Construction and validation of a predictive model for risk of hypoglycemia in neonates delivered by pregnant women with normal glucose tolerance based on Logistic regression analysis

To construct and validate a predictive model for risk of hypoglycemia in neonates delivered by pregnant women with normal glucose tolerance.

The clinical materials of 1 865 pregnant women with normal glucose tolerance and their newborns were retrospectively analyzed, and they were divided into modeling population with 1 305 cases and validation population with 560 cases according to a ratio of 7 to 3 by random number method. In the modeling population, they were divided into hypoglycemia group 91 cases and normal group 1 214 cases according to occurrence of neonatal hypoglycemia, and the clinical indexes were compared between the two groups. The indicators with statistical significance were included in the multivariate Logistic regression analysis to screen the risk factors of neonatal hypoglycemia, and a prediction model was established based on the screening results. The performance of the model was evaluated by chi-square goodness-of-fit test and receiver operating characteristic (ROC) curve, and the validation population data was included in the predictionmodel to verify the prediction efficiency of the model.

There were no significant differences in the clinical materials between the modeling population and the validation population (P > 0.05). There were significant differences in the growth of body mass during pregnancy, estimated fetal body mass, gestational weeks of delivery, number of prenatal training, delivery mode and postpartum feeding between the hypoglycemic group and the normal group (P < 0.01). Multivariate Logistic regression analysis showed that increased growth of body mass during pregnancy (OR=2.939; 95%CI, 1.941 to 6.462), lighter estimated fetal body mass (OR=1.590; 95%CI, 1.158 to 2.906), earlier gestational week (OR=1.815; 95%CI, 1.397 to 3.872), less number of prenatal training (OR=1.828; 95%CI, 1.281 to 3.045), cesarean section (OR=3.411; 95%CI, 2.196 to 5.949) and improper postpartum feeding (OR=1.529; 95%CI, 1.182 to 2.748) were the risk factors of neonatal hypoglycemia in pregnant women with normal glucose tolerance (P < 0.05). The prediction model was established according to the risk factors, the chi-square goodness-of-fit test showed no significant difference (χ²=1.619, P=0.983), the area under the curve of ROC curve was 0.890 (95%CI, 0.842 to 0.937), which indicated that the model had no overfitting phenomenon and a strong discrimination ability. The materials of the validation population were included in the prediction model for validation, and it was found that the area under the curve of ROC curve was 0.864 (95%CI, 0.808 to 0.920), the sensitivity was 86.10%, and the specificity was 82.50%.

The prediction model for risk of hypoglycemia in neonates delivered by pregnant women with normal glucose tolerance based on the indexes such as growth of body mass during pregnancy, estimated fetal body mass, gestational weeks of delivery, number of prenatal training, delivery mode and postpartum feeding has a certain application value.