2.1.1. Scope of the Metathesaurus

The scope of the Metathesaurus is determined by the combined scope of its source vocabularies. Many relationships (primarily synonymous), concept attributes, and some concept names are added by the NLM during Metathesaurus creation and maintenance, but essentially all the concepts themselves come from one or more of the source vocabularies. Generally, if a concept does not appear in any of the source vocabularies, it will also not appear in the Metathesaurus.

2.1.2. Preservation of Content and Meaning from Source Vocabularies

The Metathesaurus reflects and preserves the meanings, concept names, and relationships from its source vocabularies. When two different source vocabularies use the same name for differing concepts, the Metathesaurus represents both of the meanings and indicates which meaning is present in which source vocabulary. When the same concept appears in different hierarchical contexts in different source vocabularies, the Metathesaurus includes all the hierarchies. When conflicting relationships between two concepts appear in different source vocabularies, both views are included in the Metathesaurus. Although specific concept names or relationships from some source vocabularies may be idiosyncratic and lack face validity, they are still included in the Metathesaurus.

In other words, the Metathesaurus does not represent a comprehensive NLM-authored ontology of biomedicine or a single consistent view of the world (except at the high level of the semantic types assigned to all its concepts). The Metathesaurus preserves the many views of the world present in its source vocabularies because these different views may be useful for different tasks.

Although it preserves all the meanings and content in its source vocabularies, the Metathesaurus stores this information in a single common format. The native format of each vocabulary is carefully studied and then "inverted" into the common Metathesaurus format. For some vocabularies, this involves representing implied information in a more explicit format. For example, if a source vocabulary stores its preferred concept name as the first occurrence in a list of alternative concept names, that first name is explicitly tagged as the preferred name for that source in the Metathesaurus.

2.1.3. Need to Customize the Metathesaurus

Because it is a multi-purpose resource that includes concepts and terms from many different source vocabularies developed for very different purposes, the Metathesaurus must be customized for effective use in most specific applications. Your decisions about what to include in your customized subset(s) of the Metathesaurus will have a significant effect on its utility in your systems. Vocabulary sources that are essential for some purposes, e.g., LOINC for standard exchange of laboratory data, may be detrimental for others, such as Natural Language Processing (NLP). It can also be important to exclude a subset of the concept names found in a vocabulary source that is otherwise useful, e.g., non-standard abbreviations or shortened forms that lack face validity or produce spurious results in NLP.

The Metathesaurus contains source vocabularies produced by many different copyright holders. The majority of the content of the Metathesaurus is available for use under the basic (and quite open) terms described in Sections 1-11 and 13-16 of the Metathesaurus license. However, some vocabulary producers place additional restrictions on the use of their content as distributed within the Metathesaurus. The various levels of additional restrictions are described in Section 12 of the license. The level that applies to individual vocabularies is recorded on the UMLS Source Vocabulary Documentation page of the current UMLS release documentation and in the MetamorphoSys installation and customization program. If you already have a separate license for use of one of the source vocabularies, your existing license also applies to that source as distributed within the Metathesaurus. In some cases, you may have to request permission or negotiate a separate license with a vocabulary producer in order to use that vocabulary in a production system. There may be a charge associated with these separate permissions or license agreements.

The Metathesaurus is designed to facilitate customization. All information in the Metathesaurus is labeled as to its source(s), so it is possible to determine which concept names, attributes, and relationships come from which source vocabularies and which attributes and relationships were added during Metathesaurus construction. The labels allow you to subset the Metathesaurus by excluding information from specific source vocabularies, including those for which you do not have necessary licenses or permissions. It is also easy to exclude all source vocabularies that have particular restriction levels or all information in particular languages. In addition to identifying the source(s), restriction levels, and language of the information it contains, the Metathesaurus includes various more specific concept name flags and relationship labels that can help you to exclude content that is not relevant or helpful for particular applications.

MetamorphoSys, the installation and customization program distributed with the UMLS, makes it easy to generate custom subsets. MetamorphoSys also includes default settings that generate subsets that may be generally useful. MetamorphoSys can also be used to change the default preferred names of concepts; to change the default character set (from 7-bit ASCII to Unicode UTF8); and to include versioned vocabulary source abbreviations in every Metathesaurus file.

