Discriminative Validity of Social Responsiveness Scale (SRS) on Autism Diagnoses from a Community Study in Taiwan

Background: Researchers generally agree that autism spectrum disorder (ASD) represents one end of a spectrum of quantitative impairment that is continuously distributed in the general population.  The Social Responsiveness Scale (SRS) is a 65-item scale that characterizes quantitative impairments in social communication and repetitive behavior/restricted interests that define ASD, and therefore are expected to distinguish ASD from other developmental disabilities (DD) and development that is not affected by ASD or other DD.  We previously reported that SRS total and subscale scores distinguished significantly between ASD and other developmental disorders from a clinic-based study in Taiwan (Wang 2012); however, it is unknown similar findings can be replicated with a community-based study design.

Objectives: This study aimed to examine the discriminative validity of individual items and different cut-offs of the SRS, on ADOS, and clinician-determined ASD diagnoses.

Methods: A community-based study of autism in school children aged 6-8 was conducted in Taiwan.  Study participants were first screened using the SCQ and SRS. Children with parent-reported SCQ scores>=7 were invited for a clinical visit.  Twenty-seven boys and nineteen girls, and their primary caregiver, participated in the clinical visit which comprised the Autism Diagnostic Observation Schedule (ADOS), the Autism Diagnostic Interview-Revised (ADI-R), and a comprehensive evaluation with a child psychiatrist.  Three groups were classified based on the ADOS: ASD (n=12), subclinical (n=8), and unaffected (n=25).  The 3 groups based on clinical judgment were: ASD (n=7), other DD (n=24), and unaffected with ASD or other DD (n=15). Two cut-offs for the SRS were used to define higher versus lower risk for ASD.  As recommended in the SRS manual, a raw score of >=70 in males and >=65 in females is a cut-point that provides evidence for the presence of an ASD.  A raw score >=85 is considered “very strong evidence” of the presence of an ASD as suggested by studies from Western countries.

Results: Only 3, out of 65, items on the SRS statistically significant discriminate the 3 ADOS determined diagnostic groups; similarly, only 4 items statistically significant discriminate the child psychiatrist’s diagnostic groups.  The cut-off of >=85, but not the high risk cut-off (>=70 in males and >=65 in females), significantly discriminate diagnostic groups that were determined by the child psychiatrists (p=0.02).  However, neither of the cut-offs discriminate ADOS determined groups.  

Conclusions: Our results do not concur with findings from our clinic-based study, where we reported the SRS successfully distinguished Taiwanese children with ASDs from typical controls as well as from individuals with other psychiatric diagnoses. Possible reasons for the discrepant findings include: 1) This community-based study, with a multi-stage case identification design, yields smaller numbers of cases than our prior study, 2) Individual items on the SRS rather than the SRS full scale and domain scores were used to test the discriminative validity in this current study, and 3) Reported behavioral symptoms on participants recruited from a clinic setting may be different from those from the community.  Our findings hint sub-domains of the SRS with different cut-offs may be needed for community-based studies.