Atlas cohort builder has 2 functions (‘Cohort exit criteria’, ‘Censoring Events’) which use term ‘exit the cohort’ in the description.
But ‘Censoring Events’ also has further term ‘exclude the cohort’. It cases ambiguity in understanding of properiteis of these functions.
I think ‘Cohort exit criteria’ defines event(period) for determination of ‘cohort_end_date’ while
‘Censoring Events’ excludes persons with occurrence of listed events after index date.
Is it correct?
No, the cohort exit criteria is more of an ‘exit strategy’ where you determine how a person should ‘naturally’ end the presence in the cohort. The 2 options are to add a fixed number of days to the starting event’s start/end date, or you can create a custom drug exposure era and use the end date from the era to have the cohort end.
The censoring events are the ‘un-natural’ ends of the presence in the cohort. These events are used to kick the person out of the cohort when the specified events occur (first one that is found after the cohort entry date will be used as an exit date). You do not have to specify censoring criteria in the cohort definition.
I agree that the language for these terms should be clean up in the cohort UI. I’ll work on addressing that in the next release.
-Chris
So, if persons with some events after the index date should not be included to the cohort I have to define it as
(exactly 0 events after index) in initial event criteria while censoring events define cohort_end_date by any event concept
set (not only drug_exposure) but do not exclude these persons from cohort?
That’s correct.
The inclusion criteria will limit the people into your cohort, and you can limit them by saying 'include people that have 0 occurrences of X (could be any domain event type).
The censoring criteria is used to make people leave your cohort based on a given event, but they do not exclude anyone. The censoring criteria only applies to people that made it into your cohort.
Would it be less confusing if the initial events/inclusion critera was called ‘Enter criteria’ and those options for defining when a person leaves the cohort as ‘exit criteria’ make more sense?
May be. And exclude ‘exclude’ and ‘event prior to index’ from the note to censoring events
Thank you.
I think that it is also useful to define exit criteria as some period after index date OR any event (what will happen earlier)
You might want both, for example:
Exit at the end of the drug era of ACE inhibitors
Censor if there’s a drug exposure of (some other intervention treatment).
You can specify both exit strategy and the censoring events, because you might want to have both.
I’m not sure what you mean by the comment “exclude ‘exclude’”, but if you are referring to this note:
Note: censoring events occurring prior to cohort entry will not exclude the person from the cohort. Use inclusion criteria to filter the people that have the censoring event prior to index.
That is exactly the correct statement. Censoring events do not exclude people. Use an inclusion criteria to restrict people from entering your cohort.
1 Like