Establishment and Validation of the Nomogram Model for Mortality Risk in Patients with Dangerous Upper Gastrointestinal Bleeding

Abstract Objective A Nomogram model was established to predict the mortality risk for patients with dangerous upper gastrointestinal bleeding (DUGIB), aimed at identifying those high-risk one who required emergency treatment. Methods Collected the clinical data of 256 DUGIB patients who were admitted to Intensive Care Unit (ICU) of Renmin Hospital of Wuhan University from January 2020 to April 2022. A total of 179 patients admitted to ICU were treated as training cohort, while 77 patients who were admitted to ICU of East Campus were considered validation cohort. Multivariate logistic regression analysis was performed to determine the independent risk factors, and R packages were used to construct nomogram model. The prediction accuracy and identification ability were evaluated by receiver operating characteristic curve (ROC), C index and calibration curve. Furthermore, the clinical value of the model was evaluated by decision curve analysis (DCA). Simultaneously, the model was externally validated. Results A total of 256 patients diagnosed with DUGIB were included, among which 67 patients (19.14%) eventually died. Multivariate logistic regression analysis suggested that hematemesis, urea nitrogen level, emergency endoscopy, AIMS65 score, GBS and Rockall score were all independent risk factors. ROC curve analysis showed the area under the curve (AUC) of training cohort was 0.980 (95% CI : 0.962~0.997), while validation cohort was 0.790 (95% CI : 0.685~0.895). The calibration curves for both training and validation cohorts were then tested by Hosmer-Lemeshow Goodness ( P =0.778, P =0.516). To summarize, the Nomogram prediction model was of great clinical value. Conclusion The nomogram was proven to be an effective tool for the risk stratification of DUGIB patients, and was also helpful for the early identification and intervention, especially in ICU.