OHDSI Home | Forums | Wiki | Github

Loading claims related data to CDM

(Abhishek kumar verma) #1

We are planning to load the claims data to the CDM , we dont find any table related to claims where we can load the claims data like claimid,claim type , claim amount and the policy details like policy id , policy description etc in the CDM database .We are thinking of putting the claim amount in the cost and the claim related details in separate table , is that fine . Any thoughts where can we load these columns in CDM database?

(Ben Holmes) #2

The Cost table is a long version where you can out concepts of copay, coinsurance, deductible etc. The visit table can handle some of the claim data. There’s also a diagnosis and procedure table to link to visits. It is not structured like we normally handle our claims data but it fits better for analytical function, in my opinion.

(Qi Yang) #3

Why do you need fields like claim id, claim type, policy Id and policy description?

(Abhishek kumar verma) #4

Thanks , we need to store the claim settled date , claim approved amount ,room details , denial clause , approval remarks etc . Is it possible to store it in CDM by adding the column to the existing table ? We need this to run our use case .

(Christian Reich) #5


The OMOP CDM is for analyzing patient data. It’s not a claims management system. If you have a claims database, why would you want to OMOP it? If you intend to do observational research you need none of those details you mentioned. If you want to run an insurance you don’t need the OMOP CDM. So, feel free to add any tables and fields you want, but the standard methods and tools will not see them. In other words, you are outside the standard if you do that.

(William Johns) #6

We would like to offer our viewpoint that integrating complete and detailed claims data to the OMOP CDM along with the patient data would offer tremendous value. It seems that every element of the healthcare industry at a macro and micro level is looking to find the connection between effective care and acceptable cost. Unfortunately, a good understanding of how care and cost are related is elusive. In a scenario where observational analysis of clinical data reveals no differentiation of outcome from two alternative therapies, the story does not need to end there if we find that one therapy is much more costly than its alternative. Another scenario might find in the clinical data an innovative and effective therapeutic approach that generates high level of claim denials. Both of these findings have medical, commercial and public policy value.

We are new to OHDSI, but are very impressed with what we have found so far. Our current intention is to use the CDM as a base layer in our dataset and supplement it with our detailed financial tables. If there are other participants who are on this journey, we would be open to comparing notes.

(Christian Reich) #7


Hm. Not sure I understand. OMOP has a cost table, which can be used to calculate cost of therapy outcomes. And folks are doing that kind of work, but currently not a lot. Would be wonderful if you could contribute use cases and questions that the community could engage answeriin ng.

Claim denials are indeed not possible to study right now. And actually, I would find it tremendously valuable if we could take it further and discuss a model of claiming rules. What is allowed to claim, under what circumstances, in what combinations, with what timing. There is a huge gap here. It would help to determine what code is available with what performance characteristics (sensitivity, specificity), which right now we have no systematic way of knowing. However, unfortunately this information is completely obscure. It does exist, but not in one place, machine-readable, for all aspects of medical care. Medical coders in the hospitals know it from special consultancy groups and a myriad of PDF publications the CMS or payers put out. But again, we have no systematic access. Any good idea how to tackle that?

(Andrew Williams) #8

Welcome @Bill_Johns!
If I understand correctly, your interest in OMOP and OHDSI might have a few pieces to it.

  • standard representation of claims
  • standard representation of costs associated with claims
  • standard analytics for cost-effectiveness

The first - representation of claims - is the area where the most prior work has been done. You should be able to build on prior claims ETLs to some degree in mapping the data you have to OMOP. Since NPN seems to have claims from multiple sources, there might be needs that aren’t already covered by that prior work. Others who are bringing their data into OMOP from similarly complex sources might want to partner with you on this. See this recent thread calling for collaboration on those issues: Mapping multiple payers claims files into OMOP. And there are several very experienced ETL firms in the community that can assist you.

Re the second piece - representation of costs: @Christian_Reich’s list of improvements and call for use cases for costs is excellent. You could help drive a set of improvements like those he describes or others that your use cases require. Outcome Insights’ @Mark_Danese and @jenniferduryea led some of the prior work on costs and might be worth reaching out to.

Re the third - standard analytics: this is a great community to work with in evaluating and standardizing analytic approaches. The methods library and the workgroups on population level estimation and methods evaluation are important resources. Adding cost effectiveness packages to the Methods Library would be a great contribution!

If you and NPN want to take a leadership role beyond development of informatics and analytics, you might look to build the OHDSI community’s partnerships with important entities that need to do this kind of work. The OHDSI community’s data standardization, curation, and standardized support for best analytic practices make it an excellent partner for entities like CMS and the FDA that need reliable evidence for their regulatory decisions.

You could keep that tractable and focused by working on a clearly defined area where this need is obvious and the amount of work is manageable. For example, coverage of genomic tests. The decision last year by CMS to begin coverage of FDA-approved genomic testing panels increased access to those tests here in the US. Their uptake and benefits aren’t well understood. As noted in this paper there is a well developed model for using new coverage decisions like this to develop the evidence needed to evaluate cost effectiveness. That coverage with evidence development model has long been espoused by Sean Tunis and others at the Center for Medical Technology Policy. It is highly regarded, but one reason it hasn’t been widely adopted is that it requires a set of partners with the ability to efficiently and reliably assess cost-effectiveness. Such partners are scarce.

OHDSI might be able to fill that gap. There is excellent work in the Oncology Genomics subgroup, building out the capture of genomic panel data. There is excellent work on data curation in the Data Quality Dashboard, and ACHILLES tools and workgroups. There is excellent work on study design and analyses in the methods code libraries and workgroups mentioned above. You could complement these with work on costs and cost-effectiveness analysis to meet CMS’ evidence development needs.

In short, you might be able to develop NPN’s business by building a collaboration between CMS and OHDSI for coverage with evidence development in a narrow area like genomic tests. To investigate that possibility you might seek out other private, public, governmental, or academic groups in the community who are interested in this and have the relevant data or expertise and start a workgroup.

That’s all just food for thought - one person’s take on the opportunities a group like yours might have in collaborating with OHDSI. Good luck.