Retrieval-augmented generation in medicine: A scoping review of technical implementations, clinical applications, and ethical considerations
9 affiliations
- 1Centre for Biomedical Data Science, Duke-NUS Medical School, Singapore 169857, Singapore; Duke-NUS AI + Medical Sciences Initiative, Duke-NUS Medical School, Singapore 169857, Singapore
- 2School of Clinical Medicine, University of Cambridge, Cambridge CB2 0SP, UK
- 3Graduate School of Frontier Sciences, The University of Tokyo, Tokyo 277-8561, Japan
- 4Graduate School of Engineering, The University of Tokyo, Tokyo 113-8654, Japan
- 5Duke-NUS AI + Medical Sciences Initiative, Duke-NUS Medical School, Singapore 169857, Singapore; Division of Anesthesiology and Perioperative Medicine, Singapore General Hospital, Singapore 169608, Singapore
- 6Duke-NUS AI + Medical Sciences Initiative, Duke-NUS Medical School, Singapore 169857, Singapore; Division of Pharmacy, Singapore General Hospital, Singapore 169608, Singapore
- 7Department of Radiology and Medical Informatics, University of Geneva, 1202 Geneva, Switzerland
- 8Department of Biostatistics and Bioinformatics, Duke School of Medicine, Durham, NC 27710, USA; Duke Clinical Research Institute, Durham, NC 27705, USA
- 9Singapore Eye Research Institute, Singapore National Eye Center, Singapore 168751, Singapore; Byers Eye Institute, Stanford University, Stanford, CA 94303, USA; Artificial Intelligence Office, Singapore Health Services, Singapore 168582, Singapore
Abstract
The rapid growth of medical knowledge and the increasing complexity of clinical practice pose challenges. In this context, large language models (LLMs) demonstrate value; however, inherent limitations remain. Retrieval-augmented generation (RAG) shows potential to enhance their clinical applicability. This study reviews RAG applications in medicine. We find that research primarily relies on publicly available data, with limited use of private data. For retrieval, approaches commonly rely on English-centric embedding models, while LLMs are mostly generic, with limited use of medical-specific LLMs. For evaluation, automated metrics evaluate generation quality and task performance, whereas human evaluation focuses on accuracy, completeness, relevance, and fluency, with insufficient attention to bias and safety. RAG applications are concentrated on question answering, report generation, text summarization, and information extraction. Overall, medical RAG remains at an early stage, requiring advances in clinical validation, cross-linguistic adaptation, and support for low-resource settings to enable trustworthy and responsible global use.
Related
- Retrieval-augmented generation in biomedicine: a survey of technologies, datasets, and clinical applications
- Applications, Technical Characteristics, and Evaluation Approaches of Retrieval-Augmented Generation in Nursing: Scoping Review
- IPF-RAG: temporal-aware retrieval-augmented generation for idiopathic pulmonary fibrosis care.
- Limits of Single-Pass Retrieval-Augmented Generation for AI-Powered Cancer Care Navigation: A Comparison of Retrieval Strategies
- Integrating knowledge graphs and retrieval-augmented generation to identify clinical pharmacology gaps in infectious disease trial protocols
BibTeX
@article{yang2026retrievalaugmented,
title = {{Retrieval-augmented generation in medicine: A scoping review of technical implementations, clinical applications, and ethical considerations}},
author = {Rui Yang and Matthew Yu Heng Wong and Huitao Li and Xin Li and Wentao Zhu and Jingchi Liao and Kunyu Yu and Jonathan Chong Kai Liew and Weihao Xuan and Yingjian Chen and Yuhe Ke and Jasmine Chiat Ling Ong and Douglas Teodoro and Chuan Hong and Daniel Shu Wei Ting and Nan Liu},
journal = {Cell Reports Medicine},
year = {2026},
month = jul,
doi = {10.1016/j.xcrm.2026.102927},
url = {https://doi.org/10.1016/j.xcrm.2026.102927},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Retrieval-augmented generation in medicine: A scoping review of technical implementations, clinical applications, and ethical considerations
AU - Rui Yang
AU - Matthew Yu Heng Wong
AU - Huitao Li
AU - Xin Li
AU - Wentao Zhu
AU - Jingchi Liao
AU - Kunyu Yu
AU - Jonathan Chong Kai Liew
AU - Weihao Xuan
AU - Yingjian Chen
AU - Yuhe Ke
AU - Jasmine Chiat Ling Ong
AU - Douglas Teodoro
AU - Chuan Hong
AU - Daniel Shu Wei Ting
AU - Nan Liu
PY - 2026
DA - 2026/07/01
JO - Cell Reports Medicine
DO - 10.1016/j.xcrm.2026.102927
UR - https://doi.org/10.1016/j.xcrm.2026.102927
AB - The rapid growth of medical knowledge and the increasing complexity of clinical practice pose challenges. In this context, large language models (LLMs) demonstrate value; however, inherent limitations remain. Retrieval-augmented generation (RAG) shows potential to enhance their clinical applicability. This study reviews RAG applications in medicine. We find that research primarily relies on publicly available data, with limited use of private data. For retrieval, approaches commonly rely on English-centric embedding models, while LLMs are mostly generic, with limited use of medical-specific LLMs. For evaluation, automated metrics evaluate generation quality and task performance, whereas human evaluation focuses on accuracy, completeness, relevance, and fluency, with insufficient attention to bias and safety. RAG applications are concentrated on question answering, report generation, text summarization, and information extraction. Overall, medical RAG remains at an early stage, requiring advances in clinical validation, cross-linguistic adaptation, and support for low-resource settings to enable trustworthy and responsible global use.
SN - 2666-3791
ER -