Nomogram Predicts In-Hospital Mortality in Patients with Emergency Gastrointestinal Bleeding: A Multicenter Retrospective Study

Background Gastrointestinal bleeding (GIB) is a frequent issue encountered in emergency departments, associated with significant rates of incidence and mortality. This study aims to create and validate a reliable nomogram to predict the risk of in-hospital mortality in patients experiencing emergency GIB. Additionally, it seeks to identify the risk factors that influence mortality and to equip the emergency clinical team with a precise predictive tool. Methods This study utilized a retrospective cohort design to analyze data from patients with GIB who presented to the emergency departments and were subsequently admitted at three branches of Wuhan Central Hospital: Nanjing Road, Houhu, and Yangchunhu, from January to December 2023. Patient information was collected through the hospital's information system. The LASSO regression method was employed to identify key variables for prediction, and a nomogram was constructed using multivariate logistic regression. The model's ability to discriminate between outcomes was assessed by calculating the area under the curve (AUC). Furthermore, calibration analysis and decision curve analysis (DCA) were performed to evaluate the model's performance. Results A total of 847 patients were included, with 75 (8.85%) dying during hospitalization. In-hospital mortality was more common among elderly patients (median age 73 years vs. 65.5 years for survivors, P < 0.001). Deceased patients had lower systolic and diastolic blood pressures, higher heart rates, and higher shock indices upon emergency admission (P < 0.001). They were more likely to arrive by ambulance (P < 0.001) and classified as ESI Level 1 (P < 0.001). Additionally, they had a higher incidence of malignant tumors (P < 0.001), underwent fewer surgeries (P = 0.003), and received fewer hemostasis procedures (P < 0.001). Their total hospitalization costs were also higher (P < 0.001). Logistic regression analysis identified Ambulance ED, Shock Index > 1, ICU admission, malignancy, and hemostatic procedures as independent risk factors for GIB. ROC curve analysis showed an AUC of 0.862 (95% CI: 0.786–0.939) for the training cohort and 0.846 (95% CI: 0.787–0.904) for the validation cohort. Conclusion The developed nomogram model effectively predicts in-hospital mortality risk among emergency GIB patients, demonstrating good classification performance and clinical potential. It is recommended that this model be integrated into clinical information systems to support decision-making and optimize patient management.