A comparison of nine systems to diagnose schizophrenia.

The files of 283 hospitalized patients discharged with a diagnosis of schizophrenia, schizoaffective schizophrenia, or paranoid state were examined without knowledge of the patient's subsequent history. These patients, most of whom had originally been diagnosed by DSM-I criteria, were retrospectively diagnosed by New York Research Diagnostic Criteria (RDC), the New Have Schizophrenia Index (NHSI), the St. Louis criteria, Taylor-Abrams 1978 criteria, Schneider's first-rank symptoms (FRS), the Washington IPSS 12-Point Flexible System, Astrup's process/nonprocess distinction, and DSM-III. Kappa values measuring pairwise diagnostic agreement amont the nine systems were typically low except among RDC, DSM-III, and St. Louis criteria. Long-term followup status was not significantly predicted by the FRS, NHSI, or Taylor-Abrams criteria. Diagnosis by means of the other systems, especially the Astrup process/nonprocess distinction, was significantly correlated with followup. However, Astrup's "process" schizophrenia is not operationally defined and could not be expected to be used with the degree of interrater reliability desired by researchers. On the other hand, of the operationally defined systems, DSM-III schizophrenia has the highest correlation with followup and is thus suggested for use by researchers desiring a highly homogeneous, although narrowly defined, population. Investigators wishing to cast a wider net could use a less restrictive system such as the RDC, with or without schizoaffectives included.