Characteristics of Children Misidentified by the SCQ in a Clinic-Referred Sample

Background: The Social Communication Questionnaire (SCQ) is a checklist of ASD symptoms based on a well-validated diagnostic interview, the ADI-R. In initial validation studies, the SCQ showed promise as a screener for a research-referred sample (Berument et al., 1999). Subsequent studies using the SCQ with clinic-referred samples revealed low sensitivities and specificities in differentiating children with ASD from children with nonspectrum disorders (e.g., Corsello et al., 2007; Kochhar et al., 2010). Checklists such as the SCQ are often relied on in clinical settings to inform diagnostic decisions, especially when clinicians lack specialized training in ASD diagnosis. The SCQ is commonly used as a criterion for inclusion in research. Information is needed on the characteristics of children who tend to be misclassified by the SCQ to guide decision-making in research and clinical settings.

Objectives: Evaluate accuracy of the SCQ in a clinic-referred sample of children suspected of having ASD and characterize clinical features of children who were misclassified.

Methods: SCQ and psychometric data were analyzed for 100 consecutive referrals to an ASD specialty clinic. Participants were 4 to 18 years of age. Receiver Operating Characteristic (ROC) curves were used to determine sensitivity and specificity for children with best estimate diagnosis of ASD or nonspectrum diagnoses. Assessment data of children who were false positives or false negatives were examined to identify possible associations with being misclassified, including verbal and nonverbal IQ, adaptive skills, problem behaviors on the Behavior Assessment System for Children-2^nd edition (BASC-2), demographics, and ASD severity as measured by the ADOS calibrated severity score.

Results: The SCQ showed low sensitivity and specificity using the recommended cutoff of 15, with accuracy of classification at or around chance. Lowering the cutoff to 12 slightly improved sensitivity without further reduction in specificity. Binary logistic regression analysis demonstrated that the odds of false positives on the SCQ significantly increased as nonverbal IQ decreased and significantly increased as a child’s number of diagnoses increased. False negatives on the SCQ in our preliminary sample did not show a clear pattern, except that a higher BASC-2 Somatization t-score increased the odds of being missed.

Conclusions: Our preliminary findings affirm previous findings of low sensitivity and specificity for the SCQ in clinic-referred samples. Our analyses suggest children who were false positives on the SCQ had characteristics, such as multiple diagnoses and lower nonverbal IQ, that are accepted to be confounding factors in ASD diagnosis in general. On the other hand, those missed by the SCQ did not show a pattern of fewer ASD symptoms on the ADOS or fewer problem behaviors compared to true positives. Future studies are needed to determine whether individual parent-endorsed symptoms on the SCQ are supported in well-validated direct measures of ASD symptoms, such as the ADOS, even for children who receive nonspectrum diagnoses. This would further address whether (a) the SCQ is measuring the presence of general behavioral or developmental problems rather than ASD-specific symptoms, or (b) the SCQ reflects the overlap in behavioral symptoms of ASD and other childhood disorders.