Construct Validity of the ASRS in a Non-Clinical Diverse Sample

Background: For accurate identification of autism spectrum disorder (ASD), the selection ofappropriate, valid, and reliable measures is critical. This is particularly crucial when conducting ASD evaluations for culturally and linguistically diverse (CLD) populations. Research suggests Latino children are less likely to receive diagnosis than their White peers, and Black children are diagnosed at older ages than White children (Centers for Disease Control, 2006). ASD identification disparities have also been documented in the educational identification of ASD (Sullivan, 2013). If a child with ASD identified is late or incorrectly, they are missing critical early intervention services.

The majority of service providers use rating forms during ASD evaluations. One of the more utilized rating forms is the Autism Spectrum Rating Scales (ASRS; Goldstein & Naglieri, 2010). The ASRS measures ASD related symptoms, behaviors, and features using parent and teacher reports. Although there are several forms, the current study focuses on the ASRS Parent Form (6-18). The manual reports strong reliability and satisfactory content and criterion validity. An exploratory factor analysis (EFA) was conducted to determine construct validity. Results from the EFA suggested a three-factor structure that measures Social/Communication (SC), Unusual Behaviors (UB), and Self-Regulation (SR) as seen on the rating scales. Notably, no confirmatory factor analysis (CFA) was conducted to corroborate this.

Objectives: The purpose of this research was to investigate the factor structure of the ASRS in a diverse sample and compare said factor structure across subsamples in a non-clinical sample.

Methods: Parents (N=405) with children between the ages of 6 and 18 participated in the current study. Participants identified as Black (n=181), Latino (n=106), or White (n=181). Participants were recruited through Amazon’s Mechanical Turk. Following consent, participants completed demographic information and the ASRS. Internal consistency analyses and multiple CFA were performed.

Results: All three subscales exhibited high internal consistency in the combined sample. However, the CFA fit was poor for the combined and all three subsamples, with Black and Latino subsamples displaying the worst fit. Latent factor correlations were similar for the combined and White samples, although Black and Latino subsamples exhibited higher correlations between SR and UB (r=.95 and .93) subscales. SC was differentially related to the SR and UB subscales across subsamples. Subgroup CFAs also revealed disparities in the loading of two items across groups.

Conclusions: Results do not support the three-factor model found in the initial EFA (Goldstein & Naglieri, 2010). This may imply that the factor structure of the ASRS is not comparable across racial/ethnic groups. The high correlation between SR and UB subscales in Black and Latino subsamples may suggest that these latent factors are differentially related across subsamples. Results also suggest the social communication construct may differ across groups. Furthermore, some items may be more or less important in measuring ASD symptoms across groups. Future research specific to the ASRS should investigate measurement invariance, differential item functioning, and the role of core symptoms of ASD across various racial/ethnic groups.