A Systematic Framework for Investigating Algorithmic Bias as a Social Determinant of Health in Low- and Middle-Income Countries: A Research Protocol

Background: The rapid adoption of artificial intelligence (AI) and machine learning (ML) technologies in healthcare systems across low- and middle-income countries (LMICs) presents unprecedented opportunities for improving health outcomes while simultaneously introducing novel risks for perpetuating and amplifying health inequities. Despite growing concerns about algorithmic bias in healthcare delivery, systematic methodological approaches for investigating these phenomena in LMIC contexts remain underdeveloped. Existing research frameworks, predominantly designed for high-income country settings, inadequately address the unique socioeconomic, cultural, and infrastructural challenges that characterize LMIC healthcare systems. Objective: This research protocol presents a comprehensive, multi-phase methodological framework for systematically investigating algorithmic bias as a mechanism through which social determinants of health operate in LMIC healthcare contexts. The protocol aims to establish standardized approaches for cross-country comparative research while maintaining sensitivity to local contexts and resource constraints. Methods: We propose a sequential explanatory mixed-methods research protocol implemented across three phases over 36 months. Phase 1 involves systematic evidence mapping through scoping reviews and policy landscape analysis across 15 LMICs. Phase 2 comprises primary data collection through multi-stakeholder interviews (n=125) and in-depth healthcare system case studies in five countries (Nigeria, India, Kenya, Brazil, Bangladesh). Phase 3 focuses on framework development, validation, and refinement through pilot implementation in Ghana and Vietnam. Expected Outcomes: This protocol will yield a validated research framework for investigating algorithmic bias in LMIC healthcare systems, standardized measurement tools and indicators, evidence-based policy recommendations, and capacity building guidelines. The methodology is designed to be culturally appropriate, ethically sound, and implementable within resource-constrained settings. Significance: By providing the first comprehensive research protocol for investigating algorithmic bias as a social determinant of health in LMICs, this work establishes the methodological foundation for future empirical studies, evidence-based policy development, and international collaborative research efforts aimed at ensuring health equity in the digital health era.