Abstract
Studies of diagnostic reliability are important for quality control both in clinical work and in research. In clinical work such studies are often neglected. This is undesirable and may lead to use of inadequate procedures. Twenty-five children were included in an inter-rater reliability study of four commonly used diagnostic instruments for classification of autism and other pervasive developmental disorders (PDD). The instruments were ICD-10 research diagnostic criteria, DSM-III-R diagnostic criteria. Childhood Autism Rating Scale (CARS), and Autism Behavior Checklist (ABC). Estimates for inter-rater reliability between two experienced clinicians were calculated. The highest reliability was obtained using ICD-10 research criteria, and the lowest reliability using the CARS. Implications for clinical practice as to selection of diagnostic instruments and systems are discussed.