2.1.4. Metathesaurus Release Formats

You may select from two relational formats: the Rich Release Format (RRF), introduced in 2004, and the Original Release Format (ORF). Both are available as output options of MetamorphoSys. All Rich Release Format file names have an extension (.RRF). Original Release Format files have no extension. Both formats are described in Chapter 3 and Chapter 4 (usually abbreviated as RRF and ORF).

The Rich Release Format has a number of advantages and is the preferred format for new users of the Metathesaurus and for most data creation applications.

2.2.1. Inclusion of U.S. Standard Code Sets and Terminologies

The Metathesaurus includes terminologies and code sets mandated for U.S. use in electronic exchange of clinical and administrative health data.

2.2.2. Inclusion of Languages Other Than English

The Metathesaurus structure can accommodate translations of its source vocabularies into languages other than English. Many translations in many different languages are present in the current edition of the Metathesaurus. The Metathesaurus includes many translations of some source vocabularies, e.g., NLM’s Medical Subject Headings (MeSH) and the International Classification of Primary Care; one or a few of others, and, in many cases, only the English version. As previously explained, MetamorphoSys makes it easy to create a subset of the Metathesaurus that excludes the languages that are not relevant in a particular application.

2.3.1. Concepts and Concept Identifiers

A concept is a meaning. A meaning can have many different names. A key goal of Metathesaurus construction is to understand the intended meaning of each name in each source vocabulary and to link all the names from all of the source vocabularies that mean the same thing (the synonyms). This is not an exact science. The construction of the Metathesaurus is based on the assumption that specially trained subject experts can determine synonymy with a high degree of accuracy. Metathesaurus editors decide what view of synonymy to represent in the Metathesaurus concept structure. Please note that each source vocabulary’s view of synonymy is also present in the Metathesaurus, irrespective of whether it agrees or disagrees with the Metathesaurus view.

Each concept or meaning in the Metathesaurus has a unique and permanent concept identifier (CUI). The CUI has no intrinsic meaning. In other words, you cannot infer anything about a concept just by looking at its CUI. In principle, the identifier for a concept never changes, irrespective of changes over time in the names that are attached to it in the Metathesaurus or in the source vocabularies.

A CUI will be removed from the Metathesaurus when it is discovered that two CUIs name the same concept – in other words, when undiscovered synonymy comes to light. In these cases, one of the two CUIs will be retained, all relevant information in the Metathesaurus will be linked to it, and the other CUI will be retired.

Retired CUIs are never re-used. Each edition of the Metathesaurus includes files that detail any such changes from the previous edition. One Metathesaurus file (MRCUI.RRF and MRCUI in ORF) tracks such changes from 1991 to the present, allowing you to check the fate of any CUI that is no longer present in the Metathesaurus.

2.3.2. Concept Names and String Identifiers

Each unique concept name or string in each language in the Metathesaurus has a unique and permanent string identifier (SUI). Any variation in character set, upper-lower case, or punctuation is a separate string, with a separate SUI. The same string in different languages (e.g., English and Spanish) will have a different string identifier for each language. If the same string, e.g., Cold, has more than one meaning, the string identifier will be linked to more than one concept identifier (CUI).

2.3.3. Atoms and Atom Identifiers

The basic building blocks or "atoms" from which the Metathesaurus is constructed are the concept names or strings from each of the source vocabularies. Every occurrence of a string in each source vocabulary is assigned a unique atom identifier (AUI). If exactly the same string appears twice in the same vocabulary, for example, as both the long name and the short name for the same concept or as an alternate name for two different concepts in the same vocabulary source, a unique AUI is assigned for each occurrence. When the same string appears in multiple source vocabularies, it will have AUIs for every time it appears as a concept name in each of those sources. All of these AUIs will be linked to a single string identifier (SUI), since they represent occurrences of the same string. Unlike string identifiers, a single AUI is always linked to a single concept identifier, because each occurrence of a string in a source can only have one meaning.

AUIs appear in the RRF (.RRF files), but not in the ORF.

