Surgeon cautions against ‘blindly’ using AI for organ allocation

AI has been deployed in healthcare systems to fight insurance denials, take notes, schedule appointments, and predict outcomes.

Some think it could also help with organ transplant allocation.

The system has long struggled to efficiently and equitably allocate limited organs. For instance, CBS News found that in 2024, about 12,000 donated organs were discarded in the US. At the same time, AI tools still carry the risk of hallucinations and serious medical errors if doctors rely on them too heavily.

Kidney specialists, for example, currently rely on logistical and medical factors and use algorithms like the estimated glomerular filtration rate, which determines how well the kidneys filter waste, to determine who to prioritize for a transplant. Though the field has largely removed race as a factor in this and other algorithms, a May 2026 scientific statement from the American Heart Association found that, in the realm of heart organ transplantation, there are still “inherent biases in the allocation process that have segregated along socioeconomic and racial lines.”

Some companies and universities are now developing AI tools to provide support to doctors deciding which patients should get which donor organ.

Henry Randall is on the advisory board of one of those companies, Valiant AI. Randall is also president of the membership organization the American Society of Transplant Surgeons.

To Randall’s knowledge, no healthcare system is using AI to predict outcomes and allocate organs beyond research. But that could change in the future. Healthcare Brew talked to Randall about how he thinks AI should be used in organ transplantation and medicine broadly.

This interview has been edited for length and clarity.

What are your thoughts on the use of AI in clinical practice?

[AI] is not meant to be a standalone working product or service or tool that you can 100% rely on and think you’re going to get accurate information from it consistently. I’m an AI lover, have been since 1992…After joining the cause, I realized, oh, this is pretty dangerous, and people don’t really understand it. Let’s not use this widely until we have a critical mass of education.

People are trying to build use cases for it, and that’s the appropriate thing to do. But you also have to have systems built around it for it to be reliable. You have to have engineers who understand algorithms. You have to have data scientists who can help you mine your data and then give it to the model, and then you have to apply rigorous, real statistics to the model outcome in order for it to make sense…Blindly using it and trying to incorporate it now, which a lot of people are doing, is not the way to approach it.

The organ allocation and transplant system has come under scrutiny over the last few years. What’s a big issue facing your field right now?

Organ allocation changes from time to time, and what I mean by that is the population changes from time to time. So for instance, we’re not seeing young, healthy donors from murder as much. We’re not seeing young, healthy donors from fentanyl and opioid overdose. That population has declined because of some of the social programs that have been implemented. So now we’re looking at a population of aged individuals, people in their 50s and 60s.

But the way we look at allocation hasn’t changed to parallel the change in the social dynamics of individuals who are donating. So we’re using the same data with a different set of patients and expecting an outcome akin to the 30–40-year-old population.

Is that something you think AI could help with?

Yeah, absolutely. After we get a better idea of the population we’re working with, you can then put some of that data into these models and build constructs about what outcomes might be with this new population of patients based upon the model that you’re using. So, having people understanding the AI models, understanding the output, and then adjusting to correct for any biases, or anything else, needs to be done. But again, that [requires] education, and understanding how those tools work.