A family is currently sitting in a waiting area at a hospital, keeping an eye on the clock, and hoping that a phone will ring. They most likely picture a physician somewhere going through files, carefully weighing each patient against the others. The reality is more bizarre. The decision regarding the recipient of a donated organ, such as a liver, kidney, or heart, is made by a software system that performs calculations more quickly than the human mind could.
Every time an organ becomes available, a “match run” is created by the United Network for Organ Sharing, or UNOS, using a matching platform called UNet. It looks through all of the patients on the list, taking into account factors like blood type, tissue compatibility, organ size, medical urgency, proximity, and waiting time. A prioritized list shows up in a matter of seconds. Then, one by one, the organ procurement organization sends offers down that list until a transplant team accepts. It works well. It moves quickly. Furthermore, the majority of outsiders hardly know it operates in this manner.
This discrepancy between perception and reality is unsettling. We’ve been conditioned by popular culture to view “the list” as a straightforward line with first come, first served. However, it’s more akin to a dynamic auction of medical need, with each donated organ being recalculated from scratch. Every match run is different. For one kidney, a patient may rank third; for another, they may rank fifteenth. The patient’s uncontrollable and frequently incomprehensible circumstances shape the variables, which are always changing.
A liver-matching algorithm that was implemented in the UK in 2018 came under heavy fire after it was discovered that younger patients were routinely given less priority. The Financial Times was informed by a patient’s family that each time they voiced concerns regarding the figures, they were informed that they “didn’t understand, presumably because we weren’t doctors.” There was no procedure for appeals. There is no human override. The score was determined by the algorithm. It’s difficult to ignore the unsettling irony that a system intended to be more equitable ended up feeling incredibly arbitrary to those who were caught off guard.
In his book “Voices in the Code,” David G. Robinson discusses the American kidney allocation algorithm and provides an insightful account. He once received a call from a UNOS data scientist who wanted to know how many decimal places to use when determining patient scores. They could leave at fifteen. However, the scientist’s argument was that there is no actual medical difference between two patients based on a difference in the fifteenth decimal place. To act otherwise would be to allow math to pass for moral clarity. Robinson contends that algorithmic decision-making lacks this kind of humility, which is desperately needed.
Beneath all the technical improvements, there is a deeper concern about bias inherent in the data. According to a 2023 study that was published in the Journal of Law and the Biosciences, racial, geographic, and poverty-related social determinants of health can subtly affect algorithmic outputs even when they aren’t specifically coded in. The injustices of the system that produced the data in the first place may be replicated by machine learning models trained on historical transplant data. There are substantial gaps between algorithmic reality and the principle of equitable access, according to European legal scholars who contend that current human rights frameworks only partially address this issue.
However, there are grounds for cautious optimism. Over the years, the American kidney system has included patients, donor families, and community advocates in the policy committees that determine how the algorithm operates, in addition to doctors and data scientists. Researchers at MIT and UNOS have been working together on a framework called “continuous distribution,” which aims to replace rigid boundaries in organ allocation with more equitable and smooth scoring. This slower, more democratic method of developing algorithms might be used as a template for other high-stakes processes, such as child welfare screenings and bail hearings.
However, the underlying tension remains unchanged. At any given time, there are about 100,000 Americans waiting for transplants. Organs will never be sufficient. For the majority of them, the code that separates the living from the waiting is still unseen. One of the most important human decisions—who gets another chance at life—seems to have been delegated to systems that we haven’t even started to examine. How willing we are to continue posing difficult questions about what’s going on inside the machine will likely determine whether that proves to be prudent or careless.
