Alexandra Sturm1, Megan Kuhfeld2, Connie Kasari3, James T McCracken1,4. 1. Jane and Terry Semel Institute for Neuroscience and Human Behavior, University of California, Los Angeles (UCLA), Los Angeles, CA, USA. 2. Population Research Center, The University of Texas at Austin, Austin, TX, USA. 3. Center for Autism Research and Treatment, UCLA, Los Angeles, CA, USA. 4. David Geffen School of Medicine, UCLA, Los Angeles, CA, USA.
Abstract
BACKGROUND: Research and practice in autism spectrum disorder (ASD) rely on quantitative measures, such as the Social Responsiveness Scale (SRS), for characterization and diagnosis. Like many ASD diagnostic measures, SRS scores are influenced by factors unrelated to ASD core features. This study further interrogates the psychometric properties of the SRS using item response theory (IRT), and demonstrates a strategy to create a psychometrically sound short form by applying IRT results. METHODS: Social Responsiveness Scale analyses were conducted on a large sample (N = 21,426) of youth from four ASD databases. Items were subjected to item factor analyses and evaluation of item bias by gender, age, expressive language level, behavior problems, and nonverbal IQ. RESULTS: Item selection based on item psychometric properties, DIF analyses, and substantive validity produced a reduced item SRS short form that was unidimensional in structure, highly reliable (α = .96), and free of gender, age, expressive language, behavior problems, and nonverbal IQ influence. The short form also showed strong relationships with established measures of autism symptom severity (ADOS, ADI-R, Vineland). Degree of association between all measures varied as a function of expressive language. CONCLUSIONS: Results identified specific SRS items that are more vulnerable to non-ASD-related traits. The resultant 16-item SRS short form may possess superior psychometric properties compared to the original scale and emerge as a more precise measure of ASD core symptom severity, facilitating research and practice. Future research using IRT is needed to further refine existing measures of autism symptomatology.
BACKGROUND: Research and practice in autism spectrum disorder (ASD) rely on quantitative measures, such as the Social Responsiveness Scale (SRS), for characterization and diagnosis. Like many ASD diagnostic measures, SRS scores are influenced by factors unrelated to ASD core features. This study further interrogates the psychometric properties of the SRS using item response theory (IRT), and demonstrates a strategy to create a psychometrically sound short form by applying IRT results. METHODS: Social Responsiveness Scale analyses were conducted on a large sample (N = 21,426) of youth from four ASD databases. Items were subjected to item factor analyses and evaluation of item bias by gender, age, expressive language level, behavior problems, and nonverbal IQ. RESULTS: Item selection based on item psychometric properties, DIF analyses, and substantive validity produced a reduced item SRS short form that was unidimensional in structure, highly reliable (α = .96), and free of gender, age, expressive language, behavior problems, and nonverbal IQ influence. The short form also showed strong relationships with established measures of autism symptom severity (ADOS, ADI-R, Vineland). Degree of association between all measures varied as a function of expressive language. CONCLUSIONS: Results identified specific SRS items that are more vulnerable to non-ASD-related traits. The resultant 16-item SRS short form may possess superior psychometric properties compared to the original scale and emerge as a more precise measure of ASD core symptom severity, facilitating research and practice. Future research using IRT is needed to further refine existing measures of autism symptomatology.
Authors: Charlotte DiStefano; Rujuta B Wilson; Carly Hyde; Edwin H Cook; Ronald L Thibert; Lawrence T Reiter; Vanessa Vogel-Farley; Joerg Hipp; Shafali Jeste Journal: Am J Med Genet A Date: 2019-10-26 Impact factor: 2.802
Authors: Kristen Lyall; Mina Hosseini; Christine Ladd-Acosta; Xuejuan Ning; Diane Catellier; John N Constantino; Lisa A Croen; Aaron J Kaat; Kelly Botteron; Nicole R Bush; Stephen R Dager; Cristiane S Duarte; M Daniele Fallin; Heather Hazlett; Irva Hertz-Picciotto; Robert M Joseph; Margaret R Karagas; Susan Korrick; Rebecca Landa; Daniel Messinger; Emily Oken; Sally Ozonoff; Joseph Piven; Juhi Pandey; Sheela Sathyanarayana; Robert T Schultz; Tanya St John; Rebecca Schmidt; Heather Volk; Craig J Newschaffer Journal: J Autism Dev Disord Date: 2021-07
Authors: Phuong H Nguyen; Maku E Ocansey; Meghan Miller; Dung T K Le; Rebecca J Schmidt; Elizabeth L Prado Journal: Autism Res Date: 2019-07-29 Impact factor: 5.216