Hi there,
I just started at my institution and am in conversations with the bioinformatics specialist at the cancer center of my university. Our EHR data is in EPIC and we are interested in converting it to OMOP CDM. The bioinformatics specialist told me that her understanding is that prior to the conversion, we have to first hold all of our EHR data in a large PHI-secure cloud space.
Could someone verify that understanding is correct? If so, how do others overcome the upfront cost if the cancer center is unable to cover it?
You can store your source data in the cloud or on-prem. The location is generally dictated by the institution and in compliance with their policies and procedures.
I hope others will answer this question. Here is a white paper OHDSI’s Health Systems Interest Group wrote a couple years ago to help convince CFO’s to support an OMOP CDM conversion and implementation.
I’m unsure if OHDSI has documentation for exactly this because varies between institutions. Many institutions create another dataset in the same location as their source data. Other times a new database needs to be setup within your environment. It really depends on what your security and compliance department wants.
Do you have EHR data and Cancer registry data? Where are the data currently located? I suggest you talk to the bioinfornatics specialist and see what they are doing with their source data. Also, check with your research department. This varies between institutions and is wholly dependent on what is acceptable for your institution. The first step is to figure out what requirements they have for storage of PHI data.
Many ways to do this @Cindy_Hu! I know of several who use dbt with various databases of choice with an on prem database. (I am running a team doing this in London) Pros and cons to each - worth doing some research first!
Thanks very much Melanie! I will talk to the bioinformatics specialist about this. It is good to know that there is at least some options for keeping the data on prem.
There are a lot of resources available to you to help you get started on your OMOP journey, especially as an Epic site! If you have access to Userweb, you can take a look at the Corwell-Voyager project as well, which was critical for us to reference as we (Emory) moved from Cerner to Epic. It significantly reduced the burden of implementation for us, and thus costs.
I’m not going to hit you with the full “OHDSI Fire-hose” of information and processes, but know that we’re a very active community who want to see new sites succeed! Just one example of getting in the weeds, is to mention that the oncology workgroup will certainly be able to assist with cancer specific questions, and have a variety of standards to help sites with tumor registry data get their data ingested into the OMOP-CDM in a standardized fashion. This is once things are implemented though.
Regarding cost, to mirror Melanie’s answer, it will be drastically different depending on the cloud vendors your enterprise is tethered to. A small manageable proof of concept may assist you in determining utility. When figuring out the POC though, as Christian Reich, and many folks in the community often mention (it’s just that I’ve just heard it from Christian across workgroups a TON): “What is the use case X is solving?” Your use case might be very similar to how we started at Winship Cancer Institute at Emory University, where we focused ONLY on ingesting cancer patient data. Once Enterprise got onboard, the OMOP pipeline started to ingest ALL of the data required at scale. At first, we were just able to export data to REDCap projects for researchers in a standardized and reproducible fashion, which was what was most needed at first. Being able to define and share cohort definitions across the OMOP community was also a significant boon for us as we learned to identify patient cohorts in our data.