Mapping of UK Biobank clinical codes: Challenges and possible solutions

Area of Science:

• Biomedical Informatics
• Data Science
• Clinical Data Management

Background:

• The UK Biobank dataset is a valuable resource but contains heterogeneous and inconsistent medical terminology.
• Data cleaning, curation, and standardization are essential to make UK Biobank data usable for research.
• Integrating diverse data sources like hospital records, self-reported data, and primary care data requires significant harmonization efforts.

Purpose of the Study:

• To develop and propose a data integration pipeline for harmonizing diagnostic information within the UK Biobank.
• To evaluate methods for mapping primary care (GP) Read codes to standardized terminologies like ICD and SNOMED CT.
• To identify and report challenges encountered during the mapping and harmonization process.

Main Methods:

• Evaluation of multiple approaches for mapping GP clinical Read codes to International Classification of Diseases (ICD) and Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT).
• Comparison of mapping results, flagging inconsistencies, and assigning quality categories to assess overall mapping quality.
• Development of a curation and data integration pipeline focused on harmonizing diagnostic terms.

Main Results:

• Several approaches for mapping Read codes to ICD and SNOMED CT were evaluated.
• Mapping inconsistencies were identified and categorized by quality.
• Challenges in mapping, including lack of one-to-one ontology mapping and the need for additional ontologies, were reported.

Conclusions:

• A curation and data integration pipeline is proposed for harmonizing diagnoses from UK Biobank GP data.
• Challenges in mapping Read codes to ICD and SNOMED CT were identified, some general and some specific to automated mapping.
• Leveraging existing mappings and employing automated/manual curation can overcome mapping challenges.