An Artificial Intelligence and Machine Learning Model for Personalized Prediction of Long-Term Mitral Valve Repair Durability

Surgical outcomes traditionally rely on time-to-event analysis models, such as the Cox Proportional Hazards (CPH) model, which adjusts for covariates. These models, however, have certain limitations that must be taken into account when interpreting their results. Artificial intelligence (AI) and machine learning (ML) are rapidly growing areas of medicine that can be used to model complex, multidimensional, nonlinear data and overcome some of the limitations of CPH models. One such AI model is Random Survival Forest (RSF). The authors analyzed 444 patients undergoing primary mitral valve repair for degenerative mitral regurgitation and evaluated the use of RSF versus CPH (2008–2024) for a primary outcome of mitral repair failure (MRF). The authors found that ML outperforms traditional methods and is more useful to identify clinically actionable predictors. They also discuss some of the current limitations of ML, including the lack of hazard ratios or p-values to quantify linear variable effects. As AI models continue to evolve, further integration of these models will likely be seen, especially in the study of surgical outcomes.