Development of a model for predicting acute cerebral infarction induced by non-variceal upper gastrointestinal bleeding

To evaluate the risk factors for acute cerebral infarction(ACI) in patients with non-variceal upper gastrointestinal bleeding(NVUGIB), and construct a model for predicting ACI in NVUGIB patients. A model for predicting ACI induced by NVUGIB was established on the basis of a retrospective study that involved 1282 patients who were diagnosed with NVUGIB in the emergency department and Gastroenterology Department of Nanchang University Affiliated Ganzhou Hospital from January 2019 to December 2021. Receiver operating characteristic (ROC) curves were drawn to evaluate the sensitivity and specificity of the model and CHA2DS2-VASc score to predict ACI. Delong's test was used to compare AUCs of the present score and the CHA2DS2-VASc score. There were 1282 patients enrolled in the study, including 69 in the ACI group and 1213 in the non-ACI group. Multivariate analysis revealed that hypertension, diabetes, red blood cell (RBC) transfusion, mechanical ventilation, D-dimer, rate pressure product (RPP), somatostatin and mean platelet volume (MPV) were factors associated with ACI induced by NVUGIB. A model based on the eight factors was established, Logit(P)= 0.265 + 1.382 × 1 + 1.120 × 2 + 1.769 × 3 + 0.839 × 4–1.549 × 5–0.361 × 6 + 0.045 × 7 + 1.158 × 8(or 1.069 ×9) (X1, hypertension=1; X2, diabetes=1; X3, RBC transfusion=1; X4, mechanical ventilation=1; X5, somatostatin=1; X6, MPV(fL); X7, D-dimer(ng/l); X8, low RPP= 1; X9, high RPP = 2). The area under ROC curve of the model was 0.873, the sensitivity and specificity were 0.768 and 0.887, respectively. The area under ROC curve of CHA2DS2-VASc score was 0.792, the sensitivity and specificity were 0.728 and 0.716, respectively. Delong's test showed the area under ROC curve of the present study was significantly larger than that of CHA2DS2-VASc score. Hypertension, diabetes, RBC transfusion, mechanical ventilation, D-dimer, RPP, somatostatin and MPV were factors associated with ACI induced by NVUGIB. A model constructed based on these factors showed excellent prediction of ACI, and was superior to CHA2DS2-VASc score. However, this needs to be further validated by multi-center study with a larger sample size.