OHDSI Home | Forums | Wiki | Github

Atlas - capture exposure at frequent intervals

estimationmethods
atlas
cdm
vocabularies
webapi

(Akshay Kumar) #1

Hello Everyone,

I read the Atlas tutorial [here] (https://ohdsi.github.io/TheBookOfOhdsi/CohortDefinitions.html#ACEiUse) and watched a YouTube video of the Atlas tutorial as well.

I am trying to create a cohort in Atlas to identify patients who are diagnosed with Hypertension and have an observation period of 1 year and a follow-up of 3 years. For which, I did the below

Now from the patients identified above, I would like to see who all qualify for the event of having ACE/ARB.

Basically, Target cohort is Hypertensive patients… The event cohort is ACE/ARB

But the problem/where I struggle is how do I capture people who had continuous exposure of ACE/ARB every 60 days (until the follow-up period)… I only want these people to be in the event cohort.

For example, I would like to identify patients who have been on continuous exposure to hypertensive drugs every 60 days (during the follow-up period of 3 years).

I created something like the below but I believe it will only capture people who had ACE\ARB in the 1st 60 days (of their follow-up)…

But I want to select people who had at least one exposure of ACE\ARB in 0-60, 61-120, 121-180,181-240, etc…till the end of 3 years.


(Chris Knoll) #2

So just do that. Forget the ‘event cohort’ you can’t make a cohort definition depend on another cohort definition. But you can start with your Hypertension cohort, and then create an inclusion rule which states:

  • has at least 1 exposure to ACE\ARB between 0 days after and 60 days after
  • and has at least 1 exposure to ACE\ARB between 60 days after and 120 days after
  • and has at least 1 epxosure to … between 120 days and 180 days
  • and … repeat as needed

This will give you a cohort, but it is a very specific use case.

Instead, maybe you should think in terms of a cohort intersection, where you have your Hypertension cohort (as a sort of ‘incidence’ cohort) and a ACE\ARB exposure cohort which is constructed from ace\arb exposures allowing for a specific gap window:

Cohort 1: diagnosis of hypertension (prior/post observation doesn’t really matter here)
Cohort 2: Exposure to ACE\ARB, allowing for 30 day gaps to create continuous exposure episodes

The missing part from atlas is doing custom operations with cohorts (such as cohort intersection). You’ll have to drop into custom SQL for this, but at least the ‘hard part’ of building the cohorts is done in the tools. To get the people who have at least the 3 years of continuous exposure:

select c.subject_id -- the personID from your condition cohort
from resultSchema.cohort c
join resultSchema cohort d on c.subject_id = d.subject_id -- match on personID
where c.cohort_definition_id = @hypertensionCohortId -- your hypertension cohort ID
  and d.cohort_definition_id = @exposureCohortId -- your ACE\ARB cohort
  and d.cohort_start_date <= c.cohort_start_date and d.cohort_end_date >= date_add(d, c.cohort_start_date, 365 * 3) -- only find people who have exposure episodies starting before the diagnosis and ending after 3 years

That will give you the set of Patient IDs that have 3 years of continuous exposure. The reason why i don’t think you need to worry about ‘prior/post observation’ is that you won’t actually get anyone without the 3 years of follow up because you are inferring the followup by requiring 3 years of continuous exposure.


t