This research systematically evaluates demographic biases in medical retrieval-augmented generation (RAG) systems and proposes mitigation strategies for more equitable healthcare AI.

• Examines bias across the full RAG pipeline including retrieval, reranking, and generation components
• Identifies how systems may propagate or amplify biases related to race, gender, and socioeconomic factors
• Presents novel techniques to reduce demographic disparities while maintaining clinical accuracy
• Establishes evaluation frameworks for measuring fairness in medical QA systems

This work is crucial for healthcare organizations implementing AI, as it addresses ethical concerns around algorithmic fairness and helps ensure medical AI systems provide equitable support across diverse patient populations.

Bias Evaluation and Mitigation in Retrieval-Augmented Medical Question-Answering Systems