Modeling the Clinical Reasoning Workflow: A Dynamic, Time-Aware CDSS for the Emergency Department

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• Emergency Departments (ED) face challenges in rapidly diagnosing critical conditions.
• Timely and accurate diagnosis is crucial for effective patient management in emergency settings.
• Existing systems may not adequately integrate dynamic patient data with clinical guidelines.

Purpose of the Study:

• To develop and evaluate a dynamic Clinical Decision Support System (CDSS) for emergency departments.
• To assess the impact of a Time-Aware Retrieval-Augmented Generation (RAG) framework on diagnostic accuracy.
• To determine the system's ability to integrate evolving FHIR data with clinical guidelines.

Main Methods:

• A retrospective pilot study involving 70 cases was conducted.
• The study included critical conditions like Aortic Dissection, Myocardial Infarction (MI), Stroke, and Trauma.
• A dynamic CDSS utilizing a Time-Aware RAG framework was implemented to process patient data.

Main Results:

• The Top-1 diagnostic accuracy of the CDSS significantly improved from 69.5% at triage to 86.7% at discharge.
• The system demonstrated effectiveness in integrating dynamic FHIR data with established clinical guidelines.
• A notable reduction in diagnostic uncertainty was observed.

Conclusions:

• The dynamic CDSS powered by a Time-Aware RAG framework shows significant potential in enhancing diagnostic accuracy in emergency departments.
• Integrating real-time patient data with clinical guidelines is a viable strategy to improve emergency care.
• Further validation in larger studies is warranted to confirm these promising results.