Validity of Standardized Diagnostic Instruments for Autism Spectrum Disorders in Toddlers Recruited from a Population-Based Cohort: The Influence of Parental Concern

Friday, May 13, 2016: 5:30 PM-7:00 PM
Hall A (Baltimore Convention Center)
A. Havdahl1,2,3, S. L. Bishop4, P. Suren3, A. S. Oyen2,3, C. Lord1, A. Pickles5, S. von Tetzchner6, S. Schjolberg7, N. Gunnes3, M. Hornig8, M. Bresnahan8, W. I. Lipkin8, E. Susser9, P. Magnus3, N. Stenberg10, T. Reichborn-Kjennerud3,6 and C. Stoltenberg3,11, (1)Weill Cornell Medical College, White Plains, NY, (2)Lovisenberg Diaconal Hospital, Oslo, Norway, (3)Norwegian Institute of Public Health, Oslo, Norway, (4)Department of Psychiatry, University of California San Francisco, San Francisco, CA, (5)Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, United Kingdom, (6)University of Oslo, Oslo, Norway, (7)Mental Health, Norwegian Institute of Public Health, Oslo, Norway, (8)Columbia University, New York, NY, (9)Epidemiology and Psychiatry, Columbia University, New York, NY, (10)Oslo University Hospital, Oslo, Norway, (11)University of Bergen, Bergen, Norway

Diagnostic instruments for autism spectrum disorders (ASD) have been validated primarily in samples of children referred to ASD specialty clinics for suspected ASD, and there is little information about the performance of these instruments when used with children identified in other ways.


This study examined the validity of widely used diagnostic instruments for ASD among toddlers recruited from a population-based study, and the influence of the identification route.


Participants were 679 toddlers (age 35-47 months) who received multi-disciplinary diagnostic evaluations of ASD. We examined agreement of the Autism Diagnostic Interview-Revised (ADI-R) (Toddler algorithms) and the Autism Diagnostic Observation Schedule (ADOS) (ADOS-2 algorithms) with clinical best-estimate diagnoses. Receiver operating characteristic and logistic regression analyses were used to assess the diagnostic accuracy of the ADI-R and ADOS scores and cutoffs, and the additive value of combining the instruments. The analyses were carried out in the overall sample as well as for the comparison of toddlers identified based on 1) parental concern about ASD and 2) parent-reported behavioral signs of ASD (screening) without a specific concern about ASD.


The ADOS cutoffs showed consistently well-balanced sensitivity and specificity, whereas the ADI-R cutoffs missed a substantial proportion of toddlers ultimately diagnosed with ASD, especially among toddlers whose parents did not have a specific concern about ASD (43%). However, continuous scores from both the ADI-R and ADOS agreed well with clinical diagnoses (AUC=0.87-0.95), and contributed additively to their prediction (p<0.001). When applying modified ADI-R cutoffs, good balance between sensitivity and specificity was found also among toddlers without parental concern about ASD.


The results underscore the utility of combining standardized instruments based on parental interview and clinician observation in diagnosing ASD. The established ADI-R cutoffs had low sensitivity among toddlers whose parents did not have a specific concern about ASD. Future studies should examine the utility of adjusting classification criteria on the ADI-R and other parental report-based ASD instruments to account for the influence of parental ASD concern.