The nosology of psychotic disorders: a comparison among competing classification systems.

This study examined the concordance and predictive validity of two empirical and two operational classification systems of psychotic disorders. Latent class analysis (LCA) was applied to 16 index episode and to their corresponding 16 lifetime symptom ratings of 660 psychotic inpatients, who were also diagnosed according to DSM-IV and ICD-10 criteria. The resulting classes or diagnoses were validated against clinical (premorbid adjustment, treatment response, and course) and etiologic (morbidity risk of schizophrenia and major mood disorders) variables. LCA of index episode and lifetime ratings showed that five classes of psychotic disorders provided the best fit to the data: schizophrenia, psychosis, schizomania/schizobipolar, schizodepression, and mixed psychosis (cycloid psychosis for index episode ratings and atypical schizophrenia for lifetime ratings). The concordance between the empirical and operational classification systems was poor. Irrespective of the classification method used, the most powerful validators were the clinical ones, whereas the etiologic validators played a negligible role in validating the classes or diagnoses. Overall, DSM-IV and ICD-10 classifications had more clinical validity than empirical classifications, mainly because of circularity in operational definitions regarding treatment response and course of the disorders. With the exception of the category of schizophrenia, which showed some nosological stability across the four classification systems, the number and type of categories of psychosis were highly dependent upon the set of classification procedures. These results make clear the problems inherent in drawing well-defined boundaries between psychotic disorders.