Automated UMLS-based comparison of medical forms

Area of Science:

• Medical Informatics
• Health Data Science
• Computational Medicine

Background:

• Medical forms are highly heterogeneous across Europe, with thousands of data items in numerous systems.
• Interoperable documentation systems and harmonized forms are crucial for clinical care and research data exchange.
• Automated methods for comparing medical forms are currently lacking, hindering harmonization efforts.

Purpose of the Study:

• To develop and evaluate an automated method for comparing medical forms based on semantic annotations.
• To facilitate the harmonization of medical forms for improved data exchange and interoperability.
• To provide an open-source tool for the automated comparison of medical forms.

Main Methods:

• Developed a method to compare medical form items based on their concept codes (coded in Unified Medical Language System - UMLS) and value domains.
• Defined criteria for item identity (identical, matching, similar) based on name, concept code, and value domain.
• Implemented an open-source package (compareODM) in R for automated comparison of forms in ODM format with UMLS annotations.

Main Results:

• Applied the method to 7 real medical forms comprising 285 data items, demonstrating feasibility.
• Comparison results were visualized using grid images and dendrograms, showing clustered similar forms.
• The developed approach proved to be scalable for a large set of real-world medical forms.

Conclusions:

• Automated comparison of semantically annotated medical forms is feasible and effective.
• The developed tool and methodology support the harmonization of medical forms for better data interoperability.
• The approach is scalable and applicable to diverse medical data sources.