Machine learning improves prediction of death risk in hospitalized cirrhosis patients

Researchers employed a machine learning technique known as random forest analysis and found that it significantly outperformed traditional methods in predicting which hospitalized patients with cirrhosis are at risk of death, according to a new paper published in Gastroenterology.

This gives us a crystal ball - it helps hospital teams, transplant centers, GI and ICU services to triage and prioritize patients more effectively."

Dr. Jasmohan S. Bajaj, study's corresponding author

Key findings:

• Data analyzed from 121 hospitals worldwide, which were part of the CLEARED consortium.
• The model performed consistently across both high- and low-income countries.
• It was validated using National U.S. veterans' data and remained accurate.
• The tool maintained strong performance even when limited to just 15 key variables.
• Patients were accurately grouped into high-risk and low-risk categories, making the model scalable and clinically practical. 

Explore the model in action here: https://silveys.shinyapps.io/app_cleared/. 

This paper is one of three studies recently published on this topic in the American Gastroenterological Association's journals. One was a worldwide consensus statement on organ failures, including liver in cirrhosis patients, while the second study identified specific blood markers and complications that influence the risk of in-hospital death, focusing on liver failure biomarkers. 

"Liver disease is one of the most underappreciated causes of death worldwide - alcohol, viral hepatitis, and late diagnoses are major drivers," Bajaj said. "When someone is hospitalized, it's often because everything upstream - prevention, screening, primary care - has already failed."