Testing the construct validity of proposed criteria for DSM-5 autism spectrum disorder

To use confirmatory factor analysis to test the construct validity of the proposed DSM-5 symptom model of autism spectrum disorder (ASD), in comparison to alternative models, including that described in DSM-IV-TR.

Participants were 708 verbal children and young persons (mean age, 9.5 years) with mild to severe autistic difficulties. Autistic symptoms were measured using the Developmental, Dimensional and Diagnostic interview (3Di). The fit of the two-factor DSM-5 model, which has a social communication and a restricted, repetitive behavior (RRB) factor, was compared with that of alternative models. In one half of the sample, properties of the DSM-5 model were examined to investigate the validity of specific diagnostic criteria, informing the development of a better fitting DSM-5 model. This was then cross-validated in the remaining "hold-out" half of the sample; and its stability was tested across groups defined by age, sex, and symptom severity.

The DSM-5 model was superior to the three-factor DSM-IV-TR model. It was improved by the removal of items measuring "play and imagination" and "stereotyped and repetitive use of language." A scale measuring sensory abnormalities was added to the model, and loaded onto its RRB factor. This DSM-5 model fit well in the hold-out sample; was stable across age and sex; and fit adequately in those with clinical and sub-threshold autistic presentations.

Among higher-functioning individuals, ASD is a dyad, not a triad, with distinct social communication and repetitive behavior dimensions. As suggested in the proposed DSM-5 criteria, sensory abnormalities are part of the RRB symptom cluster.