive accuracy of these various approaches to risk assessment.

Method

A systematic review was conducted of 100 studies (79 distinct samples) of risk assessments for sexual offenders. The risk assessments procedures were grouped according to the source of the items (conceptual or empirical) and the method for combining the factors into the overall evaluation (structured/actuarial or unstructured). The measures were further classified according to the type of recidivism they were designed to predict (sexual, any violence, and any other type of offence).

Results

The risk assessments that used a structured (actuarial) approach to combining the risk factors were more accurate than unstructured professional opinion. The actuarial measures using empirically-derived risk factors and the actuarial measures using risk factors based on theory showed similar levels of predictive accuracy. In general, the measures designed for sexual recidivism were the best predictors of sexual recidivism, and the measures designed for general recidivism were the best predictors of general recidivism. The predictive accuracy of the best measures was moderately high for sexual recidivism and high for violent and general recidivism.

Structured professional judgement showed variable results: sometimes strong, sometimes weak. This type of evaluation involves reviewing a predetermined list of risk factors with the overall determination of risk being left to the professional judgement of the evaluator. On average, the predictive accuracy of structured professional judgements were similar to the predictive accuracy of scores created by counting the number of risk factors present.

Policy implications

  1. Assessment procedures are available that allow practitioners to distinguish between sexual offenders with a low risk of reoffending and those with a high risk of reoffending. Consequently, community safety can be efficiently served when responses to sexual offenders vary according to their risk level.
  2. Structured risk assessments are the most accurate and should be routinely used.
  3. Structured risk tools including dynamic (changeable) items show considerable promise. Future research should aim to develop measures that maximize both utility and predictive accuracy.

Source

For further information

R.K. Hanson, Ph.D.
Corrections Research
Public Safety Canada
269 Laurier Avenue West,
Ottawa, Ontario K1A 0P8
Phone: 613-991-2840
Fax: 613-990-8295
E-mail: Karl.Hanson@ps-sp.gc.ca

Date modified: