早产危险因素的Logistic回归及分类树分析  

Logistic regression analysis and classification tree analysis of risk factors for preterm birth

作　　者：高素红[1] 张运平[1] 刘晓红[1] 王佳楣[1] 顾岳山 张久越 周霞[4] 李庆霞[5] 张欣荣 邢艳梅 赵金凤[4] 赵书燕[5] 

机构地区：[1]北京市海淀区妇幼保健院,北京100080 [2]北京市通州区妇幼保健院,北京101100 [3]北京市顺义区妇幼保健院,北京101300 [4]北京市昌平区妇幼保健院,北京102200 [5]北京市大兴区妇幼保健院,北京102600

出　　处：《中国妇幼健康研究》2012年第5期557-559,595,共4页Chinese Journal of Woman and Child Health Research

基　　金：联合国儿童基金会资助项目（YH601-11-8）;北京市科委资助项目（z090507017709014）

摘　　要：目的应用分类树模型和Logistic回归模型进行对比分析，探索北京地区早产发生的影响因素。方法对北京地区5家妇幼保健院的1323例早产进行1：1病例对照研究，应用分类树模型和Logistic回归模型分析早产的影响因素。结果Logistic回归模型显示均衡饮食（OR=0．509）、产前检查（OR=0．233）、常住址为城镇（OR=0．555）是早产的保护性因素，受教育程度低（OR=1．674）、负性生活事件（OR=6．086）、性生活（OR=1．704）、前置胎盘（OR=11．834）、妊娠期糖尿病（OR=3．170）、妊娠期高血压疾病（OR：5．024）、早产史（OR=17．574）和胎膜早破（OR=4．083）是早产的危险因素。利用卡方自动交互检测法建立的分类树模型，共筛选出5个早产高危因素，其中胎膜早破是最重要的影响因素，其他包括妊娠期高血压疾病、膳食结构不均衡、无产前检查、文化程度低。结论孕期应进行定期的产前检查可及早发现高危妊娠，避免各种不良刺激，保持良好心态，积极预防胎膜早破。Logistic回归模型可以提供变量影响的定量解释，分类树模型可以更好地展现变量间的复杂相互作用。二者结合可更好地服务于流行病学的研究。Objective To explore the risk factors of preterm birth in Beijing by using method of classification tree and Logistic regression model. Mdthods A 1 ： 1 case-control study was performed in five maternal and child health hospitals in Beijing. Data of 1 323 preterm birth and 1 323 controls was analyzed by classification tree and Logistic regression analysis. Results Logistic regression model showed that having a balanced diet （ OR = 0. 509）, taking prenatal care （ OR = 0. 233 ） and living in towns and cities （ OR = 0. 555 ） were the protective factors of. preterm birth, while less education （ OR = 1. 674）, negative life events （ OR = 6. 086）, sexual activity （ OR = 1. 704）, placenta previa （ OR = 11. 834）, gestational diabetes mellitus （ OR = 3. 170）, hypertensive disorder, complicating pregnancy （ OR = 5. 024）, history of preterm birth （ OR = 17. 574） and premature rupture of membrane （PROM） （ OR = 4.083 ） were risk factors. Five factors were selected by classification tree set up with chi-square automatic interaction detection （CHAID）, and PROM was the most important factor, including hypertensive disorder complicating pregnancy, non-balanced diet, without prenatal care and less education. Conclusion Regular prenatal care should be taken to detect high-risk pregnancy early. Pregnant women should avoid various bad stimulations, keep good mood and prevent PROM actively. The Logistic regression model can provide quantitative interpretation of impact of variables, and the classification tree model can show the complex interactions among variables. The two data processing methods can be combined for epidemiological study.

关 键 词：早产 危险因素 分类树模型 LOGISTIC回归 

分 类 号：R714[医药卫生—妇产科学]