Automating Patient Safety Workflows: The Development and Implementation of LLaMPS, a Secure Large Language Model...

Area of Science:

• Healthcare Informatics
• Artificial Intelligence in Medicine
• Patient Safety

Background:

• Generative Artificial Intelligence (GenAI) adoption in healthcare faces challenges like data privacy and clinical relevance.
• Existing systems struggle to effectively manage and report patient safety events.

Purpose of the Study:

• To introduce LLaMPS (Large Language Model for Patient Safety), a locally deployed GenAI platform.
• To enhance patient safety event management and reporting through advanced AI capabilities.

Main Methods:

• LLaMPS utilizes a Retrieval-Augmented Generation (RAG) approach with secure, institutionally hosted Large Language Models (LLMs).
• A vector database ensures data privacy and regulatory compliance.
• Iterative development involved clinicians and patient safety experts.

Main Results:

• LLaMPS demonstrated high accuracy in incident classification.
• The platform showed improved user satisfaction among healthcare professionals.
• The system integrates automated classification, harm prediction, intelligent search, and a chatbot.

Conclusions:

• Locally controlled AI solutions like LLaMPS can significantly enhance patient safety workflows.
• LLaMPS addresses key concerns regarding data privacy and clinical integration of GenAI in healthcare.
• The platform shows potential for broader adoption in improving healthcare safety.