Abstract
The Violence Risk Appraisal Guide (VRAG) is an actuarial tool for predicting violence recidivism for convicted male serious violent offenders post-release that was first described by psychiatrists in Canada. It is a risk assessment tool that uses 12 variables to determine the violence risk associated with individual subjects, including Hare’s Psychopathy Checklist (revised). It has been described as the “gold standard” for general and violent recidivism in Canada, the US, and the UK. Because of its limitations in predicting the frequency and intensity of future violent behavior, and the costs involved, VRAG is best used within the parameters of its design and has no practical function for assisting in rehabilitation programs or in counseling.
Introduction
An overwhelming majority of the general public (95% in the US) believe that a person with a mental disorder that has been assessed as prone to violent and potentially harmful behavior towards self and others justifies legal intervention such as civil commitment, tort liability, sexual predator commitment, and employment discrimination, thus leading to many laws that prescribe such adjudication under specific circumstances.
Many of these laws depend on risk assessment to determine when an individual should be subject to these interventions. Because clinical judgments are believed to be subjective and inaccurate, there has been a definite move favoring actuarial predictive tools to improve the accuracy of such assessments. (Monahan, et al., 2000) This paper will provide an overview of one of the more widely known and used tools for predicting future violent behavior, the Violence Risk Appraisal Guide (VRAG). The history, current applications, and future developments of the risk assessment model will be discussed in brief.
Description of the Violent Risk Appraisal Guide and its history
Historically, the accurate assessment for future violent behavior has been fraught with difficulty. Mental health professionals tend to err on the side of safety, and over-predict violence risk. (Australian Institute of Criminology, 2004) A review of literature on violence prediction has led to the conclusion that psychiatrists and psychologists in their clinical judgment of the violence potential of serious violent offenders are accurate ones in three times for long-term predictions and one in two instances for short-term predictions.
The shift of paradigm from the prediction of future violence to risk assessment is significant because risk factors, a component of what is termed “dangerousness” takes into account circumstances as well as variables that may serve to indicate future violence as well as rule it out. Predicting the risk of harm being done depends on the severity, frequency, and nature of the aggression, giving importance to factors that may accurately indicate the level of risk, such as historical factors, the current environment, social support, and clinical factors such as mental disorder. Because these factors are dynamic in nature, the inclusion of these factors in predictive tools together with static, or established psychometric principles (age, gender) will provide a clearer basis for accurate risk assessment. (Kramer & Heilbrun, 2003)
Research has indicated that risk assessment based on clinical judgment rarely exceeds 33% predictive accuracy. In efforts to improve these statistics, hybrid predictive models using both clinical judgment and actuarial tools have been used. (Australian Institute of Criminology, 2004) Actuarial tools are used as opposed to clinical judgments because actuarial tools calculate objective assessments based on established numerical relationships between personal characteristics and the risk of violent recidivism.
Clinical judgments are usually the intuitive evaluation of an assessor such as a psychologist with the subjective beliefs of the assessor at the time of assessment, based on the memory of past cases. In many cases, clinical judgments have low agreement among different assessors and are often inaccurate and contrary to established findings on recidivism. It is thus considered more informal and less consistent as an actuarial tool such as the Violence Risk Appraisal Guide (VRAG), which is the most widely used predictive tool for risk assessment. It is considered the “gold standard” for structured tools of risk. (Report of the Committee on Serious Violent and Sexual Offenders, 2000)
It was only in 1966 that the importance of risk assessment gained prominence. The US ruling on Baxstrom v. Herald, which led to the release or transfer of 966 patients from maximum security hospitals based on the results of the risk assessment of each of the patients. In hindsight, the assessment proved remarkably accurate, only one-fifth of those released being re-convicted in four years, and even then mostly for non-violent offenses. Since then, there has been increasing conviction that risk assessment for future violent behavior would need a more objective and accurate method for assessment than that based solely on clinical judgment. (Dolan & Doyle, 2000)
In the context of Canada, however, it was Dr. Kenneth Gray who laid the foundation for risk assessment. Working with psychiatric patients all day, Gray was a good source of advice for managing such patients and it was in 1947 that Dr. Gray documented his methods for dealing with aggressive and violent conduct. Research at the Clark Institute of Psychiatry on the problems of sexual offenses such as voyeurism, pedophilia, and exhibitionism was based on Gray’s work and later expanded by the work of Kurt Freund and Ron Langevin on sex offender treatment and assessment. (Bloom, Webster, Hucker & De Freitas, 2005)
While the importance of risk assessment had been acknowledged much earlier, public interest in its ramifications was brought to the fore with the 1998 controversial conviction of Michael Stone for the murders of Lin and Megan Russell. The clamor for the prevention of such crimes resulted in the scramble to identify populations that represented a risk to society and who should be put in protective custody. The debate arose over the rights of the individual to personal freedom against the rights of the public in general to safety. (Feeney, 2003)
Oak Ridge is a maximum-security building, part of the Mental Health Centre at Penetanguishene, Ontario, which has long been used for the detention of male patients with severe mental illness and personality disorders. Under the pioneering work of Dr. Elliott Barker, Oak Ridge gained a reputation for its successful and positive treatment outcomes for its inmates. However, it was under the leadership of psychologist Vernon Quinsey in 1975 that organized research efforts were made into managing serious violent behavior and violence risk assessment. By 1993, Quinsey and his associates had come to the conclusion that clinical judgment was inadequate to accurately predict the recurrence of violent behavior but that it was possible to reasonably predict such events with the right tools.
