Chantal Schiltmeijer is a Strategic Account & Product Manager at Open Terminology Server, working at the intersection of clinical terminology, semantic interoperability and healthcare information architecture. With a background in nursing, she brings both clinical and technical perspectives to how healthcare information is understood, exchanged and reused. She is also a published poet and author, exploring the connections between language, context and meaning through both her professional and creative writing.
Across healthcare IT, I keep noticing something. We use the word ‘semantics’ frequently. Yet depending on who is speaking, we may be referring to a terminology system, a clinical code, an information model, a mapping, or the meaning of the information itself
These things are related. But they are not interchangeable.
At its simplest, semantics concerns meaning: what words, symbols and representations signify, and how concepts relate to one another. In healthcare IT, semantic interoperability is the ambition that information can move between systems without losing the meaning needed to interpret and use it correctly.
We agree on the definition. Do we agree on what achieving it requires?
A dictionary is not a conversation
If I send you a message saying “I’m fine,” you can look up every word. You know the grammar. You could translate it perfectly.
But have you understood me?
Perhaps I am genuinely fine. Perhaps I am disappointed. Perhaps I am ending the conversation without explaining what is wrong. The words remain the same; the interpretation depends on the circumstances around them.
Linguists would distinguish semantics from pragmatics, which concerns meaning in use. That distinction shows why knowing a definition is not necessarily understanding a message.
A dictionary matters. Without shared vocabulary, understanding becomes difficult. But shared vocabulary does not automatically give us shared understanding.
The code survived. Did the meaning?
Healthcare information is more structured than conversation, of course. And clinical terminologies are far more sophisticated than dictionaries. SNOMED CT represents clinical concepts and relationships. LOINC identifies types of observations and measurements. Information models provide structures in which clinical context can be expressed.
We need all of these.
That is precisely why I hesitate when semantics is spoken about as though it were solved by choosing a code system.
A code can identify a concept. What it cannot always tell us on its own is what was being asserted about that concept.
Take pneumonia.
Was it suspected at admission, confirmed after investigation, or recorded as a condition from a previous episode of care? Those distinctions can be represented with appropriate terminology and clinical information structures. But if information is extracted, mapped or reused without the relevant context, those distinctions may disappear.
The code might survive the journey. The clinical meaning might not.
This is not an argument against standards. Quite the opposite.
FHIR offers structures, profiles and terminology bindings that help define how information is represented and constrained. openEHR uses archetypes and templates to model clinical information. Terminology systems provide shared, computable concepts. Each addresses a necessary part of the problem.
Yet a structurally valid resource, or a code that belongs to the expected ValueSet, does not by itself prove that a transformation preserved the meaning of its source. Validation can succeed while the interpretation introduced along the way remains questionable.
The question is not whether standards support semantics. They do. It is whether we pay enough attention to what happens as information moves between implementations.
From language to understanding
In the Netherlands we speak of Eenheid van Taal — unity of language. Its ambition already includes shared meaning, not simply identical words. Still, I wonder whether it is worth making the intended outcome more explicit:
Eenheid van Begrip. Unity of understanding.
Not to replace the first idea, but to make its purpose impossible to miss.
Healthcare information rarely stays where it was created. A clinical observation enters a record, is shared with another organisation, mapped into another representation, aggregated for research or reused for decision support. At every step, a different system may ask that information to serve a different purpose.
Who ensures that the original clinical assertion and its context remain interpretable? Can we tell which terminology version was used, which mapping applied, and what was understood at the time?
If we revisit that information ten years from now, can we still reconstruct what it meant then, rather than what today’s systems assume it should mean?
Those questions become more pressing when data crosses national borders. Different languages are only the visible part of the challenge. Clinical practices, documentation conventions and implementation choices can also shape interpretation.
A question worth agreeing on
We may have agreed on the definition of semantic interoperability, without fully agreeing on what it takes to achieve and preserve it.
We have capable terminology systems, information models and exchange standards. What deserves equal attention is how we use them together, and what must remain intact when information moves.
If we truly agreed on what semantics requires, would we look differently at the infrastructure we build for healthcare interoperability — particularly when data begins to cross borders?
And would we stop asking only whether information arrived, and start asking whether its meaning arrived with it?
Leave a Reply