View article

Adequate methods of classifying patients into meaningful groups are necessary for comparing large numbers of patients or for studying the incidence, etiology, treatment, and outcome of different disorders. Clinica1 diagnosis is one classification method, but since diagnostic systems and criteria for diagnosis vary, and since no one system has been definitely validated, improved methods for classifying patients need to be developed. The development of clinically more useful classifications of patients can be approached in several ways. One of these ways is described in Chapter l and concerns WHO's programme on standardization of psychiatric diagnosis, classification, and statistics. Another is the computer-simulated diagnosis discussed in Chapter 11. It is possible, however, that current diagnostic categories, even if combined and standardized, can be improved upon and that other methods of classifying patients might yield groups with a common etiology, course, and response to treatment. For this reason, a third method for identifying groups of patients has been attempted: cluster analysis, an" empirical" method. A noteworthy feature of this type of methof is that, once items are selected and ratings made, cases are classified en tirely on the basis of statis tical criteria of their similarity. In the process of clustering, each item is given a weight equal to that of the others. Aside from the decision of which criteria of similarity should be embodied in the clus tering program selected, no clinical theory affects the process of classification.

The use of empirical methods for classifying individuals is a relatively recent development that has been made possible …