•  
  •  
 

SMU Data Science Review

Article Title

Avoiding Unnecessary Liver Transplantation: Recognizing Increased Prevalence in Alcohol-Associated Liver Disease (ALD)

Abstract

In recent years there has been an increase in liver transplantation for alcohol-associated liver disease (ALD). Unlike other etiologies of liver disease, patients with ALD that meet transplant criteria may improve with abstinence, negating the need for transplant. ALD patients on the transplant waitlist that will improve with abstinence, are not only placed at risk for harm by receiving an unnecessary transplant, but also prevent other transplant candidates from receiving a liver. This paper addresses the issue of ALD patients who were at risk for harm from receiving an unnecessary transplant by evaluating the feature importance of machine learning models that predict the probability that a patient'’s health will improve enough to be removed from the liver transplant waitlist prior to receiving a transplant. Feature importance is significant because it determines what inputs in the model drive decision behavior. The ALD diagnosis was found important in predicting if a patient would be removed from the liver transplant waitlist for condition improvement, indicating these patients are more at risk for an unnecessary transplantation. Our models find the identification of ALD as a feature of importance in predicting condition improvement highlights the increased risk of unnecessary transplantation within patients with ALD.