Transplant

We investigate how AI can support clinical decision-making in kidney transplantation, focusing on a challenging problem: should a patient accept a current organ offer, or wait for a potentially better one? Our models are trained on 20 years of data from the UK Transplant Registry and predict multiple competing outcomes simultaneously, including transplantation with patient and graft survival, removal from the waiting list, and death. Each prediction is accompanied by a patient-level explanation, detailing which factors contributed to that individual's risk estimate, enabling clinicians to verify that the model's reasoning aligns with their clinical intuition. We also quantify prediction uncertainty to distinguish cases where AI recommendations can be trusted from those requiring closer clinical review. External validation is underway with registries in the United States and Australia/New Zealand to test whether findings generalise across different healthcare systems and organ allocation policies.