Log in to MyACC
AI-Derived Risk Score for Mortality in Secondary Mitral Regurgitation
Quick Takes
- Based on the EuroSMR registry of patients with secondary mitral regurgitation undergoing transcatheter edge-to-edge repair (TEER), the AI-derived EuroSMR score improves prediction of 1-year mortality as compared with other published risk scores, with AUC 0.789.
- The risk calculator is available at eurosmr.com.
Study Questions:
What is the performance of an artificial intelligence (AI)-derived risk score in predicting 1-year outcomes of patients with secondary mitral regurgitation (SMR) undergoing transcatheter edge-to-edge repair (TEER)?
Methods:
From the EuroSMR registry (European Registry of Transcatheter Repair for Secondary Mitral Regurgitation) of patients with SMR who underwent TEER, the authors created derivation and validation cohorts to develop the EuroSMR risk score, employing machine learning. The score was based on 18 clinical, echocardiographic, laboratory, and medication parameters. The main endpoint was 1-year all-cause mortality.
Results:
The derivation cohort included 4,172 patients, and the validation cohort included 428 patients. The mean age of both groups was 74 years, and 65% of subjects were male. Median follow-up period was 1.7 years. Estimated 1- and 2-year survival rates for all subjects were 82.1% and 70.1%, respectively. The top five parameters with the strongest relationships to 1-year mortality were N-terminal pro–B-type natriuretic peptide, New York Heart Association class, hemoglobin, tricuspid annular plane systolic excursion (TAPSE), and age. The risk calculator is available at eurosmr.com.
In receiver operating characteristic (ROC) analysis, the area under the curve (AUC) for the EuroSMR score was 0.789 for 1-year mortality prediction. Other published risk prediction scores did not perform as well in ROC analysis, with AUCs as follows: EuroScore II, 0.705; MitraScore, 0.695; COAPT Score, 0.685. In the 5% of patients with the highest EuroSMR scores (mean score 74.4 points), 1-year mortality was 72.7%. The authors determined an extreme-risk score cutoff of 70.9 points. In the validation cohort, all 11 patients who were above this cutoff had died by 1.5 years after TEER.
Conclusions:
The EuroSMR score improves 1-year mortality prediction for patients with SMR who are being considered for TEER, as compared with other available algorithms.
Perspective:
Discussion of risk is a critical part of the shared decision-making process. This score could be particularly helpful in identifying extreme-risk patients for whom TEER might be futile. Further research will be needed to determine how this algorithm performs outside Europe and in racially and ethnically diverse populations.
Clinical Topics: Valvular Heart Disease, Mitral Regurgitation, Cardiac Surgery and Heart Failure
Keywords: Artificial Intelligence, Mitral Valve Insufficiency
< Back to Listings
