Overnight maternal heart rate variability for early prediction of gestational diabetes mellitus: a retrospective cohort study

Abstract A reliable early risk prediction of gestational diabetes mellitus (GDM) allows for early lifestyle modifications during pregnancy to reduce the risk of developing GDM. In this retrospective study, we developed a logistic regression machine learning model with heart rate variability (HRV) characteristics during overnight sleep in early pregnancy as predictors for GDM prediction. The study used the nuMoM2b dataset from 2748 nulliparous women in the USA who underwent a standardized home sleep test between 6 and 15 weeks’ gestation with subsequent GDM assessment at 24–28 weeks. A total of 52 overnight HRV features were analyzed alongside the baseline risk factors recommended by the National Institutes of Health (NIH). The model combining baseline and HRV features achieved an area under the receiver operating characteristic curve (AUC) of 0.73, outperforming the model using only baseline features (AUC = 0.69) and that using only HRV features (AUC = 0.65). These machine learning models all performed better than the early GDM risk assessment based on the NIH guidelines (AUC = 0.63). The findings suggest that overnight maternal HRV characteristics can be used as early predictors of GDM.