Violence Risk Appraisal Guide (VRAG)

The violence risk appraisal guide (VRAG) is an actuarial instrument that assesses the risk of further violence among men or women who have already committed criminal violence. On average, it has yielded a large effect in the prediction of violent recidivism in more than three dozen separate replications, including several different countries, a wide range of follow-up times, several operational definitions of violence, and many offender populations. It is the most empirically supported actuarial method for the assessment of violence risk in forensic populations.

The VRAG is a 12-item actuarial instrument that assesses the risk of violent recidivism among men apprehended for criminal violence. It was developed on 618 male violent offenders assessed pretrial in a secure psychiatric hospital; about half of them returned later for treatment, whereas the others were imprisoned. Most of the approximately 50 variables considered for the VRAG had predicted criminal or violent recidivism in previous research, and a few were nominated by clinicians. All variables were scored from institutional records by researchers blind to outcomes and were from four domains: childhood history, adult adjustment, referral offense details and circumstances, and assessment results. The outcome was whether, according to criminal records, there was a criminal charge for subsequent violence in an average of 7 years’ access to the community; 31% of the offenders met this recidivism criterion.

Many candidate variables predicted recidivism, but multiple regression selected the best combination for the VRAG. Several steps maximized the likelihood that the VRAG’s predictive validity would replicate— requiring that each item uniquely predict violence, ensuring the inclusion of items from all four domains, and requiring that items predict recidivism in each of several subsamples (randomly selected halves, treated and imprisoned subjects) plus the entire sample. Item weights were based on the bivariate relationship between each item and recidivism. The VRAG items in descending order of the weights are the Hare Psychopathy Checklist-Revised, elementary school maladjustment, a diagnosis of personality disorder, age (negatively related), having been separated from one or both parents prior to 16 years of age, failure on a prior conditional release, nonviolent offense history, never having married, a diagnosis of schizophrenia (negatively related), victim injury in the referral offense (negatively related), alcohol abuse, and not having a female victim in the referral offense. The VRAG can be used when as many as 4 items are missing and scored by prorating.

The VRAG predicted violent recidivism in the development sample with a high degree of accuracy— the area under the relative operating characteristic (ROC) was .76. (The ROC area is a measure of effect size equivalent to the common language effect size— the probability with which a randomly chosen violent recidivist will have a higher score than a randomly chosen nonrecidivist.)

The original sample (plus additional men who had not been released at the time) was followed again at 10 years’ average opportunity. The violent recidivism rate was .43 and the ROC area .74. There have been more than 36 replications with non-overlapping samples, and the VRAG’s average ROC area is .72—a large effect by conventional standards. Under optimal conditions (high reliability; not dropping, replacing, or modifying items; fixed and equal follow-up durations), the VRAG yields ROC areas of approximately .85. The VRAG has been shown to generalize across outcomes (number of violent re-offenses, institutional violence, very serious violence, self-reported violence, general recidivism, overall severity of violent recidivism, rapidity of violent failure), follow-up times (12 weeks to 10 years); countries (seven in North America and Europe), and offender populations (mentally disordered offenders, sexual aggressors, violent felons, developmentally delayed sex offenders, emergency psychiatric patients, wife assaulters, and juvenile offenders). Some data suggest the VRAG predicts violence among women, but there are few studies on this.

The VRAG scores range from -26 to +38; the mean in the development sample was 0.91 (SD = 12.9), and the standard error of measurement was 4.1. Each score has been associated with one of nine categories, each with a known likelihood of violent recidivism in 7 years and increasing linearly from 0% in the lowest category to 100% in the highest. There are also norms for 10 years of opportunity. Each VRAG score is associated with a particular percentile so that the violence risk of an individual assessee is evaluated according to his or her standing relative to a large sample of violent offenders. Replications of the VRAG have generally reported that the obtained rates of violent recidivism matched the predicted likelihoods for each category. If the average score of the sample is similar, the follow-up duration is approximately the same as for the norms, and the outcome is operationalized similarly.

The recommended basis for scoring the VRAG for research and individual assessment is a comprehensive psychosocial history addressing childhood conduct, family background, antisocial and criminal behavior, psychological problems, and details of offenses. Adequate psychosocial histories include more than past and present psychiatric symptoms and rely on collateral information (i.e., material gathered from friends, family, schools, correctional facilities, the police, and the courts). Scoring the VRAG is not a clinical task in its typical sense because it does not require contact between the assessor and the person being assessed. Nevertheless, compiling the required psychosocial history clearly is a clinical task, and expertise is required to score VRAG items from psychosocial histories.

The VRAG scores are static inasmuch as they do not change with time or treatment (although they might change when rescored after offenders commit further violent offenses). Current research is aimed at identifying which changes in personal characteristics or circumstances make a valid additional contribution to the assessment of violence risk based on static variables. At this point, however, no such “dynamic” variables have been identified in assessing which offenders are at risk of committing violence. There is some evidence that they might aid in predicting when violence is imminent. In concurrent testing, actuarial tools such as the VRAG have consistently outperformed structured and unstructured clinical judgment. Although requiring resources and expertise, the VRAG is the most accurate and empirically supported actuarial method for assessing the risk of violent recidivism in forensic populations.


  1. Harris, G. T., Rice, M. E., & Quinsey, V. L. (1993). Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument. Criminal Justice and Behavior, 20, 315-335.
  2. Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. A. (2006). Violent offenders: Appraising and managing risk (2nd ed). Washington, DC: American Psychological Association.
  3. Rice, M. E., & Harris, G. T. (1995). Violent recidivism: Assessing predictive validity. Journal of Consulting and Clinical Psychology, 63, 737-748.

See also: