From reviews to real-time: dynamic evidence in dentistry

Area of Science:

• Biomedical Informatics
• Artificial Intelligence in Healthcare
• Evidence-Based Dentistry

Background:

• Biomedical literature growth outpaces traditional systematic review methods.
• Systematic reviews are resource-intensive and become outdated quickly.
• A gap exists between current research and clinical practice in dentistry.

Purpose of the Study:

• To introduce Retrieval-Augmented Generation (RAG) as a methodology for dynamic evidence reviews.
• To address the need for timely and accurate synthesis of the latest research for dental professionals.
• To mitigate factual errors and hallucinations common in standalone Large Language Models (LLMs).

Main Methods:

• Retrieval-Augmented Generation (RAG) combines Large Language Models (LLMs) with a continuously updated knowledge base.
• RAG grounds AI-generated answers in verifiable, real-time retrieved sources.
• Three integration pathways for RAG are described: pre-retrieved pools, living review portals, and machine-actionable publications.

Main Results:

• RAG enables dynamic, on-demand synthesis of current evidence.
• Clinicians can receive concise, fully cited answers to complex, natural-language questions.
• This facilitates rapid access to information on material selection, healing outcomes, and procedural comparisons.

Conclusions:

• RAG presents a scalable and transparent alternative to static systematic reviews.
• The approach can significantly shorten the timeline from research discovery to clinical practice.
• RAG supports augmented intelligence, integrating AI with human critical appraisal for evidence-based dentistry.