The development of gestational diabetes mellitus (GDM), which is defined as high blood glucose levels during pregnancy, is a common complication during pregnancy and can lead to adverse outcomes for both mothers and babies.

Credit: DonnaDiavolo/getty

GDM is currently diagnosed by performing an oral glucose challenge test. This test is commonly conducted between 24 and 28 weeks of gestation. However, early detection of GDM could help to prevent or minimize the risk of complications. Now, a new study aims at leveraging nationwide electronic health records (EHRs) in order to predict GDM with high accuracy at the start of a pregnancy.

“EHR data are normally challenging to work with because of numerous confounders,” explains Eran Segal, co-corresponding author of the study. “In GDM, however, EHR data are powerful, because in several countries (including Israel) all women undergo screening for GDM during pregnancy, which results in ‘survey-like’ data that are collected across the entire relevant population.”

For their study the researchers used comprehensive records of 588,622 pregnancies from 368,351 women. They worked with a subset of these data to develop a model that could predict GDM. Several smaller data subsets, differing from the other records in date and/or geographic location, were then used to validate the model’s predictive accuracy. “The model was developed on over 2000 features and performed well,” says Segal.

While the majority of features used for developing the model are available at pregnancy initiation, some of the data are gathered throughout the pregnancy. Therefore, the researchers proceeded to derive a questionnaire-based predictor consisting of 9 questions that could be answered at the beginning of a pregnancy. “We found that a model based only on these questions performed nearly as well as the full model,” reports Segal.

the researchers proceeded to derive a questionnaire-based predictor consisting of 9 questions that could be answered at the beginning of a pregnancy

The team acknowledges that their study has several limitations, including the use or retrospective data. “Going forward, we are planning a prospective clinical trial where we will identify high-risk women and randomize them into either the current standard of care treatment or a lifestyle and dietary treatment change,” concludes Segal.