OHDSI Home | Forums | Wiki | Github

How to fill "quantity" field from "drug_exposure" table?

I am writing a ETL script to fill the “drug_exposure” table, and got stuck on the quantity field.

The documentation (OMOP CDM v5.3) states that I need to first find the drug form, so if the drug is a pill, quantity should contain the number of pills prescribed, if it is a injection, the volume and so on.

The first doubt is: what are the accepted values of “relationship_id” ? I’m currently using the values:
(‘Has dose form’, ‘RxNorm dose form of’, ‘NDFRT dose form of’, ‘Has disp dose form’,
‘Disp dose form of’, ‘Has dose form unit’, ‘Has dose form group’, ‘Has basic dose form’,
‘Basic dose form of’, ‘RxNorm has dose form’, ‘NDFRT has dose form’ )

The second doubt is, there is more than 800 values in concept_class_id = “Dose Form”. Should I go one by one, deciding which ones are “pill-like” or “liquid”? I’m hoping there is another, easier, way.

thanks!

1 Like

@djogopatrao:

The relationship_id is ‘RxNorm has dose form’. Only Dose Forms from RxNorm or RxNorm Extension count (there are some 250 of them). And no, you don’t have to go through one by one. Just go into DRUG_STRENGTH, find the drug product and see if there is a denominator_unit_concept_id != NULL. If so, it is a divisible Dose Form, otherwise it is a fixed one.

Hi Christian

Thanks for your answer. But if I can find out if the dose form is divisible using the DRUG_STRENGTH table, do I need to follow CONCEPT_RELATIONSHIP to know the dose form in order to fill the quantity column?

No, you don’t. DRUG_STRENGTH will tell you how to interpret quantity:

  • amount_denominator != NULL: fixed
  • denominator_unit_concept_id != NULL: divisible
  • both NULL: ingredient.
  • Standard concept but not in DRUG_STRENGTH: probably Drug Form, treat like Ingredient

Makes sense? We need to document that, but we are in the middle of overhauling the entire thing anyway.

I see. Thanks again!

I have other doubts on drug_strength, but I’ll create a new topic just for it.

Cheers!

t