Abstract Background/Introduction Despite guideline based indications, nearly 1/3 of patients with symptomatic severe aortic stenosis (AS) are not referred for intervention. Artificial Intelligence Decision-Support Algorithms (AI-DSAs) may be helpful to identify individuals with AS at risk for premature mortality but few have been validated for use. Purpose To evaluate an AI-DSA based on echocardiogram report data to augment the detection of severe AS within a well-resourced health care setting. Methods Blinded to clinical information, an AI-DSA trained to identify an aortic valve area (AVA) <1cm2 using minimal input data was applied to routine transthoracic echocardiogram reports from 31,141 US Medicare beneficiaries (≥ 65 years old) at a large US academic medical center (2003-2017) (Figure). Results Performance of the AI-DSA to detect an AVA <1cm2 was excellent (sensitivity 95.2%, specificity 94.8%, negative predictive value 99.8%). The AI-DSA identified 1,549 (5.0%) individuals with guideline-defined severe AS plus 463 (1.5%), 979 (3.1%), and 28,150 (90.4%) individuals with a severe AS (not meeting guidelines), moderate AS, and low-probability AS phenotypes. Five-year mortality was 73.9%, 77.1%, 71.2%, and 43.6% respectively (adjusted HR compared to low-probability AS, HR 1.40, 95% CI 1.31-1.50 [guideline severe], HR 1.31, 1.17-1.46 [non-guideline severe], HR 1.21, 1.12-1.23 [moderate], p < 0.001 for all). Overall, rates of aortic valve replacement (AVR) remained low, even in those with an AVA <1cm2 (21.9%), despite it being associated with improved survival. Findings remained consistent among outpatients, those investigated in the past decade and those with left ventricular systolic dysfunction. Conclusions Without relying on left ventricular outflow tract measurements, an AI-DSA used echocardiographic reports to reliably identify those with severe AS. These results suggest possible utility for this AI-DSA to enhance detection of individuals with the phenotype of severe AS, at risk for adverse outcomes.Figure:Graphical Abstract