Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Position: Machine Learning for Heart Transplant Allocation Policy Optimization Should Account for Incentives

Ioannis Anagnostides, Itai Zilberstein, Zachary W. Sollie, Arman Kilic, Tuomas Sandholm

TL;DR

The paper analyzes incentive misalignments in US adult heart transplant allocation and argues that purely predictive ML approaches neglect strategic behavior by centers, OPOs, clinicians, and patients. It integrates mechanism design, strategic classification, and causal inference to diagnose multiple incentive design failures, including device-based manipulation, out-of-sequence allocation, and delisting practices, supported by UNOS data. The authors present a comprehensive research agenda—audits, threshold design for open offers, continuous monitoring, counterfactual analyses for listing decisions, and RLHF-based preference elicitation—to realize incentive-aware policy optimization. They contend that embedding incentive considerations into policy design can improve efficiency, equity, and public trust, with practical implications for policymakers, clinicians, and ML researchers.

Abstract

The allocation of scarce donor organs constitutes one of the most consequential algorithmic challenges in healthcare. While the field is rapidly transitioning from rigid, rule-based systems to machine learning and data-driven optimization, we argue that current approaches often overlook a fundamental barrier: incentives. In this position paper, we highlight that organ allocation is not merely a static optimization problem, but rather a complex game involving transplant centers, clinicians, and regulators. Focusing on US adult heart transplant allocation, we identify critical incentive misalignments across the decision-making pipeline, and present data showing that they are having adverse consequences today. Our main position is that the next generation of allocation policies should be incentive aware. We outline a research agenda for the machine learning community, calling for the integration of mechanism design, strategic classification, causal inference, and social choice to ensure robustness, efficiency, and fairness in the face of strategic behavior from the various constituent groups.

Position: Machine Learning for Heart Transplant Allocation Policy Optimization Should Account for Incentives

TL;DR

The paper analyzes incentive misalignments in US adult heart transplant allocation and argues that purely predictive ML approaches neglect strategic behavior by centers, OPOs, clinicians, and patients. It integrates mechanism design, strategic classification, and causal inference to diagnose multiple incentive design failures, including device-based manipulation, out-of-sequence allocation, and delisting practices, supported by UNOS data. The authors present a comprehensive research agenda—audits, threshold design for open offers, continuous monitoring, counterfactual analyses for listing decisions, and RLHF-based preference elicitation—to realize incentive-aware policy optimization. They contend that embedding incentive considerations into policy design can improve efficiency, equity, and public trust, with practical implications for policymakers, clinicians, and ML researchers.

Abstract

The allocation of scarce donor organs constitutes one of the most consequential algorithmic challenges in healthcare. While the field is rapidly transitioning from rigid, rule-based systems to machine learning and data-driven optimization, we argue that current approaches often overlook a fundamental barrier: incentives. In this position paper, we highlight that organ allocation is not merely a static optimization problem, but rather a complex game involving transplant centers, clinicians, and regulators. Focusing on US adult heart transplant allocation, we identify critical incentive misalignments across the decision-making pipeline, and present data showing that they are having adverse consequences today. Our main position is that the next generation of allocation policies should be incentive aware. We outline a research agenda for the machine learning community, calling for the integration of mechanism design, strategic classification, causal inference, and social choice to ensure robustness, efficiency, and fairness in the face of strategic behavior from the various constituent groups.
Paper Structure (27 sections, 4 figures, 4 tables)

This paper contains 27 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Roadmap
  3. Patient feature manipulation
  4. The device game
  5. Policy optimization going forward
  6. Connection to strategic classification
  7. Open offers and out-of-sequence allocation
  8. Strategic offer rejection and performance monitoring
  9. Pitfalls in current performance monitoring
  10. Evaluation cycle gaming
  11. Strategic listing and delisting
  12. Pre-listing selection
  13. Delisting
  14. Multi-listing
  15. Value aggregation and strategic preference reporting
  16. ...and 12 more sections

Figures (4)

  • Figure 1: Strategic classification in organ allocation. Patients below the threshold (red region) incur a cost to manipulate certain features (e.g., device implantation) to cross the decision boundary and gain high priority status (green region), gaming the classifier.
  • Figure 2: Out-of-sequence allocation. While the standard policy (blue) mandates sequential offers starting from the highest priority, open offers (red) allow OPOs to bypass the entire queue and allocate the organ directly to a lower-priority candidate.
  • Figure 3: Distribution of annual accepted offers, heart transplants performed, and center acceptance rates per month among adult patients (2010--2024). Panel (a): distribution of offers. Panel (b): distribution of transplants performed. Panel (c): offer acceptance rates.
  • Figure 4: Reported center removal reasons for alive and non-transplanted US adult heart transplant candidates (2010--2024).