gms | German Medical Science

Introduction: To ensure interoperability in the healthcare sector and thus the exchange and reuse of distributed data, standards are particularly necessary. Besides the needed alignment to structural standards and harmonized data models, a consistent coding of content is crucial to be able to interpret the structurally represented data. Semantic standards can be used to this end. Although it has become common practice to speak indifferently of code systems like ICD-10, LOINC or SNOMED CT, the underlying variety of the involved vocabulary types must be considered when using them.

State of the art: Large projects such as the National Informatics Initiative (MII) in Germany are increasingly relying on the HL7^® Fast Healthcare Interoperability Resources (FHIR^®) standard for IT support of clinical research. In addition to a core data set with an extensive amount of profiled FHIR resources, the problem of effectively creating and providing referenced value sets and code systems arises to achieve. The terminology module integrated in the FHIR standard provides the interacting resources ValueSet (VS), CodeSystem (CS) and ConceptMap (CM), which make it much easier for software developers to handle terminological resources compared to previous standards.

Concept: The uniform representation of these artefacts partially obscures the variability of the underlying knowledge artefacts. The presentation will explore the main variants of coding systems or source vocabularies. We have identified five broad categories:

• Code lists: Simple enumerations of codes for entities like status, types, modes, or gender, with display terms and short definitions.
• Nomenclatures: Focus on rules for uniform entity designations, ensuring correct codes and/or designations.
• Classifications: Provide monohierarchies of classes for disjunctive partitions of entities.
• Terminologies: Offer (poly-)hierarchies of explicitly represented concepts, enabling synonymy determination.
• Ontologies: Extend terminologies, allowing computational determination of subsumption between formally defined concepts.

Implementation: As part of the Medical Informatics Initiative, working within the Service Unit Terminological Service, these differences are crucial, since we must provide adequate code systems for all use cases defined in the initiative using our terminology server.

FHIR code systems define which codes (symbols and/or expressions) exist and how they are understood. As the implementers are free to choose suitable properties it becomes a challenge to decide and harmonize property sets needed e.g. for classes within ICD-10 versus concepts within LOINC versus names within HGNC (HUGO Gene Nomenclature). Providing these resources using FHIR often requires conversions from other distribution formats like OWL or ClaML; and some can’t be expressed using enumerative CodeSystems at all.

Lessons learned: Our deliberations have shown fundamental differences in the use of vocabularies using HL7 FHIR due to the inherent differences in the underlying definitions. Our differentiation of levels of vocabularies, as shown by LOINC presenting aspects of both nomenclatures and terminologies, might have weaknesses. Regardless, we hope that the differentiation is useful to implementers for consideration when adopting a certain vocabulary in their use cases.

Acknowledgements: This work was funded by the German Federal Ministry of Education and Research (BMBF) as part of the Medical Informatics Initiative Germany, grant 01ZZ2312A.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.