Based on their findings, Harris, and colleagues published a paper that integrated the statistical data from the 618 inmates of Oak Ridge under the management of Quinsey and it became the earliest basis for what has now become known as the Violence Risk Appraisal Guide (VRAG). (Bloom, Webster, Hucker & De Freitas, 2005) In the study by Harris, Rice, and Quinsey in Penetanguishene Mental Health Centre, approximately half of the 618 subjects were returned to the facility while half were sent to prison for an average of 81.5 months, during which almost a third committed a new offense. This helped Harris and associates to identify the 12 variables that remained stable in predicting recidivism. (Kramer & Heilbrun, 2003)
The VRAG was developed from a study carried out by Rice and colleagues in a mental health institution of 618 mentally disordered patients with a history of serious violent behavior in Penetanguishene, Canada between 1965 and 1980. VRAG was originally named the Statistical Risk Appraisal Guide (SRAG) and is often used in tandem with the Sex Offender Risk Appraisal Guide (SORAG). The VRAG is an actuarial tool that will give the probability of recidivism, or the likeliness of a violent offender of committing another offense involving violence with a specific period. (MHC Penetanguishene Research Department. )
Typically, the VRAG is used to calculate the risk of recidivism in men who have already committed serious violent or sexual crimes. Monitoring of the behavior of these subjects was obtained through police reports from the files of the Royal Canadian Mounted Police. Based on hospital records of observed behavior, twelve variables were found to have consistent relationships across sample populations. Among these variables was Hare’s Psychopathy Checklist (PCL) revised in 1991 which identified psychopathy with the use of 20 criteria with the highest possible score of 40 (the numerical value for each item was 0, 1, or 2). While it was designed to identify psychopathy rather than as a risk assessment tool, it was determined that a PCL-R score of more than 30 reliably predicted violence recidivism in the forensic population. (Feeney, 2003)
Aside from the PCL-R Score, other variables included Elementary School maladjustment, age of index offense, DSM-III diagnosis of personality disorder, separation from parents by age 16, failure on prior conditional release, marital status, non-violent offense score, history of alcohol abuse, DSM-III diagnosis of schizophrenia, victim injury and female victim. An overall score determined the placement of the subject within nine categories or bins, each with a calculated risk of violent recidivism. Category 9 was the highest risk category at near 100% probability of re-offense, while those in Category 1 had an 11 percent would later be found to commit a new violent act and 42 percent for those in category 5. (Monahan, 2004)
It was found that the characteristic that was the most predictive of future violent behavior was the PCL-R score (which carried the most weight in the overall score). age of index offense (when the violent behavior was first documented), school maladjustment, and schizophrenia (DSM-III diagnosis, negatively correlated) (Report of the Committee on Serious Violent and Sexual Offenders, 2000)
Of particular importance in VRAG is the Hare PCL-R score. The PCL-R score as described by Hare measures psychopathy, or levels of personality disorder resulting in serious, potentially risky consequences for society. This includes characteristics of predatory and violent behavior and a lack of regard for the effect of behavior on others. (Australian Institute of Criminology, 2004)
The accuracy of PCL-R scores for predicting both general recidivism, violent recidivism, and sexual offending was validated through meta-analysis of studies on criminal populations, and is considered to be the best predictor for violent recidivism in North American settings and had a higher association than alcohol abuse or schizophrenia. (Kramer & Heilbrun, 2003) Based on the 1941 publication of American psychiatrist Hervey Cleckley’s “The Mask of Insanity,” Hare, a Canadian clinician with extensive experience in correctional facility settings developed a 22-item checklist as a structured clinical judgment device, later to be reduced to 20-items. The checklist comes with precise scoring instructions and standardization data. (Bloom, Webster, Hucker & De Freitas, 2005)
