Semantic Interoperability

Relies on standardized code systems to give data a universal clinical meaning.

• SNOMED CT: Encodes clinical findings, procedures, and body structures with unique concept IDs.
• LOINC: Standardizes lab tests, vital signs, and assessment scales.
• RxNorm: Normalizes drug names, strengths, and dose forms.
• ICD-10-CM: Classifies diagnoses and reasons for encounter. Without these, a 'MI' in one system remains ambiguous—is it myocardial infarction or mitral insufficiency?

Defines explicit relationships between medical concepts, not just flat lists of codes.

• Hierarchical: 'Is-a' relationships (e.g., 'Type 2 Diabetes' is a 'Diabetes Mellitus').
• Attributive: 'Has-laterality', 'has-severity' (e.g., 'severe left knee pain').
• Compositional: Post-coordinated expressions combining multiple concepts. Ontologies allow a system to infer that a patient with 'H40.11' (primary open-angle glaucoma) has a type of 'optic neuropathy' without explicit human mapping.

Provides the structural scaffolding that binds codes to clinical context.

• HL7 FHIR: Defines resources like 'Observation' and 'Condition' that carry both a code and its contextual metadata (subject, performer, effective date).
• HL7 CDA: Uses templates to constrain where and how coded entries appear in clinical documents.
• openEHR: Archetypes that define maximum data sets for specific clinical concepts. The information model answers: 'Who said this? When? About whom?'—preventing a diagnosis from floating free of its clinical context.

The formal link between a field in an information model and the specific value set of codes allowed to populate it.

• Strength: Defines whether the binding is 'required' (must use exact code) or 'extensible' (can use a local code if not in the value set).
• Value Sets: Curated lists of codes from one or more terminologies for a specific use case (e.g., 'Diabetes Mellitus Value Set' for problem list entries).
• Binding ensures a 'blood pressure' field can only contain LOINC codes for systolic or diastolic pressure, not a random lab code.

Ensures that the clinical intent of data survives transport across system boundaries.

• No Lossy Mapping: A SNOMED CT concept mapped to a proprietary local code must be mapped back to the exact same SNOMED CT concept on the receiving end.
• Semantic Equivalence: Systems must recognize that '55822004' (Hyperlipidemia) and 'E78.5' (Hyperlipidemia, unspecified) are clinically synonymous.
• Achieved through robust terminology services that maintain cross-walks and equivalence tables between code systems.

Enables machines to reason over the exchanged data for decision support.

• Allergy Checking: A system can automatically cross-reference a newly prescribed RxNorm drug code against a patient's SNOMED CT allergy codes.
• Cohort Identification: A research query can find all patients with a specific LOINC lab result above a threshold, regardless of which EHR system recorded it.
• Quality Measure Calculation: Automated eCQM engines can process standardized data to calculate performance metrics without manual chart abstraction. This is the ultimate proof of semantic interoperability: the data drives safe, automated action.