2.3.4. Terms and Lexical Identifiers

For English language entries in the Metathesaurus only, each string is linked to all of its lexical variants or minor variations by means of a common term identifier (LUI). (In the Metathesaurus, therefore, an English "term" is the group of all strings that are lexical variants of each other.) English lexical variants are detected using the Lexical Variant Generator (lvg) program, one of the UMLS Lexical Tools. As similar tools become available for other languages, they may be used to create lexical variant groups in other languages. (In the meantime, the LUI for a non-English string is really another string identifier.)

Like a string identifier, the LUI for an English string may be linked to more than one concept. This occurs when strings that are lexical variants of each other have different meanings. In contrast, each string identifier and each atom identifier can only be linked to a single LUI.

2.3.5. Uses of Concept, String, Atom, and Term Identifiers

In the Metathesaurus, every CUI (concept) is linked to at least one AUI (atom), SUI (string), and LUI (term), but can also be linked to many of each of these. Every AUI (atom) is linked to a single SUI (string), a single LUI (term), and a single CUI (concept). Each SUI (string) can be linked to many AUIs (atoms), to a single LUI (term), and to more than one CUI (concept) – although the typical case is one CUI. Each LUI (term) can be linked to many AUIs (atoms), many SUIs (strings), and more than one CUI (concept) – although the typical case is one CUI.

In the abbreviated example in Table 1, Atrial Fibrillation appears as an atom in more than one source vocabulary and has a distinct AUI for each occurrence. Since each of these atoms has an identical string or concept name, they are linked to a single SUI. Atrial Fibrillations, the plural of Atrial Fibrillation, has a different string identifier. Since the singular and plural are lexical variants of each other, both are linked to the same LUI. There is a different LUI and different SUIs and AUIs for Auricular Fibrillation and its plural Auricular Fibrillations. Since Atrial Fibrillation and Auricular Fibrillation have been judged to have the same meaning, they are linked to the same CUI.

Table 1.

Concept, Term, Atom, and String Identifiers.

All of these identifiers serve important purposes in building the Metathesaurus, in allowing efficient and accurate customization for specific purposes, and in identifying changes in its concept and concept name coverage over time.

For example, CUIs link all information in the Metathesaurus related to particular concepts. In other words, a CUI can be used to retrieve all the concept names, relationships, and attributes for a particular concept that appear in any Metathesaurus file. CUIs also serve as permanent, publicly available identifiers for biomedical concepts or meanings to which many individual source vocabularies are linked. You are strongly encouraged to incorporate CUIs in your local applications – to support data exchange and linking and to assist migration between the use of individual source vocabularies should that become necessary in the future.

2.3.6. Default Preferred Names for Metathesaurus Concepts

As a convenience for those who build the Metathesaurus, one string from one English term is designated and labeled as the default preferred name of each concept in the Metathesaurus. To avoid laborious selection among alternative terms and strings, selection of the default preferred name for any Metathesaurus concept is based on an order of precedence of all the types of English strings in all the Metathesaurus source vocabularies. Different types of strings, e.g., preferred terms, cross references, and abbreviations from each vocabulary, will have different positions in this order. The factors considered in establishing the default order of precedence include breadth of subject coverage, frequency of update, and the degree to which the source's concept names are used in regular clinical or biomedical discourse. The default order of precedence appears in MRRANK.RRF (MRRANK in ORF), and on the Source and Term Types: Default Order of Precedence and Suppressibility page of the current UMLS release documentation.

The default order of precedence will not be suitable for all applications of the Metathesaurus. The MetamorphoSys can be used to change the selection of preferred names to feature terminology from the source vocabularies most appropriate to particular user populations. For example, concept names from SNOMED CT may be preferred in clinical applications, and terminology from MeSH may be preferred in literature retrieval systems.

2.3.7. Strings with Multiple Meanings

In some cases, the same name (with or without differences in upper-lower case) may apply to different concepts, usually (but not always) in different Metathesaurus source vocabularies. In the abbreviated example that follows, the string "Cold" is a name for the temperature in one vocabulary. In another vocabulary, "Cold" is an alternate name for the "Common cold". In a third vocabulary, "COLD" is an acronym for "chronic obstructive lung disease". As a result, "Cold" or "COLD" appears as a name of more than one concept in the Metathesaurus.

