Table 1.
Synopsis of SNOMED CT's relational release format with its underlying semantics in description logics (OWL-EL) and text for the concept “Neoplasm of kidney” in the stated version.
Throughout the paper, we use the OWL Manchester syntax [28], italics for concept (class) names and boldface for relationship names (SNOMED CT linkage concepts / OWL subClassOf and object properties).
Table 2.
Stated relationships and structural pattern for the SNOMED CT concept “Fine needle biopsy of kidney (procedure)” in the stated version.
See Table 3 for sub-hierarchy abbreviations.
Fig 1.
Steps followed to tag SNOMED CT concepts by top-level concept, depth, and structural pattern.
Note that some SNOMED CT concepts belonged to several sub-hierarchies, as shown in this example (Thrombin embedded bandage, ID 412025006) and needed to be tagged twice. In SNOMED CT’s stated relationships file, sourceId denotes the concept to be defined, destinationId the target concept, and typeId the relationship, i.e., [SourceId, typeId, destinationId] triples using our notation. See Table 3 for abbreviations.
Table 3.
Distribution of structural patterns, attributes, and concepts among SNOMED CT’s 18 top-level sub-hierarchies.
Fig 2.
Distribution of structural patterns for the most populated top-level sub-hierarchies.
The x-axis contains patterns found, the y-axis the number of concepts corresponding to that pattern (in logarithmic scale). See Table 3 for abbreviations.
Fig 3.
Absolute accumulation of patterns (p) per sub-hierarchy and depth.
Orange and red indicate an accumulation of more than 10 patterns. The scale is logarithmic.
Fig 4.
Concepts (in blue) and patterns (in red) by depth in SNOMED CT top-level sub-hierarchies.
See Table 3 for abbreviations.
Fig 5.
Relative usage of patterns per sub-hierarchy and depth (c/p).
Reds indicate that, on average, each pattern is used by fewer than 100 concepts. The scale is logarithmic. See Table 3 for abbreviations.
Fig 6.
Accumulation of singleton patterns per sub-hierarchy and depth.
Red indicates that over 50 patterns are accumulated. See Table 3 for abbreviations.
Table 4.
Distribution of singleton patterns and assessed stratified sample of 50 concepts.
Table 5.
Stated relationships and role groups for the SNOMED CT concept Ossiculectomy with tympanoplasty revision (procedure).
Table 6.
Alternative, a more parsimonious model for representing Ossiculectomy with tympanoplasty revision (procedure).
Table 7.
Catecholamines, fractionation measurement, urine (procedure): Procedure concept definition, missing Specimen, but having a measurement method specified that is not reflected in the name.
Table 8.
Example of different patterns for obviously similar concepts in the Events sub-hierarchy.
Table 9.
Disease, combined with the qualifier Abnormal, as a modeling idiosyncrasy in Accessory ossification center (disorder).
Table 10.
Lymphadenopathy due to congenital toxoplasmosis (disorder): a complex concept without direct reference to the underlying etiology (Congenital toxoplasmosis).
Table 11.
Adverse effect of radiation therapy (disorder): example of uniqueness, due to non-consequential modeling of concepts that include a reference to their etiology in their definition but not in their formal representation.
Table 12.
Procedure with explicit context (situation): a high-level singleton pattern.