Back
In the last decade, the violence risk assessment has progressed. Research from across Canada, the USA, and Europe, shows that actuarial assessments provide the most accurate predictions of violence. Actuarial risk assessment compares an individual's characteristics to those of similar groups, and is routinely used to determine people's risks of medical health problems. Actuarial violence risk assessment is as accurate as many common medical assessments.
Studies also show that past behaviour predicts future violence better than current behaviour, and scores (e.g., 64%) give a more accurate assessment than categories (e.g., "high risk"). But several surveys in the United States found that forensic clinicians want to see risk assessments based on patients' current behaviour. They also want to know whether risk is "low," "moderate," or "high." These different points of view led the MHCP Research Department to embark on new area of study into risk communication.
When communicating the risk of violence, the point is to ensure that assessors and decision-makers agree on an individual's risk. Also, relatively high risk patients should be kept under relatively high security. Grant Harris, Kelly Rawson, Craig Beach, and I tested how well the Violence Risk Appraisal Guide (VRAG) is communicated, in a staff survey in the forensic division in the summer of 2001. We gave case information about hypothetical patients to 60 staff from all clinical disciplines on the clinical team. Adding a statement of the probability of future violence did improve communication of risk. Clinicians recognized that group information applied to individuals in that group, and recommended greater security for higher risk patients. When the probability of violence was not stated, clinicians used relevant case information to judge the patient's risk.
This study showed that the methods used to communicate violence risk at MHCP are effective in theory. But in practice, something goes wrong between communication of the VRAG and decisions about security in the forensic system. Oak Ridge tends to house the highest risk patients in Ontario, but dangerous offenders are sometimes supervised in the community while less risky men are detained in hospital.
The U.S. surveys showed that forensic clinicians wanted risk assessments that placed patients into categories of risk. Some people think categorizing risk as low, moderate, or high would help the Ontario Review Board (ORB) make decisions that are more in keeping with patients' risk. Should we do this in VRAG reports? If so, what should count as "low risk?" How much risk is "moderate"? Psychology research suggests that it is difficult for people to agree on what these categories mean.
Teaming up with Angela Carter, a Psychologist at MHCP, Amilynn Bryans and I studied this problem in a survey of 60 MHCP staff last winter. There was some agreement on the boundaries for the low and high categories. There was a large range in the middle, however, with a lot of disagreement. If the risk of violence was between 38% and 54%, some staff considered it to be low, while others considered it to be moderate or even high risk. In hypothetical cases, staff recommended greater security for higher risk patients, indicating that communication was effective. We were surprised to find that labeling risk "low" or "high" did not improve communication.
Furthermore, when two patient groups were compared, the risk level of the first group influenced which category staff placed the second group in. This is an important finding, because the ORB has to consider several patients each week. One patient might seem more dangerous than he otherwise would, if the last one the ORB saw had a relatively low score on the VRAG.
The undue influence of prior information was also evident when we compared risk communication using, the full VRAG report, just the VRAG results alone, or the VRAG results as part of a complete psychological assessment report. Forensic psychologists and psychiatrists from around Ontario were more conservative in their risk ratings when they read the full psychological report (i.e., they rated the patients as "riskier" than their VRAG scores).
Clinicians are often reluctant to make a recommendation about a patient if they do not have a lot of case information. But these studies show that long-term risk estimates may be more accurate when the information communicated is restricted to the violence risk assessment. Adding categories such as "low" or "high" does not improve communication, and could lead to disagreement over what these categories mean. Finally, comparing patients can make one patient appear more dangerous. These findings might help explain why MHCP clinicians understand risk well but ORB decisions do not seem to strongly reflect this understanding.
Note: This article is copyright. Readers are invited to print this article for personal and educational uses but it cannot be put to any other use without permission of the author.