Abstract: Objective The aim of this study was to construct a risk prediction model for nosocomial death in cirrhotics with portal hypertension and esophagogastric variceal bleeding (EVB). Methods A retrospective cohort study was conducted on the clinical data of 107 patients with cirrhotic portal hypertension and EVB who were admitted to our hospital between June 2018 and June 2020. All patients received transjugular intrahepatic portosystemic shunt (TIPS) therapy. The Logistic regression analysis was performed to screenindependent risk factors influencing the prognosis of patients with cirrhotic portal hypertension and EVB. A risk prediction model for nosocomial death was constructed based on these independent factors, and its predictive efficacy was verified by the area under receiver operating characteristic curve (AUC). Results 25 patients (23.4%) died and 82 patients survived in our series; the univariate Logistic regression analysis showed that Child class, portal vein diameter, the sites of bleeding, the incidence rates of hepatic encephalopathy and hemorrhagic shock in died patients were significantly different compared to in survivals (P<0.05), and the multivariate Logistic regression analysis demonstrated that the portal vein diameter (OR=2.201, 95%CI: 1.544-3.139), hepatic encephalopathy (OR=3.093, 95%CI: 1.731-5.524) and hemorrhagic shock (OR=1.101, 95%CI: 1.040-1.165) were the independent risk factors for nosocomial death (P<0.05); the C-index of the constructed nomogram model we built up by internal verification for predicting nosocomial death of patients with cirrhotic portal hypertension and EVB was 0.937 (95%CI: 0.734-0.879), witha good discrimination, and the AUC was 0.896 (95%CI: 0.796-0.958, P<0.001), with the sensitivity and specificity of the prediction model were 91.3% and 88.1%, respectively. Conclusion The early recognition of risk factors of nosocomial death in patients with cirrhotic portal hypertension and EVB is important for appropriate management of patients, and the risk prediction model we constructed might have a good predictive efficacy.

Key words: Cirrhosis, Esophagogastric variceal bleeding, Transjugular intrahepatic portosystemic shunt, Nosocomial death, Nomogram prediction model