Introduction

A number of probability indices or clinical diagnostic criteria sets have been developed for the diagnosis of melancholia or endogenous depression, although few have been empirically driven. More commonly, and as evidenced with recent dsm-iv and icd-10 criteria sets, the development process has involved literature reviews and, at the final stages, committee decisions – relying not only on the utility of descriptors but also on the imposition of cut-off points for determining caseness. Such a process clearly respects clinical wisdom but is sensitive to the composition of the committee, which may be highly conservative or adventurous, and at particular risk of preserving the personal clinical views of one or more powerful committee members. Awareness of such issues presumably encouraged those developing dsm-iv to weight empirical data. Thus, in the introduction to dsm-iv (American Psychiatric Association 1994), it is stated that, in arriving at final dsm-iv decisions, the “Work Groups and Task Force reviewed all the extensive empirical evidence and correspondence that had been gathered.… More than any other nomenclature of mental disorders, dsm-iv is grounded in empirical evidence.” The three-stage dsm-iv empirical process included (i) reviews of the published literature, (ii) reanalyses of available data sets and (iii) issue-focused field trials.

We also favour an empirically based approach to definition of clinical disorders and here report the development of an algorithm for distinguishing melancholic from non-melancholic depression.