2.3.7.1. Representation of Ambiguity in the Metathesaurus

Separate Metathesaurus files (AMBIGLUI.RRF and AMBIGSUI.RRF (AMBIG.LUI and AMBIG.SUI in ORF)) contain the LUIs and SUIs of all ambiguous terms and strings known to the Metathesaurus. See Table 2.

Table 2.

Representation of Ambiguity in the Metathesaurus.

2.3.8. Concept Names Added During Metathesaurus Construction

Although the majority of concept names present in the Metathesaurus come from one or more of its source vocabularies, some concept names are created during Metathesaurus construction. This occurs in the following circumstances:

1.

A unique name is created for a string with multiple meanings (the case explained in Section 2.3.7)

2.

A more explicit name is created when none of the source vocabulary names for a concept conveys its meaning adequately

3.

An American English variant is generated for a British spelling

4.

An equivalent basic Latin ASCII character set string is generated for a string in an extended character set, such as Unicode

Like all other concept names in the Metathesaurus, names created during Metathesaurus construction are labeled to indicate their source.

2.4.1. Basic Categories of Non-Synonymous Relationships

The Metathesaurus contains non-synonymous relationships between concepts from the same source vocabulary (intra-source vocabulary relationships) and between concepts in different vocabularies (inter-source vocabulary relationships). The Metathesaurus does not include all possible non-synonymous relationships between the concepts it contains. It includes all relationships present in its source vocabularies and some additional relationships designed to connect related concepts. In general, the relationships asserted by source vocabularies connect closely related concepts, such as those that share some common property or are related by definition. For example, a member of a class of drugs (e.g., penicillin) will be connected to the name for the class (e.g., antibiotics); a bacterial infection will be connected to the bacterium that causes it.

2.4.2. Relationship Labels

All relationships (outside the basic concept structure) in the Metathesaurus carry a general label (REL), describing their basic nature, such as Broader, Narrower, Child of, Qualifier of, etc., and are identified by their source. Most of these relationships are either directly asserted in a source vocabulary or are implied by the structure of the source vocabulary. A complete list of the general relationship labels appears in MRDOC.RRF and on the Abbreviations Used in Data Elements page of the current UMLS release documentation.

About a quarter of the relationships in the Metathesaurus also carry an additional label (RELA), obtained from a source vocabulary, that explains the nature of the relationship more exactly, such as is_a, branch_of, component_of. The Digital Anatomist vocabulary and RxNorm are examples of source vocabularies that include such relationship labels. A complete list of the additional relationship labels appears in MRDOC.RRF and on the Abbreviations Used in Data Elements page of the current UMLS release documentation.

2.4.3. Relationship Identifiers

Every relationship present in the Metathesaurus has a unique relationship identifier (RUI). The primary purpose of these identifiers is to enable easy detection of changes in relationships across versions of the Metathesaurus. The appearance or disappearance of a relationship identifier indicates a change in the relationships present in the Metathesaurus.

Some source vocabularies have their own relationship identifiers. Where they exist, these identifiers are also present in the Metathesaurus.

2.4.4. Relationship Groups

Relationship groups are source-asserted or implied associations of relationships that may be used to add meaning or clarity when multiple relationships are present. Together, each grouping may express a richer meaning than it would have ungrouped. Relationship groups can be identified in MRREL.RRF by rows that have the same AUI2 and the same numeric value for relationship group (RG). Numeric values for relationship group increase according to the number of relationship groups associated with the same AUI2. A null value indicates no grouping for the relationship is present. Relationships may be suppressible if considered obsolete, which is indicated by a value of O in the SUPPRESS field.

CUI1 | AUI1 | STYPE1 | REL | CUI2 | AUI2 | STYPE2 | RELA | RUI | SRUI | SAB | SL | RG | DIR | SUPPRESS |CVF

C0024109|A3154872|SCUI|RO|C0264408|A2957612|SCUI|has_finding_site|R14028961|994883025|SNOMEDCT_US|SNOMEDCT_US|0|Y|O||

C0231335|A2926532|SCUI|RO|C0264408|A2957612|SCUI|occurs_in|R123147138|1795540028|SNOMEDCT_US|SNOMEDCT_US|0|Y|N||

C0006255|A3104303|SCUI|RO|C0264408|A2957612|SCUI|has_finding_site|R98157815|3465258024|SNOMEDCT_US|SNOMEDCT_US|1|Y|O||

C0028778|A2873893|SCUI|RO|C0264408|A2957612|SCUI|has_associated_morphology|R98053314|3419439024|SNOMEDCT_US|SNOMEDCT_US|1|Y|O||

In this example, Relationship Group 0 groups the relationships from A2957612 “Childhood asthma” to A2926532 “Childhood” and A3154872 “Lung structure” to clarify that “Childhood asthma” is found in the lungs of children.

