Authors' reply: We think that Appleby and colleagues have misunderstood what we are saying. Of course we are aware of the methods of case ascertainment used by the National Confidential Inquiry. Our main point is exactly that made by Appleby and colleagues – that the Inquiry is not set up in a way that enables it to identify suicides following attendances at accident and emergency departments. This is because specialist mental health services in the UK do not provide comprehensive monitoring of self-harm attendances, even of those referred for a specialist opinion, and yet the Inquiry does not seek evidence directly from accident and emergency departments about attendances following self-harm.

Self-harm is closely linked to suicide, and yet self-harm services are in a disorganised and underresourced state nationally. We see this as a challenge both to national policy makers and to local service providers. The National Suicide Prevention Strategy does indeed refer to self-harm. However, we find its recommendations couched in such general terms that it is unclear how real change will come about in services hard-pressed for staff or funding.

As a first step mental health trusts should be required to provide comprehensive self-harm services to accident and emergency departments, and acute hospitals and mental health services should collaborate to monitor all attendances that follow self-harm. This action would improve local service provision for a neglected and high-risk group, at the same time as solving the National Confidential Inquiry's monitoring problem.

We disagree with the National Director for Mental Health that the evidence is not strong enough to support such a policy; it is at least as good as the evidence for the wholesale introduction of standardised risk assessment in mental health services. If further evidence is needed, then we are not sure that a study restricted to ‘mental health patients’ (and therefore presumably excluding the very people we are discussing) is the answer. It would, however, be a relatively simple matter to attempt to replicate our findings in a multi-centre prospective monitoring study at those other centres that run accurate accident-and- emergency-based clinical databases.