OMOP Common Data Model Custom Concepts
By Maxim Moinat and the Themis Working Group
The Themis Working Group convened on October 6th and December 7th 2023 to discuss the creation this convention for creating custom concepts.
Introduction
While the OMOP vocabularies are very comprehensive, it is not always possible to use concepts existing in the OMOP vocabularies. For example, when using a vocabulary that is only used in your institution or having custom defined variables. In these cases, custom concepts can be used. Custom concepts are concepts that are not part of the OMOP vocabularies, and are only used in your institution. There are two main reasons to define custom concepts in your local OMOP CDM vocabulary. The first is that they are available in your local Atlas instance, which has several use cases:
- When viewing a standard concept, you can see which custom concepts are mapped to it. This allows you to better understand what the standard concept represents in your institution.
- You can search for a custom concept and find which standard concepts it is mapped to, to include in your standard concept set.
- For studies only using your local data, you can define cohorts using custom concepts (through ‘Add attribute’->‘Add … Source Concept’).
The second reason is using the custom concepts in your ETL. By creating both the custom concept, and the ‘Maps to’ relationship (example below), we can use this in the same way as mapping other source vocabularies.
Custom concepts are only defined locally. These cannot be used for network research. Therefore it remains very important to map to standard concepts.
Conventions
It is important to follow a set of conventions when creating custom concepts, to avoid negatively impacting network studies. The conventions are as follows:
- Custom concepts have a
concept_idlarger than 2,000,000,000. This is to avoid clashes with existing concepts.† - Custom concepts are always non-standard
(
concept.standard_concept = NULL). - Custom concepts can only be used in the
_source_concept_idfields (e.g.procedure_source_concept_id) - If the field
_source_concept_iddoes not exist, add a custom column to your table (e.g. adrug_exposure.route_source_concept_id)1. If there is a wide need in the community, a proposal can be submitted to the CDM Working Group to add these fields in a future OMOP CDM version. - Always add custom concepts to a new vocabulary specifically for your source (see example below). If you have a need to add concepts to an existing vocabulary, please contact the OHDSI Vocabulary Working Group.
- Do NOT add custom concepts to the hierarchy of existing vocabularies
(either in the
concept_ancestortable orsubsumes/is arelations in theconcept_relationshiptable). This is to avoid descendant concepts to differ between sites.
In addition, it is recommended to follow these suggestions:
- For the
domain_id, choose an appropriate existing value from the domain table (‘Condition’, ‘Drug’, ‘Procedure’, etc.). Note that this determines the target table the concept can be used in. - For the
concept_class_id, it is not required to create new classes for your source vocabulary. To leave the class empty explicitly, use class ‘Undefined’. It is also allowed to reuse existing concept classes (e.g. SNOMED’s ‘Clinical Finding’ for conditions or RxNorm drug classes ‘Ingredient’, ‘Clinical Drug’, ‘Branded Drug’ etc.). - In the new vocabulary record, the
vocabulary_concept_idcan be set to 0, as this is often not used in the OMOP CDM. - Create mappings between custom concepts and standard concepts. This
can be done by adding rows to the
concept_relationshiptable, with theMaps torelation. The reverse relation,Mapped from, should also be added. This allows for easy navigation between custom and standard concepts2. The ‘mapped to’ concept should be a standard concept. - Officially,
concept_hierarchyis only for standard concepts. However, if you local use case requires this (e.g. for selection of descendants of custom concepts), the custom concepts can be added into their own, isolated, hierarchy.
Example
In this example, we will add one custom concept for the ‘DHD Diagnose Thesaurus’. This is a Dutch vocabulary, which is not part of the OMOP vocabularies. We will add the concept ‘diabetes mellitus type 1’. This concept has a mapping to the standard concept ‘Diabetes mellitus type 1 (disorder)’, concept_id 3341872.
After creating these records, we can use the custom concept in our
ETL to populate the condition_source_concept_id field.
Custom vocabulary
| Field | Value |
|---|---|
| vocabulary_id | DHD Diagnose Thesaurus |
| vocabulary_name | Dutch Hospital Data Diagnosethesaurus |
| vocabulary_reference | https://www.dhd.nl/producten-diensten/diagnosethesaurus/Paginas/diagnosethesaurus.asp |
| vocabulary_version | 2023-04-20 |
| vocabulary_concept_id | 0 |
Custom concept
| Field | Value |
|---|---|
| concept_id | 2 000 000 001 |
| concept_name | diabetes mellitus type 1 |
| domain_id | Condition |
| vocabulary_id | DHD Diagnose Thesaurus |
| concept_class_id | Undefined |
| standard_concept | NULL |
| concept_code | 0000002630 |
| valid_start_date | 2017-08-01 |
| valid_end_date | 2099-12-31 |
| invalid_reason | NULL |
Mapping relationship
| Field | Value |
|---|---|
| concept_id_1 | 2 000 000 001 |
| concept_id_2 | 3341872 |
| relationship_id | Maps to |
| valid_start_date | 2017-08-01 |
| valid_end_date | 2099-12-31 |
| invalid_reason | NULL |
| Field | Value |
| ——- | ——- |
| concept_id_1 | 3341872 |
| concept_id_2 | 2 000 000 001 |
| relationship_id | Mapped from |
| valid_start_date | 2017-08-01 |
| valid_end_date | 2099-12-31 |
| invalid_reason | NULL |
See the Source To Standard query to map a code in the source data to a standard concept.
Getting your vocabulary into the OMOP vocabularies
If you think your custom concepts are useful for others, you can submit them to the OMOP vocabularies. This could be as a supported source vocabulary (like ICD) or a new vocabulary with standard concepts (like LOINC). This is a separate process from the conventions described above. Please see the this support page.
References
1Melanie Philofsky; Implementing & adopting a customized OMOP Common Data Model; OHDSI Symposium 2021
2Melanie Philofsky; Mapping Custom Source Codes to Standard Concepts: A Comparison of Two Approaches; OHDSI Symposium 2020 †this is why custom concepts are sometimes referred to as “2B+” or “2billionaires”.