Decoding healthcare resilience for sustainable development goal 3: A machine learning analysis of global health systems

Analyzing healthcare system resilience requires dynamic, structurally nuanced tools often lacking in traditional static assessments. This study addresses this gap by developing and applying an integrated machine learning framework to global panel data (217 countries, 2000-2019) across economic, health, and environmental dimensions. The framework combines unsupervised learning — Principal Component Analysis (PCA) for dimensionality reduction and structural insight, and K-means for risk-level clustering — with supervised classification models, including Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Random Forest (RF), Classification and Regression Trees (CART), and Linear Discriminant Analysis (LDA), for predictive analysis. Key findings reveal a striking global stability in system risk distribution over two decades, suggesting significant inertia despite global initiatives. Crucially, the structural analysis component uncovers critical dimension-specific strengths and weaknesses obscured by aggregate risk scores, highlighting the need for tailored policy approaches. Furthermore, the developed classification models demonstrate exceptionally high predictive performance (SVM achieving 0.991 accuracy, precision, recall, and F1-score on test data), validating the framework's robustness as a monitoring and early-warning tool. This research introduces a novel, validated methodology for dynamic structural assessment of healthcare systems, offering richer analytical insights and providing policymakers with a powerful instrument for targeted interventions aimed at enhancing system resilience worldwide.