Pregnancy is an extraordinary journey marked by many bodily changes. One notable change is the rise in blood sugar levels, leading to a condition called gestational diabetes mellitus (GDM). It happens when the body struggles to produce or effectively use insulin during pregnancy. Several health risks of GDM highlight the critical need for accurate prediction and timely intervention. To address this, the study presents a predictive framework validated on a small real-world cohort dataset from a Brazilian public health setting. The core of this research is a composite predictive model that integrates a diverse ensemble of machine learning and deep learning algorithms. In order to enrich the training material, a function was created to generate new instances based on initial dataset records. The framework’s ability to combine the strengths of various models and leverage a meta-classifier for final predictions was rigorously tested across multiple datasets. The results demonstrate exceptional performance by achieving high AUC scores of 88.91%, 95.55%, and 98.71% on original imbalanced, balanced, and augmented datasets, respectively. Additionally, the model shows strong performance across other metrics, including accuracy, precision, recall, and F1 score. These findings validate the generalizability and robustness of the predictive framework. Furthermore, the paper outlines a practical application of this model within a remote-sensing framework in the management information system (MIS) at basic health units (BHUs). It can facilitate proactive GDM management and improve maternal-fetal health outcomes in low-resource settings. The work showcases the predictive framework’s potential to improve GDM management and maternal-fetal health outcomes.
Hassan et al. (Mon,) studied this question.