OHDSI Home | Forums | Wiki | Github

Why is Screening mammography coded as a Measurement and not a procedure

Hi, We are trying to extract information from our clinical records into the OMOP CDM and have started with breast cancer as this is a domain our team is very familiar with. We are currently trying to map the various radiological procedures, and it seems logical to use SNOMED codes. Everything seems relatively straightforward until we get to screening studies (mammography, tomo and breast MRI). Some codes from Athena are here:

mammo and breast MRI are procedures as expected, but the screening studies are measurements for some reason. There may be no technical difference between a screening and a diagnostic mammogram but our source systems clearly differentiate between these and there is a big difference if you are trying to use the data to assess the efficacy of screening programs or to identify cohorts to use in developing AI algorithms so I don’t want to throw away this valuable information.
What is the rationale for putting the screening procedures under measurement? Is it something that can be changed? I saw an old thread from 2019 on the topic of mammography (@krfeeney , @mgurley , @Eduard_Korchmar @Mark_Danese ) but it didn’t address this point. I’m new to omop but it seems like it could cause end users problems if screening is in measurement but diagnostic mammography is in procedures. Thanks!

Hello Anne,

In OMOP CDM, Measurement Domain is reserved for concepts that represent actions with diagnostic intent that have intent to obtain a single value on a determined scale (be it a numeric range, list of categories or simple Yes/No answer). If the expected diagnostic result is complicated enough to warrant it’s own entry in Clinical Data tables, like confirmed diagnosis in Condition table, the procedure should belong to Procedure domain.

Domain assignment for SNOMED concepts is fairly complicated. It would be actually very straightforward with our approach – identify key ancestors that correspond to a particular CDM domain, but is SNOMED hierarchy is multiaxial, creating ambiguity for concepts belonging to two or more conflicting categories.

Concept Screening mammography is a descendant of two important hierarchies: Screening procedures (normally Measurement, as 99% of concepts in that hierarchy fit the definition) and Diagnostic imaging procedure (normally Procedure – there is no single value to be read).

As a rule of thumb, such concepts are assigned Domain in order of precedence. Procedure has lower priority than Measurement, because latter can have more data associated with them.

I would assert that all imaging procedures should unambiguously belong to Procedure domain, regardless of screening intent.

@Alexdavv is the current lead of the Vocabulary team, adding him to the thread.

Thanks for the information - it does make me optimistic that screening mammo/mri/tomo can get put back into procedures where it used to b . It is possibly based on an assumption that the screening study only results in a single birads score. This is not the case as a mammo screen is now required to return an assessment of breast density & mri screening has alway had several parameters. We will have to use NLP to pull the measurements out of free text reports but that is a seperate problem :slight_smile:

Would you like to create an issue here Issues · OHDSI/Vocabulary-v5.0 · GitHub? :slight_smile:

Thanks for the advice - I will do that!