Large Language Models for Semantic Interoperability in Value-Based Perioperative Care

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Clinical Data Management

Background:

• Electronic Health Record (EHR) systems lack semantic interoperability, isolating valuable clinical protocols within institutions.
• Existing data models improve interoperability but do not fully address cross-institutional protocol portability challenges.
• The need for seamless data exchange is critical for advancing evidence-based medicine and patient care.

Purpose of the Study:

• To explore the potential of large language models (LLMs) in achieving semantic interoperability for clinical protocols.
• To propose a novel approach for cross-institutional protocol portability using LLMs.
• To overcome the limitations of current data models in facilitating data exchange between disparate EHR systems.

Main Methods:

• Utilizing large language models (LLMs) as a translation layer between different EHR systems.
• Leveraging existing data model frameworks as foundational structures, termed "clinical constitutions."
• Developing LLM-based translation protocols to interpret and map data semantically across institutional boundaries.

Main Results:

• LLMs demonstrate the capability to translate clinical data semantically across heterogeneous EHR systems.
• The proposed method facilitates protocol portability without necessitating the adoption of universal data standards.
• Proof-of-concept suggests significant potential for LLMs in enhancing healthcare data interoperability.

Conclusions:

• Large language models offer a viable solution to the long-standing problem of EHR semantic interoperability.
• This approach can unlock the potential of isolated clinical protocols by enabling their portability across institutions.
• Future research should focus on validating LLM-based interoperability in real-world clinical settings.