Model validation results.

Background and PurposeCirrhosis patients face high mortality risks from complications, emphasizing the need for prognostic tools to estimate risk of adverse liver outcomes based on cirrhosis etiology and hence improve disease management. This study aimed to evaluate the association between baseline clinical factors, lab values and noninvasive measurements and risk of adverse liver outcomes among patients with cirrhosis, and develop disease-specific associative models.MethodsUsing proportional hazards regression, we developed six associative models, categorized by cirrhosis etiology, after identifying significant predictors of 30-day risk of ascites, hepatic encephalopathy (HE), and variceal bleeding (VB) among cirrhosis patients, with measurements selected through LASSO regression.ResultsA total of 4045 adult patients with cirrhosis were included in the analysis from a single U.S. center retrospective cohort. The 5-year rates for ascites, HE, and VB were 31.6%, 22.9%, and 30.7%, respectively. Multivariable analyses showed that independent predictors in cirrhosis due to metabolic dysfunction-associated steatohepatitis (MASH) and viral hepatitis were: (a) ascites: albumin and international normalized ratio (INR); (b) HE: albumin, INR, total bilirubin, platelet count; (c) VB: albumin, platelet count, hemoglobin. No variables were significantly associated with outcomes in patients with alcohol-associated liver disease.ConclusionsThe newly developed models provided accurate estimates of the 30-day risks of ascites, HE, and VB in MASH or viral hepatitis-related cirrhosis using baseline variables, providing a reliable tool for identifying high-risk patients warranting intensified interventions.