孕早期妊娠期糖尿病预测模型的构建与评价

Setup and evaluation on the prediction model for gestational diabetes mellitus during early pregnancy

目的 通过分析孕10~14周数据筛选高危因素建立模型,早期预测妊娠期糖尿病的发生。方法 回顾性分析2019年1~12月于秦皇岛市第一医院建档并分娩的180例患者的临床资料,按照是否发生妊娠期糖尿病,分为妊娠期糖尿病组(90例)和正常组(90例),进行Logistic回归分析后建立预测模型。结果 模型建立为L=1.395×空腹血糖+4.628×糖化血红蛋白+2.337×霉菌性阴道炎+0.159×体脂百分比,L﹥38.11即可预测发生妊娠期糖尿病,模型的灵敏度82.2%,特异度84.4%。预测模型评价,联合预测因子ROC曲线下面积为0.906,95%置信区间为0.864~0.948,模型Brier值=0.12。结论 此研究模型可有效预测妊娠期糖尿病发生。

Objective To set up the prediction model for gestational diabetes mellitus(GDM) based on data of high-risk factors for GDM within the duration of gestation from the 10th to 14th weeks, and to early predict the development of GDM.Methods Clinical data of 180 registered and delivered patients in First Hospital of Qinhuangdao from January to December 2019 were retrospectively analyzed, the patients were divided into the GDM group(n=90) and normal glucose tolerance group(n=90) based on the occurrence of GDM, and the prediction model was set up after logistic statistical analysis.Results The model was set up as L= 1.395× fasting plasma glucose +4.628× glycosylated hemoglobin +2.337×colpitis mycotica + 0.159× percentage of body fat, L>38.11 supported predicting the GDM developed. The model was found to have the sensitivity of 82.2% and specificity of 84.4%, the areas under curve(AUC) of receiver operating characteristic(ROC) curve in 0.906, 95% confidence interval(0.864, 0.948) and Brier score of the model in 0.12. Conclusion The model supports effectively predicting the development of GDM.