OHDSI Home | Forums | Wiki | Github

How would it be possible to record a history of medication (for example) of which all the data is known except the date?

Dear experts,
I request support for this problem; thanks in advance.
Assume the following situation (CDM 5.4):
A patient goes to the emergency department on a certain day (d), and the following information is collected: less than three months ago, the patient was prescribed an antibiotic (amoxicillin, 1713332) with a certain dose and for a certain number of days, but the date of the event (drug_exposure) is unknown.
The solution I propose is:

  1. use the concept observation_concept_id = History of event within three months (1340222) + value_as_concept_id = antibiotic therapy (4085730), and observation_date = d
  2. link the previous observation event with a new drug_exposure event that registers the antibiotic-amoxicillin with its amount through the observation_event_id/ observation_event_field_concept_id fields)
    Problem: what date should be assigned to the new drug_exposure event? Obviously it couldn’t be drug_exposure_start_date = d

Another solution could be:

  1. use the concept observation_concept_id = History of event within three months (1340222) + value_as_concept_id = amoxicillin (drug, 1713332), and observation_date = d
    But there would be no possibility to indicate the dose

Do you propose any alternative?
Many thanks in advance
Mario

Hello @mpascualc,

The OMOP CDM doesn’t have conventions on how to store dose information for drugs which have an unknown date or drugs which are past events. What’s your use case for storing these data? Is dose necessary for analysis? Or do you want to store the dose because you have these data at the source?

Dear Ms. Philofsky,
It is the second situation: both the antibiotic ingredient and dose data are available in the emergency records, but not the date of prescription (only “within the last three months” is indicated). I recognize that the OMOP scheme should not be “forced” to include all the available information, since it can generate situations that, at the very least, are inconsistent. My intention was to raise this question in case there were some relationship-based mechanism to accommodate the data.
Many thanks!

Either of your #1 options will work for these data. However, the first #1 option will be most easily understood by the end user of these data since it reads “History of event within three months…antibiotic therapy”. Your other feasible option reads “History of event within three months…amoxicillin”, but the end user might ask, what is the history for “amoxicillin”? It is not quite as clear.

Even though you can shoehorn these data into the OMOP CDM, I would not bring them in at this time without a use case. The use case informs how data are inserted and retrieved from the CDM. Without this knowledge there is a high likelihood you spend resources putting them into the CDM and they will never be used. Then, later on a use case comes along for these data and you have to take these out of the CDM and then put them back in with the new method. Speaking from experience, this is complex and at high risk for error.

I completely agree with you, and I greatly appreciate your comments.

You’re welcome!

t