Combining Information From Multiple Sources in the Diagnosis of Autism Spectrum Disorders Using the New ADI-R Algorithms for Toddlers From 12 to 47 Months of Age

Background: Diagnostic instruments such as the Autism Diagnostic Interview-Revised (ADI-R) and Autism Diagnostic Observation Schedule (ADOS) are intended to be used together, and researchers have made a systematic attempt to evaluate how information from those two instruments should be combined together (Risi et al., 2006).  However, the ADI-R algorithm used in the previous study was a “2 years to 3 years and 11 months algorithm” and previous studies have shown that this algorithm resulted in relatively poor sensitivity and specificity for the comparison between toddlers and preschoolers with Autism Spectrum Disorders (ASD) and those with nonspectrum disorders (NS) (Lord, Storoschuk, Rutter, & Pickles, 1993; Ventola et al., 2006).  Thus, in this study, we used newly developed ADI-R algorithms that have shown better predictive validity than the preexisting algorithm (Kim & Lord, in prep).

Objectives: The purpose of this study is to propose standard criteria for the combined use of the ADI-R and ADOS to diagnose cases of ASD for toddlers from 12 to 47 months of age.

Methods: Analyses were conducted using a dataset of the ADI-R and ADOS psychometric scores for 731 children aged from 12 to 47 months (515 cases with ASD; 142 with NS; 74 from children with typical development).  Since the new algorithms were developed separately for different groups of children divided by their chronological age and language level, the cases were divided into three groups to examine different standard criteria using information from the ADI-R and/or ADOS.  The three groups were: all children from 12 to 20 months of age and nonverbal children from 21 to 47 months; children with single words from 21 to 47 months; and children with phrase speech from 21 to 47 months.  The criteria tested in this study were: meeting the cutoff scores for ASD on 1) both the ADOS and ADI-R; 2) either the ADOS or ADI-R; 3) the ADOS alone; 4) the ADI-R alone.

Results: For all three groups of children, the first criteria of meeting the cutoffs for both the ADOS and ADI-R resulted in well balanced, high sensitivity and specificity (ranging from 71-88% sensitivity and 70-92% specificity by developmental cells).  When the second criteria (meeting the cutoffs on either the ADOS or ADI-R) was used, sensitivity was very high (ranging from 96-98%) but specificity was relatively poor (48-73%).  Similarly, when the ADOS was used alone, sensitivity was very high (ranging from 94-98%) but specificity was poor (59-79%).  When the ADI-R was used alone, sensitivity and specificity were well balanced but both were lower than those of the first criteria.

Conclusions: ASD diagnostic criteria using combined information from both the ADI-R and ADOS better reflect clinical judgments of ASD than any single instrument for toddlers under 4 years of age.