OHDSI Home | Forums | Wiki | Github

The prescribed drugs records beyond the Observation Period


Can we put in Drug_Exposure drugs, which are expected to be administered in the future?
Please see the explanation below.

We put source data on medications ordered and not directly linked to an drug administration to CDM Drug_Exposure with type concept_id equal 38000177 (‘Prescription written’ drugs).
These information hasn’t been excluded due to the our intention to keep all possible information. There are dates in future and here we allow future dates within a year.

The ‘Prescription written’ drug records dates were not included in Observation Period. And because of this, we have records in CDM Drug_Exposure having dates outside valid Observation Period. There are relatively few such records.

We’ve suggested that it is better to take these records too. Since we have a small dataset with two cancer diseases and it is important which medications and the line of therapy were prescribed for treatment or palliative care.
If necessary, the researcher may later exclude these records by drug_type_concept_id value.

I like this solution but please write any thoughts about it.
Whether it is acceptable and whether it is a good solution or not?
@Dymshyts @nzvyagina @ericaVoss @clairblacketer @Patrick_Ryan @Natalya_Gopkalo @Alexandra_Orlova @TBanokina

(Dmytry Dymshyts) #2

In short:
There are some drug entries after the date of datebase extraction.
Some are reallу scheduled in the future, some are just typos.
So, we put in drug_exposure prescription drug entries within a year after the latest date in the database.

(Tatiana) #3

Hi @Maria_ya,

My thoughts:

  1. The latest version of OMOP CDM v6 allows to have events outside of an observation_period;
  2. You mentioned that you pull only ordered drugs that are not indicated as administered.

If you have an information about the ordered drugs beyond the dataset year and the date on which the drug is expected to start - it seems as a good option to store such records.

P.s. if your dataset will get new data, you will need to check if the previously ordered drugs were administered.


Hi @TBanokina, thank you for the thorough and detailed answer!

Yes, these drugs are not indicated as administered. More than 45% records from source Medication Order table are not presented on source Medication Administration table (containing data on medications and amounts administered). Meanwhile when orders the same then almost all dates in these two tables coincide too. And there is no one record in Medication Administered table which contains order canceled in Medication Order table.

And yes if dataset will get new data we will check this point using drug_type_concept_id.