The Interrater Reliability of the Autism Diagnostic Interview-Revised (ADI-R) in Clinical Settings

Abstract

Background: The Autism Diagnostic Interview-Revised (ADI-R) is considered a first choice assessment tool in autism spectrum disorder. Nevertheless, despite its wide use in psychiatric practice and recommendations by various clinical guidelines, its interrater reliability has predominantly been confirmed in research settings by specially trained, research reliability interviewers. The reliability of ADI-R assessments among clinicians has not yet been established. Therefore, this study examined the spontaneous interrater reliability of the ADI-R in a naturalistic clinical multicenter setting. Sampling and Methods: Ten video-recorded ADI-R administrations were rated by 5 different raters each from a pool of 11 raters affiliated to 8 different clinical sites. Results: The interrater reliability for the 12 diagnostic criteria operationalizing autism spectrum disorders according to DSM-IV/ICD-10 in the ADI-R algorithms ranged between G(q,k) (analogous to intraclass correlations) = 0.96 and 0.99 for reciprocal social interaction, 0.96 and 1.00 for communication, and 0.91 and 0.97 for repetitive and restricted behavior. Reliability of diagnostic classification was K Cohen 0.83. Conclusions: The findings endorse the psychometric properties of ADI-R in terms of interrater reliability previously reported from research settings and support their generalization to common clinical settings. Limitations of this study include an unbalanced sample composition.