Risk Factors and Prediction Models for Early and Long-Term Mortality Following Left Ventricular Assist Device Implantation

Introduction: Changes in the United Network for Organ Sharing (UNOS) allocation system, coupled with declining donor availability, have contributed to a rising proportion of durable left ventricular assist device (LVAD) implantations, particularly among patients with prior cardiac surgeries and those requiring multiple concomitant intracardiac procedures. As the complexity of LVAD candidates increases, understanding peri/postoperative determinants of survival becomes critical. We sought to identify risk factors associated with short- and long-term mortality following LVAD implantation. Methods: We analyzed 22,467 adults (18–80 years) who underwent LVAD implantation between January 2013 and June 2023 using data from the INTERMACS registry. A FasterRisk-derived risk score was developed for 90-day mortality prediction, and a Cox proportional hazards model was constructed to evaluate long-term post-implant survival. Results: Predictors of 90-day mortality included advanced age, preoperative extracorporeal membrane oxygenation (ECMO), LVAD exchange, prior cardiac surgery, and concomitant surgery. A 90-day mortality score incorporating pre-implant ECMO, prior cardiac surgery, concomitant surgery, and age demonstrated c-statistic 0.69 in training and 0.67 in testing. The Cox regression model identified advanced age, preoperative ECMO, previous cardiac surgery, planned concomitant RVAD, concomitant surgery, IABP, CVA and postoperative dialysis as significant predictors of long-term mortality (c-index 0.72 training; 0.70 testing). Conclusions: A cross-validated machine learning-based risk model for 90-day mortality following LVAD implantation and a Cox regression model for long-term survival may support improved risk stratification and perioperative management in LVAD patients.