Artificial Intelligence and Machine Learning for Predicting 1-Year Mortality after Transcatheter Aortic Valve Replacement: A Systematic Review and Meta-Analysis

Abstract

The prediction of mortality after transcatheter aortic valve replacement (TAVR) is important in risk stratification, selection of patients, and the joint decision-making process. Conventional surgical and TAVR-specific risk scores have poor predictive ability, while artificial intelligence (AI)/machine learning (ML) models can provide better performance in capturing the complex nature of interdependencies between variables. This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 protocol. PubMed, Scopus, and Google Scholar were systematically searched from inception to September 2025 for papers describing AI/ML models used in predicting mortality at 1 year following TAVR. The eligible papers reported either area under the receiver operating characteristic curve (AUC)/C-statistics with 95% confidence intervals (95% CIs) or had enough data for calculation. Papers that lacked comparable outcome measures, did not report on discrimination, or did not employ AI/ML methods were excluded from further analysis. The pooled estimates were estimated using the random-effects model. The risk of bias and applicability assessment was performed using PROBAST criteria. Four retrospective studies comprising 1,732 patients who underwent TAVR were included. ML-based gradient boosting and clustering were used as methods. The overall pooled estimate of AUC for 1-year mortality prediction was 0.76 (95% CI: 0.71–0.81; p < 0.01) with considerable heterogeneity (I 2  = 64.76%, p = 0.04). Sensitivity analysis with the leave-one-out method showed consistent results for pooled estimates. AI/ML models proved their ability in predicting 1-year mortality following TAVR with moderate discrimination ability. Future research requires larger sample sizes in prospective studies.