The Under-Identification of Autism in Females: A Review and Analysis of Sex-Based Scoring Differences Observed in Autism Diagnostic Observation Schedule (ADOS) Module 3

Emerging research highlights sex differences in autism presentation, raising questions about the validity of clinician-administered diagnostic tools like the gold-standard Autism Diagnostic Observation Schedule (ADOS). This study examines whether clinicians’ perceptions, potentially biased toward male-typical presentations, contribute to sex bias in ADOS items and the under-identification of autism in females. In a sample of 813 children, multidimensional item response theory graded response models were employed to analyze differential item functioning (DIF) of items in ADOS Module 3 (the version commonly administered to school-aged children with fluent spoken language). Six items showed significant sex bias. On four social communication items, females tended to be underscored, meaning rated as showing fewer autistic features (e.g., more emotion talk) than males with equivalent levels of autism latent traits. In contrast, females tended to be overscored on two items related to restricted or repetitive behaviors and interests (i.e., more finger mannerisms, specific interests, or repetitive behaviors). The impact of sex bias on overall scores was small, ranging from 0.10 to 0.28 points per item, and approximately 0.41 points on the overall test score. This analysis joins three previous reports of specific ADOS items that function differently in females than males, with converging evidence identifying two problematic items (D2 and D4). Overall, DIF was small and unlikely to affect ADOS classification. We speculate that the direction and consistency of DIF in social communication items might reflect underlying phenomena (viz., masking, clinician bias) that may relate to female underdiagnosis more generally.