Deploying OMOP
A high-level guide to taking your cohort from raw data to the OMOP Common Data Model
Deploying OMOP for a cohort follows four steps, based on the workflow used by the National Data Curation Team (NDCT). This page gives an overview of each step. Step-by-step instructions, code and templates are in the Deploying OMOP Guide in the enTRUST members' portal.
The four steps
Jump to the step you are at
Clean the data
Data dictionary, profiling, cleaning rules
Map the data
Catalogue, ATHENA, Usagi, expert review
Extract, Transform and Load
Database, design, load, data quality
Maintain, share and reuse
Updates, versions, contributions
Before you start
Make sure you are a TRUST Member, so you can use OMOP Concepts Search and the members' guidance.
Decide who will review mappings. A clinician or domain expert who knows the cohort is ideal.
Plan where your OMOP database will run. PostgreSQL is recommended.
Clean the data
Get the dataset into a consistent, documented state before anything is mapped.
Build a data dictionary: every variable, its type, its allowed values and what it means.
Profile the data to find missing values, outliers and inconsistent formats.
Apply and record cleaning rules, such as ranges, allowed categories and date formats.
Map the data
Match each variable and answer value to a valid, standard OMOP concept, then have domain experts confirm the result.
Search OMOP Concepts Search first. Another cohort may already have mapped the same term.
No suitable Singapore mapping? Search ATHENA, and use Usagi to map a long list of terms in bulk.
Have a clinician or domain expert review the mappings.
Send missing or questionable mappings to NDCT.
Extract, Transform and Load
Extract your data, transform it into the OMOP CDM v5.5 tables, load it, then check its quality.
Set up an OMOP CDM v5.5 database and load the OMOP vocabularies.
Design the process with WhiteRabbit and Rabbit-in-a-Hat, then build and run it.
Run the OHDSI Data Quality Dashboard and fix issues until you reach a score of at least 95%.
A hands-on tutorial for this step is coming soon in the Members' Portal.
Maintain, share and reuse
Keep your CDM current and share what you have mapped so the next cohort can build on it.
Refresh the OMOP vocabularies and the Singapore harmonised OMOP vocabulary when new releases come out.
Version your mappings and transformation code, and rerun the quality checks after each change.
Contribute confirmed mappings back to NDCT through the NDCT common mailbox.
Multimodal data
Genomic data mapping
(coming soon)
Imaging data mapping
(coming soon)