In 1954, the globe’s inaugural successful organ transplant occurred at Brigham and Women’s Hospital, involving a kidney provided from one twin to the other. At that time, a collective of medical professionals and researchers had accurately theorized that the recipient’s antibodies were unlikely to reject an organ from a genetically identical twin. One Nobel Prize and several decades later, improvements in immune-suppressing medications heightened the feasibility of and interest in organ transplants. Presently, more than 1 million organ transplants have been conducted in the United States, more than any other nation in the world.
The remarkable scale of this accomplishment was made feasible by advancements in organ matching systems: The initial computerized organ matching system was launched in 1977. Despite continuous innovations in computing, medicine, and matching technology over the years, over 100,000 individuals in the U.S. are presently on the national transplant waiting list, with 13 individuals perishing each day while awaiting an organ transplant.
Most computational research in organ allocation centers on the preliminary stages, when patients on the waiting list are prioritized for organ transplants. In a new study introduced at the ACM Conference on Fairness, Accountability, and Transparency (FAccT) in Athens, Greece, scholars from MIT and Massachusetts General Hospital concentrated on the concluding, lesser-explored phase: organ offer acceptance, where an offer is made and the physician at the transplant facility determines for the patient whether to accept or decline the offered organ.
“We were not particularly surprised, but we were obviously let down,” co-first author and MIT PhD student Hammaad Adam states. Utilizing computational models to evaluate transplantation data from over 160,000 transplant candidates in the Scientific Registry of Transplant Recipients (SRTR) between 2010 and 2020, the researchers discovered that doctors were generally less inclined to accept liver and lung offers for Black candidates, leading to additional challenges for Black patients during the organ offer acceptance procedure.
For livers, Black patients exhibited 7 percent lower chances of offer acceptance compared to white patients. Regarding lungs, the gap widened significantly, with Black patients having 20 percent lower likelihood of offer acceptance than white patients with equivalent characteristics.
The findings do not necessarily indicate clinician bias as the primary factor. “The major takeaway is that even if certain factors validate clinical decision-making, there may be clinical conditions we did not account for, that are more prevalent among Black patients,” Adam clarifies. If the waiting list does not consider specific patterns in decision-making, it might create barriers in the process, even if the procedure itself is “unbiased.”
The researchers also highlight that significant variability in offer acceptance and risk preferences among transplant centers complicates the decision-making process. Their FAccT paper refers to a 2020 publication in JAMA Cardiology, which found that candidates listed at transplant centers with lower offer acceptance rates face a heightened risk of mortality.
Another noteworthy discovery was that an offer was more likely to be accepted if the donor and candidate shared the same race. The paper describes this pattern as “concerning,” given the historical disparities in organ procurement that have curtailed donations from racial and ethnic minority groups.
Previous research conducted by Adam and his associates has sought to bridge this gap. Last year, they compiled and published Organ Retrieval and Collection of Health Information for Donation (ORCHID), the first multi-center dataset detailing the performance of organ procurement organizations (OPOs). ORCHID encompasses a decade of OPO data and aims to promote research that tackles bias in organ procurement.
“Conducting meaningful work in this domain requires time,” notes Adam, who mentions that the overall project on organ offer acceptance took years to complete. To his knowledge, only one study so far examines the correlation between offer acceptance and race.
Although the bureaucratic and highly interdisciplinary characteristics of clinical AI projects may deter computer science graduate students from engaging in them, Adam dedicated himself to the project throughout his PhD under the mentorship of associate professor of electrical engineering Marzyeh Ghassemi, who is affiliated with the MIT Jameel Clinic and the Institute of Medical Engineering and Sciences.
To graduate students keen on exploring clinical AI research initiatives, Adam suggests that they “liberate [themselves] from the cycle of publishing every four months.”
“I found it liberating, honestly — it’s acceptable if these collaborations require time,” he reflects. “It’s challenging to avoid that. I made a conscious decision a few years prior, and I was content focusing on that work.”
This endeavor was backed by funding from the MIT Jameel Clinic. It also received partial support from Takeda Development Center Americas Inc. (successor in interest to Millennium Pharmaceuticals Inc.), an NIH Ruth L. Kirschstein National Research Service Award, a CIFAR AI Chair at the Vector Institute, and various National Institutes of Health programs.