The Reliability of Psychiatric Diagnosis

No observation can be incorporated satisfactorily in a body of knowledge unless its reliability, however defined, is also known. In psychiatry, the concept of reliability has in great measure been transplanted from clinical psychology, but its optimal application to meet psychiatric needs requires some attention. For the purposes of psychiatry, reliability can conveniently be measured as the amount of agreement between different observers examining the same subjects (or a comparable series of subjects). The results may be expressed as the percentage of subjects in which agreement on the relevant variables is reached, or, in the case of two or more closely matched series being employed, by comparing the reported incidence of the phenomena under study. In one or other form this simple principle has been applied in a number of reliability studies, of which the majority have been concerned with the vital question of the reliability of nosological diagnosis (Ash, 1949; Mehlman, 1952; Hunt, Wittson and Hunt, 1953; Seeman, 1953; Foulds, 1955; Schmidt and Fonda, 1956; Pasamanick, Dinitz and Lefton, 1959), while Reid and Finnesinger (1952) have contributed an interesting semantic analysis of the concepts involved.