Utility of the SCQ and RBS-R to Confirm Caseness of Adolescents and Adults in ASD Research

Background: Better tools are needed to screen and diagnose adolescents and adults with ASD (IACC, 2013). The Autism Diagnostic Interview-Revised (ADI-R; Rutter et al., 2003a) may be useful to establish “caseness,” but is cost and time-intensive and often infeasible for parents of older individuals. Less is known about the predictive validity of the Social Communication Questionnaire Lifetime and Current forms (SCQ-L, SCQ-C; Rutter et al., 2003b) in older samples. The original SCQ validation (Berument et al., 1999) included individuals from 4-40 years, but did not report findings separately by age group. In a study of adults with ID, both SCQs yielded good sensitivity (79-89%), but low specificity (SCQ-C=66%; SCQ-L=48%; Sappok et al., 2015). No study has examined the predictive validity of the SCQ in adolescent or adult ASD samples spanning the full range of cognitive ability.

Objectives: This study aims to 1) compare sensitivity and specificity of the SCQ-L, SCQ-C and ADI-R and 2) assess whether combination with the Repetitive Behavior Scale-Revised (RBS-R; Bodfish et al., 2000) improves predictive validity.

Methods: Participants were drawn from a study focused on improving ASD diagnostic methods for adolescents and adults. Analyses included 49 individuals with diagnoses of ASD and 12 participants with non-ASD diagnoses (Age M=19.70, SD=3.43; NVIQ M=79.54, SD=32.57) who had an ADI-R, SCQ-C, SCQ-L and RBS-R completed at a single time point. ADI-R sensitivity and specificity were based on standard algorithm cutoffs and CPEA criteria (Lainhart et al., 2006); for both SCQ forms, cutoff=15. ROC analyses were used to derive cutoffs for SCQ-C, SCQ-L and RBS-R.

Results: For both groups, the SCQ showed moderate to strong correlations with the ADI-R and IQ. The ADI-R yielded the best balance of sensitivity (78-88%) and specificity (67-71%) using standard or CPEA criteria, respectively. The SCQ-L offered comparable sensitivity (76%), but poorer specificity (58%); a cutoff of 16 yielded somewhat better specificity (67%) without significant reduction in sensitivity (73%). In contrast, the SCQ-C yielded superior specificity (92%), but poor sensitivity (47%); lowering the cutoff to 13 reduced specificity (67%) and only improved sensitivity to 61%. Combining SCQ-L and SCQ-C did not improve predictive validity over either form alone. Although not intended as a screener, an RBS-R cut-off of 10 yielded results comparable to the SCQ-L (sensitivity=71%, specificity=58%). Requiring participants to meet on the SCQ-L or RBS-R provided the same balance of sensitivity (78%) and specificity (67%) as the standard ADI-R. Combining the SCQ-C and/or the RBS-R resulted in good specificity (>83%), but sensitivity remained below 62%.

Conclusions: Combining the SCQ-L and RBS-R yielded comparable sensitivity and specificity to the ADI-R, suggesting this combination may be a useful alternative to confirm “caseness” in adolescent and adult ASD research. In a sample of more diverse cognitive abilities, SCQ-C provided higher specificity, but lower sensitivity than studies of individuals with ID (Sappok et al., 2015); similar results with the SCQ-L were achieved, albeit with a lower cutoff. Future directions include replicating results in an independent sample and exploring whether combination with the ADOS further improves predictive validity.