Roaldset et al. Reference Roaldset, Hartvig and Bjørkly[3] recently described a prospective validation study of the V-Risk-10 among patients discharged from two acute psychiatric units. They concluded that the V-Risk-10 is a “valid and useful screen for violence risk after discharge”, because the test can categorize patients into groups with a lower and higher chance of committing aggressive and violent behaviours. However, we believe that Roaldset et al.’s claim that the V-Risk-10 is a useful screening tool requires further clarification.

Not all tests that can distinguish between high and low risk groups are useful for screening. The World Health Organisation (WHO) and others have published guidelines outlining when screening is worthwhile [4,1].

First, the WHO criteria suggest that a specific diagnostic test should be available to follow a sensitive but nonspecific screening procedure like the V-Risk-10. In other areas of medicine, screening tests are followed by a specific diagnostic test. Well known, among many examples, are mammography and cervical screening, which are followed by a biopsy, antibody tests for HIV which are followed by a Western Blot test and Mantoux screening for tuberculosis. In the case of violence risk assessment, no such further test exists. Without a test that is more specific for violence, fewer than half of those identified by the V-Risk-10 will actually go on to be aggressive or violent. Therefore a significant proportion of nonviolent patients will receive unnecessary interventions and will be stigmatised as dangerous. Notwithstanding unnecessary treatment, such labelling has been shown to have adverse consequences, for socially neutral medical conditions such as hypertension Reference Macdonald, Sackett, Haynes and Taylor[2]. Being labelled as dangerous is likely to have more adverse consequences for mentally ill patients.

Second, the WHO contends that a useful screening test should be for an important health problem for the individual or for the community. We accept that serious acts of violence meet this threshold of importance, but in the recent study there were as few as 17 incidences of more severe violence among 367 patients in the 1 year period of follow-up. Using this base rate and the sensitivity and the sensitivity provided by the authors, the positive predictive value for serious violence can be calculated to be under 8%, indicating that 92% of predictions of more serious violence will be proven incorrect.

Finally, according to the WHO, a useful intervention should be available to justify screening. As is well known, there are no highly effective or specific treatments that prevent future violence.

To date there is no empirical evidence showing that risk assessment can actually reduce harms in clinical practice. If the principles of evidence-based medicine are to be applied in mental health, currently there is little role for violence risk assessment.