CREDIBILITY AND CLUSTER

The medical clinician has to make decisions about ( 1) management and therapeutics and (2) diagnosis. Diagnosis, that is, the classification of a particular patient, is necessitated by the need for communication among clinicians (and between clinician and patient). Reference to the literature of therapeutics and record keeping are facilitated. And, no doubt, some deep psychological need is satisfied by the construction and use of diagnostic schemata. Whatever the justification, clinicians make diagnoses, and they would like to know how to make better diagnoses. The use of computers is advocated as one way to make better (and perhaps faster) diagnoses. However, utilizing a computer requires, at the least, a formal description of the (a ) diagnostic process. I believe it is safe to say that a fully satisfying theory of medical diagnosis does not yet exist. It is the purpose of this paper to review where we now stand with regard to the construction of such a theory. The concepts employed have come from many sources. Many of these are cited in an earlier review by the author (Turner, 1965) and in the excellent papers by Ledley (1965) and Lusted (1965). The new work by Feinstein (1967) should also be consulted. In addition, the monumental monograph on cluster analysis by Sokal and Sneath (1963) contains an exhaustive review and bibliography of work on that part of our subject.