Relationship Group 1 groups the relationships from A2957612 “Childhood asthma” to A3104303 “Bronchial structure” and A2873893 to indicate that “Childhood asthma” is characterized by obstruction of the bronchi.

In SNOMEDCT_US, relationship groups are source-asserted. The SRUI field of MRREL.RRF contains the SNOMEDCT-asserted unique identifier assigned to the relationship. Please refer to the SNOMED CT Document Library for a detailed description of relationship groups in SNOMED CT.

In MeSH and MedlinePlus, relationship groups are source-implied by a Descriptor (SDUI) and Qualifier that appear under the same MeSH Mapped To Heading. Example from MeSH:

<HeadingMappedToList>
   <HeadingMappedTo>
    <DescriptorReferredTo>
     <DescriptorUI>D012694</DescriptorUI>
     <DescriptorName>
      <String>Serine</String>
     </DescriptorName>
    </DescriptorReferredTo>
    <QualifierReferredTo>
     <QualifierUI>*Q000031</QualifierUI>
     <QualifierName>
      <String>analogs &amp; derivatives</String>
     </QualifierName>
    </QualifierReferredTo>
   </HeadingMappedTo>
</HeadingMappedToList>

This association is represented in UMLS as a grouping of mapped_to and has_mapping_qualifier relationships that connect the AUIs corresponding to the DescriptorName and QualifierName strings to the AUI of the MeSH Mapped To Heading.

CUI1 | AUI1 | STYPE1 | REL | CUI2 | AUI2 | STYPE2 | RELA | RUI | SRUI | SAB | SL | RG | DIR | SUPPRESS |CVF

C0002776|A3879704|SDUI|RO|C0067636|A0207764|SDUI|has_mapping_qualifier|R148279824||MSH|MSH|1||N||

C0036720|A0115503|SDUI|RN|C0067636|A0207764|SDUI|mapped_to|R148155946||MSH|MSH|1||N||

The relationship group links the relationships to clarify that A0207764 “N-acetyl-4-nitrophenylserinol” is mapped to A0115503 “Serine” and has mapping qualifier A3879704 “analogs & derivatives”.

2.5.1. Kinds of Attributes

The Metathesaurus includes concept attributes, atom attributes, and relationship attributes.

Concept attributes are added during Metathesaurus construction and apply to all names of a concept. For example, the Semantic Types "Pathologic Function" and "Finding" are attributes of the concept with the preferred name "Atrial Fibrillation" and are applicable to any atom connected to that concept.

Atom attributes come from a particular source vocabulary. Some of them are of general interest; others are relevant only to a particular source vocabulary. For example, the definition "Disorder of cardiac rhythm characterized by rapid, irregular atrial impulses and ineffective atrial contractions" is an attribute of the atom Atrial Fibrillation that comes from the Medical Subject Headings (MeSH). It may be one of several definitions connected to names of this concept, because the Metathesaurus includes all definitions provided by any of its source vocabularies. Although this particular definition comes from MeSH, it might well be useful in Metathesaurus applications that otherwise do not use MeSH. In contrast, the date an occurrence of a string (an atom) was added to a source vocabulary applies only to that specific atom. The utility of specific atom attributes will vary considerably for different applications of the Metathesaurus.

Relationship attributes come from a particular source vocabulary and describe special characteristics of particular relationships in that source, e.g., refinability.

The majority of attributes are distributed in MRSAT.RRF and MRSAT in the ORF. In these files, each row contains the name of the attribute, the source of the attribute, and the value of the attribute, in addition to all appropriate identifiers. There are separate files for selected attributes such as the Semantic Types (MRSTY.RRF and MRSTY in the ORF) and the definitions (MRDEF.RRF and MRDEF in the ORF).

