Liver Severity Score-Based Modeling to Predict Six-Week Mortality Risk Among Hospitalized Cirrhosis Patients With Upper Gastrointestinal Bleeding.
Academic Article
Overview
abstract
BACKGROUND: Patients with cirrhosis who have gastrointestinal bleeding have high short-term mortality, but the best modality for risk calculation remains in debate. Liver severity indices, such as Child-Turcotte-Pugh (CTP) and Model-for-End-Stage-Liver Disease (MELD) score, are well-studied in portal hypertensive bleeding, but there is a paucity of data confirming their accuracy in non-portal hypertensive bleeding and overall acute upper gastrointestinal bleeding (UGIB), unrelated to portal hypertension. AIMS: This study aims to better understand the accuracy of current mortality risk calculators in predicting mortality for patients with any type of UGIB, which could allow for earlier risk stratification and targeted intervention prior to endoscopy to identify the bleeding source. METHODS: In a large US single-center cohort, we investigated and recalibrated the model performance of CTP and MELD scores to predict six-week mortality risk for both sources of UGIB (portal hypertensive and non-portal hypertensive). RESULTS: Both CTP- and MELD-based models have excellent discrimination in predicting six-week mortality for all types of bleeding sources. However, only a CTP-based model demonstrates calibration for all bleeding, regardless of bleeding etiology. Median predicted 6-week mortality by CTP class A, B, and C estimates a risk of 1%, 7%, and 35% respectively. CONCLUSIONS: Our study corroborates findings in the literature that CTP- and MELD-based models have similar discriminative abilities for predicting 6-week mortality in hospitalized cirrhosis patients presenting with either portal hypertensive or non-portal hypertensive UGIB. CTP class is an effective clinical decision tool that can be used, even prior to endoscopy, to accurately risk stratify a patient with known cirrhosis presenting with any UGIB into low, moderate, and severe risk groupings.
