·糖尿病临床流行病学·

　　【摘要】目的 对产生自北京社区人群的糖尿病(DM)简易决策树模型和糖尿病及糖尿病前期(DM+PDM)简易决策树模型进行外部验证。方法 北京地区无糖尿病病史的年龄20岁以上的社区居民2989名，进行75gOGTT。按照WHO 1999年糖尿病和糖尿病前期诊断标准，绘制受试者工作特征(ROC)曲线验证DM和DM+PDM简易决策树模型的表现。结果 DM和DM+PDM简易决策树模型ROC曲线下面积分别为0.624 (0.596, 0.652)和0.648 (0.630, 0.667)，敏感性分别为66.3%和67.7%，特异性分别为58.5%和61.9%。结论 DM和DM+PDM简易决策树模型在北京社区人群中具有中等程度的敏感性和特异性，可以用于社区糖尿病前期和糖尿病高危人群的初步筛查。

　　【关键词】 2型糖尿病;糖尿病前期;筛查;验证;决策树

External validation of decision trees for detecting type 2 diabetes and prediabetes ZHOU Xiang-hai, ZHANG Xiu-ying, LUO Ying-ying, et al. Peking University People’s Hospital, Beijing100044, China.

Corresponding author: Ji Li-nong, E-mail: jiln@bjmu.edu.cn

　　【Abstract】 objective To externally validate decision trees for detecting previously undiagnosed diabetes (DM) and diabetes and prediabetes (DM+PDM) developed from a Beijing community population. Methods A diabetes survey was conducted in 2989 individuals aged 20 years or older living in communities in Beijing.75g glucose tolerance test was performed to define diabetes and prediabetes according to 1999 WHO criteria. Receiver operating characteristic curve (ROC curve) was plotted to determine the performance of the decision trees for detecting DM and DM+PDM . Results The area under the ROC curve of the decision trees for detecting DM and DM+PDM were 0.624 (0.596, 0.652) and 0.648 (0.630, 0.667), respectively. The decision tree for detecting DM gave a sensitivity of 66.3% and a specificity of 58.5%. The sensitivity and specificity of the decision tree for detecting DM+PDM were 67.7% and 61.9%, respectively. Conclusion The decision trees for detecting DM and DM+PDM have moderate performance and can be used for detecting previously undiagnosed diabetes and prediabetes in Beijing community residents.

　　【key words】Diabetes mellitus, Type 2; prediabetes; screening; validation; decision tree