基于决策树方法的2型糖尿病患者痰浊证与临床指标关联模式的建立

Establishment of the Associated Model between Turbid Phlegm Syndrome and Clinical Indicators in the Patients of Diabetes Type 2 Based on Decision Tree Method

目的采用决策树的数据挖掘方法建立2型糖尿病患者痰浊证与临床常规检测指标间的关联模式。方法采用多中心临床流行病学调查方法,在全国5家三级甲等医院共收集249例合格2型糖尿病病例,综合分析基本信息、中医四诊信息、临床常规检测指标。在t检验、非参数检验、Pearson相关分析基础上,进一步采用决策树的数据挖掘方法建立痰浊证与临床常规检测指标间的关联模式。结果 249例患者中有106例为痰浊证,占42.57%。以尿素氮、白细胞、平均红细胞体积、超敏C反应蛋白、红细胞、甲状腺素6个核心指标建立了痰浊证决策树模型,10倍交叉验证得到模型的灵敏度为75.47%、特异度为76.22%,检测总正确率为75.90%。结论决策树模型可以清晰、直观的进行2型糖尿病患者痰浊证的判断,在证候客观化研究中显示了一定的优势。

Objective To establish the associated model between turbid phlegm syndrome and clini-cal routine indicators in the patients of diabetes type 2,using data-mining method of decision tree.Methods The multi-central clinical epidemiological investigation was adopted.Two hundred and forty-nine cases of diabetes type 2 were collected from 5 Three-A hospitals.The basic information,the information of four di-agnostic methods of TCM and clinical routine indicators were analyzed comprehensively.On the basis of the t test,nonparametric test and Pearson correlation analysis,the data-mining method of decision tree was adopt-ed further to set up the association model between turbid phlegm syndrome and clinical indicators.Results Of 249 cases,1 06 cases（42.57%）were differentiated as turbid phlegm syndrome.Six core indicators inclu-ding urea nitrogen,white blood cells,average red blood cell volume,hypersensitive C-reactive protein,eryth-rocyte and thyroxin were used to establish decision tree model of turbid phlegm syndrome.Ten times cross validation indicated that the model sensitivity was 75.47%,the specificity was 76.22%and the total accura-cy was 75 .90%.Conclusion The decision tree model clearly and intuitively determines turbid phlegm syn-drome of diabetes type 2 in the patients and displays a certain advantage in syndrome objectification study.