Advanced aI algorithms-driven agent based simulation for policy improvement in hard-to-place deceased donor kidneys
Richard Threlkeld1,2, Lirim Ashiku2, Cihan H Dagli2, Casey Canfield2, Krista Lentine3, Henry Randall3, Tom Levanos4, Rich Rothweiler4, Lindsey Speir4, Gary Marklin4.
1Axiom AI, Orlando, FL, United States; 2Missouri University of Science and Technology, Rolla, MO, United States; 3Saint Louis University, St Louis, MO, United States; 4Mid America Transplant, Saint Louis, MO, United States
Introduction: The current method for assigning high-risk deceased donor kidneys, reliant on manual efforts by organ procurement organizations, leads to delays and decreased organ quality. A new digital simulation tool aims to identify these kidneys faster, proposing improvements to the allocation process and enabling policy evaluation before implementation.
Method: An interactive digital simulation tool functions as a research instrument for evaluating kidney allocation policies and their impact on the acceptance of deceased donor kidneys. Users can manipulate or modify policy parameters within the tool to observe how they affect the ultimate decision outcomes and policies. Artificial intelligence (AI) models have been incorporated to provide insights into the likelihood of organ utilization and to pinpoint transplant centers that are more inclined to accept and transplant kidneys at risk of going unused.
Results: Figure 1 presents the initial graphical user interface encompassing the current framework of kidney allocation as viewed from the OPO perspective, involving agents, kidney information, ‘hard to place’ kidneys, and ‘accelerated placement’ transplant centers. This tool gauges overall policy performance of OPO efforts aimed at enhancing kidney utilization. The efficacy of the digital simulation tool is assessed through Key Performance Parameters related to kidney utilization, discard, and equity. Additional metrics considered include allocation duration, occurrences of out-of-sequence allocation, and the projected time from procurement to transplant.
Conclusion: The digital simulation tool forecasts a heightened utilization of deceased donor kidneys when implementing early engagement in accelerated placement for out-of-sequence allocation, thereby endorsing modifications in high-risk kidney allocation policies. Future endeavors will encompass gradual enhancements to the simulation and AI models, especially for data-driven transplant centers that demonstrate a readiness to transplant high-risk kidneys.
The work was supported by Mid America Transplant. We thank our colleagues and collaborators at Saint Louis University who have provided critical insight into the kidney transplant process, Krista Lentine, Mark Schnitzler, and Henry Randall.
[1] Artificial Intelligence, Expedited Placement, Organ Allocation, Kidney Allocation