Associations between health system indicators and key public health outcomes using machine learning on WHO country level data

This study examines how data analytics and machine learning (ML) techniques can be applied to country-level health indicators to generate descriptive and predictive insights into public health systems. While existing literature highlights the role of large-scale digital data in improving disease surveillance, policy design, and predictive health analytics, many low- and middle-income countries continue to rely on more limited datasets. This raises important questions regarding the extent to which structured, aggregated data can support similar analytical objectives. Using publicly available data from the World Health Organization (WHO) DataDot platform, this study explores relationships between key health system indicators, such as the “Universal Health Coverage (UHC) Service Coverage Index”, healthcare workforce density, and “Domestic general government health expenditure (GGHE-D) as percentage of general government expenditure (GGE)”, and major public health outcomes, including immunization coverage, noncommunicable disease (NCD) mortality, and life expectancy. A correlation analysis and ML models (Random Forest, XGBoost, and LightGBM) were used to detect patterns and evaluate predictive relationships. The findings identify descriptive associations and predictive patterns between selected health-system indicators and public health outcomes. Moreover, feature importance analysis was also conducted to emphasize the relative predictive contribution of interaction terms and multiple health system indicators. A conceptual framework is also proposed to illustrate how data analytics can inform public health system monitoring and decision-making. The framework is presented as a theoretical model to guide future empirical research. Overall, findings of the study are considered associational rather than causal.