OHDSI Home | Forums | Wiki | Github

Adding second painkiller (2 overlapping drugs for 90+ days) - as Atlas cohort criterion


(Vojtech Huser) #1

We need a consult for a Atlas cohort definition.

Imagine a scenario where one pain killer is not enough for Osteoarthritis (OA) pain and a researcher needs to go after severe osteoarthritis (to compare patients with knee replacement; T) with truly severe cases of OA(as C). IC stands for Inclusion Criterion.

Option 1a (eras)
Initial event : NSAID painkiller (set1 of ingredients) is defined and using DRUG_ERA
IC1: second stronger painkiller (set2 of ingredients; not overlapping with set1)
see http://atlas-demo.ohdsi.org/#/cohortdefinition/1774831
IC2: for Dx prior index (or at index)

Option 1b (eras but different order (winner so far)
(variant of 1a but: strong painkiller is initial event and mild painkiller is in IC1
(we would prefer to have index date the date of the intensive therapy

Option 2 (dx as index)
Index event is onset of Osteoarthritis (Dx onset index time).
IC1 and IC2 for painkillers

Option 3 (drug exposure)
well, eras would be best but did others have similar problem ever?

Has any one encountered 2 concurrent drug criterion before and what was your approach? (or any advice you have for us would be appreciated)

In fact, alternatively, we would enumerate one large superset of ingredients and we would look for 2 concurrent ingredients from the superset. (I would think currently either not supported by Atlas or requires power user help) @Chris_Knoll ?


(Chris Knoll) #2

If you want your cohort start date to be when they initiated the intensive therapy, then it makes sense to have your initial events based on the strong pain killers.

If you are saying that the membership in the cohort depends on having a prior mild pain killer, then that’s your inclusion criteria.

If you want an overlap of 90 days, then you’ll say that the mild pain killer era ends 90 days after the major pain killer starts. Note: this is now using future time for your inclusion criteria (ie the person must exist for at least 90 days after index) so if you have any survival analysis, you run the risk of immortal time bias. This may be acceptable depending on your analysis.

note: this will only consider the mild/major pain killers at the ingredient level: you may want all mild pain killers get grouped together into ‘mild painkiller era’ and the major pain killers get grouped together into ‘major pain killer eras’ and then find the overlap with the combined ingredients. This type of operation is beyond the scope of Atlas capabilities.


t