Utility of cluster analysis in classification of mixed neuropsychiatric patients

This study examines the hypothesis that reliable neuropsychologically-based subgroups can be found in a mixed neuropsychiatric patient sample (N = 600) and that the subtype membership will be associated with external diagnostic and case history variables. A modified Halstead-Reitan Neuropsychological Test Battery was employed. The sample was randomly divided into two groups of 300 each. Results show three reliable cluster analytically-based groups that differ as a function of level, but not pattern of performance. External validation failed to show the cluster groups to be significantly related to any of the neurodiagnostic or case history variables, except that the severely impaired group had a greater preponderance of organic diagnoses and a longer history of hospitalisation. Caution is warranted in using configural analysis of standard neuropsychological test batteries in an effort to identify subtypes in a mixed neuropsychiatric patient population.