The Toddler Version of the Autism Diagnostic Interview-Revised (Toddler ADI-R): New Algorithms

Background:   The ADI-R is a standardized, semistructured, investigator-based interview for caregivers of autistic individuals, which provides diagnostic algorithms for the ICD-10 definition of autism (World Health Organization [WHO], 1992) and DSM-IV (American Psychiatric Association [APA], 1993).  The interview focuses on the three domains of functioning – language/communication; reciprocal social interactions; and restricted, repetitive, and stereotyped behaviors and interests. The Toddler-ADI-R exists in a research version, which includes additional questions about early development and symptom onset with total 132 questions.  The algorithms for the ADI-R have been used for children administered Toddler-ADI-R since the Toddler-ADI-R does not provide its own algorithms.  

Objectives: In an effort to develop the Toddler-ADI-R algorithms appropriate for younger children, the sensitivity and specificity of individual items have been explored in differentiating children with ASD clinical diagnoses from children with other nonspectrum disorders and children with typical development.  Exploratory and confirmatory factor analyses, building on recent work on the ADI-R, will be described as well.   

Methods: Analyses were conducted using a dataset of ADI and psychometric scores for 624 cases children aged 10 to 63 months (mean age=31.87).  259 cases were derived from children with autism; 156 from children with PDD-NOS; 127 from children with various non-ASD developmental delays (DD); 82 from children with typical development (TD).  ADI item distributions were compared by diagnosis, and items with high sensitivity and specificity were identified.  These items will be included in exploratory and confirmatory factor analyses for the development of the Toddler-ADI-R algorithms.

Results: We identified 37 items that showed high specificity and sensitivity.  There were12 items on which fewer than 20 % of children with autism scored 0 and not more than 20 % of children with DD and TD scored 2 or 3.  25 additional items were identified with more lenient criteria.  29 of these 37 items are included in the standard ADI-R algorithm (though as “ever” or most abnormal 4 -5 codes).  Several items not included in the original ADI-R algorithms showed high specificity and sensitivity for the children in the present study: sharing others’ pleasure and excitement, non-speech vocalization, elicited vocal imitation, greeting, social crying, affection, initiation of appropriate activities.

Conclusions: The items identified here will be included in the analyses for the development of the new Toddler-ADI-R algorithms to improve the diagnostic validity of the instrument.