Development and External Validation of an Echocardiography-Based AI Model for Predicting Aortic Stenosis Progression.

BACKGROUND: Aortic stenosis (AS) is the most common valvular heart disease in older adults, necessitating careful medical surveillance and timely intervention. However, current fixed surveillance intervals often fail to account for the variability of disease progression. This results in inefficient use of resources and potential delays in treatment for rapid progression of disease. OBJECTIVES: The purpose of this study was to develop and externally validate models for fixed-horizon prediction of progression from mild or moderate to severe AS. METHODS: A retrospective cohort study analyzed structured echocardiographic report data from 2 independent large tertiary centers (derivation: 9,247 visits; external validation: 3,873 visits). Three models were developed using 21 routinely available echocardiographic and demographic variables to predict progression to severe AS at 1-, 3-, and 5-year horizons. Model performance was evaluated using the area under the receiver-operating characteristic curve (AUROC) and the F1 score. RESULTS: The models demonstrated consistent discrimination across internal testing and external validation. Internal test set AUROCs were 0.90 (95% CI: 0.86-0.93), 0.91 (95% CI: 0.88-0.93), and 0.91 (95% CI: 0.87-0.95) for the 1-, 3-, and 5-year prediction horizons, respectively. External validation AUROCs were 0.91 (95% CI: 0.90-0.93), 0.88 (95% CI: 0.87-0.90), and 0.88 (95% CI: 0.87-0.90), respectively. CONCLUSIONS: Models based on routinely available structured echocardiographic data predicted progression from mild or moderate to severe AS and showed consistent performance in an independent external cohort. These findings support the potential use of structured-data risk prediction to inform personalized surveillance strategies for patients with AS.