National Institute of Corrections
You are not signed in! To post comments and participate in discussions you need to sign in or create a free account.
Forensic Risk Assessment: A Meta-review

by Jay P. Singh and Seena Fazel (2010). Criminal Justice and Behavior, 37(9), 965-988.

Over 80 years ago the Chicago sociologist Ernest W. Burgess conducted the first violence risk assessment study, “Factors determining success or failure on parole.”  In 2009, a thorough search of different sources found 6,093 risk assessment-related studies, three times as many as just a decade before. Naturally, one consequence of these escalating research efforts is an interest in the validity of the methods used in these studies.

In this article, British psychiatrists Jay Singh and Seena Fazel engage in a meta-review of “the review literature on risk assessment.”  Overall, Singh and Fazel report on the coverage over a 15-year period (1995-2009) of previous meta-analyses and systematic reviews of the methods and results of risk assessment studies. In this article, the authors observe, “Forensic risk assessment was deigned to mean literature that was concerned with calculating the probability that antisocial behavior or criminal, violent, or sexual offending will occur. Articles were considered systematic or meta-analytic reviews if they descriptively or quantitatively summarized the findings of previous studies that had been located through the use of a predetermined search strategy.”

The sample for this meta-review included nine systemic reviews and 31 meta-analyses. The studies reviewed in this article included the following: Five studies comparing actuarial and clinically-based measures, 10 comparing the predictive validity of two or more instruments, six studies examined “the predictive validity of a single-risk assessment scheme,” and 12 studies assessed the impact of risk factors on offender recidivism. The 40 systemic reviews and meta-analyses in this article comprised 2,232 studies that covered six themes: comparisons of actuarial tools with measures that used unstructured and structured clinical judgment, comparisons of individual tools with each other, comparisons of predictive validity in cross-national research, comparisons of tool validity with different sample characteristics, comparisons of predictive validity produced by different study designs, and comparisons of relative effect sizes for individual recidivism-related risk factors.

Ten findings emerged from this meta-review:

  • Mixed evidence was found comparing the accuracy of actuarial and clinically based tools. Five meta-analyses found that actual measures had higher predictive validity, while one found no difference.
  • None of the 126 risk assessment tools were “better than any other.”  Ten studies comparing Psychopathy Checklist measures with other instruments found only “nonsignificantly different rates of predictive validity.” Few meta-analyses are available on measures other than Psychopathy Checklists, which were found to be “valid predictors of both institutional and community recidivism” in five studies.
  • Mixed evidence was found comparing the predictive validity of measures used in the United States and non-American countries.
  • Mixed evidence was found concerning the comparative validity of measures used with different genders. Eleven studies reported no difference between men and women, while a few others reported that either men or women did better than the other.
  • Mixed evidence was found concerning the comparative validity of measures used with different racial or ethnic groups.
  • Clear evidence of predictive validity was lacking in terms of psychiatric versus correctional samples.
  • Studies assessing recidivism used definitions stressing a range of different criteria, including rearrest, reconviction, reincarceration, nonaggressive conduct, general aggression, physical; violence, verbal aggression, and property destruction.
  • Mixed evidence was found on the affect of the length of follow-up periods on effect size. Four studies found no variation based on length of follow-up period, while some found that longer follow-up periods resulted in larger effect sizes.
  • Eleven studies reviewed risk factors for recidivism and found that static and dynamic risk factors were equally associated with reoffending. Interestingly, different risk factors were reported having direct association with recidivism.
  • Mixed evidence was also found of the moderating effects of age of study participants, base rates of recidivism, the study’s time orientation (retrospective versus prospective), types of new offenses (general versus violent), and information sources for risk assessment scores (case files, interviews, or self-report).

For further information, contact Seena Fazel, University of Oxford, Department of Psychiatry, Warneford Hospital, Oxford, UK OX3 7JX, (email)  

Posted Fri, Mar 11 2011 4:51 PM by Tracey Vessels


Be the first to comment on this article!
You must sign in or create an account to comment.
Brought to you by:
National Institute of Corrections
U.S. Dept. of Justice | 320 First Street | Washington, DC 20534 | 800.995.6423

This blog is funded by a contract from the National Institute of Corrections, U.S. Department of Justice. Points of view or opinions stated in this document are those of the authors and do not necessarily represent the official position or policies of the U.S. Department of Justice.