Child Characteristics As Moderators of Parent-Clinician Agreement on Autism Symptoms

Background: Gold standard diagnosis of autism spectrum disorder (ASD) is based upon both parent report of child history/behavior and thorough observational assessment by a trained clinician. Marked disagreement between these reporters could result in misdiagnosis, in which parents fail to initiate assessment of a child with unrecognized ASD or clinicians fail to confirm valid parent concerns. Although parent and clinician assessments do often correlate with one another, the degree of correspondence likely varies according to child characteristics. However, these factors have not been identified, and improved understanding of child or family characteristics influencing symptom agreement could identify families at highest risk for misdiagnosis.

Objectives: Our goal was to explore factors moderating the degree of correspondence between (1) parent report of ASD symptoms on the Autism Diagnostic Interview, Revised (ADI-R), and (2) clinician assessment of ASD symptoms on the Autism Diagnostic Observation Schedule (ADOS). Putative moderators included child age, sex, IQ, adaptive behavior, and comorbid social-emotional difficulties.

Methods: Children and adolescents (ages 4 to 18 years; N = 79) with ASD or a twin with ASD were recruited through two studies encompassing four research sites across the US. Parents completed the ADI-R, CBCL, and Vineland-2, and youth completed the ADOS Module 3 and a standardized cognitive assessment (either DAS-II or WASI). To address parent-clinician correspondence, we created a series of linear regression models predicting ADOS scores. In each model, we entered ADI-R total score, one putative moderator, and their interaction as predictors. Next, significant predictors and their interactions were entered into a new regression model to more directly compare their relative contributions in the prediction of ADOS scores.

Results: The first set of models revealed main effects of ADI-R scores (ß=6.76, p<.001), adaptive behavior (ß=-.41, p=.05), and CBCL behavior problems (ß=.51, p<.01) in the prediction of ADOS scores. Adaptive behavior (ß=.82, p=.08) and behavior problems (ß=-2.22, p<.001) also interacted with ADI-R scores, indicating that they moderated parent-clinician correspondence on ASD symptoms. Interactions were such that parents and clinicians had better agreement when children had higher levels of adaptive behavior and fewer behavior problems. The subsequent model including ADI-R scores, adaptive behavior, behavior problems, and their interactions was significant (F(6, 58)=9.28, p<.001) and accounted for 46% of the variance in ADOS scores. Within this model, the interaction of ADI-R scores and behavior problems emerged as the strongest predictor (ß=-1.87, p=.02), underscoring the notion that child behavioral symptoms moderate the extent to which parents and clinicians report similar levels of ASD symptoms. See Figure 1.

Conclusions: Comorbid psychological difficulties appear to increase the discrepancy between parent and clinician assessment of ASD symptoms. Such comorbidities are quite common, and this discrepancy likely complicates the diagnostic process and affects families’ reception of an ASD diagnosis. More fully characterizing these and additional moderators influencing reporter agreement will be important in improving diagnostic accuracy and family experiences, and future analyses of ongoing data collection will expand to address additional factors such as family socio-economic status, race/ethnicity, and medical comorbidities.