We thank Professor Fazel and Dr Wolf for their thoughtful letter and interest in our article. ^{Reference Quinlivan, Cooper, Meehan, Longson, Potokar and Hulme1} Well, at least we agree on one thing: randomised controlled trials could help to clarify the role of risk assessment scales in the management of people who have harmed themselves. However, we do not think that future research should investigate scores ‘with or without additional clinician input’. Risk should not simply be a score or a colour on a traffic light system. In isolation, such ratings may be worse than useless, especially if they distract clinicians from engaging with their patients. An alternative design might be to investigate what risk scales add to ‘assessment as usual’.

Just to be clear, we are not suggesting that global clinician or patient judgement should be used in preference to rating scales to predict future self-harm. None of the measures is fit for this purpose in our view. Global risk assessments by clinicians have been found to perform poorly in previous studies. ^{Reference Kapur, Cooper, Rodway, Kelly, Guthrie and Mackway-Jones2} With colleagues in Australia some of us are currently engaged in a systematic review of this very issue. We think that Fazel & Wolf's observation that there was less variability in the scores on the scales than in the clinician ratings is an interesting one. The problem was that in our study the scales actually performed, on average, a bit worse than the global measures. There may be circumstances when scales are useful, of course. For example, as an aide memoire for new staff or as measures of change.

How might we explain the fact that risk scales performed even less well than unvalidated single-item clinician and patient measures? The obvious explanation is that the risk scales themselves were not very good. But there are other possibilities too. For the clinician ratings, we acknowledged in our discussion that the centres all had a special interest in self-harm and that predictive performance might be different in other hospitals. We also mentioned that clinicians might subliminally have used items on the scales to derive their global assessments. Fazel & Wolf's idea that patients were also taking such factors into account as the assessment progressed is intriguing and certainly worth exploring in future studies. It could be that we gain useful new insights by asking our patients how they make their own judgements of risk.

There have been a number of large systematic reviews recently all pointing in the same direction. ^{Reference Chan, Bhatti, Meader, Stockton, Evans and O'Connor3,Reference Quinlivan, Cooper, Davies, Hawton, Gunnell and Kapur4} In their editorial on our paper, Owens & Kelley highlighted the poor to worthless performance of all measures. ^{Reference Owens and Kelley5} They also asked whether risk assessment scales were on their way out. Over to readers of the BJPsych …