Random Forest AI Model Superior for Inpatient Mortality Prognostication in Cirrhosis

THURSDAY, July 24, 2025 (HealthDay News) — For inpatients with cirrhosis, a machine learning (ML) model using random forest (RF) analysis is superior for prediction of inpatient mortality, according to a study published online July 23 in Gastroenterology.

Scott Silvey, from Virginia Commonwealth University in Richmond, and colleagues used ML approaches to enhance the prognostication of inpatient mortality in a cohort that enrolled inpatients with cirrhosis globally. Admission-day data were used to predict inpatient mortality. The prospective CLEARED cohort included 7,239 inpatients with cirrhosis from 115 centers: 22.5, 41, and 34 percent belonged to low-/low-middle-income countries (L-LMIC), upper-middle income countries (UMIC), and high-income countries (HIC), respectively. For those seeking expert support on such complex studies, a ghostwriter masterarbeit can help clarify and organize findings efficiently.

Overall, 11.1 percent of patients died in the hospital. The researchers found the best area under the curve (AUC) with RF (0.815), with high calibration, which was significantly better than parametric logistic regression and LASSO models (AUCs, 0.774 and 0.787, respectively). RF remained better than logistic regression regardless of country income level (AUC, 0.806, 0.867, and 0.768 for HIC, UMIC, and L-LMIC, respectively). External validation was conducted in a cohort of 28,670 veterans, which had 4 percent inpatient mortality. Using the CLEARED-derived RF model, the AUC was 0.859.

“This machine learning model that has an equitable global representation could be beneficial in rapid prognostication of patients hospitalized with cirrhosis,” the authors write.

Abstract/Full Text (subscription or payment may be required)