Post-Coordination

The fundamental operation of joining two or more primitive concepts from a reference terminology to create a compound clinical expression. For example, combining the SNOMED CT concepts for 'fracture' and 'left femur' to represent 'fracture of left femur' when no single pre-coordinated code exists. This preserves compositional semantics without requiring the terminology to enumerate every possible clinical permutation.

Post-coordinated expressions use defining attributes to refine a base concept's meaning. Common qualifiers include:

• Laterality: left, right, bilateral
• Severity: mild, moderate, severe
• Anatomical site: proximal, distal, lateral aspect
• Temporal context: acute, chronic, recurrent
• Subject of record: family history of, past history of

Each qualifier is linked to the base concept through a formal relationship type defined by the ontology's description logic.

Post-coordination follows a strict formal grammar that governs how concepts can be combined. In SNOMED CT, this is defined by the Compositional Grammar Specification, which specifies:

• Valid attribute-value pairs for each concept hierarchy
• Cardinality constraints on how many qualifiers can be applied
• Domain and range restrictions ensuring only semantically valid combinations

This syntactic rigor prevents nonsensical expressions like 'fracture of the liver' from being constructed.

Pre-coordinated concepts are single codes representing complete clinical ideas (e.g., SNOMED CT 263102004 'Fracture of shaft of left femur'). Post-coordination assembles these on demand. The trade-offs:

• Pre-coordination: faster retrieval, simpler queries, but combinatorial explosion of codes
• Post-coordination: expressive flexibility, smaller terminology footprint, but requires normalization and subsumption reasoning at query time

Most production systems use a hybrid approach, pre-coordinating common patterns and post-coordinating edge cases.

A critical challenge in post-coordination is determining when two structurally different expressions represent the same clinical meaning. Description logic classifiers are used to compute:

• Semantic equivalence: Do expression A and expression B subsume each other?
• Canonical form generation: Reducing an expression to its simplest, standardized representation
• Redundancy detection: Identifying and removing logically implied qualifiers

This normalization is essential for reliable clinical decision support and cohort identification.

Post-coordinated expressions are stored as structured tuples rather than flat codes, typically using:

• Triple stores with RDF/OWL representations for graph-based reasoning
• Closure tables that precompute all subsumption relationships for fast hierarchical queries
• Expression repositories that index and deduplicate common post-coordinated patterns

Querying requires concept expansion to include all subsumed post-coordinated variants, which can significantly impact query performance if not properly indexed.

The opposite design philosophy where every possible complex concept is assigned a unique, pre-existing code within the terminology. While post-coordination combines atoms at runtime, pre-coordination creates a single identifier for 'Laparoscopic emergency appendectomy'.

• Advantage: Simplicity and consistency in data entry
• Disadvantage: Leads to combinatorial explosion, making the ontology unmanageably large
• SNOMED CT uses a mixed model, pre-coordinating common concepts while allowing post-coordination for rare edge cases

An atomic concept that refines the meaning of a primary clinical concept by specifying a characteristic such as severity, laterality, or course. Qualifiers are the building blocks of post-coordinated expressions.

• Severity: 'Mild', 'Moderate', 'Severe'
• Laterality: 'Left', 'Right', 'Bilateral'
• Clinical Course: 'Acute', 'Chronic', 'Recurrent'
• Finding Context: 'Family history of', 'Absent'

A qualifier is always linked to a primary concept via a defining relationship.

The formal syntactic rules governing how atomic concepts can be combined into a valid post-coordinated expression. This grammar prevents nonsensical combinations like 'Fracture of the liver'.

• Defines domain constraints for each relationship type
• Specifies which qualifiers are allowed for which concept hierarchies
• SNOMED CT Expression Constraint Language provides a machine-readable grammar
• Ensures the resulting expression is clinically valid and logically consistent

The underlying semantic framework that defines the permitted attributes and relationships for each top-level hierarchy in an ontology. It acts as the schema for post-coordination.

• Clinical Finding hierarchy: Permits attributes like Finding site, Severity, Clinical course
• Procedure hierarchy: Permits Method, Procedure site, Using device
• Pharmaceutical product hierarchy: Permits Has active ingredient, Has dose form

Each hierarchy's concept model strictly limits which qualifiers can be applied.

A formal, machine-readable syntax for defining intensional subsets of clinical concepts using post-coordinated logic. ECL enables dynamic cohort building without enumerating every code.

• << 73211009 |Diabetes mellitus| : 363698007 |Finding site| = 113331007 |Structure of endocrine system|
• Used in clinical decision support rules and quality measure definitions
• Allows queries that adapt automatically as the ontology evolves
• Eliminates the need for manually maintained code lists

A software component that transforms a post-coordinated expression into its canonical, fully-defined form by applying description logic classification. This ensures semantic equivalence can be computed.

• Resolves redundant or nested qualifiers into a standard structure
• Detects whether an expression is subsumed by an existing pre-coordinated concept
• Enables subsumption testing: Is 'Severe pneumonia of left lung' a kind of 'Lung disease'?
• Critical for maintaining data quality in systems that accept post-coordinated input