Machine Learning-Driven Comparative Effectiveness Research: Utilizing Random Forest and XGBoost to Predict Patient Response and Adverse Drug Reactions to First-Line Antidiabetic Therapies

• Abilly Elly

  Texas University

  Author

Diabetes mellitus remains a significant global health challenge, with first-line antidiabetic therapies exhibiting variable patient responses and adverse drug reaction (ADR) profiles. Traditional comparative effectiveness research methods often fail to capture the complex, non-linear relationships between patient characteristics and treatment outcomes. This study addresses this gap by developing and validating machine learning models—specifically Random Forest and XGBoost—to predict patient response and ADRs to first-line antidiabetic therapies. Utilizing a comprehensive dataset of patient clinical and demographic features, we employed ensemble learning techniques to enhance predictive accuracy and model interpretability. The XGBoost model achieved superior performance with an accuracy of 97.24% and an AUC of 0.9117, outperforming traditional classifiers. SHAP analysis identified glucose levels, age, and BMI as the most influential predictors of treatment response, while pharmacovigilance data revealed significant associations with psychiatric ADRs. This research demonstrates that ensemble machine learning approaches can provide robust, interpretable frameworks for personalized diabetes management, offering clinicians actionable insights for treatment selection and ADR prevention. The findings have significant implications for clinical decision support systems and precision medicine initiatives in endocrinology.

• PDF

Copyright (c) 2026 Abilly Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.