Estimating the impact of missed cases on the accuracy of autism screening tools

A common problem for assessing psychiatric screening tools is that initial diagnostic accuracy estimates are conducted on cross-sectional data and later found to be lower than expected when longitudinal analyses are conducted. This article uses prevalence estimates to identify potentially missed cases and adjust diagnostic accuracy metrics. To display this approach, we meta-analyze and assess 27 population-level autism screening studies identified via an umbrella review and contrast four population adjustments (none, national, U.S.-centric, and world prevalence). Studies with no positive screen adjustments displayed adequate sensitivity (.75), but poor sensitivity resulted when applying national (.52), world (.50), and U.S.-centric (.33) population adjustments. We also address missing positive screen cases due to the common problem of attrition (Sheldrick et al., 2023). Positive screen adjustments suggested a range of sensitivity impacts ranging from very poor (.32) to good (.89). Across analyses, positive predictive value estimates also ranged from .20 to .73 depending on population and missed case assumptions. The strengths and benefits of blending epidemiology and psychometric perspectives to identify screening tools that are weaker than expected are discussed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).