OHDSI Home | Forums | Wiki | Github

CDM change proposal: Drug era: add route_concept_id column

CDM Builders
#1

Dear CDM builders,

This is to initiate a discussion on the proposal to add a new column to DRUG_ERA table.

The proposal is described here:
http://www.ohdsi.org/web/wiki/doku.php?id=documentation:next_cdm:eraroute

Some drug ingredients, like neomycin, can be taken oraly and localy (otic, ophtalmic). Such drug eras are medically distinct. The table should distinguish somehow between them.

  • If you fully agree - just like this first post.

  • If you agree with the idea in general, like this first post. (and
    provide modifying comments to the proposal)

  • If you disagree (reply indicating oposition) and discuss the details why it is a bad idea (or burden).

#2

I disagree. The DRUG_ERA table is intended to store periods of continuous
exposure to active ingredients. It is not intended to tease out
differences in dose or formulation.

RxNorm clinical drug concepts, which are stored in the DRUG_EXPOSURE table,
do provide a mechanism via a CONCEPT_RELATIONSHIP in the vocabulary to link
to the dose/form for each product. But, there is not one concept for
‘oral’, rather there are dozens of formulations that one could reasonably
bundle as oral (e.g. oral tablet, oral capsule, oral solution, etc.).

If your specific analysis requires grouping concepts by dose/form, I
recommend you pull the appropriate records from DRUG_EXPOSURE and then
build up the eras as a custom script. The cohort definition in ATLAS
already enables this to be done, in case you are looking for guidance on
how to do it if it’s not being performed directly in ATLAS itself.

#3

Because I can’t resist, I disagree with your disagreement. :grinning:

Swallowing levofloxacin for a week is a world of different exposure to the drug than putting drops in your ear, and an analysis that ignores that distinction is likely to reach, shall we say, different conclusions than one that accounts for it.

But I don’t think this requires a change to drug_era. Since the exposure is keyed to drug_concept_id, and RxNorm hierarchy works well here, my recommendation is to roll up to SCDF for most biological exposure applications. That way you capture the most important distinctions in extent of exposure. (If a particular query really just cares about neomycin in the neighborhood (e.g. the outcome is anaphylaxis), then the query can expand an IN to SCDFs.)

The biggest exception to this I can think of offhand is that your principal use case is product safety, you may want to keep drug_era and dose_era down at package-level codes, so you can get that granularity in queries where you need it.

#4

The DRUG_ERA table is intended to store periods of continuous exposure to active ingredients. It is not intended to tease out differences in dose or formulation.

The sentence claims that formulation is out of scope, but would drug route be also out of scope. Oral route vs. rectal route vs. intra-eye injection? (e.g., bevacizumab systemically vs. into an eye only)

If drug_concept_id has all the meaning (in DRUG_EXPOSURE), why do we have the column route_concept_id in the exposure table. One could say it is redundant. (and maybe that is why it is optional there and not required).

But we have some non-dispensation drug data and for those me may have drug_route data.

If our CDM justification logic allows route_concept_id in the DRUG_EXPOSURE table, I see absolutely no reason why the same justification is not applied to the same idea in the DRUG_ERA.

If all my rows in drug_exposure have route data (optionally), I see no reason to drop that information in the era table. I think neomycin oraly and systemically is a different “cup of coffee” than neomycin in my eye locally.

In other words I am proposing extending the scope of the era table.
I am arguing that dose and formulation (topical gel, topical oil, topical spray are out of scope for era but I propose to to bring drug route (“apply on skin” (and not inject into muscle)) into the DRUG_ERA scope. (extend the scope for one of those three medical terms (dose vs. formulation vs. route).

Especially because it is a derived table and we have the knowledge to derive it well or copy if from the exposure table. It will increase the utility of the era table for analysts.

t