A review of the DSM-III-R criteria for autistic disorder.

The objective of this paper is to review the psychometric properties of the new DSM-III-R criteria for autism. Five data sets were evaluated according to a set of methodological criteria. The results indicate that the DSM-III-R criteria for autistic disorder have, on average, very good sensitivity, but much lower specificity. The implications of this are (a) greater numbers of children diagnosed as autistic; (b) greater numbers of children misdiagnosed as autistic; and (c) greater heterogeneity among samples of autistic children. In essence, the DSM-III-R criteria act more like screening tests than diagnostic criteria. Conceptual and methodologic issues in the evaluation of diagnostic criteria are discussed.