【摘要】 目的 构建慢性阻塞性肺疾病（chronic obstructive pulmonary disease,COPD）患者筛检模型。方法 采用多阶段分层随机抽样的方法，抽取贵州省≥40岁的常住居民4 587名，对其进行问卷调查、体格检查及肺功能检查。经过单因素分析初步筛选模型纳入变量，经多因素logistic回归确定最终纳入变量。分别应用logistic回归（logistic regression,LR）、随机森林（random forest,RF）、支持向量机（support vector machine,SVM）构建COPD患者筛检模型，使用受试者工作曲线下面积（area under the curve,AUC）评价模型效果。使用delong法检验模型之间AUC的差异。结果 根据多因素logistic回归分析结果，本研究将年龄、14岁前经常咳嗽、哮喘、每日吸烟量（支）、烹饪燃料与排风、有害气体暴露6种因素纳入LR、RF、SVM模型。三种模型训练集AUC分别为73.64%、87.14%、73.30%，测试集AUC分别为76.10%、70.96%、76.08%，均具有较好的筛检效果。Delong法结果显示，三种模型的筛检效果在训练集与测试集均存在一定差异。结论本研究通过年龄、哮喘等6个简单变量建立经济、快捷且有效的COPD患者筛检模型。更多还原

【Abstract】 Objective To establish a screening model for patients with chronic obstructive pulmonary disease(COPD). Methods By using the method of multi-stage stratified random sampling, 4 587 permanent residents ≥ 40 years old in Guizhou Province were investigated by questionnaire, physical examination, and pulmonary function examination. Variables to be in-cluded into the model were screened by univariate analysis and then further screened by multivariate Logistic regression. Lo-gistic regression(LR), random forest(RF) and support vector machine(SVM) were used to construct the screening model of COPD patients, and the area under the curve(AUC) was used to evaluate the effect of the model. Delong method was used to test the difference of AUC between models. Results According to the results of multivariate Logistic regression analysis, age,frequent cough before 14 years old, asthma, daily smoking, cooking fuel and exhaust, and harmful gas exposure were included in LR, RF and SVM models. The AUC of the three model training sets were 73.64%, 87.14%, and 73.30%, respectively,and the AUC of the test set were 76.10%, 70.96%, and 76.08%, respectively, all of which had good screening results. The results of Delong method showed that the screening effects of the three models were different between the training set and the test set. Conclusion This study established an economical, rapid, and effective screening model for COPD patients through six simple variables such as age and asthma.更多还原