An empirical perspective on subtypes in schizophrenia: A cluster-analytic approach to classification.

摘要

Schizophrenia is a label used diagnostically to classify a very heterogeneous population of mental patients. For years, attempts have been made to reduce this heterogeneity, defining more homogeneous subtypes based upon symptomatology or historical factors. Such efforts, however, have met with limited success as there is often as much heterogeneity within subtypes as between. A more fruitful approach to subtyping may be to use statistical techniques to uncover "naturally" occurring groups. The current study attempts to uncover such "naturally" occurring subtypes utilizing cluster analysis within a stepwise validation procedure to test the resultant subtypes.;Subjects were 531 patients from hospitals and residential settings with a primary diagnosis of schizophrenia from a previous multi-institutional study (Paul, 1987, 1988). Variables included measures of both overt behavior in a time-sampling framework and traditional symptom rating scales. Cluster analyses were run on the data to derive clusters of patients. These clusters were then passed through a stepwise validation process that included, along with derivation, replication, external validation and cross-validation. Multivariate techniques including MANOVA, discriminant function analysis and factor analysis were utilized within this stepwise process to provide progressively more stringent tests of the clusters. Five clear and consistent clusters were found including "normal", paranoid, low functioning/disoriented, defect state and acutely agitated clusters. These clusters were discussed in relation to other subtypes of schizophrenia from the literature. Potential treatment implications and direction for future research were discussed.