A numerical value was assigned by clinical experts to each variable or characteristic in each of the 800 study cases based on clinical records and psychosocial histories, which included family background, antisocial behavior, childhood conduct, psychological problems, and offense details. The data also included testimonies or information from significant third parties including schools, family, and friends as well as authoritative sources such as the police and court records to confirm or refute information gathered directly from the study subject. (MHC Penetanguishene Research Department. )
Psychometric qualities of the Violent Risk Appraisal Guide
It would be of considerable importance to be able to predict the likelihood of violent behavior in the future in mentally ill patients. In a UK study to determine an appropriate tool for accomplishing this aim, the predictive validity of VRAG was tested over two years. In the short term, less than 12 months, VRAG demonstrated an AUC of 0.85 for both violent offenses and general offenses. Over the longer term, (>1 year) VRAG predictiveness was on par with other actuarial models in terms of AUC. (Snowden, Gray, Taylor & Macculloch, 2007)
Consequently, VRAG as a tool for predicting violent recidivism has been established by several studies in the US and Canada, including two meta-analyses of several replications of violence recidivism studies. While widely used in Canada and the US, it has yet to be validated in a New Zealand sample population of serious offenders. (Bakker, O’Malley & Riley, 2003).
Many assessors consider VRAG an important tool for predicting future violence for subjects with mental disorders, although its applicability for the general population has yet to be established because the psychometric principles on which it is based was derived from a study of convicted male violent offenders who have histories of past violent behavior. VRAG is also designed to predict violent recidivism post-release rather than violent behavior during incarceration or institutionalization. It is important to emphasize that the use of VRAG should be within the confines of the task for which it was designed, which is predicting future violent behavior. It would not be appropriate for use as the sole basis for planning interventions or assessment of the effectiveness of interventions in place. (Kramer & Heilbrun, 2003)
VRAG was also established by the Ontario Lieutenant Governor’s Review Board as the model by which risk assessment should be conducted for the management and incarceration of violent offenders. It was discussed in the publication of the special research committee that VRAG, duly modified depending on clinical opinion and circumstances, would be the basis for any legal decisions regarding the treatment of serious violent offenders. Quinsey and associates strenuously argued against any modification of the VRAG scores, claiming that it compromised the efficacy of the tool. (Bloom, Webster, Hucker & De Freitas, 2005)
Quinsey, Harris, Rice, and Cormier described the Violence Risk Appraisal Guide in 1998 that was partially validated in Scotland and the UK as a risk assessment tool. It was found that VRAG was valid as a predictor of general and violent offending although it did not provide a prediction as to the frequency of future offensive behavior in Scotland. It was found that in general, VRAG was not a valid predictor of sexual recidivism for sexual offenders. (Risk Management Authority, 2007)
Uses of the Violent Risk Appraisal Guide
The role of violence risk assessment in criminal as well as civil psychiatric practice is to determine whether the release of a person accused of a violent crime or suspected of having a mental disorder will put the general public in danger of harm’s way. It is also the basis by which a judge may determine whether a convicted person would best benefit from a rehabilitative rather than a mainstream correctional facility.
