Assessment of Symptom Severity in Siblings with Autism Spectrum Disorder: Comparing Parental Report Versus Direct Observation Methods Using Multilevel Modeling

Background: Despite high heritability estimates for Autism Spectrum Disorder (ASD), so far, only a few inherited genes have been linked to an increased risk for the disorder. This failure to identify inherited autism genes raises important methodological questions concerning our ability to clearly define and/or accurately measure the ASD phenotype. The most common approaches to the assessment of ASD symptoms include a direct observation of the child and a semi-structured interview of the parent. However, to date, these two approaches have never been contrasted among multiplex families (i.e., families with more than one child with ASD). 

Objectives: This study explores issues related to the assessment of the ASD phenotype in multiplex ASD families. The specific objectives of this study are (a) to estimate the “between” (shared) versus “within” (unique) family variance in ASD symptom severity across two assessment methods, direct child observation and parental report; and (b) to examine  the effect of child and family-level covariates on ASD symptom severity across the two assessment methods.

Methods: Data came from the Canadian Genetics of Autism Study investigating genetic mechanisms of autism susceptibility. The sample consisted of 110 multiplex families who had two children diagnosed with ASD (N=220). All children were assessed using the Autism Diagnostic Interview – Revised (ADI-R; parental report) and the Autism Diagnostic Observation Schedule (ADOS; direct observation). Multilevel regression was used to examine the between-family and within-family variance in ASD symptom severity across the two assessment methods, parental report and direct observation.

Results: Results show that for both methods – parental report (ADI-R) and direct observation (ADOS) – most of the variance in ASD symptom severity is at the child-level (i.e. within families). However, the proportion of variance at the family-level (i.e. between families) is substantially higher for ADOS scores (20.23%) compared to ADI-R scores (11.84%); the reverse pattern is true for the proportion of variance explained at the child-level. Additional results indicate that variables such as the child’s age and sibling-pair age discrepancy influence the estimation of child scores and familial correlations of ASD symptom severity.

Conclusions: This study provides the first empirical evidence suggesting the presence of the “sibling de-identification” phenomenon in multiplex ASD families. Specifically, study findings show that when reporting on the symptom severity of two of their children with ASD using the ADI-R, parents tend to “maximize” differences and “minimize” similarities between siblings (relative to differences observed on the ADOS). The implications of these findings are discussed within the larger context of collecting and evaluating phenotypic information for genetic research.