OHDSI Home | Forums | Wiki | Github

More ambiguous domains - need your help

Friends:

In developing the domain assignment heuristic, I am running into the following situations (all in CPT4 right now):

  • Therapeutic Drug Assays: These are situations where the level of a drug is measured in the patient. This is useful for drugs where dosing is tricky and you therefore titrate the dose based on the concentration. Examples are:
    • 80157 “Carbamazepine; free”
    • 80154 “Benzodiazepines”
    • 80182 “Nortriptyline”
    • 80185 “Phenytoin; total”

Obviously, they are measurements. But the question is: Are they also drug exposures? Can we assume that the patient was drugged with the stuff that is now measured, or not?

  • Documentation of things. There are a whole lot of CPT4 codes which essentially codify the documentation of something. Examples are:

    • 3014F Screening mammography results documented and reviewed (PV)
    • 3006F Chest X-ray results documented and reviewed (CAP)
  • Often they are documenting diagnoses or providing test results:

    • 3092F “Major depressive disorder, in remission (MDD)”
    • 3088F “Major depressive disorder, mild (MDD)”
    • 3040F “Functional expiratory volume (FEV1) less than 40% of predicted value (COPD)”
    • 3048F Most recent LDL-C less than 100 mg/dL (CAD) (DM)

Same question for these? My assumption is right now that these CPT4 codes are Observations, but they also have a mapping to the very thing that is documented, which is an imaging test (Measurement), a diagnosis (condition) or another test (Measurement).

I would err on under-interpreting the data, particularly in the early days of the domain mapping. In particular, I don’t think procedures should be mapped unless there is a compelling case. A lot of these are documenting the process of care, and/or the quality of care. And their primary use is in that context. We still have prescription data, imaging procedures, and diagnoses to find the these concepts. If we see an assay without the drug, what does that mean? Or the mammography review without the procedure? As an extreme example, do we put in a cardiovascular diagnosis code for somebody having a stent placed?

But diagnoses and test results in the CPT (and HCPCS) codes – those seem like they should be moved to conditions or measurements. There is a long history in administrative data of trying to force results into the codes. The Medicare PQRS program does this. There are laboratory values reported along with claims data for erythropoietin (very useful in the Medicare dialysis program). So, there is definitely a precedent for parsing the conditions and labs into their proper places.

@jenniferduryea has put a more detailed discussion of these issues under our discussion about HCPCS level 2 codes that has a lot more detail on closely related concepts (what to do with “procedure” codes): HCPCS details

1 Like

I agree with Mark. I think it gets messy as you really don’t know the date
of the actual event (e.g., drug exposure). So would lean towards
under-interpreting.

1 Like
t