Prognostic value of gastrointestinal endoscopy combined with hematocrit in acute non-variceal upper gastrointestinal bleeding

Objective:To investigate the prognostic value of gastrointestinal endoscopy combined with hematocrit (HCT) in acute non-variceal upper gastrointestinal bleeding (ANVUGIB).Methods:A total of 366 patients with ANVUGIB risk admitted to Chengdu Seventh People’s Hospital from August 2019 to August 2022 were selected and divided into ANVUGIB group(100 cases)and non-ANVUGIB group(266 cases)according to whether ANVUGIB occurred or not.Clinical data, gastrointestinal endoscopy performance and HCT level between the two groups were compared, and LASSO regression analysis screened the relevant factors affecting the occurrence of ANVUGIB.The influential factors of ANVUGIB were analyzed and modeled by multi-factor logistic regression.The nomogram of the logistic regression model was drawn in R language, Bootstrap method was used for internal validation, receiver operating characteristic(ROC) curve was used to analyze all correlated factors and the value of the nomogram in predicting ANVUGIB, and decision curve analysis(DCA)was used to verify the clinical utility of all correlated factors and nomogram.Results:Red blood cell count and HCT were lower in the ANVUGIB group than in the non-ANVUGIB group, urea nitrogen was higher than that in the non-ANVUGIB group, and the patients with gastric cancer, peptic ulcer, vascular nudity and blood clot adhesion were more than those in the non-ANVUGIB group (P< 0.05).Six predictor variables were screened by cross-validation method to obtain the optimal solution, logistic regression analysis showed that red blood cell count (OR=0.408), HCT(OR=0.322), gastric cancer (OR=6.829), peptic ulcer (OR=8.771), vascular nudity (OR=9.461), and blood clot adhesion (OR=9.829) were all influential factors for the occurrence of ANVUGIB(P< 0.05).The risk capacity index(C-index)of the nomogram of the logistic regression model was 0.962, with good predictive accuracy.Internal validation using Bootstrap showed that the predictive model was well calibrated; ROC curve showed that the AUC for predicting ANVUGIB was 0.962, with a sensitivity of 80.93% and a specificity of 96.24%, and the AUC for predicting ANVUGIB using the nomogram was greater than that of all correlated factors, and the net gain rates were higher than those of all correlated factors(P< 0.05).Conclusion:The occurrence of ANVUGIB is affected by red blood cell count, HCT, gastric cancer, peptic ulcer, vascular nudity, blood clot adhesion, etc.The nomogram of the logistic regression model has high predictive value for the occurrence risk of ANVUGIB.