Development and evaluation of a predictive model of upper gastrointestinal bleeding in liver cirrhosis.

backgroundUpper gastrointestinal bleeding (UGIB) is a prevalent and severe complication of cirrhosis, often resulting from esophagogastric variceal bleeding (EVB). This condition poses significant life-threatening risks. Once bleeding occurs, the risk of recurrent episodes substantially increases, further compromising liver function and worsening patient outcomes. This study aims to identify risk factors for UGIB in cirrhotic patients using clinical examination data and to develop a non-invasive predictive model to improve diagnostic precision and efficiency.

methodsBased on the inclusion and exclusion criteria, the study included 140 cirrhotic patients hospitalized at the First Affiliated Hospital of Nanjing Medical University between June 2022 and May 2023, who experienced UGIB within six months after discharge. These patients were compared with 151 cirrhotic patients hospitalized at the same hospital during the same period, who were discharged within six months without experiencing UGIB. General characteristics of the patients during hospitalisation, laboratory parameters on admission, and liver and spleen stiffness were retrospectively collected, and a retrospective case-control study was conducted. All patients were randomly assigned to the training and validation sets in a ratio of 7:3. Independent factors associated with UGIB were identified by univariate analysis, multivariate logistic regression analysis, and stepwise regression analysis, on the basis of which a predictive model was developed. The model's performance was assessed via receiver operating characteristic (ROC) curve and decision curve analysis (DCA) and was compared with established prognostic models, including the Child-Pugh and MELD scores.

resultsThe study analyzed 291 patients with cirrhosis, of whom 208 were allocated to the training set and 83 to the validation set. Independent predictors were identified, and predictive models were constructed using multivariate logistic regression analysis, and stepwise regression analysis in the training set, followed by validation in the validation set. The stepwise regression analysis identified ascites, spleen stiffness, albumin, fibrinogen, total cholesterol, and total bilirubin as independent predictors of UGIB (P < 0.05). These variables were incorporated into the predictive model. The area under the curve (AUC) for UGIB prediction was 0.956 in the training set and 0.909 in the validation set, demonstrating strong predictive performance. Furthermore, comparative analysis using ROC and DCA demonstrated that the developed model outperformed established scoring systems, such as the Child-Pugh score and the MELD score.

conclusionAscites, spleen stiffness, albumin, fibrinogen, total cholesterol and total bilirubin as independent predictors of UGIB in cirrhotic patients.