Comparison of the Original and Revised ADOS Algorithms in a Children's Community Mental Health Clinic

Background: As the incidence of ASDs has increased, so have referrals for diagnostic assessment in our children's community mental health clinic. Beginning in 2006, children with suspected ASD have been assessed using the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000), which has been scored using the original and revised (Gotham et al., 2006) algorithms.

Objectives: In a community-based clinic, to examine how the original and revised ADOS algorithms compare in the classification of children relative to contemporaneous clinical diagnoses, and to compare the new ADOS module divisions.

Methods: Since 2006, ten consecutive referrals for suspected ASD were seen for diagnostic assessment. Diagnosis was based on clinical best judgement by an experienced clinician using DSM-IV criteria, a structured diagnostic interview, and the ADOS. Clinical diagnosis was independent of ADOS scores. Children ranged in age from 3 to 10 years (nonverbal IQ > 70, with at least single words).

Results: There was 60% agreement between clinical diagnosis (ASD vs. Non-ASD) and the original algorithm, but it increased to 90% with the revised algorithm. Sensitivity was 63% for the original algorithm, whereas it was 100% for the revised algorithm. Specificity was 67% for both algorithms. Further comparison of the revised algorithm to clinical diagnosis (Autism vs. ASD vs. Non-ASD) revealed 100% congruence for Module 1–Some Words, Module 2–Younger, and Module 3, but only 50% for Module 2–Older. The noncongruent cases included one child with ASD whose high RRB score put him into the Autism range, and another child with ADHD, ODD, and LD who had sub-threshold ADI-R scores despite being in the Autism range on the ADOS.

Conclusions: These preliminary data suggest that the revised ADOS algorithm may show better agreement than the original algorithm with contemporaneous clinical diagnosis in our community-based clinic.