2.5.2. Attribute Identifiers

Each occurrence of each attribute within the Metathesaurus is assigned a unique attribute identifier (ATUI). The appearance or disappearance of ATUIs signals changes in the content of the Metathesaurus, thus ATUIs assist the efficient production of a complete change set for each new version of the Metathesaurus. ATUIs appear only in the RRF, not in the ORF.

2.6.1. Characteristics of the Current Metathesaurus

There are discrete Metathesaurus files for:

• The names and sizes of every Metathesaurus file (MRFILES.RRF and MRFILES in ORF)
• The names and size range of every Metathesaurus data element (MRCOLS.RRF and MRCOLS in ORF)
• The possible values for selected data elements that contain a finite set of abbreviated values (MRDOC.RRF only). Note: Eventually this file will include values for every data element that contains a finite set of abbreviated values.
• The source vocabularies in the Metathesaurus (MRSAB.RRF and MRSAB in ORF)
• The LUIs and SUIs for terms and strings that are known to be ambiguous, that is, to have multiple meanings (to be linked to multiple concept identifiers) within the Metathesaurus (AMBIGLUI.RRF and AMBIGSUI.RRF in RRF and AMBIGLUI and AMBIGSUI in ORF)
• The order of precedence of vocabulary source and term types that is used to compute the default preferred concept name for each concept in the Metathesaurus (MRRANK.RRF and MRRANK in ORF). Note: MetamorphoSys can be used to change this order.

MRCOLS, MRDOC, MRSAB, and MRRANK contain data that do not appear in the actual Metathesaurus content files. The others are computable from the Metathesaurus content files. They are pre-computed and provided in separate files as a convenience to users.

2.6.2. Changes Between the Current Metathesaurus and the Previous Version

Each version of the Metathesaurus contains a set of files that summarize changes from the previous version.

CHANGE/MERGEDCUI.RRF in the RRF (CHANGE/MERGED.CUI in the ORF) documents cases in which two discrete concepts in the previous version of the Metathesaurus are now considered to be synonyms.

CHANGE/MERGEDLUI.RRF in the RRF (CHANGE/MERGED.LUI in the ORF) documents cases in which two discrete terms in the previous version of the Metathesaurus are now identified as lexical variants of each other, based on the current version of luinorm (the program used to compute them).

Three files contain the CUIs, LUIs, and SUIs for Metathesaurus concepts, terms, and strings that appeared in the previous version, but are not in the current version (CHANGE/DELETEDCUI.RRF, CHANGE/DELETEDLUI.RRF, CHANGE/DELETEDSUI.RRF in the RRF and CHANGE/DELETED.CUI, CHANGE/DELETED.LUI, CHANGE/DELETED.SUI in the ORF).

Note: Future versions of the Metathesaurus change files will provide for relationships and attributes in the RRF only. The generation of these files is dependent on the relationship and attribute identifiers (RUI and ATUI) introduced in the 2004AA version of the Metathesaurus.

2.6.3. Historical CUIs

The retired CUI file (MRCUI.RRF in RRF and MRCUI in ORF) includes all CUIs present in any previous version of the Metathesaurus, but not in the current version. In general, the file maps the retired CUI to one or more current CUIs.

2.7.1. Word Index

2.7.2. Normalized Word Index

2.7.3. Normalized String Index

2.7.4. Word Index Programs

The programs that generate these indexes are written in Java. They may be of use to system developers who are developing their own interfaces to the UMLS data or for other purposes. Chapter 6 includes information about these and other lexical programs provided with the UMLS Knowledge Sources.

2.9.1. Representation of Content View Metadata in RRF

Higher level information about each Content View is included as a concept directly in the Rich Release Format (RRF) files as described below.

STR = <Content View Name>, e.g. ‘MetaMap NLP View’
CODE = NOCODE
TTY = CV
SAB = MTH
SAUI, SCUI, SDUI = null

2.9.2. Extracting Content Views

SELECT * FROM MRCONSO WHERE TTY='CV';

SELECT ATN, ATV FROM MRSAT WHERE CUI='C1700357';

SELECT * FROM MRCONSO WHERE BITAND(CVF, 256) <> 0;

2.10.1. Representation of Mappings in the Metathesaurus

Inter-source mapping data is represented using the following specifications (there may be exceptions, e.g. for map sets that have not been updated recently):