Diagnostic reliability in a study of schizophrenic (SC) subjects from Costa Rica

Diagnostic reliability is particularly relevant in psychiatric genetics research. We report here the final diagnoses of 129 patients recruited for a genetic study of SC in the Costa Rican population. Subjects had discharge diagnoses of SC in the years 1994-2001 and were interviewed by a trained psychiatrist using the DIGS. DIGS, FIGS and medical records were reviewed by two best estimators who arrived at individual and consensus diagnoses using the DSMIIIR, RDC and DSMIV. Diagnostic differences between DSMIIIR and DSMIV were minimal. Major differences were observed between DSMIIIR and RDC systems. Using DSMIIIR, 61% of subjects received a consensus diagnosis of SC, but only 28% were SC by RDC. By RDC, 61% were diagnosed as schizoaffective (SA), versus 18% by DSMIIIR. DSMIIIR Inter-rater reliability (IRR) was highest for diagnosis of bipolar disorder (k = 0.85) and lowest for SA bipolar (k = 0.56), while RDC IRR was the highest for SC (k = 0.65) and lowest for SA bipolar (k = 0.54). Optimal IRR when using the DSMIIIR system was shown when SC and SA were combined into one category (k = 0.81 for SC/SA, 0.85 for bipolar). This most likely reflects the ambiguity in determining overall time course of affective and psychotic symptoms in persons who have had periods of affective and non-affective psychosis. DSMIIIR reliability was very good in distinguishing persons who had had non-affective psychoses from persons who only had affective psychoses.