Development and validation of prediction models for gestational diabetes treatment modality using supervised machine...

Area of Science:

• Reproductive Medicine
• Endocrinology
• Medical Informatics

Background:

• Gestational diabetes mellitus (GDM) requires timely treatment to manage glycemic control effectively.
• Clinicians face challenges in selecting optimal GDM treatment modalities, including medical nutrition therapy (MNT) and pharmacologic agents.
• Predicting the necessary GDM treatment modality is crucial for timely and effective patient management.

Purpose of the Study:

• To investigate the predictability of GDM treatment modality using clinical data available at various pregnancy stages.
• To compare the efficacy of different machine learning models in predicting the need for pharmacologic GDM treatment.
• To develop a clinically applicable model for risk-stratifying GDM patients for treatment selection.

Main Methods:

• A population-based cohort of 30,474 pregnancies with GDM was analyzed.
• Clinical data from electronic health records were extracted at four timepoints: preconception to LMP, LMP to diagnosis, at diagnosis, and one week post-diagnosis.
• Ensemble machine learning (super learner) and LASSO regression were employed to predict the need for pharmacologic treatment beyond MNT.

Main Results:

• The super learner model using all predictor levels demonstrated high predictability (C-statistic: 0.934 in discovery, 0.815 in validation).
• A simpler logistic regression model, using timing of diagnosis, fasting glucose, and early glycemic control, showed comparable predictability (C-statistic: 0.825 in discovery, 0.798 in validation).
• Predictability was higher at GDM diagnosis and significantly higher one week post-diagnosis.

Conclusions:

• Clinical data effectively predict GDM treatment modality, particularly one week after diagnosis.
• Population-based, clinically oriented models can support algorithm-based risk stratification for GDM treatment.
• These models have the potential to inform timely treatment decisions and enhance GDM management.