Development and internal validation of a clinical score to predict neonatal hypoglycaemia in women with gestational diabetes.

Abstract

Gestational diabetes (GD) is a risk factor for neonatal hypoglycaemia (NH), but other factors can increase this risk. To create a score to predict NH in women with GD. Retrospective study of women with GD with a live singleton birth between 2012 and 2017 from the Portuguese GD registry. Pregnancies with and without NH were compared. A logistic regression was used to study NH predictors. Variables independently associated with NH were used to score derivation. The model's internal validation was performed by a bootstrapping. The association between the score and NH was assessed by logistic regression. We studied 10216 pregnancies, 410 (4.0%) with NH. The model's AUC was 0.628 (95%CI: 0.599-0.657). Optimism-corrected c-index: 0.626. Points were assigned to variables associated with NH in proportion to the model's lowest regression coefficient: insulin-treatment 1, preeclampsia 3, preterm delivery 2, male sex 1, and small-for-gestational-age 2, or large-for-gestational-age 3. NH prevalence by score category 0-1, 2, 3, 4, and ≥5 was 2.3%, 3.0%, 4.5%, 6.0%, 7.4%, and 11.5%, respectively. Per point, the OR for NH was 1.35 (95% CI: 1.27-1.42). A score of 2, 3, 4, 5 or ≥6 (versus ≤1) had a OR for NH of 1.67 (1.29-2.15), 2.24 (1.65-3.04), 2.83 (2.02-3.98), 3.08 (1.83-5.16), and 6.84 (4.34-10.77), respectively. Per each score point, women with GD had 35% higher risk of NH. Those with ≥6 points had 6.8-fold higher risk of NH compared to a score ≤1. Our score may be useful for identifying women at a higher risk of NH.