FairAlloc: Learning Fair Organ Allocation Policy for Liver Transplant.

Area of Science:

• Healthcare informatics
• Transplantation research
• Machine learning in healthcare

Background:

• Liver transplantation is vital for end-stage liver disease but faces challenges in equitable organ allocation.
• Current policies struggle to balance patient outcomes with fairness across diverse demographic groups.
• Scarcity of donor organs necessitates efficient and fair allocation systems.

Purpose of the Study:

• To introduce FairAlloc, a learning-based framework for optimizing liver organ allocation.
• To integrate group and individual fairness metrics into organ allocation decision-making.
• To enhance both equity and efficiency in the organ allocation process.

Main Methods:

• Formulated organ allocation as a machine learning ranking problem.
• Incorporated group fairness (across race, gender) and individual fairness objectives.
• Evaluated the FairAlloc framework using real-world data from the Organ Procurement and Transplantation Network (OPTN).

Main Results:

• FairAlloc improved group fairness by up to 37.9% and individual fairness by up to 39.9%.
• The framework maintained competitive performance in key post-transplant outcomes like graft failure and survival rates.
• Demonstrated superior fairness metrics compared to six baseline allocation methods.

Conclusions:

• FairAlloc offers a novel, fairness-aware decision-making framework for organ allocation.
• The proposed system has the potential to significantly improve equity in healthcare.
• This approach advances the field of healthcare informatics by integrating fairness into algorithmic decision-making for organ transplantation.