OHDSI Home | Forums | Wiki | Github

Two data sources - Usability of Atlas features

estimationmethods
atlas
workinggroup
patientprediction
cdm

(Selva Muthu Kumaran Sathappan) #1

Hello Everyone,

We have two data sources (Hospital 1 and Hospital 2). I mean EHR data from 2 hospitals. We would like to retain them in the same database.

Can you let me know whether the below scenarios are possible ?

For example,

  1. Will I anyway be able to distinguish data that they are from different sources? Is there any indicator or field in db that can help me figure out this?

Ex: Will I be able to run a query to find people whose Age > 50 only in Hospital 2 datasource.

  1. When I generate a cohort based on say (“Metformin”), will I get to know the number of people from Hospital 1 and Hospital 2 separately. I don’t want the output to be as 100 records. Instead I want as 50 and 50. Is this even possible? In Atlas UI, Will I get to see two reports (Attrition view/per data source. In total 2 attrition view for 2 data source)

  2. Let’s say I would like to compare data from both the sources. Does Atlas allow me to do this?

  3. Can you let me know whether this is possible or not? If it’s not, is there any other way to do this?

  4. This is not exactly related. But will I be able to create “odds ratio” in Atlas. I mean I didn’t see them in the Existing features under “Features” section.

I see that Atlas has the dropdown features to list down the datasources but would like to know whether the above scenarios are even possible and can you direct me to a resource in OHDSI forum if any?

Thanks
Selva


Randomized data - Atlas Cohort Generation
(Chris Knoll) #2

If you register those data souces as 2 separate CDMs, you would add 2 records to the source table (one for your Hospital 1, other for Hospital 2), and you will then generate cohorts and cohort reports on each source directly, and then by nature of the separate sources, you will be able to distingusih the data from one source vs the other. These sources will remain separated, and you will not be able to merge/combine cohort reports across sources (as the different sources are separate and should not be mixed).

If you want to combine them, then you could introduce a third source ‘Combined Hosiptails’ and run the same cohort and cohort reports on the third source and you will see those patient level data together. I can not tell you if it is appropriate to combine Hospital 1 and Hospital 2 data together into the same CDM. You must make that determination yourself.

Based on my proposed configuration, the output from Source 1 will be from hospital 1, and source 2 will be from hospital 2.

They will be on different sources, so you can generate the reports on a per-source basis.

The cohort generation result screen does show the cohort counts for each source, however there is not a way to see the inclusion reports (ie: the attrition report) ‘side by side’ across sources.

Atlas will show certain summary statistics across different sources, but in other cases (where it’s the convention to not show data side-by-side across datasources) you will only see one source at a time. Of course, if there is enough need in the community, the UI can be enhanced to allow for this.

Using a language like R, you can import data from each data source into an ‘in memory dataframe’ and do any merging/slicing/dicing you want with the data.

You could use generated statistics (such as counts of outcomes) to create an odds ratio by hand, but there isn’t currently a way to take something like an incidence rate from 2 different sources and combine them into a incidence ratio.


t