-In the beta phase of the 11th revision of International Classification of Diseases (ICD-11), the World Health Organization (WHO) exposes ICD-11 content through a collection of web services. The emerging Common Terminology Services 2 (CTS2) standard provides a common model and semantics for representation, interchange, and federation of terminological resources. We hypothesize that the CTS2 standard can provide service layer of standardization that could potentially aid in the interoperability among authoring applications for the ICD-11 revision. This paper examines the existing WHO ICD-11 content services from the perspective of the CTS2 standard. The content in the WHO ICD-11 content services was mapped to the Code System Catalog, Code System Version Catalog, Entity Description and Association models in the CTS2 specification. It proposes mappings for existing ICD-11 properties and suggests additional CTS2 properties may be important for ICD-11. The mapping effort was used to develop a prototype of the CTS2 Services Wrapper for the ICD-11. In conclusion, the CTS2 standard is useful in exposing ICD-11 content representation through predictable and familiar services.
Keywords—ICD-11; CTS2; Medical Classification; Data Standards; Biomedical Ontologies
I.
INTRODUCTION
The International Classification of Diseases (ICD) is an
international data standard presently maintained and curated by
the World Health Organization (WHO) [
To facilitate the ICD-11 revision, WHO developed a
content model to present the knowledge that underlies the
definitions of an ICD entity. The content model is composed of
three layers: a foundation layer, a linearization layer, and an
ontological layer [
The Common Terminology Services 2 (CTS2) [
The objective of this paper is to describe the existing WHO ICD-11 content services using the CTS2 standard specifications. We profiled the WHO ICD-11 content services using the information models of Code System Catalog, Code System Version Catalog, Entity Description and Association from the CTS2 specification. We developed the mappings for the elements between the two service models and identified the CTS2 elements that are missing from the WHO services but important for the ICD-11 content representation and services.
We also developed a prototype of the CTS2 Wrapper Services for ICD-11.
A Unified Resource Identifier (URI) is commonly used to
name and identify a resource, and is treated as a core principle
of Semantic Web Linked Open Data by Tim Berners-Lee [
A RESTful, resource-oriented service exposes a URI for every piece of data the client might need to interact with. A welldesigned URI structure/scheme would allow users to easily explore and invoke ICD content through an implementation of the RESTful services. Recently, an ICD URI scheme was proposed for naming and supporting web services by WHO.
The base URI of http://id.who.int has been adopted, with http://id.who.int/icd/schema as the prefix for the vocabulary terms that are related to ICD classification efforts maintained by WHO and http://id.who.int/icd/entity for the foundation entities related to ICD concepts. Title, Latest Version, Version Title, Latest Version, Version
Fig. 1. A HTML view of the RDF rendering of the individual foundation entity “Acute myocardial infarction (http://id.who.int/icd/entity/1334938734)”. (icd prefix: http://id.who.int/icd/schema/; skos: http://www.w3.org/2004/02/skos/core#.)
CTS2 is an OMG specification [
The CTS2 specification specifies the information models for the terminological artifacts. These artifacts include: Code System Catalog – metadata about code systems (ontologies, code sets, thesauri, classification systems, etc.) (See Fig. 2) Code System Version Catalog - metadata about specific versions of code systems.
Entity Description - a set of entity (aka. “class”, “category”, “concept”, “predicate”, “property”, “term”, “individual”) identifiers known to the service along with information about which code system versions make assertions about these identifiers and what they say.
Association - sets of “semantic” assertions about entity identifiers, in which the entity identifier may play the role of subject, predicate (verb) or object in the assertions. Value Set Catalog - metadata about sets of entity identifiers (value sets) that have been grouped for some purpose. Value Set Definition - information about how value sets are constructed. • •
Concept Domain Catalog - a catalog of abstract “concept domains” that represent a collection of possible meanings. Map Catalog - a catalog of “maps” - collection of rules that allow human or machine assisted transformation between the codes in one value set or code system and those in a second.
The CTS2 specification also supports a number of functional areas including the Read, Query, Update, History, Maintenance, and Temporal services.
Fig. 2. Code System Catalog Information Model.
information models of Code System Catalog, Code System Version Catalog, Entity Description and Association. We identified the mappings of the properties between the two service models and suggested the CTS2 properties that are important for ICD-11 but currently missing from the existing service implementation.
We identified three code systems from the existing WHO ICD content services: 1) ICD-11 Foundation (http://id.who.int/icd/entity) that corresponds to the Top Level of the Foundation category in Table 1; 2) ICD-11 Morbidity (http://id.who.int/release/11/morbidity) that corresponds to the Top Level of the ICD-11 Linerizations without minor version in Table 1; and 3) ICD10 (http://id.who.int/release/10) that corresponds to the Top Level of the ICD10 without minor version. Table 2 shows the mappings of the properties between the ICD code systems and the CTS2 Code System Catalog model. We also suggested 5 properties shown in red (codeSystemName, resourceType, resourceSynopsis, rights and entryState) that may be important for describing the ICD code systems. versioning information, so the same URI is used for both the code system and the code system version for now. 2) ICD-11 Morbidity Beta (http://id.who.int/release/11/beta/morbidity) is the current version of the code system ICD-11 Morbidity; 3) ICD10 has two version: ICD10 2010 (http://id.who.int/release/10/2010) and ICD10 2008 (http://id.who.int/release/10/2008), in which ICD10 2010 is the current version of the code system ICD10.
