Stream: argonaut
Topic: Race encoding
Keith Boone (Aug 09 2016 at 11:05):
Also posted to Argonauts list:One of my engineers was reviewing how to support multiple race, and we ran across this in the Argonaut specification:
4. Race extension *NOTE: For this IG this extension may repeat. It is erroneously limited to a single occurance in the FHIR specification.
However, the Race extension referenced is to a CodeableConcept, which already supports capture of multiple codes.
Since the purpose of the Race extension is to capture the concept of race associated with the patient, it seems correct that there is ONE us-core-race extension element, which may contain multiple codes to represent the single concept describing the patient’s race.
The only reason I would see to allow more than once concept is to address the impedance mismatch between EHR granularities in the capture of race information, so that one could readily find the OMB five codes necessary, even though more granular codes in the CDC set role up.
Keith
P.S. Has anyone yet addressed the distinction between CDC Race code set and HL7 V3 codeset required in the extension?
Josh Mandel (Aug 09 2016 at 12:11):
I'm not sure that representing multiple distinct races in a single CodeableConcept is a correct use -- if that were the case, FHIR would never use a repeating CodeableConcept ¡ anywhere, right? And what would you populate in text?
Abbie Watson (Aug 09 2016 at 12:14):
We've also been having discussions around this, but are still at the research-and-collecting-data stage. One thing we're worried about is the mismatch between politically-defined groups and biologically-defined groups. In our case, we're currently working on mammography and breast-cancer risk analysis; and the racial differences in incidences of BRCA1 and BRCA2 mutations.
Caucasians, African-Americans, and Ashkenazi Jews are all known to have different mutation rates; yet to just say 'White, Black, Other' as the CDC Race and Ethnicity Code Set specifies has all sorts of problems.... Western Europeans have different rates than Eastern, so 'White' is an erroneous category; African seems straightforward until you get into North Africa; and Ashkenazi Jew isn't even tracked, and could be recorded as either Middle Eastern, African, or Eastern European, or just statistically lost in 'Other'. It's a mess. And the CDC Code Set arguably contributes to the problem.
I'm tentatively gravitating towards the stance that there's a professional responsibility to use a granular allele-based encoding of some sort (although I'm not clear yet what that would be), and a legal responsibility to report via CDC Race code or OMB five codes, and to do a mapping between the two. So, perhaps another reason to have multiple extensions with which to encode race is if you're working with genomic data and tracking allele groups?
Abbie Watson (Aug 09 2016 at 12:14):
This autumn I'll have the opportunity to run this question by some bioethicists and professors of biomedical informatics. Will try to report back what I find out; but it will be a few months yet.
Keith Boone (Aug 09 2016 at 17:48):
@Josh Mandel The issue is whether Race is a single concept or multiple concepts. If someone asks me my race, I could express that conceptually by saying that I'm "white with some small bit of American Indian". That's the concept. The codes describing that are separate. There is no code for that particular combination, but these codes and this code system are meant to be combined to express multiple races. The quoted text is what you would see in text.
Keith Boone (Aug 09 2016 at 17:49):
@Abigail Watson You probably want to look at the CDC Race and Ethnicity codes in general. They have that level of granularity. It's the OMB rollups that doesn't have it at that level of detail.
Keith Boone (Aug 09 2016 at 17:51):
For what it is worth, genetic observations about race and ethnicity from genetic tests belong in observation, not it patient demographics. They are based on clinical, rather than socio-legal definitions. If you want clinical accuracy, look to clinical data, not demographics.
Joel Schneider (Aug 09 2016 at 21:20):
At NMDP, we sometimes use the term self-identified race & ethnicity (SIRE). People who check more than one box could be classified as multi-race.
This paper contains an analysis of relationships between self-reported categories and racial admixture in the US (see the extended PDF for supplemental material): http://dx.doi.org/10.1016/j.ajhg.2014.11.010
Lloyd McKenzie (Aug 15 2016 at 14:23):
@Keith Boone Multiple Codings in a CodeableConcept are essentially translations. They're not intended to capture multiple distinct types of things. If you want to say I'm Asian and I'm Hispanic, those are distinct concepts. If you want to say "I'm an Asian-Hispanic", you could say that's one concept too, but it's not clear that a code of "Asian" is a proper representation of "Asian-Hispanic" - in general the coding should encompass the concept, not be narrower than it.
Joel Schneider (Aug 17 2016 at 21:06):
The 2002 update to the OMB guidelines on race & ethnicity defines five racial categories and two ethnic categories:
http://archives.hud.gov/offices/cpd/systems/idis/library/raceethnicity.pdf
In those guidelines, "Hispanic or Latino" is an ethnic category, not a race category. They allow an optional multiple race response, but not a multiple ethnicity response.
Joel Schneider (Aug 17 2016 at 21:06):
(deleted)
Joel Schneider (Aug 17 2016 at 21:07):
(deleted)
Brett Marquard (Aug 18 2016 at 02:16):
Where do you see multiple ethnicity response isn't allowed?
Joel Schneider (Aug 18 2016 at 14:22):
The multiple ethnicity constraint in that OMB guideline is on page 4, in the second paragraph following the list of ethnic categories.
Probably common sense, since the two ethnic categories are "Hispanic or Latino" and "Not Hispanic or Latino".
Later on, it also recommends the ethnicity question should precede the race question.
Robert McClure (Aug 18 2016 at 14:51):
@Brett Marquard Joel is correct that when reporting at the OMB level (five race types and two ethnicity types) the expectation is a single code for ethnicity. I believe it is acceptable to report more than one high level Race type but I'd need to check. NOTE - I say that in explaining how OMB has said R&E is to be reported for census use. I'm not sure CMS has defined what conformance they require. As you know C-CDA made some changes to allow that multiple DETAILED race concepts can be exchanged and I thought we had made a change to allow the same for Ethnicity (so more than one sub-type of a "hispanic heritage" could be captured. I'm not sure if any US gov. entity supports such reporting, let alone encourages it.
@Keith Boone Josh and Lloyd describe my firm understanding that code able concept can only care the meaning of one idea and that a translation is of that one idea.
BTW, This whole idea of race v ethnicity seems to be craziness that I've never understood. Because of that I suspect it will always feel awkward. I could say more but I'll let that stand.
Joel Schneider (Aug 18 2016 at 15:49):
Our analysis indicates ethnicity and ancestry are somewhat independent of each other, but not completely independent. Taken as a whole, the Hispanic category ends up looking like a broad north/central/south American population.
Last updated: Apr 12 2022 at 19:14 UTC
