TRAI: An AI-Driven Mobile Application to Reduce the Gap Between Triage and Care

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Informatics
• Natural Language Processing

Background:

• Healthcare workflows face challenges in efficiency and patient engagement.
• Artificial intelligence (AI) offers potential solutions for streamlining processes.
• Patient triage and information gathering can be optimized with advanced technology.

Purpose of the Study:

• To present a novel AI-powered chatbot for patient triage and information gathering.
• To leverage large language models (LLMs) for context-aware and adaptive patient conversations.
• To enhance clinical insights and generate structured summaries for physicians.

Main Methods:

• Developed an AI chatbot integrating physician prompts, triage notes, and EHR data.
• Utilized LLMs for dynamic refinement of clinical insights through iterative questioning.
• Ensured compliance with healthcare data standards and FHIR integration for EHR data.

Main Results:

• Achieved high performance (4.0-5.0/5.0) in documentation quality, assessment accuracy, patient education, and management planning across ten scenarios.
• Demonstrated alignment between user ratings and independent hepatologist assessments.
• Reduced redundant history-taking and improved previsit workflows through EHR integration.

Conclusions:

• LLM-driven patient-facing chatbots are viable for improving clinic efficiency and patient engagement.
• The developed system enhances diagnostic accuracy and patient-provider communication.
• The chatbot effectively bridges gaps between triage and care, streamlining healthcare workflows.