Designing a Healthcare Co-Pilot with Generative AI

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Healthcare professionals require efficient access to clinical information.
• Existing systems may not cater to specific linguistic and contextual needs.

Purpose of the Study:

• To design, test, and evaluate an AI co-pilot for Spanish-speaking healthcare professionals.
• To improve the retrieval of textual information from clinical notes and Electronic Health Records (EHRs).

Main Methods:

• Utilized a Retrieval-Augmented Generation (RAG) architecture.
• Integrated state-of-the-art Large Language Models (LLMs).
• Developed and evaluated a co-pilot for Spanish-speaking clinical contexts.

Main Results:

• Achieved an F1 score of 0.82 on a dataset of 3,500 yes/no questions.
• Demonstrated effective information retrieval for clinical data.
• Validated the co-pilot's performance in the target domain.

Conclusions:

• The developed AI co-pilot is effective for Spanish-speaking healthcare professionals.
• The RAG-based system enhances efficiency in accessing EHR data.
• The methodology provides a framework for domain-specific AI tool development.