Cluster analysis in diagnosis.

The purpose of this paper is to survey the usefulness of cluster analysis in the special case of diagnoses. This complex topic is restricted, however, to the application on laboratory characteristics, separately or in connection with clinical data. The article is subdivided into three parts: (a) the fields of a possible use of cluster analysis, detection of diseases or subgroups of diseases, and data reduction by detection of structures; (b) a brief mathematical description of hierarchical and partitioning classification techniques (as a crucial point, the problems associated with these methods are discussed); (c) a critical review of 24 publications of the past 10 years concerning cluster analysis and diagnoses.