Module-specific diagnostic accuracy of ADOS-2 in real-world clinical referral populations: an updated systematic review and HSROC meta-analysis

Abstract

Abstract

Background The Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), is widely used in the diagnostic evaluation of autism spectrum disorder (ASD); however, its diagnostic performance in real-world clinical referral populations remains heterogeneous, particularly across modules and clinical contexts. This systematic review and meta-analysis evaluated the module-specific diagnostic accuracy of ADOS-2 using hierarchical meta-analytic modeling and examined sources of heterogeneity in updated evidence clinical studies. Methods A systematic search of PubMed/MEDLINE, Scopus, and Web of Science was conducted from January 2021 to February 2026, with additional screening of reference lists. Studies were included if they evaluated ADOS-2 diagnostic accuracy in real-world clinical referral populations, used DSM- or ICD-based clinical best-estimate diagnosis as the reference standard, and reported extractable 2×2 data. Diagnostic accuracy was pooled using a hierarchical summary receiver operating characteristic (HSROC) model with a bivariate random-effects approach. Module-specific analyses (Toddler Module, Modules 1–2, Module 3, Module 4) and meta-regression were performed to examine heterogeneity. Results Ten studies were included in the qualitative synthesis, and six provided extractable 2×2 data for quantitative pooling. Overall pooled sensitivity was 0.88 (95% CI: 0.83–0.92) and pooled specificity was 0.74 (95% CI: 0.68–0.80), with an HSROC AUC of 0.86. The Toddler Module showed the highest diagnostic performance (sensitivity 0.92; specificity 0.88; AUC 0.94), whereas specificity decreased in Modules 3 and 4. Meta-regression identified module level, psychiatric referral setting, and adult samples as significant contributors to reduced specificity. No significant publication bias was detected. Conclusions ADOS-2 demonstrates high overall sensitivity but variable specificity across modules in real-world clinical referral populations. Reduced specificity was more commonly observed in higher ADOS-2 modules, which are typically administered to verbally fluent adolescents and adults with greater psychiatric complexity.