OMOP vocabulary mapping as a process
Synthetic data. Not submission-ready. Not a copy of official IG mapping tables. Licensed vocabularies are not republished here.
The problem
OMOP’s analytical value comes from standard concept ids, not from storing English text in event tables. Official CDM v5.4 text for many source-value fields says, in substance: map the source to a standard concept in the standardized vocabularies, and keep the original code for reference. The mechanic of that mapping is a process, not a one-line CASE statement.
A working example
A practical, official-tooling-shaped sequence (no vocabulary content included here):
- Inventory source codes as they actually appear (system + code, or local text if there is no code). For the synthetic AE extract, the “code” is only verbatim text — that already tells you the mapping will be a coding step, not a code-to-code lookup.
- Load the official OHDSI standardized vocabularies for the CDM version you run, through the official Athena / OHDSI vocabulary process your organization uses. Do not paste concept ids from a slide deck.
- Map source codes to standard concepts using documented relationships (“Maps to” and related official relationships), then review the leftovers. Official OHDSI materials emphasize review; unmapped or incorrectly mapped codes bias every downstream study.
- Write both fields:
*_source_value(and*_source_concept_idwhen you have one) plus*_concept_idfor the standard concept. Official table descriptions treat these as a pair. - Version the vocabulary. A concept id is only meaningful relative to a vocabulary release. Record it.
The synthetic CONDITION_OCCURRENCE example on the FHIR to OMOP page stopped after step 4’s source-value half: condition_concept_id was left NULL on purpose. Filling it without a licensed/official vocabulary run would have been fiction.
WhiteRabbit / Rabbit-in-a-Hat are official-community tools for scanning a source and designing an ETL. Data Quality Dashboard is listed on the CDM v5.4 page as a large official-community check suite. Use them; this site does not replace them.
The gotcha
Mapping display names instead of codes (“upset stomach” to whatever Google suggests) is not an OHDSI vocabulary mapping. Different sites will choose different concepts, and network studies will not be comparable — the problem the official FHIR-to-OMOP IG says it is trying to reduce.
Some vocabularies require licenses (MedDRA is the obvious example; see MedDRA levels). If a tool or vendor documents that they cannot redistribute a vocabulary, do not scrape it into your repo to “complete” a mapping table.
When not to use this
- Do not publish concept-id crosswalk files that contain licensed content.
- Do not mix CDM versions (5.3 vs 5.4 vs later) in one pipeline without reading the official change documentation.
- Do not assume a FHIR code system URI automatically equals an OMOP vocabulary_id. Official IGs and vocabulary relationship tables are the authority.
Process steps summarize public OHDSI CDM v5.4 and tools pages. No Athena dump, concept id, or licensed vocabulary was downloaded or reproduced. The synthetic NULL concept ids were produced in DuckDB 2.0.0 as a deliberate non-mapping.