OHDSI Home | Forums | Wiki | Github

Cohort Pathway Analysis - ATLAS


(Shinyoung Ju) #1

Hi smart peeps,

I was using Cohort Pathways function within Atlas, and I was wondering if you know what it means by “Target cohort”. We already have the number of people who have been included in the analysis as “persons with pathways vcount”, but was not sure what exactly “target cohort count” could mean.

Thanks in advance for your great support! :slight_smile:


(Chris Knoll) #2

The pathway analysis will only report on people that have at least 1 event cohort occurrence within the target cohort’s start and end date. So you could have some people that are in your cohort, but did not have ANY of the cohort events you specified in your pathway analysis.

Therefore, the ‘target cohort count’ = number of people in your cohort, and ‘persons with pathways count’ = number of people from your target cohort that had at least 1 path.

In case you are wondering ‘why?’: it is possible that you could have a large cohort where only 5% of people have at least 1 path. If we were to represent the 95% in the visualization, you’d have this 95% hole in the graph, with 5% of the remainign space dedicated to the actual data contained in the pathway analysis. We decided that showing that hole is not useful, and instead annotated the result with ‘People in Cohort vs. People in Pathway Results’.


(Shinyoung Ju) #3

Hi Chris,

Thanks for your great answer!
I partly understood your logic.
I still have a question: I already selected the events as the entry criteria when I built the target cohort. For instance, I crated a cohort and set the index date as the first prescription of the drugs (events). Therefore, target cohort only contains those who had at least one prescription of the events of interest. So what I thought was that the results should contain all the patients from the target cohort.
Is there any minimum/default duration of the therapy you specified during the development of this tool?

Thank you!!!


(Chris Knoll) #4

Sounds like a bug: if your target cohort starts at the earliest event of any of the events you are specifying for your event cohorts, then it stands to reason that you should see 100% of the people in your target cohort represented in the pathway analysis. Could you open an issue on github.com/ohdsi/webapi?

(Shinyoung Ju) #5

Thanks for your answer!
I will open an issue on GitHub page.

Really appreciate your support.