Specificity of the Social Responsiveness Scale When Used with Children with Other Diagnoses

Background: Child characteristics not specific to Autism Spectrum Disorders (ASD), such as behavior problems and IQ, have been shown to influence scores on ASD screening instruments in children with ASD (e.g., Corsello et al., 2007; Hus et al., 2012) . Studies examining whether child characteristics affect screening measure scores in children with non-ASD diagnoses have been limited to relatively small, heterogeneous samples (e.g., Charman et al., 2007; Warren et al., 2011). Non-ASD-specific influences on ASD screeners may contribute to false positives that result in inappropriate referrals to specialty ASD clinics when used in clinical settings and erroneous inclusion of children with non-ASD diagnoses in research studies.

Objectives: To explore relationships between non-ASD-specific child characteristics and scores on the Social Responsiveness Scale (SRS; Constantino & Gruber, 2005), a widely used parent-completed ASD screening questionnaire, in a sample of children with non-ASD diagnoses.

Methods: Participants were 161 children with non-ASD diagnoses recruited for a research study. Clinicians blind to previous diagnoses made clinical diagnoses based on information from comprehensive assessments for ASD including questionnaires, the ADI-R, Vineland-II, ADOS and cognitive testing. Diagnoses were grouped into: Intellectual Disability, Anxiety Disorders, ADHD, Language Disorders and Other. Using these clinician diagnoses as the “gold standard,” specificity for the overall sample and individual diagnostic groups was calculated for T-score (60 or 76) and raw score (65 for females/70 for males or 85 for both) cut-offs recommended by the SRS authors for use in different settings. T-Tests were used to compare characteristics of children who did and did not meet cut-offs. Logistic regression was used to determine best predictors of meeting SRS cut-offs; binary predictors grouping children according to common divisions of standard scores (below average/average; borderline to clinical range of concern/normal) were used to compute odds ratios.

Results: Overall specificity ranged from 29-70%. Specificity was lowest when using the ASD cut-off (raw score=65/70) recommended for general population screening and for groups with Intellectual Disability, ADHD, or Other diagnoses. Children meeting cut-offs had more impaired social and expressive communication skills on the Vineland-II, more behavior problems on the Child Behavior Checklist, and lower IQ (p<.001) than children who did not meet cut-offs. For children with Internalizing or Externalizing scores in the borderline/clinical range, the odds of meeting the cut-off was 5.46 and 3.52 times the odds of children in the “normal” range. Post-hoc analyses indicated that scores in the borderline/clinical range on the Anxious-Depressed, Withdrawn-Depressed, Attention Problems, or Syndrome Scale significantly predicted meeting the ASD cut-off on the SRS (odds ratios of 5.53, 4.64 and 3.58, respectively).

Conclusions: Consistent with previous reports in ASD samples, non-ASD-specific characteristics, particularly behavior problems, have considerable effects on SRS scores in non-ASD samples and reduce specificity of this screening instrument. Clinicians and researchers should exercise caution when using the SRS as a screening instrument in populations of children with elevated emotional and behavioral difficulties. Results will be further discussed for different diagnostic groups and from the perspective of using the SRS as a dimensional measure of ASD-related symptoms or social-communication impairments.