Violence risk assessment has not been shown to reduce violence

To the Editor

Among the thousands of papers about violence risk assessment only two studies have found any evidence for any reduction in violence as a result of redirecting treatment resources from low-risk to high-risk patients. These two studies, cited in a recent commentary (Roaldset, 2015), comprise the totality of the evidence base for the use of violence risk assessment in mental health.

Abderhalden and colleagues (2008) asked nurses on four randomly allocated wards to conduct twice-daily violence risk assessments during days 1–3 of each admission. They found that with this intervention the number of severe aggressive events on these wards dropped by 41%. This was compared to five control wards where the drop was only 15%. Unfortunately, the randomisation process had resulted in the four experimental wards being those with much higher rates of pre-study violence. They were also four wards where each ward manager perceived aggression prior to the study as being a ‘big or very big problem’, while only two of the five control wards’ managers rated aggression to be this problematic. By the end of the study, the fall in the rates of aggression in the intervention wards only resulted in those wards achieving the rate already evident in the control wards. Given these features of the study it is very hard to confidently attribute the apparent improvement in the rates of aggression to the experimental intervention, rather than a Hawthorne effect or simply regression to mean.

In the van de Sande study (2011), the two experimental wards rated every patient every day with two risk assessment scales and then used another two risk assessment scales weekly. Staff on the experimental wards also received ‘ongoing clinical supervision … provided by a clinical nurse specialist supported by a risk assessment expert panel’. At the end of the intervention period, the two experimental wards did see a reduction in violent episodes compared to the two treatment-as-usual wards, though after correction for the number of patient days a fall in the number of violent patients did not reach statistical significance.

These two studies represent the sum-total of the evidence for the harm-reducing effects of violence risk assessment. The results of these studies, taken together with the other problems of risk assessment that Roaldset (2015) acknowledges, suggest that violence risk assessments cannot lead to better clinical practice.