Evidence-based AI clinical decision support system for acute burn care and complex reconstruction.

Area of Science:

• Artificial Intelligence in Medicine
• Surgical Decision Support Systems
• Burn Care Research

Background:

• Burn surgery demands rapid, evidence-based decisions for optimal patient outcomes.
• Existing clinical decision support tools lack burn-specific, transparent, and citation-traceable guidance.
• Time-critical surgical care necessitates reliable AI assistance.

Purpose of the Study:

• To develop and evaluate BurnRAG, the first AI-driven clinical decision support system tailored for burn surgery.
• To provide transparent, citation-traceable evidence synthesis for the entire burn care continuum.
• To enhance decision-making in acute resuscitation, operative management, and reconstructive procedures.

Main Methods:

• Developed BurnRAG, a retrieval-augmented generation (RAG) system with a hierarchical RAPTOR architecture.
• Utilized a curated corpus of 4683 peer-reviewed burn surgery articles (2000-2025).
• Evaluated performance on 10 expert-designed clinical scenarios using semantic similarity, G Score, and citation verification.

Main Results:

• BurnRAG achieved high evidence-alignment with a mean semantic similarity of 0.94 and G Score of 0.97.
• The system synthesized an average of 6.8 sources per query with 100% citation accuracy.
• Top performance was noted in acute surgical decision-making, including escharotomy and fluid resuscitation.

Conclusions:

• BurnRAG is the inaugural burn-specific AI system offering citation-traceable evidence synthesis.
• It provides transparent guidance for triage, operative planning, and reconstruction in burn care.
• BurnRAG lays the groundwork for integrating specialty-focused AI in surgical practice.