Machine Learning Model Predicts Short-Term Mortality in Patients with Acute-On-Chronic Liver Failure

Strategic Insights -

Researchers from Cedars-Sinai Medical Center have developed a machine learning model that can predict short-term mortality in patients with acute-on-chronic liver failure (ACLF) with high accuracy. The model, which utilizes a large ICU cohort with detailed clinical information, was developed to enhance effective management of ACLF patients. According to the study, the model was able to accurately predict 30-day mortality in patients with two or more organ failures, outperforming existing predictive scores.

Key Takeaways:

  • The study developed two predictive machine learning models, CatBoost ACLF (CBA) and Random Forest ACLF (RFA), which showed robust calibration and accuracy in predicting short-term mortality in ACLF patients with two or more organ failures.
  • The CBA model had the greatest accuracy in the NACSELD cohort (area under curve [AUC] of 0.87), while the RFA model performed best in the EASL-CLIF cohort (AUC of 0.83).
  • The models were explained by SHAP score analysis, yielding a rank list, and the top twelve predictors were selected.
  • Both simplified models demonstrated similar performance (CBA model: AUC 0.89, RFA model: AUC 0.81) and significantly outperformed contemporary scoring systems, including CLIF-C ACLF and MELD 3.0.
  • The models were validated in both internal and external cohorts, and a simple-to-use online tool was created to predict mortality rates.
  • The study concluded that the developed predictive models can provide valuable insights for clinicians to make informed decisions in the management of ACLF patients.

Statistics:

  • 5,994 patients with cirrhosis were admitted to ICU during the study period.
  • 1,511 patients met NACSELD criteria, and 1,692 met EASL-CLIF grade II or higher criteria.
  • The CBA model had an AUC of 0.87 in the NACSELD cohort, while the RFA model had an AUC of 0.83 in the EASL-CLIF cohort.
  • The simplified models demonstrated an AUC of 0.89 for the CBA model and 0.81 for the RFA model.
  • The models were validated in internal and external cohorts, with a validation coefficient of 0.85.

Sources:

  • Predictive machine learning model in intensive care unit patients with acute-on-chronic liver failure and two or more organ failures. Clinical and Molecular Hepatology, 2025, 31(4): 1355-1371. (Clinical and Molecular Hepatology - http://www.e-cmh.org)
  • NewsRx. Study Findings on Liver Failure Published by Researchers at Cedars-Sinai Medical Center (Predictive machine learning model in intensive care unit patients with acute-on-chronic liver failure and two or more organ failures). Gastroenterology Week. November 3, 2025; p 1138.