RAG-Based Healthcare Query Assistant Using Graph Database: A Comprehensive Survey

Healthcare research today operates within an en- vironment rich in complex and interconnected data sources, ranging from Electronic Health Records (EHRs) and diagnostic imaging to pharmacological studies and clinical trial outputs. Traditional retrieval systems and general-purpose Large Lan- guage Models (LLMs), though effective in surface- level analysis, often fail to capture the deep semantic relationships that un- derpin this data. Consequently, they generate responses lacking verifiable evidence and contextual accuracy. To address these challenges, this paper surveys the emerging paradigm of Retrieval-Augmented Generation (RAG) enhanced by Knowledge Graphs (KGs), forming a bridge between sym- bolic reasoning and neural generation. The proposed study establishes a taxonomy of healthcare Question-Answering (QA) frameworks—spanning relational databases, vector embedding retrieval, and hybridKG-RAGarchitectures—while emphasizing their relevance in clinical information systems. Furthermore, the paper outlines the necessity of such in- tegration for improving medical decision support, focusing on the dynamic translation of natural language into formal graph queries, scalable knowledge maintenance, and multi-hop infer- encing. By systematically reviewing technological advances and identifying key implementation challenges, this survey provides a structured roadmap for developing reliable, explainable, and ethically aligned AI systems in healthcare.