AI Model Reduces Risk of Canceled Liver Transplants by 60%

Researchers at Stanford Medicine have developed a groundbreaking machine learning model that predicts the viability of donor livers in real time. This innovation could significantly reduce the number of canceled liver transplants, potentially slashing the rates by 60%. The model is designed to assess whether a donor is likely to pass away within the critical time frame of 30 to 45 minutes, which is essential for organ transplantation after circulatory death.

The challenge stems from the fact that many potential donors are identified only after life support is removed, leading to uncertainty about the timing of death. If death occurs too late, the liver may no longer be suitable for transplantation, resulting in wasted resources and missed opportunities. According to Kazunari Sasaki, MD, a clinical professor of abdominal transplantation and the senior author of the study, “By identifying when an organ is likely to be useful before any preparations for surgery have started, this model could make the transplant process more efficient.”

Addressing the Organ Shortage

The need for liver transplants continues to grow, with many individuals suffering from end-stage liver disease. Currently, the demand for organs exceeds the available supply, but advancements in techniques such as normothermic machine perfusion are helping to bridge this gap. This method maintains organs at optimal temperatures and oxygen levels during transport, making it feasible to use organs from donors who have experienced circulatory death.

Historically, most liver donations have come from brain-dead donors, but there is a notable increase in those from circulatory death. “The number of liver transplants keeps going up because of donation after circulatory death, and the waitlist is getting smaller,” Sasaki added. He envisions a future where every patient in need of a liver transplant might receive one from a deceased donor.

The research findings, published in The Lancet Digital Health, highlight the importance of acting quickly in the transplantation process. If the donor’s time of death occurs more than 30 minutes after blood flow to the organs ceases, the liver’s viability is compromised. Approximately half of potential donors die within this critical time frame, with many transplants being canceled due to late deaths.

Enhancing Predictive Accuracy with AI

The innovative model developed by Stanford’s research team utilizes a variety of clinical data to predict the time of death. Key metrics include age, body mass index, blood pressure, heart rate, and neurological assessments. By analyzing this information, the model achieves an accuracy rate of 75%, surpassing the average surgeon’s judgment, which has an accuracy rate of 65%.

The research team conducted extensive comparisons of different machine-learning algorithms to determine which was most effective in predicting death times. The selected algorithm demonstrated its superiority not only in accuracy but also in its ability to deliver reliable predictions even when certain medical record information was missing.

In addition to improving transplant outcomes, the model aims to minimize missed opportunities when death occurs unexpectedly within the viable time frame but without prior preparations for transplantation. Both the model and surgeon judgments had a similar missed opportunity rate of just over 15%, but researchers are optimistic about further enhancements. Sasaki stated, “We are now working on decreasing the missed opportunity rate because it is in the patients’ best interest that those who need transplants receive them.”

The research team includes contributors from several leading institutions, such as Duke University School of Medicine, Cleveland Clinic, and the International University of Health and Welfare. Efforts are underway to adapt the model for use in heart and lung transplants, expanding its potential impact on organ donation and transplantation.

As advancements in artificial intelligence progress, the potential for this predictive model to revolutionize liver transplantation and improve patient outcomes continues to grow.