An early prediction model for gestational diabetes mellitus created using machine learning algorithms.

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Public Health

Background:

• Gestational diabetes mellitus (GDM) poses risks to maternal and fetal health.
• Early identification of high-risk pregnancies is crucial for timely intervention.
• Existing diagnostic methods may be invasive or delayed.

Purpose of the Study:

• To identify key risk factors for GDM in early pregnancy.
• To develop and validate a machine learning (ML) model for GDM prediction.
• To enhance early diagnosis and intervention strategies for GDM.

Main Methods:

• Retrospective analysis of demographic and clinical data from 942 pregnant women.
• Application of a stacking ensemble ML model for GDM prediction.
• Evaluation of model performance using ROC analysis (AUC=0.89) and validation on an independent dataset.

Main Results:

• Significant GDM predictors include age, pre-pregnancy BMI, and medical history (GDM, diabetes, macrosomia, hypertension).
• The ML model achieved an AUC of 0.89, demonstrating robust predictive capability.
• The model effectively identified high-risk individuals, with strong exclusion performance at a 28.53% threshold.

Conclusions:

• An ML-based model accurately predicts GDM risk using non-invasive early pregnancy data.
• This facilitates early identification of high-risk populations before invasive testing.
• The model supports optimized clinical intervention and resource allocation for GDM management.