MedicalResearch.com Interview with:
Simone Ullrich, PhD
Senior Lecturer in Forensic Mental Health
Violence Prevention Research Unit
Queen Mary University of London
Medical Research: What is the background for this study? What are the main findings?
Dr. Ullrich: There are currently thought to be more than three hundred risk assessment instruments used by professionals such as psychiatrists, psychologists, and probation officers to assess the risks of violence and sexual offending among psychiatric patients, prisoners, and the general population. In some mental health services the hospital does not get paid unless staff have carried out a risk assessment on their patients. Producing risk assessment instruments has become an ‘industry’ and new instruments are being produced annually, on every form of violence and criminal activity. The Queen Mary research group believe that none of these instruments have any advantage over those produced before. Furthermore, their best predictions for future violence get 30% wrong.
Professor Coid and colleagues believe that no further progress can be made because researchers have been too obsessed with predicting the future of whether a patient will be violent rather than looking for the causes of why they become violent. All previous studies have used special statistical techniques which are designed to measure predictive accuracy. The Queen Mary research group say there is nothing wrong with being accurate or measuring accuracy, but there is no point in trying to develop new instruments which can never improve on getting it right more than 70% of the time. It may be helpful to know that your patient has a high or low risk of being violent if you release them from hospital, but this is not going to tell you what you should do to stop them being violent. Furthermore, if the risk assessment says that their risk is high then it is likely that you will not release them. The problem is that professionals will always play safe and, although there is a good chance (around 30%) that they are totally wrong, the patient will not be released. This is probably one of the most important reasons why patients are staying longer and longer in secure mental health services. These instruments achieve little more than making healthcare professionals risk averse.
The National Institute for Health Research (NIHR) funded a study where 409 male and female patients who were discharged from medium secure services in England and Wales were followed up after release into the community. They received assessments with two ‘state of the art’ assessment instruments, the HCR-20 which aims to guide clinicians in their assessment of violence, and the SAPROF, another instrument aimed to guide clinicians on which factors protect patients from becoming violent. Both instruments were developed on the basis of predictive statistics. Measures were taken with these instruments prior to release into the community, then after 6 and 12 months following discharge.
Information on violence was gathered via individual case notes and a search of the police national computer. By 6 months following discharge, 54 (14%) had committed a violent act, between 6 and 12 months 43 (13%) had been violent.
The authors used two methods to investigate the associations between these risk/ protective factors and violence. They first tested the standard approach of risk assessment for the factors that occurred in the past 6 months which were then used to statistically predict violence in the following 6 months (predictive model). They then used a second approach which looked at the co-occurrence of the risk/ protective factors and violence within the same 6 month time window (causal model).
Using the traditional approach and looking at accuracy, the predictive model produced statistical coefficients of low size, suggesting that the risk and protective factors were poor in identifying who would be violent and who would not. Because many associations between the factors and violence were weak, few appeared useful in identifying those which should be targeted to manage future violence. Surprisingly, symptoms of major mental disorder did not show an association with violence, even though most of the patients in the study suffered from major mental disorder. It might have been expected that some patients would relapse, with more symptoms, leading to violence.
When the researchers used a causal approach aiming to confirm which risk and protective factors resulted in violence, the findings were very different. Symptoms of major mental disorder, the patients’ living condition, and whether they were taking medication were highly important factors. Secondly, the effects of risk and protective factors on violence were much bigger using the causal approach. For example, the effects of violent thoughts and ruminations, being in an unstable life situation, were about 3 times stronger using the causal model. The effects of being under stress and unable to cope were more than 4 times stronger than using the traditional predictive approach. They concluded that the causal approach was much better in identifying the key factors that need to be considered in the assessment and management of violence.