For the individual entities in each code system version, we used the CTS2 Entity Description model for the profiling. Table 4 shows the mappings of the properties between the individual entities and the CTS2 Entity Description model. We also suggest four additional CTS2 properties shown in red (describingCodeSystemVersion, ancestors, descendants and entityType) that may be important for describing an individual ICD-11 entity.
As mentioned in the section above, the CTS2 Association represents sets of “semantic” assertions about entity identifiers, in which the entity identifier may play the role of subject, predicate (verb) or object in the assertions. In fact, the properties “subjectOf”, “predicateOf”, “targetOf”, “parents”, “children”, “ancestors” and “descendants” in the CTS2 Entity Description model (see Table 4) are such sets of assertions utilizing the Association model. Specifically, the CTS2 Association Directory model is used for the semantics asserted for the property “subjectOf” and the CTS2 Entity Directory model is used for the semantics asserted for the rest of four properties.
Based on the mappings identified and the new properties
suggested in the section above, we have initiated a prototype
implementation by building a CTS2 services wrapper for
ICD11. The wrapper takes a URI from existing WHO ICD content
services as input, and extracts RDF-based content (by content
negotiation) using the Jena API [
Note that the CTS2 Development Framework is a development kit for rapidly creating CTS2 compliant applications. The CTS2 Development Framework provides a REST web application, along with URL routing and parameter handling. The CTS2 Development Framework has been successfully used to build a NCBO BioPortal Wrapper service that provide a layer of standardized CTS2 signatures to existing biomedical ontologies available in BioPortal.
Table 5 shows a set of service signatures proposed for the implementation of an ICD-11 CTS2 Services Wrapper. The wrapper services cover those terminological artifacts presented in the section III. The detailed examples of each service for the prototype implementation can be accessed at:
http://informatics.mayo.edu/rest2/project/icd11/cts2/doc.
First, we found that ICD-11 Foundation, as a code system
from the CTS2 perspective, does not have a versioning
mechanism in the existing WHO services. The existing WHO
service implementation was based on a specific use case
assumption that all the linearization releases (e.g., ICD-11
Morbidity) should refer to versionless foundation entity URIs
so that different linearizations created at different times could
be anchored by the foundation entity URIs. However, when we
consider the entire life cycle of ICD-11 content authoring, the
versioning mechanism is a critical component for the authoring
applications that would need to access version specific
foundation entities or history of changes. We propose that
WHO may consider a model similar to that used by SNOMED
CT which allows for both fine-grained and coarse versioning.
Specifically, revisions to the SNOMED CT are released twice
a year and marked by the release date, e.g., the version
20140131. In a fine-grained level in the Release Format 2
(RF2) of SNOMED CT [
Second, we identify three code systems and their corresponding code system versions from the existing WHO services, while we also find that these important terminological artifacts are generally underspecified. In addition to identifying the property mappings, we suggest 5 properties from the CTS2 Code System Catalog model and 9 properties from the CTS2 Code System Version Catalog model for describing these ICD11 terminological artifacts. For example, we suggest property “rights” for the artifacts because it will make the WHO rights statement explicit . Such statement will be critical for providing the guidance to the user community for the use of ICD-11 that is critical to the users. We also suggest the properties “defaultLanguage” and “supportedLanguage” for the ICD-11 because the multilingual support is one of major goals for the ICD-11 revision and eventually at least the six official United Nations languages will need to be supported.
Third, those ICD properties (such as Body System, Causal Agents, etc.) that will have values from external terminologies (e.g., SNOMED CT, ICF) are mapped to the CTS2 property “subjectOf”, which can be elegantly described by the Association Directory model. Note that the existing WHO services have not exposed the external identifiers for those values annotated by external terminologies. For example, as shown in Fig. 1, the property icd:bodySite in the foundation entity “Acute myocardial infarction” has the value “Myocardium structure (body structure)” in a textual string. The original unique identifier from SNOMED CT (i.e., http://snomed.info/id/74281007) is not exposed in the existing services. In addition, CTS2 uses the Entity Directory model to describe the semantics for the properties “parents”, “children”, “ancestors” and “descendants” in the CTS2 Entity Description model. We found that the existing WHO services do not have a mechanism to retrieve the transitive closure for parent (i.e., skos:broaderTransitive) or child (skos:narrowerTransitive) relationships to get all ancestors or descendants for a specific ICD entity. This is an area that would need to be enhanced in the future.
Fourth, CTS2 is designed to address a broad range of
requirements within the ontology and terminology community.
In addition to the specification about the four kinds of
terminological artifacts addressed by this paper, CTS2
specifications also have the notions of Value Set Catalog,
Value Set Definition and Map Catalog. Although the existing
WHO services have not exposed any such terminological
artifacts, these artifacts (e.g., value sets) have been actively
utilized in the ICD-11 revision process. Actually, a collection
of value sets (including Body Systems, Anatomical Sites,
Functioning, etc.) had been developed and utilized in support
of the ICD-11 revision. In a previous study, we developed and
evaluated approaches to value set extraction from SNOMED
CT for the ICD-11 anatomy use case [
In summary, CTS2 standard specification is a very useful tool for profiling the ICD-11 terminology content. The outcome produced by this profiling would potentially enrich the WHO ICD-11 content representation and enable a robust and standards-based content services to better support the ICD11 revision applications.