Application of cluster analysis to investigate neuropsychological heterogeneity in psychiatric and neurological patients

Abstract

Many neurological and psychiatric disorders are associated with heterogeneous levels and patterns of neuropsychological function. Efforts have been made to develop empirically based meaningful subtypes within that heterogeneity utilizing techniques such as cluster analysis. Identification of reliable subgroups may have important implications for treatment and management. In adults, major efforts have been made in the areas of schizophrenia and closed head injury, where clustering has produced stable subgroups with good external validity. Of particular interest in schizophrenia is a “neuropsychologically normal” subgroup consisting of individuals with definitive diagnoses of schizophrenia but who perform normally on most cognitive tests. Among adults with closed head injury, we have identified subgroups with near-normal cognitive function, a subgroup with moderate impairment with a pattern of normal and abnormal test scores, and a subgroup with generalized global impairment. In the case of schizophrenia, cluster membership has been associated with MRS-derived measures of membrane repair and neuron count, as well as with age, education, and neurological comorbidity, while in closed head injury it has been shown to be associated with various demographic factors including age at which injury occurred, employment status, and months between time of injury and assessment. Cluster analytic approaches have also been taken to classify adults with learning disability and health problems of homeless veterans. The findings attest to the external validity of different types for the cluster solutions involving these conditions and support the use of quantitatively based classification in their assessment.