External Validation of Multiple Risk Prediction Models for Gestational Diabetes Mellitus

Abstract

Objective: To externally validate gestational diabetes mellitus (GDM) prediction models in a Chinese population and assess their performance in early pregnancy risk stratification. Methods: Six GDM prediction models identified from a systematic review were externally validated using data from 1,385 pregnant women in a tertiary hospital in China. Model performance was evaluated in terms of discrimination (C-statistic), calibration (calibration slope and intercept), and clinical utility (decision curve analysis). Results: Among 1,385 women, 661 were diagnosed with GDM. All models showed decreased discrimination compared with the original studies, with the area under the curve (AUC) ranging from 0.693 to 0.751. All models underestimated risk in high-risk individuals, and most models demonstrated relatively stable net benefit, indicating potential suitability for early pregnancy risk stratification. Conclusion: Existing GDM prediction models exhibit variable performance in Chinese populations. Further recalibration and impact assessment are recommended.