Background/purpose: Machine learning has been used to predict procedural risk in patients undergoing various medical interventions and procedures. One-year mortality in patients after Transcatheter Aortic Valve Replacement (TAVR) has a wide range (from 8.5 to 24% in various studies). We sought to apply machine learning to determine predictors of one year mortality in patients undergoing TAVR. Methods/materials: A retrospective study of 1055 patients who underwent TAVR (Jan 2014–June 2017) with one-year follow up was completed. Baseline demographics, clinical, electrocardiography (ECG), Computed Tomography (CT) and echocardiography data were abstracted. Variables with near zero variance or ≥50% missing data were excluded. The Gradient Boosting Machine learning (GBM) prediction model included 163 variables and was optimized using 5-fold cross-validation repeated 10-times. The receiver operator characteristic (ROC) for the GBM model was calculated to predict one-year mortality post TAVR, and then compared to the TAVI2-SCORE and CoreValve score. Results: Among 1055 TAVR patients (mean age 80.9 ± 7.9 years, 42% female), 14.02% died at one year. 78% had balloon expandable valves placed. Based on GBM, the ten most predictive variables for one-year survival were cardiac power index, hemoglobin, systolic blood pressure, INR, diastolic blood pressure, body mass index, valve calcium score, serum creatinine, aortic annulus area, and albumin. The area under ROC to predict survival for the GBM model vs TAVI2-SCORE and CoreValve Score was 0.72 (95% CI 0.68–0.78) vs 0.56 (95%CI 0.51–0.62) and 0.53 (95% CI 0.47–0.59) respectively with p < 0.0001. Conclusion: The GBM model outperforms TAVI2-SCORE and CoreValve Score in predicting mortality one-year post TAVR.
- Machine learning
- Transcatheter Aortic Valve Replacement
ASJC Scopus subject areas
- Cardiology and Cardiovascular Medicine