FAIR Data Models

FAIR Data Models

FAIR Data Models allow data owners to specify if there dataset uses a particular data model e.g. OMOP CDM. It also allows standard users to search and filter datasets by data model used. 

By default each hub has four curated data models which the hub administrators can enable or disable these as required, and the data model framework allows hub owners to add their own data models. 

Curated models

The curated data models are detailed below:

Name	Description	Enabled/Disabled	Validation
Asset	Data model for a catalogue entry that is not a dataset	Disabled	No
Custom	For all datasets that do not use a recognised data model 'Custom' does not appear as a search and filter option for standard users	Enabled	No
OMOP	Data model for datasets using OMOP CDM	Enabled	Yes - the default OMOP data model has a four level validation framework.  See below for more details
SDTM	Data model for datasets using SDTM	Enabled	No

Hub administrators can enable and disable the curated data models via the PATCH /datamodels/{code} endpoint. They can add their own data models via the POST/datamodels endpoint. 

Data model validation

Hub administrators have the option of adding validation to a data model, this allows data owners to confirm that their metadata conforms to the expected standard. 

By default the OMOP FAIR data model comes with a four level validation framework:

Validation	Description
Basic	All OMOP CDM dictionaries are present, non-OMOP dictionaries may exist in dataset
Bronze	All OMOP CDM dictionaries are present and contain expected columns, non-OMOP dictionaries and columns may exist
Silver	All OMOP CDM dictionaries and columns are present, no additional dictionaries or additional columns allowed, column types are not checked.
Gold	All OMOP CDM dictionaries and columns are present, no additional dictionaries or additional columns are allowed, column types must be correct.

The detail of these validations can be retrieved from the GET datamodels/{code}/validations/{data_model_validation_code} endpoint. We recommend users reveiw these before developing their own validations.