Improving Heart Transplant Outcomes via Statistical Modeling of Mortality

Abstract

Each year, the Organ Procurement and Transplant Network (OPTN) transplants only about one-third of potential donor hearts, often due to concerns about organ function or difficulties matching organs to suitable patients in time. To address this underutilization and improve post-transplant outcomes, a logistic regression model trained on data from the Scientific Registry of Transplant Recipients (SRTR) was used to predict one-year post-transplant mortality. A Cox proportional hazards model was also fit to estimate patient mortality risk while on the waitlist. This model was integrated into a discrete-event simulation model reflecting OPTN allocation policies, using SRTR patient and organ data from 2006–2018. The simulated allocation policy would determine the highest priority candidate whose risk from waiting exceeded the risk of post-transplant one-year mortality. The simulation showed approximately a 33% increase in hearts transplanted per donor and a 90% reduction in average waiting time. These results suggest that predictive modeling can help optimize heart allocation and could be incorporated into current organ evaluation workflows to speed up and improve matching decisions.