Making Better Forensic Decisions With Fewer Resources

      At MHCP, the Violence Risk Appraisal Guide (VRAG) is well known and available on most patients with a history of violence. The VRAG measures the risk of future violence and has been validated in over 50 studies (www.mhcp-research.com/ragreps.htm). However, past studies by the MHCP Research Department showed that decision makers failed to use the VRAG – the Ontario Review Board (ORB) was just as likely to release the highest risk men as the lowest.

      Tribunals like the ORB rely heavily on advice from clinicians. Surveys conducted at MHCP show that clinicians understand the VRAG and apply it correctly in hypothetical cases. On the other hand, clinicians are influenced by arguments that risk changes a lot over time (even though there is scant evidence that changeable risk factors have been identified). Clinicians also want to adjust risk scores for factors that are unique to individual patients, but research shows this results in poorer prediction of violence.

      What forensic clinicians do and say about violence risk is important because tribunals closely follow clinical advice. Drs. Shari McKee, Grant Harris, and Marnie Rice of the MHCP Research Department recently re-evaluated whether the ORB transfers patients out of maximum security in accordance with violence risk. Unfortunately, they found that the ORB is still almost as likely to transfer the most dangerous patients as lower risk patients.

      Dr. McKee and her colleagues found that clinical team recommendations for release are more consistent with violence risk, but only a little. Clinicians were more likely to recommend transfer for patients with lowest VRAG scores than patients with the highest scores. But clinicians were also more likely to recommend transfer for patients with greater insight and medication compliance, and those who requested a transfer. These factors do not indicate lower risk. As a result, apart from the extreme ends of violence risk, riskier patients were more likely to be recommended for transfer. There is optimism, though, because clinicians did use the VRAG score to make riskappropriate decisions in hypothetical cases.

       As expected, ORB decisions were more strongly influenced by clinical recommendations than anything else. Perhaps the best way to improve ORB decisions, then, is to improve the information that clinicians give the board.

      In a second study, Dr. McKee found that patients’ dishonesty, selfishness, and rule-breaking predicted violence outcomes in addition to violence risk. A third study showed that clinicians believed assaultiveness to be the most important patient characteristic to consider. These variables might be truly dynamic risk factors. The solution that Dr. McKee proposed is to use a formula that gives most weight to the VRAG score, but also takes into account recent behavior, the clinical team’s opinion, and administrative constraints e.g., pressure on forensic beds).

      Recently, the public’s attention has been turned to how the ORB decided to release a man reported to be a former MHCP patient (“Ontario Tories demand review of prisoner- release rules,” Ottawa Citizen, December 06, 2007). Let us examine how that decision might have been made by a formula suggested by McKee, Harris, and Rice, called the Forensic Institutional Decisions or Recommendations Scheme (FIDORS).

      The media reported that the patient’s actuarial risk of violent recidivism was 24 percent; the lowest possible VRAG score yielding this risk gives the patient 60 points on FIDORS. Having no institutional misbehaviour adds 20 points, and assuming maximum pressure on forensic beds adds 40 more points. The media reported that the patient committed murder in the past, which adds 0 points, but if we assume the patient took maximum steps towards his reintegration, then 20 points are added. The clinical team evidently agreed that maximum security was not required, which adds 20 points. Thus, the FIDORS score is 160 for this patient if we assume that all unknown items actually work in the patient’s favour. This score would place a patient in a minimum security hospital. The media reported that, in fact, the ORB gave this patient an absolute discharge with no restrictions or treatment conditions. He was recently re-arrested for assault.

      FIDORS is one illustration; other tools could be empirically developed to ensure that scores on a valid risk assessment tool are consistently combined with other risk-relevant and legally mandated considerations. Clinicians would contribute the information needed to score the tool, and the ORB (or a policy-based procedure that eliminates the cost of the ORB) would assign the placement according to the score, with limited and specified exceptions (e.g., for patients who have made specific threats of violence).

      Using automated decision tools like FIDORS throughout the forensic system would permit patients to be placed at security levels that offer the greatest public safety in an appropriate balance with their need for treatment and reintegration into society.