OHDSI Home | Forums | Wiki | Github

If we have order information and administration information for drugs should we bring in both rows?

We have several sources of information for our drug exposure table, they can be divided into drug orders, drug administration records and drug dispensing records.
There is overlap between these tables - for example all administrations will have a drug order associated with them.

We are trying to decide between two options:

  1. Should we bring in all the information for every drug exposure (i.e. one row for the order, AND one row for the administration)?

We can use the drug_type_concept_id to indicate whether the provenance is an EHR order or EHR administration, however if users do not use this filter then it may seem like a patient has received a drug twice. I am also curious whether the network studies take drug_type_concept_id into account?

  1. Should we bring in only one row per drug exposure. We would preferentially bring in administration information when we have it (as this is more accurate to what was actually delivered) and then order information when this is not available.

Happy to provide more info or examples to make this clearer! :slight_smile:


Most of the time (as far as I know) studies do not take status into account and, unless you impose a gap between drug exposures in your study, you can indeed accidentally count the same drug exposure twice. So a very good option is #2 where you have one row, which, I believe, is also the most common practice.

great, thank you!