分类树模型与Logistic回归在儿童高血压预测中的应用

Application of classification tree and Logistic regression analysis model on prediction of children ' s hypertension

目的应用分类树模型与Logistic回归模型分析郑州市儿童青少年血压的影响因素,为青少年高血压早期干预提供依据。方法对2010年郑州市7~17岁学生4 400人体质调研数据进行分析,运用分类树模型及Logistic回归对儿童高血压影响因素进行探讨,建立预测模型并运用ROC曲线比较2种模型优劣。结果郑州市儿童高血压患病率较低,为6.5%,男女血压偏高发生率分别为12.4%,15.9%,差异无统计学意义（χ^2=0.74,P〉0.05）;肥胖与超重儿童高血压发生率大于正常儿童,差异有统计学意义（χ^2=16.14,P〈0.05）;Logistic回归显示超重肥胖与血压关系密切,超重青少年患高血压风险是正常体重的2.118倍（OR=2.118,95%CI=1.492~3.007）,肥胖青少年患高血压的风险为正常体重的6.933倍（OR=6.933,95%CI=5.183~9.273）;分类树模型和Logistic回归均显示超重、肥胖是儿童高血压主要影响因素。结论分类树模型对儿童高血压预测效果较好,可作为Logistic回归模型的补充。

Objective To explore the risk factors of hypertension in Zhengzhou by using method of classification tree and Logistic regression analysis model. Methods Using data mining techniques in the data from the survey on students constitution and health in Zhengzhou in 2010, the classification tree and Logistic regression analysis prediction model were evaluated, ROC curve was used to compare the two models. Results The prevalence rate of children hypertension was low in Zhengzhou city, the preva- lence of hypertension had no statistical significance by gender （P〉 0.05} ;Compared to the normal weight, overweight and obese children and adolescents had elevated blood pressure, with statistical significance （P〈 0. 05）. The Logistic regression analysis showed that the important risk factors of essential hypertension were BMI, the OR value of the overweight group with hypertension was 2.118 times as the normal weight, the OR value of the obesity group with hypertension is 6.933 times than the normal weight, Classification tree model and Logistic regression showed that overweight and obesity was the main influencing factors of children hy- pertension. Conclusion The use of classification tree model seems to improve the prediction effect, can be used to supplement the Logistic regression model.