OHDSI Home | Forums | Wiki | Github

DRUG_EXPOSURE conventions


Then put the Ingredient into the drug_concept_id field, and the amount into the quantity field.


@Christian_Reich, and Ingredients should have units in drug_strength then?


Correct. It’s being put in place. Next release.

In this case quantity field could contain mixed up information. For example, there can be drug that was mapped to ingredient and there was no other information but qty in source data (let’s say 5 pills), and also the same drug with qty 5, but with dosage info (let’s say 30 mg). At the same time it may be not possible to find RxNorm concept with exact dosage (30 mg) if there is no such concept. What to do in such case?


If you don’t know the product, and all you are putting in is Ingredient, the quantity is about the compound matter, not some units. So, in your case 30 mg. Make sure you put in the total of the entire exposure. So, if there is a daily dose of 30 mg over a month you need to put in 900 mg.

@Christian_Reich, all
we have cases similar to this

when we can map only to Clinical drug form or Ingredient levels, so in this case units will be defined by the drug_strength table. And we have some cases when the drug dosage can be represented both in MG and Units. In drug_strength we have MG defined, while in our source data there’s Interantional Units ([iU] or [U]), and exact dosage corresponds to the units indeed.
@Alexdavv, can you please provide examples.

  1. nystatin.
    mg is predefined as a unit in DRUG_STRENGTH table. In RxNorm there are no any Clinical Drug Components with mg.

  2. bacitracin/ bacitracin zinc.
    mg is predefined as a unit in DRUG_STRENGTH table. In RxNorm there is just only one Clinical Drug Component with mg - bacitracin zinc 0.5 MG/MG which is used just for one single Clinical Drug - bacitracin zinc 0.5 MG/MG / Polymyxin B Sulfate 10 UNT/MG Topical Powder.

  3. Penicillin G (almost the same case like bacitracin).

And rather important that in general clinical practise these drugs are measured mostly (sometimes always) in iU. The main question: what units should we use for figures which would be inserted to the DRUG_EXPOSURE.quantity field? If source data contain all the quantities in iU should we convert them to mg and basically how?


Right now, we allow one unit. The ETL would have to convert.

But the vocab team should help by putting out the conversion factors. Let’s figure that out.

What about the other case:

We have drug administration with the relative units like 20 mg/kg (of patient weight). And we don’t have the weight of the patient to calculate the real dosage.

So we map this to the Clinical Drug Form. And there’s no way to reflect the dosage in drug_exposure table.

You really don’t have the dose, then.

But I understand that this sometimes is all you got. And for some drugs the per kg is the important information. Here, I am with @schuemie’s “let’s just add a field”. This is a piece of information we cannot store right now without re-defining the content of other fields.

@Dymshyts: Can you put a proposal into the CDM Github? Something like “dose_per_kg” with allowable values NULL (usual) and Y.

You might consider being more flexible on the units. The denominator can be weight, body surface area, and maybe 1 or 2 other things I can’t recall at the moment.


True. In that case we could add a “alternative_denominator_unit_concept_id”.

That column name would not fit in an oracle database.

Urgh. Forget the denominator then. alternative_unit_concept_id?

We (at least here) call the normalized dose the “dose basis”. So maybe dose_basis_unit_concept_id? Or maybe that’s thinking too focusedly . . .

What about ‘rate’? so exposure_rate and exposure_rate_unit? Seems like we’re talking about 2 things possibly happening when a patient is exposed to a drug: they get a dose of something or they are exposed at a rate of something. I found this discussion here, tho it was describing radiation, but the idea might apply to our drug exposure table:

So, this way, you can have someone exposed to the same mass of a drug as another person, but the rate could be different between them. Maybe we can standardize the units to either by time (to a day) or by weight (to a gram),

looks like the discussion goes to the relative units only

@Christian_Reich, another example where we have problems is when the substance can be present both in iU and mg.
and conversion is not so easy in this case.

and it’s not so easy to convert.

@Alexdavv, I know you tried something, can you share?

Not a bad idea.

Yes, but we already were talking about two things before: Absolute unit based (=dry) dose like in tablets, and divisible dose (=liquid). Now we are talking a third thing: the first derivation thereof. We haven’t made those distinctions canonical, except when we want to calculate daily doses (e.g. for DOSE_ERA records). That’s when we need the complex @Klaus-like logic. This will add to that.

We could try figuring it out. But won’t be easy on the ETLers, because instead of just dumping NDC codes they will have to start getting into the detail.

Correct, but I would definitely not open the door for that. In contrast, in the long run we should provide means to convert. The conversation here is about what dose is actually given. Not how it can be differently represented.

That is the idea, @alexdaw. We will fix it.

+1 for @bailey’s idea

And this would be the concept_id for the dose_unit_source_value field, correct?

I’ve added a proposal to create a dose_unit_concept_id field for the dose_unit_source_value field here

@Bailey I’ve changed the field name that you proposed to stay in line with naming conventions. Keep the concept_id name the same as the source_value name

@Christian_Reich @clairblacketer Would this be considered a bug fix since the source value didn’t have a corresponding concept id field?