An autonomous AI agent for knowledge and data cooperation in ED clinical decision support.

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Informatics
• Emergency Medicine

Background:

• Medical knowledge and clinical practice are interconnected but integrating dynamic data for real-time guidance remains a challenge, particularly in emergency departments (EDs).
• Existing systems struggle to autonomously distill updated medical knowledge from evolving clinical data to inform immediate patient care decisions.
• Bridging the gap between vast medical knowledge bases and the fast-paced, data-rich environment of the ED is crucial for improving healthcare delivery.

Purpose of the Study:

• To develop an autonomous AI agent capable of integrating established medical knowledge with dynamic clinical data for enhanced emergency care.
• To create a hybrid knowledge graph by combining medical knowledge graphs and real-time clinical data.
• To leverage large language models (LLMs) for efficient knowledge extraction and semantic mapping within the AI agent.

Main Methods:

• Developed an autonomous AI agent integrating medical knowledge graphs and dynamic clinical data into a hybrid graph exceeding 800,000 nodes.
• Utilized large language models (LLMs) for automated knowledge extraction and semantic mapping to dynamically select relevant information.
• Implemented specialized tools powered by the hybrid graph for emergency department (ED) recognition, prediction, and decision-making.

Main Results:

• Achieved significant average performance improvements over state-of-the-art baselines: 23.13% in ED triage, 13.05% in drug-drug interaction detection, 1.58% in readmission prediction, and 5.47% in medication recommendation.
• Demonstrated superior performance across all evaluated task categories, including patient recognition, risk prediction, and treatment recommendations.
• The AI agent effectively synergized established medical knowledge with dynamic clinical data to enhance emergency care processes.

Conclusions:

• The developed autonomous AI agent provides an effective framework for integrating medical knowledge graphs and dynamic clinical data in emergency care.
• This approach significantly improves key performance metrics in the ED, including triage accuracy, drug safety, and predictive capabilities.
• The study highlights the potential of AI-driven hybrid knowledge graphs to revolutionize clinical decision-making and optimize patient outcomes in high-acuity settings.