New Machine Learning-Based Model More Accurately Predicts Liver Transplant Waitlist Mortality

Media Contact: Caroline Laurin
Phone: (703) 299-9766
Email: media@aasld.org

Alexandria, VA – Data from a new study presented this week at The Liver Meeting Digital Experience® – held by the American Association for the Study of Liver Diseases – found that using neural networks, a type of machine learning algorithm, is a more accurate model for predicting waitlist mortality in liver transplantation, outperforming the older model for end-stage liver disease (MELD) scoring. This advancement could lead to the development of more equitable organ allocation systems and even reduce liver transplant waitlist death rates for patients.

Use of MELD-Na scores ─ which determine the severity of liver cirrhosis─ have shown success in the past in predicting waitlist mortality for liver transplant patients. Could newer models that utilize advances like machine learning be more effective? To find out, researchers in Michigan conducted a study to develop neural network (NN) models that more accurately predict waitlist mortality for liver transplant (LT) patients. NNs are a type of machine learning inspired by biological neurons and used to calculate input data and extract meaningful patterns.

"The MELD-Na score, which consists of four variables--serum total bilirubin, INR, creatinine and sodium level—was designed to predict the severity of liver disease. Its clinical utility in liver allocation has been validated by many studies. However, still, the current MELD-Na score-based allocation model has lots of limitations," says the study’s co-author, Shinji Nagai, MD, a transplant surgeon at Henry Ford Hospital and Henry Ford Cancer Institute in Detroit. "We have seen many liver cirrhosis patients whose MELD scores were low but suffered from life-threatening complications due to liver cirrhosis and actually could not have a chance of a liver transplant." Advances in big data analysis and machine learning may improve prediction models and benefit patients waiting for a liver transplant, prompting the need for this study, he adds.

The study used data from the United Network for Organ Sharing’s Organ Procurement and Transplantation Network (OPTN/UNOS) registry, which includes detailed records for 194,299 patients waitlisted for liver transplantation from Feb. 27, 2002 to Dec. 31, 2018. The researchers used subsets of data to create four separate NN models. Models were constructed to predict mortality at four different timeframes: 30, 90, 180 and 365 days. They excluded patients who received liver transplantations before the outcome timeline, those with liver cancer, those who received MELD exceptions and those waitlisted for combined organ transplants other than kidney.

Then, they combined the Liver Data and Liver Wait List History files in the OPTN/UNOS registry to select a total of 44 variables, including recipient characteristics, trend of liver and kidney function during waiting time, UNOS regions and registration year. They did not include age, ethnicity and gender in the NN model to avoid assigning waitlist priority based on these demographic factors. For each NN model, they split the data using random sampling into training, validation and test datasets in a 60:20:20 ratio. They assessed model performance using evaluation metrics Area Under Receiver Operating Curve (AUC-ROC) and Area Under Precision-Recall Curve (PR-AUC).

The 90-day mortality NN model outperformed the MELD score in both assessments, and also outperformed the MELD score for recall (sensitivity), negative predictive value (NPV) and F-1 score. The 90-day mortality NN model specifically identified more waitlist deaths with higher sensitivity, but the MELD score performed better for specificity and precision.

The study also compared performance metrics of the models by breaking the test dataset into multiple subsets based on ethnicity, gender, region, age, diagnosis group and year of listing. The 90-day mortality NN model significantly outperformed MELD scoring across all subsets of the data for predicting waitlist mortality.

"In the future, if these advanced technologies are introduced into the liver allocation system, liver waitlist ranking would better reflect patients’ medical urgency and this should lead to lower waitlist mortality," says Dr. Nagai.

Dr. Nagai will present these findings at The Liver Meeting Digital Experience™ during the Presidential Plenary: Liver Transplantation session on November 14 at 9 AM ET. The corresponding abstract "Use of Neural Network Models to Predict Mortality/Survival Among Patients on the Liver Transplant Waitlist" can be found in the journal, HEPATOLOGY.

About the AASLD
AASLD is the leading organization of clinicians and researchers committed to preventing and curing liver disease. The work of our members has laid the foundation for the development of drugs used to treat patients with viral hepatitis. Access to care and support of liver disease research are at the center of AASLD’s advocacy efforts.