Analysis of Race and Sex Bias in the Autism Diagnostic Observation Schedule (ADOS-2)

• DOI: 10.1001/jamanetworkopen.2022.9498
• Publication Date: 2022-04-26T15:30:19Z
• Authors: Luther G. Kalb; Vini Singh; Ji Su Hong; Calliope Holingue; Natasha N. Ludwig; Danika Pfeiffer; Rachel Reetzke; Alden L. Gross; Rebecca Landa
• Keywords: Differential item functioning

<h3>Importance</h3> There are long-standing disparities in the prevalence of autism spectrum disorder (ASD) across race and sex. Surprisingly, few studies have examined whether these arise partially out systematic biases Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), reference standard measure ASD. <h3>Objective</h3> To examine differential item functioning (DIF) ADOS-2 items sex race. <h3>Design, Setting, Participants</h3> This is a cross-sectional study children who were evaluated for ASD between 2014 2020 at specialty outpatient clinic located Mid-Atlantic region US. Data analyzed from July 2021 to February 2022. <h3>Exposures</h3> Child (Black/African American vs White) (female male). <h3>Main Outcomes Measures</h3> Item-level harmonized algorithm items, including social affect (SA; 10 items) repetitive/restricted behaviors (RRBs; 4 items), 3 modules. Measurement bias was identified by examining DIF test (DTF), within graded response, response theory framework. Statistical significance determined likelihood ratio χ²test, series metrics used magnitude DTF. <h3>Results</h3> A total 6269 (mean [SD] age, 6.77 [3.27] years; 1619 Black/African [25.9%], 3151 White [50.3%], 4970 male [79.4%]), included this study. Overall, 16 140 diagnostic (11%) had significant DIF. For race, 8 DIF, 6 which involved SA. No single showed consistently all Most with greater difficulty poorer discrimination compared children. sex, 5 split SA RRB. However, hand mannerisms evidenced algorithms, generally difficulty. The only moderate large 2 items: (among female children) repetitive interests children). overall estimated effect on DTF not large. <h3>Conclusions Relevance</h3> These findings suggest that does widespread measurement or raise some concerns around underdetection warrant further research.