Questions
Clinical Meaning Questions
Last reviewed: November 2026
Eight common questions about clinical meaning, clinical documentation, terminology, and AI.
Is semantic interoperability the same as clinical meaning?
No. Semantic interoperability concerns whether different systems can exchange information and interpret it consistently using shared meanings and representations.Clinical meaning asks a related but different question: whether the information, including the clinical intent and context established during the encounter, remains accurately represented when it is recorded, transformed, and used by other systems.
For example, a record may be transferred successfully and contain correctly mapped codes, yet lose the distinction between a positive and a negative finding. The exchange may be technically successful, while an important part of the clinical meaning is no longer represented.
Semantic interoperability addresses consistency of meaning across systems. Computable clinical meaning examines what those systems can legitimately establish and act upon from the resulting representation.
Does SNOMED CT capture negation?
Yes, but not through a single, universal mechanism.SNOMED CT includes precoordinated concepts that represent negative findings. Its Situation with Explicit Context model also supports the representation of clinical findings with contextual information, including whether a finding is known to be present or absent.
However, negation is not simply a property that can be added to any clinical concept. SNOMED CT guidance also recommends that negation be handled within the electronic health record’s information model, rather than relying solely on negated terminology concepts.
This distinction matters because the clinical meaning of a finding depends not only on the concept itself, but also on how its context is represented and preserved.
A common problem occurs when data extraction systems identify the main clinical concept but fail to preserve the information that establishes whether it was present, absent, suspected, or uncertain.
The important question for an institution is therefore not simply whether SNOMED CT supports the representation of negative findings. It is whether the implementation preserves the original finding context when information is extracted, transformed, and transferred to another system.
Is an empty field the same as a negative finding?
No. These represent different information states.“No chest pain” is a negative clinical finding. It indicates that chest pain was assessed and the patient denied experiencing it.
An empty field means that no information was recorded in that field. It does not establish whether the symptom was assessed, denied, or simply not documented.
These situations may look identical in a poorly designed or incomplete electronic record, but they are not clinically equivalent.
When another system processes the record, missing information may be interpreted as an absent finding if the data model or processing logic does not preserve the distinction.
This can lead to an incorrect clinical interpretation. A system may treat something that was never established as though it had been explicitly ruled out.
The problem is not necessarily a clinical decision. It may arise from how the information was documented, structured, or transformed.
Why is a coded diagnosis not the same as a clinical statement?
A diagnosis code alone does not tell the whole clinical story.
To understand a clinical statement, we also need to know whether the condition is present or absent, suspected or confirmed, current or historical, and whether it belongs to the patient or a family member.
Consider three statements:
• The patient does not have a myocardial infarction. • Myocardial infarction is suspected. • The patient had a myocardial infarction ten years ago.
All three relate to the same clinical concept, but they describe three very different clinical situations.
A code also has consequences beyond describing a diagnosis. It can place a patient on a clinical register, trigger recalls and alerts, affect insurance decisions, and, in some cases, reach licensing authorities, such as those that issue driving licences.
These actions can happen without anyone reading the original clinical note.
This is why a coded concept must retain its clinical context. Otherwise, systems may act on information that does not accurately represent the patient's situation.
Why do ambient AI scribes omit information?
Because recording what was said is not the same as identifying what matters clinically.
For example, a patient may briefly mention that she stopped taking a medication. The clinician notices this information and intends to discuss it later. The AI system may consider it less important than other parts of the conversation and leave it out of the note.
The resulting note may be accurate in everything it includes and may look complete. However, an important clinical detail is missing.
This makes omissions particularly difficult to detect. A fabricated statement can be read and questioned. A missing statement cannot be seen in the note.
Omission is therefore one of the most important failure modes to examine when evaluating AI-generated clinical documentation.
What does provenance mean in clinical documentation?
Provenance describes the origin and history of information in a clinical record.It helps establish who or what generated a statement, whether it was subsequently reviewed or modified, and how it became part of the final record.
For example, was a statement entered directly by the clinician, generated by an AI system, or drafted by AI and subsequently edited by the clinician?
This distinction matters when reviewing the quality and safety of clinical documentation.
If information is missing from a note, provenance may help establish which systems and people contributed to the final document. However, knowing the source does not necessarily prove why the information was omitted.
Without adequate provenance, it becomes more difficult to reconstruct the documentation process, investigate errors, and identify recurring system-level problems.
A record should therefore preserve not only what was documented, but also sufficient information about how its content was produced and modified.
Who is accountable when a system drafts a clinical note?
Responsibility for clinical documentation depends on the applicable professional, institutional, and regulatory framework.In many settings, the clinician who reviews and signs a note is expected to verify its accuracy before it becomes part of the clinical record. However, the legal and professional implications of that attestation vary across jurisdictions and institutional policies.
AI changes the documentation process because the person approving the note may not have composed its original content.
A clinical note may now pass through several stages:
- Generated by an AI system.
- Reviewed by a clinician.
- Edited or amended by one or more people.* Approved and signed by an authorised professional.
These stages are related, but they are not interchangeable.
A signature identifies an act of attestation. It does not, by itself, establish who generated each statement, what was changed, or whether the original source of the content can be reconstructed.
This creates an important governance question: can the institution distinguish between the generation, review, modification, and approval of clinical documentation?
Without that distinction, investigating documentation errors and identifying where they occurred becomes more difficult.
→ Chapter 11 – When the data producer is no longer human
How is clinical AI readiness assessed?
Clinical AI readiness should be assessed by examining what an institution can demonstrate, rather than asking the institution to rate itself.
The assessment used in this practice examines 18 metrics across three dimensions, answered independently by three groups: Governance, Clinical Practice and Technical Infrastructure.
A score above the second level requires supporting evidence. An institution cannot receive a higher score based on its claims alone. It must provide a specific document or record, such as a technical specification, a query result, an audit record, or a contract clause.
Ten of the 18 metrics are assessed across two different groups. Each group receives questions about the same subject, worded from its own perspective.
The differences between their answers are measured and reported alongside the overall score.
For example, an institution may report that it has strong governance controls, while clinicians or technical teams describe a different reality.
A high score combined with a large difference between the groups may indicate that the controls exist in institutional documents but are not consistently reflected in practice.
This difference is important because an assessment based on one person's answers cannot reveal it.
This is reported as the Divergence Index.
Clinical meaning is not generated by AI. It is captured, governed, and made computable.