Overview

In this section, we discuss a range of topics that were considered in the design of the SNOMED CT MRCM, including:

• How the MRCM will be authored and quality checked;
• How the MRCM will be used to support the use cases described in Chapter 2;
• How the MRCM will be versioned when corrections or enhancements are made; and
• How the MRCM can be extended and adapted for use with SNOMED CT extensions;

The following sections discuss each of these topics in more detail.

Authoring and Quality Checks

The MRCM has been designed to include minimal redundancy, to make the authoring and maintenance of these rules less error-prone. For example, by using separate reference sets to represent the attribute domains and ranges, each range does not need to be repeated for each of the attribute's domains.

However, some controlled redundancy has been introduced to help support efficient implementation of the MRCM. The following attributes, in particular, include redundant information, and should be automatically populated rather than directly authored:

In addition, the quality of the international SNOMED CT MRCM is expected to be checked using the following mechanisms:

• Manual review, supported by the inclusion of human-readable terms for each concept identifier in the reference sets;
• Automated testing, including checks for the following:
  • All MRCM reference sets conform to the associated Descriptor Template;
  • All concept identifiers refer to active concepts in the relevant SNOMED CT international edition;
  • All concept identifiers refer to concepts from a value set appropriate for that field;
  • All parsable strings (e.g. expression constraints) are syntactically valid (based on the associated ABNF syntax) and refer to concepts that are active in the relevant SNOMED CT international edition;
• Feedback from direct use by IHTSDO staff in the SNOMED CT authoring tools;
• Feedback from the SNOMED CT member, vendor and user communities.

Supporting Use Cases

Another important consideration in the design of the MRCM is how it can best be used to support the range of anticipated use cases. In the following section, we describe the ways in which the MRCM design supports each of the use cases presented in Chapter 2.

Precoordinated Content Development

As described in Section 2.2, one of the key use cases for the SNOMED CT MRCM is to assist in the consistent authoring and validation of SNOMED CT concepts. With this in mind, the following features were included to optimize the MRCM for use in precoordinated content development:

• The |MRCM domain reference set| and the |MRCM attribute domain reference set| include attribute domains defined in terms of the lowest common supertype permitted in the inferred view (i.e. after classification of the concept being authored);
• The |MRCM domain reference set| includes the proximal primitive constraints and mandatory defining relationships required to support proximal primitive modelling in each domain (in the stated view);
• The |MRCM domain reference set| includes a general domain template, which can be used for authoring any content in the given domain using either proximal primitive modelling or proximal parent modelling authoring methods. Please note that this general domain template may be further constrained to support the specialized modelling requirements of specific subhierarchies in the given domain.
• The |MRCM domain reference set| and the |MRCM attribute domain reference set| use domains defined in terms of a single top-level hierarchy. This enables concepts being authored to be tested for subsumption with each domain concept, using techniques that are relatively simple to implement (e.g. using a transitive closure table).
• The |MRCM attribute range reference set| specifies the valid set of values (rangeConstraint) for each attribute used in the authoring of precoordinated content.
• The |MRCM attribute range reference set| also includes a set of attribute rules that incorporate the domain, grouping, cardinality and range constraints together into an expression constraint, to support the efficient implementation of content validation.
• The ruleStrengthId attribute in the |MRCM attribute domain reference set| and the |MRCM attribute range reference set| allows both mandatory rules (resulting in errors) and optional rules (resulting in warnings) to be specified.
• The contentTypeId attribute in the |MRCM attribute domain reference set| and the |MRCM attribute range reference set| allows the scope of each rule to be restricted to a particular type of SNOMED CT content, including:
  • All SNOMED CT content – for rules which apply to both precoordinated and postcoordinated content;
  • All precoordinated SNOMED CT content – for rules which apply to all precoordinated content (but not necessarily to postcoordinated content);
  • New precoordinated SNOMED CT content only – for rules which apply to newly authored precoordinated content, which may not necessarily be valid for all existing content; and
  • All postcoordinated SNOMED CT content – for rules which apply to all postcoordinated content (but not necessarily to precoordinated content).

Expressions, Constraints and Queries

Section 2.3 describes using the SNOMED CT MRCM to assist the authoring and validation of SNOMED CT postcoordinated expressions, SNOMED CT expression constraints, and SNOMED CT queries. These languages allow one or more focus concepts to be defined, which when tested against a domain in the MRCM enables possible attribute refinements, cardinalities and valid ranges to be found.

