Retrospective analysis of risk factors for non-variceal upper gastrointestinal bleeding and construction of a nomogram prediction model

Aim: By analyzing the clinical data of patients with non-variceal upper gastrointestinal bleeding (NVUGIB), the independent risk factors for NVUGIB were found, and a risk prediction model was initially constructed. Methods: This retrospective analysis collected patients hospitalized in Laizhou City People's Hospital from January 2020 to January 2022. According to whether the patients had NVUGIB during hospitalization, they were divided into a bleeding group of 173 cases and a control group of 121 cases. We collected the medical records of the two groups, including general conditions, disease conditions, medication conditions, and laboratory test indicators. The independent risk factors of NVUGIB were screened by univariate and multivariate logistic regression analysis, and a prediction model was initially constructed. The nomogram was developed using R language. the establishment of a regression equation model was based on the above risk factors: logit (P) = -8.320 + 0.436 * history of peptic ulcer + Helicobacter pylori infection * 0.522 + use of anticoagulant and antiplatelet drugs * 0.881 + 0.583 * increased leukocyte count + prolonged international normalized ratio (INR) * 0.651 + hypoproteinemia * 0.535. By using receiver operating characteristic curves, area under curve and Hosmer-Lemeshow test, the discrimination and calibration of the model was evaluated, and a calibration curves were plotted. Results: Univariate and multivariate regression analysis identified that history of peptic ulcer, Helicobacter pylori infection, use of anticoagulant and antiplatelet drugs, increased leukocyte count, prolonged INR and hypoproteinemia were risk factors for NVUGIB. Those risk factors were used to construct a clinical predictive nomogram. The calibration curves for NVUGIB risk revealed excellent accuracy of the predictive nomogram model. The unadjusted C-index was 0.773 [95% CI, 0.515-0.894]. The area under the curve was 0.793982. Decision curve analysis showed that the predictive model could be applied clinically when the threshold probability was 20 to 60%. Conclusions: A history of peptic ulcer, Helicobacter pylori infection, use of anticoagulant and antiplatelet drugs, increased leukocyte count, prolonged INR, and hypoproteinemia may be independent risk factors for NVUGIB. Furthermore, this study initially established a risk prediction model for NVUGIB and developed a nomogram. It was verified that the model had good differentiation ability and consistency, andcould provide a practical reference for clinical work.