In 2025, a CBS News investigation found that nearly one in three kidneys donated for transplant in the United States are never transplanted at all — discarded instead, despite tens of thousands of patients waiting on the national list. It’s a strange kind of scarcity: not a shortage of donated organs so much as a systemic failure to move the organs that exist to the patients who need them, fast enough and well-matched enough to actually use them. A 2024 New York Times investigation went further, describing the current organ allocation system as being in outright “chaos,” documenting how a supposedly rigid, rules-based priority queue has been increasingly bypassed through informal, ad hoc allocation decisions that skyrocketed in frequency as organ procurement organizations scrambled to avoid the ischemic-time deadlines that make a viable organ turn unusable. It’s a problem that, on its face, has almost nothing to do with medicine — and increasingly, researchers building the actual fix agree.
The Scientific Foundation
Every donor organ carries a hard, unforgiving countdown clock the moment it’s removed from a donor’s body. According to Health Resources and Services Administration figures, kidneys survive 24 to 36 hours outside the body, a pancreas 12 to 18 hours, a liver 8 to 12 hours, and a heart or lung as little as 4 to 6 hours before the tissue degrades past viability. Matching a specific organ to a specific compatible recipient, verifying logistics, and physically transporting that organ — sometimes across a continent — all has to happen inside that shrinking window, which is precisely the kind of hard-deadline, resource-constrained routing problem that logistics and operations research fields have spent decades optimizing in entirely different industries.
The current U.S. system for kidneys has traditionally relied on something close to a rigid priority queue based on medical urgency and geographic proximity, working through a sequential list of potential recipients one offer at a time. But a 2025 arXiv analysis found that this rigid, sequential process is increasingly bypassed through a mechanism called “out-of-sequence allocation,” where organ procurement organizations skip the formal queue entirely and allocate directly to a specific hospital when they judge that traversing hundreds of potential offers risks losing the organ to ischemic time. That same analysis found the rate of out-of-sequence kidney allocations rose from about 2 percent in 2020 to 18 percent in 2023 — a system increasingly improvising around its own official rules because the rules themselves were too slow for the actual physical constraints involved.
The Cross-Domain Connection
The genuinely novel synthesis now underway pulls directly from operations research, machine learning, and dynamic matching theory — fields with a long history in trucking fleet routing, e-commerce fulfillment, and ride-share dispatching, but only recently applied seriously to organ allocation. A 2025 paper published in Transplantation Direct, led by researchers including Elijah Pivo at MIT’s Institute for Data, Systems and Society, developed a new simulation algorithm that reduced the computational time needed to simulate an entire year of kidney allocation policy from more than six hours down to about 15 seconds — a speedup that, for the first time, allowed the Organ Procurement and Transplantation Network’s policy committee to test thousands of different allocation policy variants through multiobjective optimization rather than manually evaluating a handful of options, and to build an interactive tool letting committee members explore trade-offs directly.
Separately, Northwestern Engineering professor Sanjay Mehrotra’s team has developed an adaptive algorithm specifically targeting kidney discards, aiming to improve allocation efficiency using the same organs and infrastructure already in place, without requiring any new medical technology — purely a smarter matching and sequencing layer over the existing system. And a 2026 paper on heart transplant matching borrows a technique called “learning potentials for dynamic matching,” directly adapting mathematical frameworks originally developed for general two-sided matching markets, the same theoretical territory that underlies ride-share driver-passenger matching and labor market job-matching platforms.
What Remains Undemonstrated
The honest caveat is that better algorithms alone haven’t yet closed the gap, and the field has documented real risks in how these tools get built. A 2026 medRxiv study examining match-run data from 2021 to 2024 found that formal documentation of efforts to avoid kidney discards through out-of-sequence attempts remained rare even as informal bypassing of the sequential queue increased sharply, suggesting the data needed to properly evaluate newly designed “expedited allocation pathways” doesn’t yet fully exist. A 2026 arXiv paper focused specifically on heart transplant allocation argues that machine learning-driven policy optimization needs to explicitly account for incentives, warning that optimizing purely for aggregate outcome metrics like a concordance index can create policies that look statistically strong on average while producing genuinely poor incentive structures for hospitals and organ procurement organizations in practice — a reminder that a mathematically elegant matching algorithm can still fail if it doesn’t account for how the humans and institutions inside the system will actually respond to it. As of early 2026, the specific weight parameters for the “continuous distribution” allocation framework already deployed for lungs, and under consideration for hearts, kidneys, livers, and other organs, remained unfinalized for hearts specifically.
Why It Matters
The stakes here are about as direct as public health stakes get: every discarded, viable kidney represents a patient on a waitlist of over 100,000 people who didn’t get an organ that existed and was medically usable. Unlike most of the interventions this field usually covers, this isn’t a request for a new drug, device, or biological therapy — it’s a request for better logistics software applied to organs and infrastructure that already exist, which makes it one of the rare cases where a meaningfully large improvement in survival outcomes could come from computer science and operations research rather than new medical technology at all.
The Human Dimension
There’s something almost uncomfortable about the fact that the same mathematical tools optimizing how quickly a package reaches a doorstep could, applied to the right problem, mean the difference between a donated kidney reaching someone in time or being quietly discarded. It’s a reminder that the last mile problem, so familiar from commercial logistics, is sometimes also a matter of life and death — and that fixing it doesn’t always require a scientific breakthrough, just the will to bring an underused toolkit to a system that has been running, for years, on rules too rigid for the reality it’s trying to serve.
Sources:
1. “Adaptive Algorithm Aims to Reduce Kidney Discards and Improve Transplant Access,” Northwestern Engineering, 2026 — https://www.mccormick.northwestern.edu/news/articles/2026/04/adaptive-algorithm-aims-to-reduce-kidney-discards-and-improve-transplant-access/
2. “Position: Machine Learning for Heart Transplant Allocation Policy Optimization Should Account for Incentives,” arXiv, 2026 — https://arxiv.org/pdf/2602.04990
3. “Modernizing the Design Process for US Organ Allocation Policy: Toward a Continuous Distribution Policy for Kidneys,” Transplantation Direct, 2025 — https://pmc.ncbi.nlm.nih.gov/articles/PMC12377324/
4. “Newly designed expedited allocation pathways cannot be expected to rely on data that does not currently exist,” medRxiv, 2026 — https://www.medrxiv.org/content/10.64898/2026.01.06.26343389.full.pdf
5. “A Simulation-Optimization Framework To Improve The Organ Transplantation Offering System,” arXiv — https://arxiv.org/pdf/2204.11623
6. “Learning Potentials for Dynamic Matching and Application to Heart Transplantation,” arXiv, 2026 — https://arxiv.org/pdf/2602.08878
7. “Optimizing the Transport of Organs for Transplantation,” Expert Systems with Applications, 2026 — https://www.sciencedirect.com/science/article/abs/pii/S0305054824004064
8. “Smart match: revolutionizing organ allocation through artificial intelligence,” Frontiers in Artificial Intelligence, 2024 — https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2024.1364149/full
Idea originated at artificialideas.org. Article researched and written by Claude Sonnet 4.6. Published at artificialideas.org.
Artificial Ideas 