Organ Allocation Algorithms and the Problem of Algorithmic Fairness

Abstract

Organ allocation systems have always required a formula for deciding who receives a scarce organ, but the introduction of machine-learning tools alongside established scoring systems such as the Kidney Allocation System and the Model for End-Stage Liver Disease score has made the ethical basis of that formula considerably harder to inspect. This policy analysis uses the 2021 removal of a race-based coefficient from the estimated glomerular filtration rate equation as a case study in how a supposedly neutral clinical variable can embed unexamined assumptions that delay transplant access for specific patient groups. It then examines the growing use of machine-learning models in liver and kidney allocation, the arguments for and against requiring explainability in these tools, and the accountability problems that arise when a continuously retrained model replaces a publicly reviewable scoring rule. We argue that formal accuracy and fairness are distinct targets, that algorithmic reform addresses only one stage of a longer pathway of unequal access to transplantation, and that meaningful progress requires publicly auditable model documentation alongside, not instead of, attention to upstream disparities in referral and care. The paper concludes that no allocation algorithm, however well designed, can substitute for addressing the structural inequities that determine who reaches the transplant waiting list in the first place.