We were interested in the paper by Powell et al (Reference Powell, Geddes and Deeks2000) concerning suicide in psychiatric hospital in-patients. We conducted a 10-year review of the same phenomenon in Irish psychiatric hospitals (Reference Corcoran and WalshCorcoran & Walsh, 1999). Although this was not a case-control study in the sense of Powell et al, the findings were very similar, particularly in respect of the high proportion of suicides (70%) among in-patients who died ‘off the premises’ although ‘on the books’. The Irish suicide rate among acute in-patients (i.e. those in hospital for less than one year), at 319 per 100 000 of population (v. 118 per 100 000 for long-stay patients), is higher than that reported by Powell et al (137 per 100 000 admissions).

However, the validity of calculating inpatient suicide rates on the basis of an admissions denominator is questionable. It ignores the length of time during which each individual is at risk during a hospital year, and is based on admissions rather than on the individuals contributing to these admissions. Many patients may have had several admissions in any given year. We approached the problem by the person-year method which takes into account not only the number of admissions, but also their mean length of stay and other data which were available to us from the National Psychiatric In-Patient Reporting System. Our paper discusses the imperfections of this approach.