Establishment of a risk prediction model for acute upper gastrointestinal rebleeding in elderly patients / 中华危重病急救医学

ABSTRACT

Objective:To establish a prediction model of acute upper gastrointestinal rebleeding (AUGIRB) for elderly patients by combining and applying multiple indicators.Methods:A retrospective observational study was conducted. The clinical data of 161 elderly patients (age ≥ 65 years old) who suffered from acute upper gastrointestinal bleeding (AUGIB) and who were hospitalized in Shanghai General Hospital from July 2018 to December 2020 were recorded. The patients were divided into the rebleeding group (31 cases) and the non-rebleeding group (130 cases) according to whether gastrointestinal rebleeding occurred. Univariate analysis was adopted to screen AUGIRB-related risk factors and Logistic regression analysis was used to screen independent predictors of AUGIRB so that a predictive model was constructed. Based on the area under the curve (AUC) of the receiver operator characteristic curve (ROC curve), the predictive ability of the prediction model for AUGIRB was evaluated, the optimal cut-off value was determined, and the odds ratio ( OR) and its 95% confidence interval (95% CI) were calculated. Bootstrap resampling technology was used to validate the predictive ability of the model. Results:Univariate Logistic analysis showed that oral anticoagulant drugs, oral antiplatelet drugs, albumin (ALB), platelet count (PLT), Glasgow-Blatchford bleeding score (GBS), D-dimer, fibrinogen (FIB), and international normalized ratio (INR) all had a significant effect on the occurrence of AUGIRB among elderly patients. Multivariate Logistic regression analysis showed that the oral antiplatelet drugs ( OR = 11.150, 95% CI was 1.888-65.852, P < 0.05) and GBS score ( OR = 2.503, 95% CI was 1.523-4.114, P < 0.05) were the independent risk factors of AUGIRB among elderly patients, while the ALB ( OR = 0.764, 95% CI was 0.626-0.932, P < 0.05) and FIB ( OR = 0.065, 95% CI was 0.011-0.370, P < 0.05) were the protective factors of AUGIRB among elderly patients. The AUC of the above four indexes joint prediction model was 0.979. The verification results of the model showed that the consistency index (C-index) of the model was 0.986. Conclusion:The prediction model fitted in this research has a high prediction accuracy and it also has a certain reference value for the judgment of elderly patients who suffer from AUGIRB.