妊娠期糖尿病危险因素的Logistic回归模型初探

Logistic regression analysis of risk factors for gestational diabetes mellitus

目的:新诊断标准下,调查分析妊娠期糖尿病(GDM)的相关因素,建立临床评分体系,初步构建GDM危险因素Logistic回归模型。方法:按2010年国际妊娠合并糖尿病研究组织(IADPSG)推荐的GDM的新诊断标准,通过病例-对照研究分析GDM发生的影响因素,建立Logistic回归模型,采用接受者工作特征曲线(ROC)和Hosmer-Lemeshow拟合优度检验评价模型。结果:妊娠年龄、孕前体重指数(BMI)、孕期补铁、糖尿病(DM)家族史、多囊卵巢综合征(PCOS)史、孕期锻炼6项因素进入回归模型。当累积分值达到6.3分时,Youden指数最大,为0.640,此时灵敏度为83.4%,特异度为80.6%,阳性预测值为81.0%,阴性预测值为83.1%,诊断准确度为82.0%。ROC曲线下面积为0.875。Hosmer-Lemeshow拟合优度检验P=0.91。结论:妊娠年龄、孕前BMI、孕期补铁、糖尿病家族史、PCOS是GDM发生的危险因素,而孕期锻炼是GDM的保护因素,初步构建了一种简便易行且具有较高诊断效能的回归模型。

Objective：To set up a clinical scoring system and build a logistic regression model of gestational diabetes mellitus （GDM） risk factors under the new diagnostic criteria, through investigating and analyzing the related factors of GDM. Methods：A case-control study was to establish a logistic regression model according to the new criteria for diagnosis of GDM recommended by 2010 international pregnancy with diabetes study groups （IADPSG）. Receiver operating characteristic curve （ROC） and Hosmer-lemeshow goodness-of-fit test were used to evaluate the model. Results：Six factors were added into the regression model. When the cumulative score reached 6.3, Youden index became the largest of 0.640. The sensitivity was 83.4%, the specificity was 80.6% ,positive predictive value was 81.0% ,and negative predictive value was 83.1% ,the diagnostic accuracy was 82.0%. The area under the ROC curve was 0. 875. Hosmer-Lemeshow goodness-of-fit test P was 0.91. Conclusions：Pregnancy age ,body mass index before childbirth, iron during pregnancy, family history of diabetes, history of polycystic ovary syndrome were risk factors while exercise during pregnancy was a protection factor for GDM. A simple and efficient regression model was built.