In cases where the subject’s behavior becomes suspect, clinicians need to provide an expert opinion on whether a person should be removed from the general population because of the risk of violence or harm. In many cases, the expert opinion of the clinical psychiatrist will decide the extent to which a person’s right to freedom will be curtailed and the capability of a person to stand trial and face the charges being levied. For each assessment, the danger of an inaccurate assessment can lead to serious emotional and physical harm or trauma for the subject and may render the clinician open to legal problems. (Bloom, Webster, Hucker & De Freitas, 2005)
The value of risk assessment for mental health professionals from a legal standpoint is in providing a basis for the involuntary commitment or detention of those who are considered at risk for future violent behavior because of mental disorder or illness. It also determines the risk for child abuse as well as assuring workplace protection and safety. In the case of criminal law, risk assessment provides a legally acceptable basis for determining concerns for bail applications, parole as well as sentencing recommendations. Risk assessment further indicates how far confidential communication can be legally breached to prevent the commission of harm on the self and others. (Australian Institute of Criminology, 2004)
Limitations of VRAG
There has been increasing concern, especially in the UK, over the question of how to assess the risk of violent recidivism in subjects with severe personality disorders. Typically, the focus of risk assessment research has been on validating the accuracy of risk prediction variables in static actuarial predictors in large, heterogeneous populations. VRAG is considered to be one of two of the most widely accepted hybrid models for violence risk prediction in North America, mostly because of its inclusion of Hare’s Psychopathy Checklist – Revised (PCL-R) as one of the variables for predicting future violent recidivism. (Dolan & Doyle, 2000)
The main objections to VRAG have been its limitations in terms of predicting the nature, type, severity, and imminence of violence risk as well as its value in assisting clinicians in managing the presence of risk. It has also been recently suggested that risk variables not included in the 12 characteristics in the VRAG but which have been shown to be linked to violence (i.e. threatening behavior and prior history of violence) should be ignored in the assessment, which would be negligent on the part of the assessor and will make risk management less effective because important dimensions of the risk would be left out in the assessment. (Report of the Committee on Serious Violent and Sexual Offenders, 2000)
Because of the many variables that are involved in the populations under study, some concerns have been expressed regarding the consistency of the accuracy of VRAG over time. While the population may vary in terms of some characteristics (such as age or offender diagnosis), the variables that predict violent recidivism remain constant. Another concern is the necessity of the subject’s willingness to provide the information as well as the amount of skill, time, and manpower required to arrive at an accurate VRAG score, which can be significant, averaging about 60 hours per case study. While the former is a non-issue because subjects under these circumstances have no choice but to give the information sought, the latter warrant some concern, although the value of the results often justifies the expense.
The VRAG is recommended primarily to ensure that the subjects with the highest risk for violent recidivism receive the appropriate level of restriction, supervision, and treatment efforts in order to protect the subject as well as the public in general. The use of VRAG as an index for risk in forensic populations enables custodial and clinical assessors to more efficiently appropriate resources for each segment of the forensic population. (MHC Penetanguishene Research Department. )
The VRAG, however, is applicable only to a specific group of the population, its accuracy is compromised when used outside the forensic population (serious violent or sexual offenders). Based on a study of subjects with mental disorders such as schizophrenia, the VRAG consistently indicated that schizophrenia was negatively correlated to recidivism. This would indicate the area under the receiver operator characteristic (ROC) curve of.76 found for VRAG has been validated only for the typical sample population of serious offenders. (Bakker, O’Malley & Riley, 2003)
The ROC is one of several statistical measures for evaluating the accuracy of predictive tools in violence risk, based on the area under the curve (AUC) data. ROCs are independent of a given population’s base rates of violence. (Dolan & Doyle, 2000)
Based on the findings of the study that led to VRAG, a diagnosis of schizophrenia indicates that compared to other diagnoses of psychopathic disorders in the PCL-R, the likelihood of violence is lower. The study also showed that those who committed violent crimes against females were less likely candidates for recidivism, as are those who committed the greatest injuries i.e. leading to death. (Report of the Committee on Serious Violent and Sexual Offenders, 2000)
One barrier to the use of VRAG is the costs involved in using such risk assessment tools which makes it impractical for use in most clinical scenarios. (Monahan, et al., 2000) To illustrate the economic aspects of risk assessment, an estimate of the actual costs was tabulated. In cases of non-sexual violent offenders, the costs of a VRAG, clinical interview, and Personality Projective can run a tab of up to $1000 per subject (3-4 hours). For an adult sex offender, risk assessment using VRAG or its companion tool SORAG. PCL-R, Intellectual function, Multiphasic Sex Inventory, Personality testing, STATIC-99, and sexual history can cost up to $2000 for an 8-10 hour session. (Johnson County Certified Providers, 2007)
The Use the VRAG in Counseling
The skills necessary for valid risk assessment include the ability of the clinician to correctly classify the particular character of the inappropriate or dangerous behavior of the subject and subsequent actions based on well-established and relevant principles. Assessors must also possess the training to react and respond in a timely and appropriate manner in cases of high-risk situations i.e. threats of suicide. In cases when disclosure of information may be necessary, the trained psychiatrist will be able to decide the extent to which such information should be given, typically to prevent the doing of serious bodily harm to others. (Bloom, Webster, Hucker & De Freitas, 2005)
The 12 item of the VRAG makes use of ROCs to analyze the data of forensic populations. It was found that in several populations of serious violent offenders, the predictive accuracy of VRAG for violent recidivism was about.73 AUC for up to 10 years, even under restrictive definitions of recidivism. This accuracy, however, was compromised when applied to a population of pedophiles. (Dolan & Doyle, 2000) This would indicate that in the counseling setting, the VRAG is of value only in tandem with other methods of intervention based on the results of a VRAG assessment.
