Updating allocation algorithms could help donor hearts reach transplant patients who need them most

Receiving a heart transplant is a matter of life and death for many patients. Each time a heart becomes available, a “match run” is created to generate a list of transplant candidates, ranked by an algorithm based on medical urgency, geography and pediatric status. Unfortunately, deceased organ donors are very rare in the United States, so much so that some patients are not even placed on waiting lists because it is too unlikely that a heart will be available for them.

A research team led by experts at the University of Chicago Medicine has developed a new risk score designed to predict the likelihood that a patient will die without a heart transplant. This innovation aims to address the limitations of the current system based on 6-status therapy, by providing a more precise and equitable approach to prioritizing candidates based on medical emergency. An article detailing the development and initial validation of the proposed US Applicant Risk Score (US-CRS) is published in JAMA.

“The goal is to identify the sickest patients. Every candidate on the list is sick and needs a heart transplant, but some may wait slightly longer than others,” said William F. Parker, MD, Ph.D., assistant professor of medicine. and Public Health at UChicago Medicine and senior author of the article. “This new risk score improves on the status quo. Among other things, it does not depend as much on the treatment decisions of individual doctors.”

Integrate objective physiological measures

Instead of looking only at what treatments patients were receiving — the measure currently used to allocate deceased donor hearts — Parker and his team also looked at clinical and laboratory measures associated with end-stage heart failure, such as heart rate levels. molecules in the blood associated with the liver. and renal failure. They chose an initial list of variables based on the current French Candidate Risk Score (French-CRS) model, which incorporates more clinical measures than the current US system, and added others that they had identified as important.

By combining these variables into a single risk score (the US-CRS), researchers analyzed data from more than 16,900 adult heart transplant candidates from the US Heart Allocation System to determine the relationship between US-CRS and mortality within 6 weeks of their implementation. the waiting list for a transplant.

“We initially expected that advanced deep learning and machine learning (which is more like a black box algorithm) would lead to a much more accurate model, but it turned out that a fairly standard was fairly accurate in predicting mortality,” Parker said. . “It just shows that we chose good variables to measure. The resulting model has the advantage of being easy to understand.”

The team compared their new US-CRS model to the current heart allocation system in the United States and found that it was much better at accurately predicting mortality in patients who did not receive heart transplants in the 6 weeks.

Parker highlighted the contributions of the entire research team, particularly first author Kevin Zhang, MS, a data scientist at UChicago Medicine.

“He did an awful lot of data science,” Parker said. “He’s the first person in the country to put all of these lab values ​​together. That’s the reason we were able to do this work.”

Transforming scientific knowledge into policy improvements

The US-CRS will have to go through many additional rounds of validation and committees before being adopted. This would also only be part of the upcoming continuous distribution algorithm for core allocation. Parker and colleagues are already working on grant proposals for follow-up research aimed at improving cardiac allocation overall. Their primary goals include designing a fair and resilient continued distribution system for deceased donor hearts that alleviates health inequities.

In the long term, Parker believes this work could provide an important foundation as researchers, technology experts and policymakers tackle the broader problem of using algorithms to distribute limited health care resources. health.

“Organ transplantation gives us the opportunity to address this issue in a very practical way,” Parker said. “We hope that the lessons we discover can be generalized.”