SIMPL-D: Scalable and Inclusive Machine Learning Prediction for Diabetes Using Sparsely Overlapping Datasets

Machine learning has revolutionized disease prediction in healthcare by uncovering intricate patterns beyond the capabilities of traditional methods. This advancement has marked the beginning of a new era in personalized and predictive medicine, transforming the anticipation and management of health outcomes. However, the integration of these models into clinical practice faces significant challenges, particularly due to their reliance on datasets from singular institutions. Such limitations hinder their applicability in diverse healthcare environments with disparities in resources, data collection protocols, and diagnostic practices. In this study, SIMPL-D (Scalable and Inclusive Machine Learning Prediction for Diabetes Using Sparsely Overlapping Datasets), we tackle these challenges within the context of diabetes prediction. By integrating two highly disparate datasets-the CDC Diabetes Health Indicators dataset and the Pima Indians Diabetes Dataset-we develop predictive models that generalize effectively across varied healthcare settings. These models eliminate the need for standardized operational procedures, enabling scalable and robust disease prediction. Leveraging machine learning techniques such as Random Forest, CatBoost, and Multi-Layer Perceptrons, our approach achieves high accuracy and robustness while prioritizing inclusivity and practicality. By demonstrating the feasibility of integrating sparsely overlapping datasets, our methodology advances equitable healthcare outcomes and establishes a foundation for broader applications of machine learning in diverse clinical environments.