Brown and colleagues ^{Reference Brown, Chhina and Dye1} rightly point out that there are minimal reference data against which psychiatric intensive care units (PICUs) can measure their own performance. We welcome their study, which describes the clinical activity of seven such units in England. What is particularly striking is their finding that there was a huge variability in prescribing practices between the units studied, which reached statistical significance in 14 of the 16 prescribing measures described. This included a tenfold variation in the rate of combination antipsychotic prescribing (P<0.001) and a ninefold variation in the rate of high-dose antipsychotic prescribing (P<0.005). The authors acknowledged that the rate of high-dose prescribing may have been underestimated owing to potential calculation errors.

As pointed out by Brown and colleagues, PICU patients are at a particularly high risk of neuroleptic malignant syndrome. Therefore, it is difficult to justify deviating from the evidence base for the particular conditions being treated, and practices such as combination prescribing and high-dose prescribing should be avoided if at all possible.

We question Brown et al's assertion that despite the wide variation in practice, and the potential calculation errors, their results are robust enough to serve as reference data for clinical governance purposes. Certainly, if these results are to be used as reference points, it needs to be clear which results should be used, i.e. the best results (e.g. 6% rate of combination prescribing) v. the combined percentages (23% rate of combination prescribing overall). Given the high variability between the units which participated in the study, perhaps other PICUs should be comparing themselves against the best results achieved, rather than the average.

A 6% rate of combination prescribing and a 2% (albeit an underestimate) rate of high-dose prescribing seem like standards that all PICUs should aspire to. Our experience is that such rates may well be achievable. We have achieved rates of 13% combination antipsychotic prescribing and 0% high-dose prescribing without any increase in our rate of violence (abstract in publication). We hope that the study performed by Brown and colleagues serves as a stimulus for further research and debate on the important issue of maintaining evidence-based practice, even when treating the most severely ill patients.