Fair Matching in Dynamic Kidney Exchange

Abstract: Kidney transplants are sharply overdemanded in the United States. A recent innovation to address organ shortages is a kidney exchange, in which willing but medically incompatible patient-donor pairs swap donors so that two successful transplants occur. Proposed rules for matching such pairs include static fair matching rules, which improve matching for a particular group, such as highly-sensitized patients. However, in dynamic environments, it seems intuitively fair to prioritize time-critical pairs. We consider the tradeoff between established sensitization fairness and time fairness in dynamic environments. We design two algorithms, SENS and TIME, and study their patient loss. We show that the there is a theoretical advantage to prioritizing time-critical patients (around 9.18% tradeoff on U.S. data) rather than sensitized patients. Our results suggest that time fairness needs to be considered in kidney exchange. We then propose a batching algorithm for current branch-and-price solvers that balances both fairness needs.