I was disappointed by Szmukler's (Reference Szmukler2001) negative response to Dolan & Doyle's (Reference Dolan and Doyle2000) excellent review of attempts to measure the risk of violence in psychiatric patients. His pessimism about the practical application of structured risk assessment results from a misunderstanding of the way in which these instruments may be used. First, he emphasises the low baseline. Of course, we do not know the baseline, as the information has never been collected accurately in this country. The 6% in 6 months cited in the letter derives from retrospective ratings by clinicians. Studies in the USA show that rates of violence by psychiatric patients may rise threefold if self-report is supplemented by official records and by the account of a key informant. In any case, the low baseline is irrelevant. Most risk assessment tools were developed on high-risk populations, usually people who had already committed serious offences. They were not designed to be applied to all patients. If these instruments were applied to those 6% of patients with a record of recent violence, would they assist future management by identifying those who were at highest risk of a repetition? I do not know the answer, but it is a sensible and important question.

The second problem is an unrealistic expectation of what such instruments can do. They are a supplement to good clinical practice, not a substitute for it. In many parts of Canada, for example, the Violence Risk Appraisal Guide (Reference Harris, Rice and QuinseyHarris et al, 1993) is administered to all patients admitted to a medium secure unit. The results do not dictate future management but, like the results of any other investigation, they inform it. I have no good answer to the question of why we are not evaluating this practice in the UK. When one concentrates on small, high-risk populations, the resource arguments lose this force. With medium secure admissions in this country averaging > 18 months in duration, at £ 100 000 per patient per annum, a few hours of a psychologist's time is neither here nor there.

It is unfortunate that British psychiatry has been slow to look at the application of these instruments, on which much basic research has already been done. The situation is even more depressing because National Health Service Trusts, fearful of lawsuits, are insisting on the introduction of standardised risk assessments. In most cases, they are not of proven value. Trusts in different areas use different instruments, making evaluation more difficult. The measures are likely to be employed indiscriminately, becoming just another form to be filled in. Would it not be better if this initiative were driven by clinicians, and underpinned by a sound methodology for evaluation?

Psychiatry must not persist in assuming that violence, an uncommon complication of mental disorder, is unimportant because of its rarity. Reforming the Mental Health Act (Department of Health, 2001) illustrates that concern about violence dominates the thinking of politicians in this area. It is unlikely that they are going to lose votes by overstating the level of risk associated with psychiatric patients, so the profession is going to have to come up with something better than bland reassurance. We are in a weak position, so long as we lag behind North America and parts of Europe, both in our use of existing risk-assessment technology, and in research into violence. The paper by Dolan & Doyle should stimulate us to make up some of that lost ground.