Prediction model for gestational diabetes mellitus using the XG Boost machine learning algorithm.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Obstetrics and Gynecology

Background:

• Gestational diabetes mellitus (GDM) poses risks to maternal and fetal health.
• Accurate prediction of GDM is crucial for timely intervention and improved outcomes.
• Traditional logistic regression (LR) models have limitations in predicting complex diseases like GDM.

Purpose of the Study:

• To develop and evaluate an extreme gradient boosting (XG Boost) machine learning (ML) model for GDM prediction.
• To compare the predictive performance of the XG Boost ML model against a traditional logistic regression (LR) model.
• To assess the clinical utility of the developed XG Boost ML model for GDM risk stratification.

Main Methods:

• A case-control study involving pregnant women, divided into training (n=735) and testing (n=190) sets.
• Application of the XG Boost ML model to identify significant predictors from 33 variables.
• Evaluation of model performance using Area Under the Receiver Operating Characteristic Curve (AUC), Hosmer-Lemeshow (HL) test, calibration plots, and Decision Curve Analysis (DCA).

Main Results:

• The XG Boost ML model, utilizing 20 predictors, achieved an AUC of 0.946 and a predictive accuracy of 0.875.
• The traditional LR model, with four predictors, yielded an AUC of 0.752 and a predictive accuracy of 0.786.
• Both models demonstrated good calibration, but DCA indicated a net clinical benefit for the XG Boost ML model in guiding treatment decisions.

Conclusions:

• The XG Boost ML model exhibits significantly better predictive discrimination for GDM compared to the traditional LR model.
• The established XG Boost ML model demonstrates high accuracy and clinical utility for predicting GDM risk in pregnant women.
• This study highlights the potential of advanced ML techniques for enhancing GDM prediction and management.