The nature of pragmatic research is to examine clinical practice in the manner it happens - that is both its weakness (for example, not using research-standardised diagnostic interviews or detailed cognitive testing) and its strength. The MMSE has been in use in both clinical and research settings since 1975 as a tool for cognitive assessment and Drs Kripalani and Poongan are correct in stating the unreliability of a single cutoff point for any diagnosis. In our study we examined MMSE scores only of those patients with a diagnosis of dementia, and other than commenting on the rise in total number of referrals made no comment on the underlying referral rate.

We agree with Dr Willis that the MMSE, assuming that it measures an actual underlying ‘cognitive ability’ where the intervals between adjacent scale values are indeterminate, is an ordinal rather than an interval or ratio scale and corresponding tests should be used. We are pleased to note that our data still show a significant move towards earlier referral in dementia.