December 2004

Using Survey Results to Improve the Validity of the Standard PsychiatricNomenclature

Author Affiliations

Author Affiliation: Department of Psychiatry,Washington University School of Medicine, St Louis, Mo.


Copyright 2004 American Medical Association. All Rights Reserved.Applicable FARS/DFARS Restrictions Apply to Government Use.2004

Arch Gen Psychiatry. 2004;61(12):1188-1194. doi:10.1001/archpsyc.61.12.1188

  Measuring the validity of psychiatric diagnoses is still an unsolved problem. Yet, revisions of the Diagnostic and Statistical Manual of Mental Disorders and of chapter V of the International Classification of Diseases are now under way, with the hope of improving the validity of the current systems. This article suggests data that could be used to assist in this goal. This article has 3 objectives. (1) To show that although the validity of the interview protocols used in collecting epidemiologic survey data has not itself been proven, the data banks they have collected are well suited to raising questions about the validity of the existing diagnostic nomenclature. This is the case because they faithfully operationalize the current nomenclature in large interview studies of diverse general populations. (2) To show the kinds of changes that appropriate analysis of these data may suggest as ways to improve the validity of the nomenclature. (3) To show how suggested changes that emerge from such analyses should be tested to learn whether they actually improve validity before they are implemented. The data sets from large epidemiologic studies have hardly been tapped for testing the validity of the current nomenclature. It is feasible to use them for this purpose because they are in the public domain and because they assess the presence or absence of each of the criteria in the manuals before applying the manuals’ algorithms for combining them to make a diagnosis. Thus, these data banks allow exploration of the effects of combining and splitting diagnoses, of omitting criteria or reweighting them, and of choosing altered algorithms with respect to age at onset, number of symptoms, and duration of episodes. Assessing the consequences of these alterations can be tested by applying some of the criteria of Robins and Guze and Kendell.