It has been suggested that the VRAG in combination with other predictive, both static and dynamic, models that complement the variables of an actuarial tool will increase its predictive accuracy across populations and cultural contexts. In particular, actuarial predictors are considered static, while clinical judgment is considered dynamic. A combination of actuarial tools and clinical judgment is believed to take into account the idiosyncrasies of the individual patient, especially those outside the normal forensic population. (Bakker, O’Malley & Riley, 2003)
Of the 12-item VRAG scheme, the algorithm employed placed heavy emphasis on the PCL-R, childhood maladjustment, and employability characteristics. Such information can be elicited during a counseling/interview session. Mental illness was weighted in the negative. In terms of interpretation, a high positive VRAG score is considered highly likely to commit violent offenses in the future. An expanded version of the VRAG, which deals mainly with serious sexual offenders, is the Sex Offender Risk Assessment Guide (SORAG). (Bloom, Webster, Hucker & De Freitas, 2005)
Conclusion
Predicting how people will act in any given situation is problematic because it is almost impossible to take into consideration the many factors and variables that come into play that will influence a person’s actions and reactions. However, when a person is violent and becomes a source of potential menace to the general population, it becomes necessary to determine whether it is safe for them to be out in society. On the other hand, the rights of the individual must be taken into account before the decision for legal intervention (involuntary commitment or incarceration) is made. The role of risk assessment is to predict the extent to which a person will indulge in violence based on several variables, including mental health and personal history.
It is popularly believed that it is the clinical judgment of psychiatrists or psychologists which determine whether it is safe to let a person who has been convicted of a violent offense roam free after a certain period, but recent developments have established that a well-designed actuarial predictive tool combined with clinical judgment is more accurate in predicting violence recidivism, or the potential of a violent offender of succumbing to violent behavior in the future. The “gold standard” in Canada, many US states, and the UK for such actuarial tools are the Violence Risk Appraisal Guide (VRAG).
However, it has also been established that the parameters of VRAG are limited to determining violence recidivism to the population of convicted male violent offenders post-release in Canada, the US, and the UK. Its use in other cultural settings has not been conclusively validated. Moreover, its predictive accuracy for sexual recidivism of sexual offenders is comparable to other, simpler actuarial tools.
As with any tool, the VRAG was designed for a specific purpose, and in that area, it is so far the most accurate. Its use in the counseling setting is limited to the convicted male violent offender with a mental disorder or illness and its value to the counselor will be the same as employing PCL-R. Because of the costs involved in acquiring a VRAG assessment, it would not be cost-effective to use it for any other purpose than to determine violence recidivism post-release.
References
Australian Institute of Criminology. (2004) Risk assessment by mental health professionals and the prevention of future violent behavior. Canberra, ACT: Bernadette McSherry.
Bakker, L., O’Malley, J. & Riley, D. (2003) Storm warning: Statistical models for predicting violence. Web.
Bloom, H. Webster, C., Hucker, S. De Freitas, K. (2005) The Canadian contribution to violence risk assessment: history and implications for current psychiatric practice. The Canadian Journal of Psychiatry, 50(1), 3-11.
Dolan, M. & Doyle, M. (2000) Violence risk prediction: Clinical and actuarial measures and the role of the Psychopathy Checklist. The British Journal of Psychiatry, 177, 303-311.
Feeney, A. (2003) Dangerous severe personality disorder. Advances in Psychiatric Treatment, 9, 349–358.
Johnson County Certified Providers. (2007). Web.
Kramer, G.M. & Heilbrun, K. (2003) A decade of advances in risk assessment: implications for corrections. Web.
MHC Penetanguishene Research Department. (n.d.) The Violence Risk Appraisal Guide: a brief summary. Web.
Monahan, J. (2004) Forecasting harm: the law and science of risk assessment among prisoners, predators, and patients. Web.
Monahan, J., Steadman, H.J., Silver, E., Applebaum, P.S., Grisso, T., Mulvey, E.P., et al. (2000) Developing a clinically useful actuarial tool for assessing violence risk. The British Journal of Psychiatry, 176, 312-319.
Report of the Committee on Serious Violent and Sexual Offenders. (2000). Web.
Risk Management Authority. (2007) Risk assessment tools evaluation directory. Web.
Snowden, R.J., Gray, N.S., Taylor, J. & Macculloch, M.J. (2007) Actuarial prediction of violent recidivism in mentally disordered offenders. Psychological Medicine, 37, 1539-1549.