Identifying Key Factors in Pregnancy Complications Among Hypothyroid Women: An Ensemble Machine Learning Approach
Thyroid dysfunction is a prevalent endocrine disorder in pregnancy, second only to gestational diabetes, and is associated with adverse outcomes such as gestational hypertension, preeclampsia, preterm birth, and fetal developmental deficits. While studies have explored these relationships, gaps persist due to diagnostic variability and limited sample sizes. This study leverages advanced machine learning techniques to identify key factors associated with four adverse pregnancy outcomes in hypothyroid women: premature rupture of membranes, fetal distress, macrosomia, and meconium stained amniotic fluid. Using ensemble-based algorithms—Histogram Gradient Boosting, LightGBM, and XGBoost—alongside SHAP analysis, we developed interpretable predictive models from anonymized hospital data spanning nine years. Missing data were addressed through Multiple Imputation by Chained Equations (MICE), which also generated multiple datasets to enhance model robustness and performance. Our models achieved weighted-average precision scores above 0.90 with a focus on minimizing false negatives due to its implication in medical context. Particularly, we achieved strong predictive performance for premature rupture of membranes and meconium stained amniotic fluid. These findings highlight the transformative potential of AI in prenatal care, offering novel insights into the factors driving adverse pregnancy outcomes and demonstrating the ability to predict and mitigate these conditions effectively. By advancing the integration of AI into clinical medicine, this study paves the way for improved maternal and fetal health outcomes.
Keywords: Pregnancy, Hypothyroidism, Fetal Distress, Macrosomia, Meconium Stained Amniotic Fluid, Premature Rupture of Membrane, Ensemble Machine Learning
Karl Luigi Loza Vidaurre1,2, Yicheng Zhou3, and Jinjin Li1,2,4
1National Key Laboratory of Advanced Micro and Nano Manufacture Technology, Shanghai Jiao Tong University, Shanghai, 200240, China
2Department of Micro/Nano Electronics, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China
3Department of Biobank, Shanghai First Maternity and Infant Hospital, Tongji University, Shanghai, 201240, China.
4State Key Laboratory of Biocatalysis and Enzyme Engineering, School of Life Sciences, Hubei University, Wuhan, China