Measuring autism with the ADOS-2 using a bifactor model

Abstract

The measurement of autism characteristics can be challenging due to variability of social impairments and restricted and repetitive behaviors or interests (RRBs). Psychometrically strong measures such as the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) can improve our capacity for thorough autism assessment. The conceptualization of the ADOS-2 has been shaped by research exploring the structure of its items, which evaluate autism traits associated with social affect and RRBs. Continuously refining our understanding of these items and their relations to other characteristics, such as cognition, is crucial for more accurate autism assessment and diagnosis. This study used data from a sample of 188 school-age children with mostly average cognitive functioning referred for clinical autism evaluations to (1) test the dimensionality of the ADOS-2, Module 3 (appropriate for children with relatively higher verbal ability), across two sets of items (i.e., algorithm only, algorithm with three non-algorithm RRB items) using confirmatory factor analysis (CFA) and (2) examine the relations of cognition to the dimensions tested in the ADOS-2. A bifactor model, featuring a general autism trait and two subfactors (RRB and Social Affect), provided superior fit for algorithm-only and algorithm with three non-algorithm items. Cognitive functioning was not significantly related to the general or specific factors in the model with only algorithm items. While the findings support the validity of the ADOS-2, it may not fully capture RRBs among children referred for autism. This study enhances our understanding of the ADOS-2, highlighting the utility of a bifactor model for characterizing its dimensionality, measuring autism traits with minimal cognitive influence, and identifying its limitations in assessing RRBs.