皖南医学院学报

目的:对消化性溃疡(PU)患儿并发消化道出血(GIB)的影响因素进行分析，并构建其贝叶斯网络模型。方法:回顾性选取2021年12月～2025年4月于安徽省儿童医院儿童消化科接受治疗的155例PU患儿临床相关资料，根据是否发生GIB，将其分为GIB组和非GIB组。收集两组临床资料，采用多因素Logistic回归筛选PU患儿并发GIB的危险因素，并建立PU患儿并发GIB的贝叶斯网络模型，对其预测效能进行验证。结果:155例PU患儿中GIB发生率为21.29%(33例)，多因素Logistic回归分析显示，活动溃疡分期、腹痛、采用非甾体抗炎药物、采用抗血小板药物以及感染幽门螺杆菌等为PU患儿并发GIB的危险因素(P<0.01);构建PU患儿并发GIB的贝叶斯网络模型，活动溃疡分期、腹痛、采用非甾体抗炎药物、采用抗血小板药物以及感染幽门螺杆菌与PU患儿并发GIB均存在直接关系; PU患儿并发GIB的贝叶斯网络模型的ROC曲线下面积为0.873(95%CI:0.800～0.946)，灵敏度为84.8%，特异度为82.0%，校正曲线的预测值和实际值基本拟合。结论:基于活动溃疡分期、腹痛、非甾体抗炎药使用、抗血小板药物使用及幽门螺杆菌感染等危险因素构建的贝叶斯网络模型，对PU患儿并发GIB具有良好的预测效能，可为临床早期识别高危患儿提供参考。

Objective: To analyze the influencing factors of gastrointestinal bleeding( GIB) in children with peptic ulcer( PU),and construct a Bayesian network model for predicting the risks of GIB.Methods: The clinical data were reviewed in 155 children with PU who were treated between December 2021 and April 2025 in the Department of Pediatric Gastroenterology of our hospital.The cases were divided into GIB group and non-GIB group according to the occurrence of GIB.The clinical data of the two groups were collected.Multivariate logistic regression was used to screen the risk factors of GIB in children with PU,and a Bayesian network model for GIB in children with PU was established.Finally,the predictive efficacy of the constructed model was validated.Results: In the 155 children with PU,33( 21. 29%) developed GIB.Multivariate logistic regression analysis showed that active ulcer,abdominal pain,use of non-steroidal anti-inflammatory drugs or antiplatelet drugs,and Helicobacter pylori infection were risk factors for GIB in children with PU( P<0. 05).A Bayesian network model for GIB in children with PU was constructed,which indicated that there were direct relationships between ulcer stage,abdominal pain,use of non-steroidal anti-inflammatory drugs or antiplatelet drugs,Helicobacter pylori infection and GIB in children with PU.The area under the ROC curve of the Bayesian network model for GIB in children with PU was 0. 873( 95% CI: 0. 800-0. 946). The sensitivity and specificity of the model was84. 8% and 82. 0%,respectively,and the predicted and actual values of the calibration curve were basically fitted. Conclusion: The Bayesian network model constructed based on the risk factors,including active ulcer stage,abdominal pain,history of non-steroidal anti-inflammatory or antiplatelet drug use,and Helicobacter pylori infection,has good predictive efficacy for PU children complicated with GIB,and this model can provide a reference for early identification of high-risk children in clinic.