OHDSI Home | Forums | Wiki | Github

OHDSI cohort definition: how to constrain drug durations


(Chante K) #1

Hello OHDSI community! We are in the process of building our cohort definition studying antibiotic use for acne (titled “Antibiotics for acne”). We were hoping to get some guidance regarding how to limit drug durations separated by less than 28 days to be considered part of one drug course. Has anyone come across this in their cohort building? Thank you!


(Anna Ostropolets) #2

If you use drug_era, your events are already collapsed if the interval between them is less then 30 days.
It works on the ingredient level, so you don’t need put all drugs in your Karimkhani- antibiotics test set, just the ingredients; days supply criterion will be transformed into era length.


(Chris Knoll) #3

Hi, this is done by:

Cohort Entry Events: select drug_exposures of interest (make a concept-set of antibiotics for acne).
Specify that you will use all events per person.

Inclusion Criteria: Apply these only if you want to restrict those Cohort Entry events in some way, ie: person had a prior acne diagnosis, for example.
Specify that you will use all events per person.

Cohort Exit: this one is conditional based on your data.

  • Do you have accurate drug_exposure end date? then set your cohort entry event to persist for: fixed duration relative to initial event, specifying the offset from the event’s end date, plus 0 days after (this will use the drug_exposure’s end date).
  • Do you not have accurate drug_exposure end dates, but want to assume each exposure should assume to last 28 days? then specify event persists for: fied duration relative to initial ivent, specifying the offset from the events start date, plus 28 days after.

For Cohort Eras: specify a gap window of 28 days. Any drug exposure you identified will be ‘collapsed’ together into eras, separated by the 28 day gap between drug exposures.

Note, this allows a person to have multiple periods of acne exposure in your cohort. In your analysis, you can restrict it to just using the first cohort era per person, if that is desired.

@aostropolets offers a good suggestion about just using the drug_era table, but the benefit of the records there being at the ingredient level could be an advantage or disadvantage. The advantage is that it’s easy, and you just query directly into the table (tho, you must remember to collapse overlapping drug_eras in the case you pick multiple ingredients to look for, and a person has overlapping drug_eras for the different ingredeints).

The disadvantage is that you don’t have control over the gap size (all eras were built off the same 30 day gap window). You also aren’t able to start your cohort during the middle of the drug_era. If you need that, it would be better to select drug_exposures and built cohort eras from that.


(Chante K) #4

Thank you so much @aostropolets and @Chris_Knoll!


t