Magnitude of rater differences in assessment scales for schizophrenia

The magnitude of rater differences, instead of interrater reliability, in the assessment scales of schizophrenia has rarely been investigated and was therefore addressed in this study.Thirty-six patients with schizophrenia were independently assessed by 4 expert physicians, using clinical rating scales including the Positive and Negative Syndrome Scale (PANSS). The scores obtained by the physician in charge (PIC), who had a long close contact with the patients, served as the referent answer for the purpose of this study. The scores rated by the other 3 non-PIC psychiatrists, who had a first formal examination with them, were evaluated for percentage deviance from the referent answer.The results showed that the PIC raters endorsed the numerically highest score in 20 (56%) of the 36 patients, whereas they rated the lowest in only 2 (6%) in the PANSS total score. The non-PIC assessors on the average underrated the PANSS total score by 10%, and such a tendency of underestimating the severity was noted across other clinical scales. Furthermore, the PANSS total score by one of the non-PIC physicians was deviant from the referent answer by at least 20% in 15 (42%) of 36 instances. Importantly, this magnitude of deviance was noted in the context of an intraclass correlation coefficient of 0.92.This unique investigation disclosed clinically pertinent differences among raters, even under an excellent interrater reliability. The magnitude of differences described herein seems to be an underestimation, and the baseline scores by the independent new raters might need to be corrected for those by the PICs.