The two main concerns of Dr Shah relate to the standard hospital incident form used in the study and lack of definition of ethnicity. We accept the potential methodological problems associated with the use and retrospective analyses of routine ‘untoward’ incident data.

However, we do have two specific comments on Dr Shah's letter. First, our study was a large-scale investigation involving 1515 ‘untoward’ incidents on 14 general wards within our trust over a 3-year period. In view of the large number of incidents analysed we believe it is unlikely that unrecorded incidents or inaccurately recorded ethnic background of some patients would have significantly influenced the findings. In the statistical analyses the patients were only classified into two groups: ‘White’ and ‘Black’ (i.e. ‘African–Caribbean’ and ‘African’). Second, the main findings were broadly similar to those of a previous large-scale study of 165 medium secure unit patients at the Bethlem Royal Hospital (Reference Gudjonsson, Rabe-Hesketh and WilsonGudjonsson et al, 2000). In that study the ethnic background of the patients was obtained from the patient register rather than from the incident forms (Reference Gudjonsson, Rabe-Hesketh and WilsonGudjonsson et al, 1999).