Automatic detection and resolution of measurement-unit conflicts in aggregated data

Area of Science:

• Data Science
• Bioinformatics
• Clinical Research Informatics

Background:

• Measurement unit conflicts pose significant challenges in integrative research, particularly in clinical meta-analysis.
• The increasing prevalence of multi-national collaborations, novel measurement instruments, and Linked Open Data exacerbates these unit conflicts.

Purpose of the Study:

• To develop a generic approach for encoding measurement units in a machine-readable format.
• To detect datasets containing mixed measurement units and automatically convert conflicting units to a standardized unit for study-specific needs.

Main Methods:

• Leveraged existing ontologies and standards for scientific data representation and measurement unit definition.
• Developed a Semantic Web Service-based approach for flexible measurement-unit harmonization.
• Implemented automated detection and conversion of unit conflicts.

Main Results:

• Successfully harmonized measurement units in a cardiovascular patient cohort dataset.
• Automatically classified patients into clinical phenotypes despite variations in units (e.g., mmHg vs. cmHg for blood pressure).
• Demonstrated the efficacy of the semantic approach in resolving unit integration problems.

Conclusions:

• Semantic standards and frameworks can effectively automate the detection and resolution of measurement unit integration problems.
• The proposed approach enhances data interoperability and facilitates accurate analysis in complex research domains.