Establishment Of Risk Prediction Model For Rebleeding In Patients With Acute Non-Varicose Upper Gastrointestinal Bleeding

Objective: To explore the risk factors of rebleeding in patients with acute non-variceal upper gastrointestinal bleeding(ANVUGIB),and establish a visual rebleeding prediction model,in order to provide scientific reference for screening high-risk groups and primary prevention strategies.Methods: Through the inpatient case system,we screened the patients who were definitely diagnosed as ANVUGIB by endoscopy from October 2018 to October 2022,and collected the general information,past history,medication,symptoms,laboratory tests and prognosis of the patients.Through strict inclusion and exclusion criteria,544 patients who were diagnosed as ANVUGIB and successfully stopped bleeding immediately after endoscopic hemostasis treatment in the hospital from October 2018 to October 2022 were diagnosed.The case data of ANVUGIB patients during hospitalization were systematically reviewed and collected.A patient questionnaire was developed to record the basic information,clinical characteristics and laboratory examination indicators of ANVUGIB patients in detail.According to the rebleeding situation of patients after treatment,they were divided into bleeding group(39 cases)and non-bleeding group(504 cases).Data collection was carried out by Excel software,and sorting and analysis were carried out by SPSS26.0 and R4.0.3 software.Qualitative data were described by the percentage of cases and analyzed by chi-square test.Continuous variables were described by the standard deviation of the mean and analyzed by independent sample t test.The selection variables were performed by Lasso regression,and the logistic regression model was used to construct a nomogram risk prediction model.The Harrell’s C statistic was used to calculate the consistency index(C-index),and the ROC curve was used to evaluate the accuracy of the model.The clinical applicability was evaluated by DCA curve and CIC curve.Results:1.Among the 544 patients with ANVUGIB,39 patients had rebleeding,with an incidence of 7.2%.Among them,there are 100 women,accounting for 18.4%,and 444 men,accounting for 81.6%.The average age was 58.66 ± 16.07 years old.2.Compared with the baseline data of the two groups,the two groups had statistically significant differences in smoking,cerebral infarction,non digestive tract tumors,type 2 diabetes and peptic ulcer history(P<0.05).3.Compared with the laboratory test indexes of the two groups,the differences in PT,INR and PLT between the two groups were statistically significant(P<0.05).4.Lasso regression variable screening: log is selected this time(λ)= 0.000292,punish the partial likelihood function.Nine variables were screened: age(β= 0.002).Smoking(β=-0.053).Cerebral infarction(β= 0.071),non-digestive tract tumo(r β= 0.046).History of peptic ulce(r β=-0.021).Blood transfusion treatmen(t β= 0.052)、 PT(β= 0.024)、INR(β= 0.199)、PLT(β=-0.003)。5.Logistic regression prediction model and multivariate analysis showed that cerebral infarction,alcohol consumption,non digestive tract tumor,type 2 diabetes,digestive tract ulcer,PT,INR,PLT had statistical significance(P<0.05).Cerebral infarction,alcohol consumption,non digestive tract tumor,type 2 diabetes,digestive tract ulcer,PT,INR are risk factors,and PLT is protective factor.LASSO-logistic regression model is in the validation set.The area of AUC in the validation set of LASSOlogistic regression model was 0.928(0.903-0.948),the sensitivity was 97.44%,the specificity was 81.55%,the positive predictive value was 94.2%,and the negative predictive value was 87.3%.6.Evaluation of clinical applicability: As shown in the DCA curve,the threshold of 0.15-0.95 has the greatest clinical net benefit.The clinical impact curve also shows that within the threshold probability range,the expected rebleeding patients are always more than the actual recurrence patients.Conclusion: This study found that the rebleeding rate of patients with ANVUGIB was 7.2%.The model results showed that patients with history of cerebral infarction,history of alcohol consumption,history of non digestive tract tumor,history of type 2 diabetes,history of digestive tract ulcer,increase of PT,increase of INR and decrease of PLT were high-risk groups for rebleeding events,and targeted intervention measures should be taken to prevent and control high-risk groups.The risk of rebleeding in patients with acute non-varicose upper gastrointestinal bleeding is predicted by the nomograph model,which makes the risk factors visible and provides a more intuitive reference for clinicians.