Examining the Stability of the Autism Diagnostic Interview-Revised In the Autism Genome Project Sample of Children 4 to 18 Years

Background: The Autism Diagnostic Interview-Revised (ADI-R) is a standardized semi-structured interview for caregivers of children and adults used in assessing autism spectrum disorder (ASD). The instrument focuses on behaviour in three areas - reciprocal social interaction, communication and language, and restricted and repetitive, stereotyped interests and behaviors. Earlier published works have indicated different numbers of quantitative traits using the ADI-R items.

Objectives: The objective of this study is to examine the stability of the ADI-R in subgroups of children with Autism Spectrum Disorder and to derive psychometrically sound quantitative traits from the ADI-R that can be used in genetic analyses.

Methods: Data came from the Autism Genome Project (AGP), a collaborative genetics research project studying genetic mechanisms of autism susceptibility. The fixated interests and repetitive behaviours (FIRB) items were selected from the repetitive, stereotyped interests and behaviors and items related to it. The social communication  (SOCCOM) items were selected from the algorithm items for reciprocal social interests and non-verbal communication. The sample consisted of 4237 individuals with autism or ASD, selecting one random individual aged between 4 and 18 years per family from the combined Phases I and II data sets. The sample was divided into two equal samples – Exploratory and Confirmatory - using random selection. The analyses consisted of (a) examining the factor structure using the exploratory samples, assessing the fits of selected models using the confirmatory samples and (b) finally, the measurement invariance of the best-fitting factor models with the full sample consisting of a random sample of verbal subjects equal to the number of non-verbal subjects in the sample were tested in a confirmatory factor analysis (CFA) framework.

Results: The results indicate that: (a) CFA indicated that the best fitting models for the data are a 2 factor solution for FIRB (insistence on sameness, IS and repetitive stereotyped behaviours and unusual sensory interests, RSMB; CFI=0.943,TLI=0.924,RMSEA=0.038) and a 4-factor solution for SOCCOM (joint attention, JTATT; social interactions, SOCINT; peer interactions, PEERINT; and nonverbal communication, NVCOMM; CFI=0.921,TLI=0.907,RMSEA=0.50). The means of the non-verbal group are higher than that of the verbal group for factors of the SOCCOM and FIRB except for IS and mean scores were higher for autism than ASD on SOCCOM factors. (b) Tests of measurement invariance showed that the final factor solutions measured the same constructs in the groups. However, levels of intensity of behaviour were different by verbal status group for both the FIRB and SOCCOM.

Conclusions: The constructs measured by the FIRB related items and the SOCCOM algorithm items of the ADI-R are the same in both verbal and nonverbal subjects in the AGP sample. There are differences in levels of FIRB as well as the SOCCOM factors between the two groups (verbal vs. nonverbal individuals) in the AGP sample. Differences in levels of FIRB and SOCCOM factors can be attributed to different levels of intensity of behaviour between the groups. Thus, the use of the ADI-R quantitative traits in genetic analysis should take into account the differences between verbal groups.