The optimizations described in section 6.3.1 for use in SNOMED CT concept authoring, can also make the authoring and validation of SNOMED CT expressions, expression constraints and queries more efficient. In addition to these general authoring features, the |MRCM domain reference set| also includes a general domain template, which is designed to support the authoring of any postcoordinated content in the given domain. Once again, this general domain template may be further constrained to support specialized modelling requirements for a specific use case. The rules included in the |MRCM attribute range reference set| also allow some attributes to have a larger range of possible values when used to author postcoordinated content (compared to the authoring of precoordinated content).

Please note that SNOMED CT expression constraints and queries used on a substrate that includes only precoordinated content must use the MRCM rules that apply to |All SNOMED CT content| or |All precoordinated SNOMED CT content|, while expression constraints and queries used on a substrate that includes postcoordinated content may use the less restrictive rules (e.g. broader ranges) that apply to |All postcoordinated SNOMED CT content| (in additional to the rules that apply to |All SNOMED CT content|).
When authoring SNOMED CT expression constraints and queries, the cardinalities specified in the |MRCM attribute domain reference set| should be used to determine sensible cardinality constraints to apply to specific attribute refinements. In particular, a cardinality constraint in an expression constraint or query should be the same, or stricter than the cardinality for the corresponding attribute in the MRCM rules.

Natural Language Processing

In a similar way to how the MRCM can be used for postcoordinated expression authoring, the MRCM can also be used to support Natural Language Processing (NLP), as described in section 2.4. When terms used in the free text are associated with a particular SNOMED CT concept, this concept can be tested for subsumption against the domains in the |MRCM domain reference set|. If the concept is found to belong to one of the domains, then the |MRCM attribute domain reference set| can be used to determine its possible attributes, and the |MRCM attribute range reference set| used to determine these attribute's possible values. These rules can be used to indicate possible ways in which this concept may be linked (via suitable SNOMED CT attributes) to concepts found in the surrounding text.

Optimizations to the MRCM rules may also be adopted by NLP implementations to simplify the testing of membership in a potential range. For example, each range constraint could be split into its separate subhierarchies to allow simple techniques, such as a transitive closure table, to be used.

Terminology Binding to Information Models

Section 2.5 describes using the MRCM to suggest possible SNOMED CT terminology bindings for an information model, to ensure that the bindings are consistent (at least at a high level) with the SNOMED CT concept model, and to support the design of expression templates.

The features designed to support SNOMED CT content authoring can similarly be used to support these terminology binding purposes. For example, the domainTemplateForPostcoordination could be adopted and specialized to support the terminology binding requirements of a specific information model. Similarly, new use-case specific MRCM reference sets could be developed, which restrict the available attributes and ranges that may be used in a specific implementation scenario.

Versioning

As mentioned in Requirement D.5, the concept model must be able to be changed between releases to fix identified issues and enhance future releases.

The RF2 reference set format used in the design of the MRCM provides a standard way of versioning the MRCM and retaining a history of changes, to enable the concept model to evolve over time. Using this standard approach, the effectiveTime field is used to reflect the date on which each rule version comes into effect, while the active field is used to indicate whether a rule is active or inactive. A new version of an MRCM rule may be created to update the mutable fields in the reference set, while changing immutable fields requires the rule to be inactivated, and a new rule to be created.

When a new or updated rule is introduced to the MRCM, it may take a period of time before all existing content conforms to the new rule. In these situations, a content type of |New precoordinated SNOMED CT content only| can be used, to ensure that all new content obeys the rule. Over time, as the old content is updated to reflect the new rule, the content type may be changed to either |All precoordinated SNOMED CT content| or |All SNOMED CT content|. This approach can be used to ensure that all new content added to SNOMED CT conforms to high quality concept model rules, even though reviewing and correcting existing content may take additional time. 

As described in Requirement D.6, the international MRCM rules defined by the IHTSDO must be able to be extended and adapted by organizations developing SNOMED CT extensions, to support the concept model requirements of their extension content.
To support this requirement, SNOMED CT extension developers will be able to copy the international MRCM rules defined by the IHTSDO into one or more extension MRCM reference sets and edit these reference sets to add new rules and restrict or extend the existing rules as required. New MRCM reference sets created for a SNOMED CT extension must be designed to ensure consistency and data integrity between editions. Please note that if no changes to the SNOMED CT concept model are required to support an extension, no additional MRCM reference sets need to be created, as the international MRCM can be reused by the extension.

The |MRCM Module Scope Reference Set| is designed to allow extension developers to specify which MRCM reference sets should be applied to the content in each module. The |MRCM Module Scope Reference Set| should therefore be referred to, to determine whether an extension is reusing the international MRCM, or using a customized extension MRCM to support its content.