A New Approach to Assessing Risk of Violence

Researchers from Queen Mary University of London have called for the abandonment of standard approaches for investigating the risk of violence among psychiatric patients and prisoners in all future studies due to findings that show these approaches to be inaccurate.

Researchers from Queen Mary University of London (QMUL) have called for the abandonment of standard approaches for investigating the risk of violence among psychiatric patients and prisoners in all future studies due to findings that show these approaches to be inaccurate. The study team has proposed a completely new approach to risk assessment for future violence in a study published in the November 10, 2015 issue of PLOS One.

Current methods for identifying violence risk rely on Structured Professional Judgement (SPJ), which is routinely administered in mental health and criminal justice settings with moderate accuracy at best and no evidence of an ability to help prevent violence. These approaches rely on assessing risk factors that may be linked to, but don’t necessarily cause, violence, such as young age, male gender, lower social class, and having previous violent convictions.

In the study, the QMUL investigators were unable—using two standardized SPJ instruments—to identify risk and protective factors that they suggest must be targeted in interventions for discharged patients with severe mental illness. Thus, they suggest that predictive methods be abandoned if the aim is to progress from risk assessment to effective management. Their suggested replacement approach is based on identifying risk factors with clear causal associations with violence.

The study authors write that psychiatrists, psychologists, and probation officers currently use more than 300 risk assessment instruments to predict risks for violence and sexual offense. They add that producing these instruments has become an “industry” in which new instruments are produced annually. Their concern with so many being produced is that none were found to have any advantage over any other, with predictions for future violence incorrect 30% of the time at best.

“Researchers have become too obsessed with predicting whether a patient will be violent in the future, rather than looking for the causes of why they become violent,” said lead author Jeremy Coid, MB ChB, MD (Lond), FRCPsych, M. Phil. Dip. Criminol, Professor of Forensic Psychiatry at the Wolfson Institute of Preventative Medicine of QMUL. “While it is helpful to know that a patient has a high or low risk of being violent if you release them from hospital, this is not going to tell you what you should do to stop them being violent. It is more important to know which factors are causally related because these are the factors that must be the targets for future treatments and management interventions if the aim is to prevent violence happening in the future.”

For the prospective cohort study, 409 male and female patients were followed up with at baseline, 6 months, and 12 months following release into the community from medium secure services in England and Wales. Measurements were taken at eat time point using the Historical, Clinical and Risk-20 items version 3 (HCR-20v3) and Structural Assessment of Protective Factors (SAPROF). Information on violence was obtained via the McArthur community violence instrument and the Police National Computer. These “state-of-the-art,” standard approaches were poor in identifying people who would or would not be violent.

However, findings were substantially different when a causal approach was used to confirm which risk and protective factors resulted in violence. Symptoms of major mental disorder, patients’ living conditions, and whether patients are taking medications were strongly predictive of future violence. Violent thoughts, instability, and the inability to cope with stress were also three to four times stronger using the causal model than using the traditional predictive approach.

“The future direction should be to identify risk factors that have causal relationships with violent behavior and not those which predict violent behavior,” said Coid. “Risk factors, such as being young, male, of lower social class, with many previous violent convictions, may be good predictors; however, none of these factors are truly causal.”