摘要
目的基于最优子集法建立肠道准备预测模型,量化评估影响肠道准备的预测变量。方法选取2017年1月至2019年10月在广东省中医院总院消化内科行结肠镜检查的门诊及住院的患者。分别运用向前逐步和最优子集法在训练集中建立logistic回归模型,并在测试集中评估两个模型的ROC曲线下面积(area under the ROC curve,AUC)、净重新分类改善指数(net reclassification index,NRI)、综合判别改善指数(integrated discrimination improvement,IDI)等指标。结果研究共纳入了455例患者,其中病例组195例,对照组260例。经最优子集模型筛选出的4个预测变量为高纤维饮食、糖尿病、便秘史、术前活动,AUC值0.873(0.819~0.926)。两个模型间的AUC值、NRI、IDI均差异无统计学意义,但最优子集模型更为简洁。结论最优子集法模型能更简洁、有效地筛选出影响肠道准备的因素,根据这些因素建立的列线图能为医护人员提供有效的指导。
Objective To establish a prediction model for intestinal preparation based on the optimal subset method and to quantitatively evaluate the predictive variables affecting bowel preparation.Methods The patients visiting the outpartient department and hospitalized in Guangdong Provincial TCM Hospital from Jan.2017 to Oct.2019 were selected, who received colonoscopy. The logistic regression model was established in the training set using the forward stepwise and optimal subset method.the AUC(Area under the ROC Curve),NRI(Net Reclassification Index)and IDI(Integrated Discrimination Improvement)were evaluated in the test set.Rresults A total of 455 patients were included in the study,including 195 in the case group and 260 in the control group.The four predictors selected by the optimal subset model were high fiber diet,diabetes,constipation history and preoperative activity,and the AUC value was 0.873(0.819-0.926).There was no statistical difference in AUC value,NRI or IDI between the two models,but the optimal subset model was more concise.Conclusion The optimal subset method model can screen out the factors affecting bowel preparation more concisely and effectively.The nomogram established based on these factors can provide effective guidance for medical staff.
作者
李健民
刘添文
符思远
李叶
林燕凤
张北平
LI Jian-min;LIU Tianwen;FU Si-yuan;LI Ye;LIN Yan-feng;ZHANG Bei-ping(Department of Gastroenterology,Guangdong Provinicial Hospital of Chinese Medicine,Guangdong 510140,China;不详)
出处
《中国实用内科杂志》
CAS
CSCD
北大核心
2020年第3期231-236,共6页
Chinese Journal of Practical Internal Medicine
基金
广东省中医院院内专项(YN2016ZWB02)
广东省自筹经费类科技计划项目(2017ZC0169)。
关键词
肠道准备
便秘
最优子集法
预测模型
bowel preparation
constipation
optimal subset method
prediction model
