A nomogram to predict gestational diabetes mellitus: a multi-center retrospective study.

Abstract

While gestational diabetes mellitus (GDM) poses great threat to the health of mothers and children, there is no standard early prediction model for this disease yet. This study developed and evaluated a nomogram for predicting GDM in early pregnancy. Overall, 1824 pregnant women were randomly divided into the training and internal validation sets in the ratio of 7:3, with additional 1604 pregnant women for external validation. Multivariate logistic regression analysis was used to develop a prediction model for GDM, and a nomogram was utilized for model visualization. Risk factors in the prediction model involved age, pre-pregnancy body mass index, reproductive history, family history of diabetes, creatinine level, triglyceride level, low-density lipoprotein level, neutrophil count, and monocyte count. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision clinical analysis (DCA). The area under ROC curve (AUC) value of the model was 0.804 for the training set, and similar AUC values were obtained for the internal (0.800) and external (0.829) validation sets, verifying the stability of the model. The calibration curves showed that the probabilities of GDM predicted by the nomogram highly correlated with the observed frequency values. The DCA curves indicated that the prediction model is clinically useful, thus potentially aiding early pregnancy management in women.