广西地区妊娠期糖尿病的危险因素分析及其风险预测模型的建立

摘要

Objective:To explore the risk factors for patients with gestational diabetes mellitus (GDM) in Guangxi,and to establish a risk model for the early diagnosis of the patients.Methods:A retrospective analysis was conducted in 578 pregnant women in Guangxi from 8 hospitals including the First Affiliated Hospital of Guangxi Medical University,Maternal and Child Health Hospital of Nanning,Liuzhou,Qinzhou,Yulin,Guigang,Wuzhou and Yuzhou.The risk factors for GDM were analyzed by univariate and multivariate Logistic regression.The Logistic regression model was evaluated using Homer-Lemeshow goodness of-fit test and receiver operating characteristic (ROC).Results:Age,pre-pregnant genital tract inflammatory,hormone used before pregnancy,family history of diabetes,fasting plasma glucose level and white blood cells counts were independent risk factors for GDM (P＜0.05).The area under the ROC curve (AUC) for the evaluation model was 0.768 (95％ CI:0.716-0.820).The sensitivity of Logistic regression model for GDM was 59.26％,specificity was 89.58％,positive predictive value was 61.54％ and negative predictive value was 88.66％.Conclusion:Age,pre-pregnant genital tract inflammatory,hormone used before pregnancy,family history of diabetes,fasting plasma glucose level and white blood cells counts were independent risk factors for GDM.The risk model might serve as an effective tool for predicting GDM in pregnant women in Guangxi and guide the clinical practice.%目的:分析广西地区妊娠期糖尿病(GDM)的危险因素,建立风险预测模型.方法:回顾性分析2015年7-9月在广西医科大学第一附属医院、南宁市妇幼保健院、柳州市妇幼保健院等8所医院进行常规产检的578例广西籍贯孕妇的临床资料.采用单因素及多因素Logistic分析GDM的独立危险因素.建立模型,采用Homer Lemeshow拟合优度检验和受试者工作特征曲线(ROC)对该模型进行评估,并随机抽取20％的研究对象,根据临界点判断模型准确性.结果:年龄、生殖道炎症、妊娠前服用激素、糖尿病家族史、妊娠前空腹血糖和白细胞计数均为GDM的独立危险因素(均P＜0.05).所建立的Logistic回归模型,拟合优度检验P=0.408,ROC曲线下面积(AUC)为0.768(95％ CI:0.716～0.820),灵敏度为59.26％,特异度为89.58％,阳性预测值为61.54％,阴性预测值为88.66％.结论:年龄、生殖道炎症、妊娠前服用激素、糖尿病家族史、妊娠前空腹血糖和白细胞计数为GDM的独立危险因素;基于危险因素建立的风险预测模型对于广西籍贯孕妇发生GDM有较高的预测价值,可用于指导临床实践.