Developing a scalable FHIR-based clinical data normalization pipeline for standardizing and integrating unstructured...

Area of Science:

• Health Informatics
• Clinical Data Standards
• Natural Language Processing in Healthcare

Background:

• Electronic Health Records (EHRs) contain vast amounts of unstructured clinical data.
• Standardizing this data is crucial for interoperability, research, and clinical decision support.
• Existing methods often struggle to efficiently process and normalize diverse EHR information.

Purpose of the Study:

• To design, develop, and evaluate a scalable clinical data normalization pipeline.
• To standardize unstructured EHR data using the HL7 Fast Healthcare Interoperability Resources (FHIR) specification.
• To enable portable EHR-driven phenotyping and large-scale data analytics.

Main Methods:

• Established the NLP2FHIR pipeline, incorporating a natural language processing (NLP) engine with an FHIR-based type system.
• Integrated a module for structured data incorporation and a module for content normalization.
• Evaluated FHIR modeling capability on core clinical resources (Condition, Procedure, MedicationStatement, FamilyMemberHistory) using Mayo Clinic's unstructured EHR data.

Main Results:

• Developed 30 mapping rules, 62 normalization rules, and 11 NLP-specific FHIR extensions for the NLP2FHIR pipeline.
• Identified key elements for integrating structured data into clinical resources.
• Achieved high F-scores (0.69-0.99) for unstructured data modeling across various FHIR elements.

Conclusions:

• The NLP2FHIR pipeline effectively models unstructured EHR data and integrates structured elements.
• This work provides standards-based tools for clinical data normalization, essential for phenotyping and analytics.
• Offers insights for future FHIR specification development in handling unstructured clinical data.