OHDSI Home | Forums | Wiki | Github

Questionable HCPCS codes in Procedure domain


(Rebecca) #1

I’ve come across a number of HCPCS/CPT4 codes that are assigned domain_id = Procedure, but I don’t believe they are procedures.

A few examples - shouldn’t these be in the drug domain?
90471 - Immunization administration (includes percutaneous, intradermal, subcutaneous, or intramuscular injections); 1 vaccine (single or combination vaccine/toxoid)
90472 - Immunization administration (includes percutaneous, intradermal, subcutaneous, or intramuscular injections); each additional vaccine (single or combination vaccine/toxoid) (List separately in addition to code for primary procedure)
S9490 - Home infusion therapy, corticosteroid infusion; administrative services, professional pharmacy services, care coordination, and all necessary supplies and equipment (drugs and nursing visits coded separately), per diem
96365 - Intravenous infusion, for therapy, prophylaxis, or diagnosis (specify substance or drug); initial, up to 1 hour

Examples - shouldn’t this be in the visit domain?
99282 - Emergency department visit for the evaluation and management of a patient, which requires these 3 key components: An expanded problem focused history; An expanded problem focused examination; and Medical decision making of low complexity. Counseling and/o

(Anna Ostropolets) #2

I’d say that although these concepts seem to be related to drugs, but they don’t specify the drugs at all, which makes it impossible to map them to anything in the Drug domain or to fit the structure of the Drug domain. S9490 has a sort of ingredient but also doesn’t work well: it can only live as a non-standard concept without mapping as it represents a broad group of ingredients (corticosteroids include prednisone, prednisolone and so on).
Why does the domain matter? And would it be more useful for you if these concepts were in the Drug domain but were non-standard?

(Christian Reich) #3


Two things:

  1. Should we resume discussing the question of allowing ATC as standard Drug Concepts?
  2. We need to map those Visit HCPCS and CPT4s to the Visit concepts. Will do.

(Anna Ostropolets) #4

I somehow feel that we haven’t even started an open discussion :slight_smile:
So here’s the summary of what I was thinking about that:

  • There is a gap in Drug domain coverage when it comes to ETL: we don’t have standard concepts representing groups of drugs (say, corticosteroids). So if source data have terms like that, we can map them to 0, create 2bil concepts or go against the rules and put ATCish codes. None of these solutions is elegant/support data standards/promote network research/is easy to query/preserve the meaning etc.
  • We can’t allow ATC to sit on the same level with RxNorm in terms of terminology structure. ATC concept is not an ingredient, clinical or branded drug, it’s a group. Moreover, the name ‘classification concept’ itself already means that these concepts should not be used to store the data but to query/group/classify the data.
    Although I have some thought about the way it may be done from a terminology perspective, I would like to hear from the guys who use ATC in their work.

@Rijnbeek, @abedtash_hamed, @schuemie, @ericaVoss, and others, your comments are highly appreciated.

(Christian Reich) #5

Well, easy: In Condition, we have the full hierarchy all the way up to the top dog “Clinical Finding”. No reason we cannot do that with Drugs. So, you can have a DRUG_EXPOSURE record “corticosteroid”, if that’s all we know about the patient and let this classification come in as a standard concept.

Problem is this: Classifications currently don’t follow the rule that we can have only one standard concept carrying the meaning. Instead, we can have many overlapping similar classifications in parallel. @Frank wrote a paper about the consequences that has in practice: These classifications don’t agree with each other. So, which corticosteroid do we pick?

If we were to extended the Drug Domain all the way up to the top concept “Drug”, we would have to take on one classification system and make that standard. ATC is a good candidate. Are we ready for that? Is ATC comprehensive? Unambiguous? Of high enough quality?

It actually is a good time, as FDB is rarely used, GPI was never implemented, NDF-RT went bust last year and VA Class has been dead even longer than that. The successor MED-RT hasn’t been implemented. Also, ATC is international. We could just abolish drug classifications altogether and create a combined ATC-RxNorm-RxNorm Extension universe.

(Rebecca) #6

Hi Christian and Anna, thanks for your replies! It seems I opened a larger discussion than I even intended to :slight_smile:

Re Anna’s question: "Why does the domain matter? And would it be more useful for you if these concepts were in the Drug domain but were non-standard?"

My use case is that I would like to use the domain to inform which table I map a HCPCS code into during ETL. So I would prefer for these concepts to be in the Drug domain but non-standard, so that I don’t map vaccines into my Procedure table.

(Rebecca) #7

Is there a better way to decide which table to map a HCPCS code into than to use the domain_id?

(Patrick Ryan) #8

We should always use the standard vocabulary concept domain_id to define which table source data will be placed. This is the only way to ensure that data are standardized in one singular, unambiguous location. If different organizations may different subjective choices of where to place each code, then no standardized analysis will be able to consistently retrieve this data element when trying to generate reliable evidence.

(Dmytry Dymshyts) #9

I think the reason is DRUG_ERA, we build it using Standard concepts.

(Christian Reich) #10

True. But there is no reason we would not continue doing that. Yes, we’d lose the ATCs, but there are not that many, and if somebody wants to aggregate all drugs to, say, ATC3 they can do that. In COHORT.

(Dmytry Dymshyts) #11

Doing what?

(Christian Reich) #12

Rolling up to Ingredient.