OHDSI Home | Forums | Wiki | Github

Resources for the local use of the OMOP-CDM

I would like to know if there is already a place to share experiences, tips, code, and tools related to using the OMOP-CDM for local purposes, rather than for federated purposes.

We agree that federated analyses are more powerful for answering research questions than local analyses. We also understand that most efforts around the OMOP-CDM are focused on facilitating federated research.

However, based on our experience—and likely the experiences of others—there are many cases where the OMOP-CDM and its surrounding ecosystem have been beneficial for local purposes, such as data quality assessment, harmonization of local data, and local data analysis.

Moreover, I believe that encouraging the use of the OMOP-CDM for local purposes can help onboard users for federated research. For example, a researcher might begin by using OHDSI tools on their local database to address a local research question, and then realize how straightforward it is to scale the same question to a federated research context.

If there is already a forum for this type of discussion, please let me know. If not, I would like to initiate this discussion here.

1 Like

Our Experience:

FinnGen is the largest academic-industry partnership in Finland, encompassing genetic data and multiple health registry data from over 500,000 individuals. Several pharmaceutical companies and academic institutions use this data in our secure research environment (SRE). We have dedicated teams at FinnGen focused on making our users’ lives as easy as possible by maintaining the SRE, preparing the data, and providing analysis and visualization tools.

The OMOP-CDM has benefited us in the following ways:

  • Health registry data comes in various tables and formats. We have transformed (ETL’d) them into the OMOP-CDM. This allows users to bypass the need to understand each format or figure out how to combine them.
  • The Data Quality Dashboard (DQD) helped us identify a few issues that escaped our raw data quality checks.
  • ATLAS is used daily by our non-technical users to create cohorts without needing coding skills. Finnish researchers can continue using familiar source codes, while international researchers benefit from the standardized OMOP concepts.
  • We have built custom tools that use Atlas cohorts as inputs to create analyses based on standard concepts, source concepts, or custom genetic data. For example, a user can create a pair case-control cohort in Atlas and run a GWAS with just a few clicks.
  • Most importantly, as new registries are added to FinnGen, we can seamlessly incorporate them into the OMOP-CDM without having to modify all our tools to accommodate the new data, which is a tremendous time saver.

Our Tips, Codes, Tools:

To make the OMOP ecosystem more compatible with the use of source concepts, we had to develop some hacks, codes, and tools. Some we learned from the community, and others we developed ourselves. Most of these can be helpful to others, and while some may already exist, we couldn’t find them, which is one of the motivations for starting this conversation.

  • As part of FinOMOP, we created several custom vocabularies following the Concept & ConceptRelationships methodology. This allows us to use source concepts in ATLAS and other tools. See Skuharuskaya et al.
  • In fact, there are so many custom vocabularies that we created an R tool (FinOMOP/ROMOPMappingTools) and a repository (FinOMOP/FinOMOP_OMOP_vocabulary) to manage them. Unfortunately, the repository we use in FinOMOP to maintain our vocabularies is currently private, but we will open it soon. For now, there is a demo version available: FinOMOP/FinOMOP_OMOP_vocabulary_test.
  • We extended the Concept & ConceptRelationships methodology to include descendants in the Ancestor tables, making it much easier to create source concept sets in ATLAS.
  • We added person counts (PC) and descendant person counts (DPC) for source codes to Achilles. This allows us to view these values in ATLAS we follower this.
  • We created some R functions to get source concepts as covariates in the FeatureExtraction package. This makes the Hades packages compatible with source concepts. See FINNGEN/HadesExtras.
  • We built a Shiny app to take cohorts from ATLAS (or other sources, such as text files) and run Hades or custom analyses with just a few clicks. See FINNGEN/CohortOperations2 and our showcase at OHDSIGlobal2024.

Hello @Javier!

You should join the Healthcare Systems Interest Group (HSIG)! We are hosting a speaker’s series focusing on OMOP for non-OHDSI use cases. We had two speakers this fall and will continue the series after the global symposium. We’d love to have you share your work! Please reach out if you are interested.

A not very well kept secret, the HSIG will reveal our new website at the global symposium. The website will house the many artifacts created by the health systems group and the OHDSI community specific to supporting health systems on their OHDSI or observational health journey.

2 Likes

(thanks @MPhilofsky, I just learnt about that in yesterdays call, you talked it just before me. We are doing very much the same and would like to share our experience and learnt from others. Lets have a chat in OHDSI2024)

1 Like

Looking forward to connecting, collaborating and learning more at the OHDSI Symposium next week :slight_smile:

t