Agenda Setting for Health Equity Assessment Through the Lenses of Social Determinants of Health Using Machine Learning Approach: A Framework and Preliminary Pilot Study Publisher



Ramezani M^{1, 2, 4} ; Mobinizadeh M³ ; Bakhtiari A^{1, 2} ; Rabiee HR⁴ ; Mostafavi H² ; Olyaeemanesh A³ ; Fazaeli AA^{1, 2} ; Atashi A⁵ ; Sazgarnejad S^{6, 7} ; Mohamadi E² ; Takian A^{1, 2, 8} Show Affiliations
Source: BioData Mining Published:2025

Abstract

Introduction: The integration of Artificial Intelligence (AI) and Machine Learning (ML) is transforming public health by enhancing the assessment and mitigation of health inequities. As the use of AI tools, especially ML techniques, rises, they play a pivotal role in informing policies that promote a more equitable society. This study aims to develop a framework utilizing ML to analyze health system data and set agendas for health equity interventions, focusing on social determinants of health (SDH). Method: This study utilized the CRISP-ML(Q) model to introduce a platform for health equity assessment, facilitating its design and implementation in health systems. Initially, a conceptual model was developed through a comprehensive literature review and document analysis. A pilot implementation was conducted to test the feasibility and effectiveness of using ML algorithms in assessing health equity. Life expectancy was chosen as the health outcome for this pilot; data from 2000 to 2020 with 140 features was cleaned, transformed, and prepared for modeling. Multiple ML models were developed and evaluated using SPSS Modeler software version 18.0. Results: ML algorithms effectively identified key SDH influencing life expectancy. Among algorithms, the Linear Discriminant algorithm as classification model was selected as the best model due to its high accuracy in both testing and training phases, its strong performance in identifying key features, and its good generalizability to new data. Additionally, CHAID in numeric models was the best for predicting the actual value of life expectancy based on various features. These models highlighted the importance of features like current health expenditure, domestic general government health expenditure, and GDP in predicting life expectancy. Conclusion: The findings underscore the significance of employing innovative methods like CRISP-ML(Q) and ML algorithms to enhance health equity. Integrating this platform into health systems can help countries better prioritize and address health inequities. The pilot implementation demonstrated these methods’ practical applicability and effectiveness, aiding policymakers in making informed decisions to improve health equity. © The Author(s) 2025.