Observational Health Data Sciences and Informatics

Proposing person: Vojtech Huser (someone initially posted this to the wiki (not indicating the source), I got assigned to it)

Forum thread: http://forums.ohdsi.org/t/omop-cdm-and-clinical-trials/2109

Use Case(s): 1. observational trial collects CRF data but also wants to digest any EHR updates (e.g., AllOfUs project)

2. multi site observational trial decides to use OMOP CDM as the schema for their pooled data

3. sponsor must provide trial data to FDA in SDTM format (and has those currently in OMOP CDM)

Possible problems:

1. convention for: person_id is in trial identified by NCT0000123456 and consented on date X

2. convention for: CRF data is mapped to concepts. And concepts dictate where it goes. E.g., hematocrit CRF question goes into MEASUREMENT; others go into OBSERVATION.

3. if CRF question does not map to any concept, what do we do?

3. convert OMOP CDM <Dx data> into SDTM

4. convert OMOP CDM <any data> into SDTM

Many organizations using the OMOP CDM for observational data are also charged with the production and handling of clinical trials data. One common format (that required for FDA submission) is the CDISC Study Data Tabulation Model.

OMOP is fundamentally an observational data model, so the ability to handle data derived from clinical trials may not be a priority. However, the community should consider whether the integration of such data into the CDM would add value. Most of the standard clinical trial domains can be represented in the current CDM, albeit with extensive use of the concept_relationship table. There may also be hybrid approaches to representing clinical data in the CDM while maintaining the study information (eg. subject metadata) separately.

If inclusion of CT data is a priority for community members, it would be worth exploring the SDTM and its relation to the CDM in further detail.