Journal of Clinical Hepatology
This study is based on 177 cases of liver cirrhosis esophageal and gastric variceal bleeding (EGVB) patients admitted to the First Affiliated Hospital of Xi'an Medical University from 2018 to 2023. By univariate and multivariate analysis, 8 independent predictive factors (RBC, ChE, ALP, Alb, TT, main portal vein diameter, sequential treatment, first-line prevention) were selected, and a new Logistic prediction model was constructed. A nomogram was drawn to achieve bedside evaluation. The model's AUC reached 0.928, significantly better than traditional scores such as MELD (0.603) and Child-Pugh (0.650), with the optimal cut-off value of 0.607, sensitivity and specificity both at 0.817. After Bootstrap internal validation, the model has good calibration and strong stability. The indicators used in the model are all routine tests and ultrasound examination items, non-invasive, easy to obtain, and highly acceptable to patients.
This model provides a simple, accurate, and scalable rebleeding risk stratification tool for liver cirrhosis EGVB patients, helping to identify high-risk populations early (such as those with prolonged TT and widened portal vein), and timely strengthen follow-up or intervention. At the same time, it proves that sequential treatment and first-line prevention are strong protective factors, indicating that attention should be paid to standardized endoscopic prevention and treatment strategies. Compared with invasive HVPG measurement or expensive and complex AI models, this model balances practicality and scientific rigor, filling the gap of convenient rebleeding prediction tools in grassroots clinical practice. Although there are limitations of single-center retrospective studies, it lays a foundation for subsequent multi-center prospective validation and integration of new parameters such as liver stiffness.
Objective: To establish a new noninvasive, simple, and convenient clinical predictive model by identifying independent predictive facto…
