OHDSI Home | Forums | Wiki | Github

Suggested new mappings (relationships) for vocabulary - how to submit?


(Sulev Reisberg) #1

I’m currently working on a large set of national EHR data and trying to map them to OMOP CDM. All the diagnoses are given as ICD-10 codes, but approx 5% of them do not have a corresponding mapping to standard concepts. So, I’ve prepared a custom mapping for them, but I can’t see a reason why they shouldn’t be part of the standard mapping. For instance:

45606791 ICD-code “Epistaxis” has currently no mapping, but should have a relationship to 4096682 Standard concept “Bleeding from nose, epistaxis”

However, it has been unclear to me to where to submit or git commit such mappings or what is the relevant “board” to review these mappings…?


(Roger Carlson) #2

http://athena.ohdsi.org/search-terms/terms?page=1&pageSize=15&query=R04.0

ICD10 (45606791) does not have such a mapping, but ICD10CM (35211270) does.


(Sulev Reisberg) #3

Thanks @roger.carlson

I think there should be also mapping for ICD10.

However, there are also other examples that are not given in ICD10CM either, e.g Z25.1:

45581017 Z25.1 - Need for immunization against influenza
should map to (currently no map)
44784283 - Requires influenza virus vaccination

So take my question more general… commonly used codes need a mapping. What is the process to add them?


(Roger Carlson) #4

Okay. Sorry. I thought you might just be limiting yourself to ICD10. It’s much smaller than ICD10CM.

Maybe try here: https://github.com/OHDSI/OMOP-Standardized-Vocabularies/issues


(Don Torok) #5

If your source data is from the United States you should be using ICD10CM. The other ICD10 vocabulary is from the World Health Organization.


(Sulev Reisberg) #6

Yes, my data is from Europe.


(Anna Ostropolets) #7

Of course, they should. And actually @Dymshyts is working on that now. The reason why ICD10 is covered not as well as ICD10CM is simpe: the need in ICD10CM mappings is more prominent.

To comment, propose or report an issue please go to:


(Dmytry Dymshyts) #8

Yep, updated and fixed ICD10 mapping is expected to be released in a week or so.
I’ll let you know when it’s done


(Sulev Reisberg) #9

Good! Can I also participate in this any way?


(Dmytry Dymshyts) #10

Actually, all the mappings are already sitting internally just waiting for some final checks and release.
We checked concepts you mentioned before and they are mapped correctly, but no one can see them yet:(
If you want to participate you can review the mappings we’ll release, especially those you’re interested in your study.
Let me know, if it’s convinient for you to grab them from Athena, or we can send you the table with the mappings


(Sulev Reisberg) #11

Well, probably it is better to sit and wait… Hope that it is coming soon.


(Dmytry Dymshyts) #12

Happy to tell that the new version of ICD10 mapping is released.


#13

Proposed Remappings of condition id.xlsx (26.7 KB)

This email thread is similar to an issue I found when doing a conversion of an EMR database to the OMOP CDM (5.3.1) and is related to mapping condition_source_concept_id to OMOP Standard condition_concept_id. As a background, our source data uses ICD coding for conditions (spans a period when ICD9CM was used, then the hospital transitioned to ICD10CM)

During the conversion process:

  1. condition_source_concept_id is assigned by looking up the ICD code in the CDM ‘concept’ table and taking its corresponding concept_id
  2. condition_concept_id is assigned by looking up the condition_source_concept_id in the ‘concept_relationship’ table and taking the concept id that it maps to (relationship = ‘maps to’)
    Out of the ~1700 most frequent diagnosis codes in the database representing 92% of the diagnoses, I found 55 codes that had relationship mappings which, although not necessarly wrong, seemed less than optimal. Searching Athena (this is a WONDERFUL resource), I found different SNOMED concept_ids that seemed a better match to the original diagnosis description. 47 of these are already Standard OMOP terms, eight are Non-standard, but since they are SNOMED terminology, could they be upgraded to Standard? I have attached these concept ids (exisiting and proposed mappings) as an xlsx file for the Vocab group to review and deterimine if they can be used in lieu of the current mapping. I am relatively new to OHDSI and there probably are rules or some history that I am unaware of about Vocab mappings. Apologies in advance if I’m rehashing some old ground.
    P.S. Sending this a few hours before I go on vacation, when I’ll have very limited email access. In case there are any questions about my posting, I may not be able to respond until early Jan. @mintmountain89 is working with me on this, though, and she might be able to pipe in.

(Dmytry Dymshyts) #14

@CSung,
Thanks for your initiative,
I noticed that you mapped some to non-standard concepts that is not permissible in OMOP.
Other mappings looks reasonable.
We’re working on ICD10CM and ICD9CM mapping improvement now, So probably we already fixed what you find as well, but those mappings sit internally and wait untill the whole file is done.
@Polina_Talapova is curating this work.
She’ll compare what we have internally with your work and comes back to you soon.


#15

OK, thanks for letting me know. For those mapped to non-standard, I couldn’t find a Standard SNOMED code that was as close to the original ICD text description of the condition among the Standard Codes. For example ICD10CM R74.8 “Abnormal levels of other serum enzymes” is mapped to “Enzyme measurement”, which loses a lot of the meaning of the condition. A non-standard SNOMED code exists for “Abnormal level of other serum enzymes”. Do you sometimes “promote” a non-standard term to standard?


(Polina Talapova) #16

Thanks for your work!

1). As Dmytry mentioned, mappings to non-standard concepts are not permissible in OMOP. Thus, to avoid this in some cases we map concepts either to the Measurement or Observation domain. These domains allow to represent any values (e.g.‘Abnormal’), using relatiobnship_id ‘Maps to value’ (in Athena - ‘value_as_concept map’)

R74.8 35211422 Abnormal levels of other serum enzymes
Maps to 4046263 Enzyme measurement
Maps to value 4135493 Abnormal

2). Also we should inform you, that in a case, when the name of ICD concept (independently to the iteration) contains ‘condition AND condition’, it directly means ‘OR’. Because these conditions don’t occur simultaneously (you can check it there https://www.icd10data.com or there https://icd.who.int/browse10/2016/en). That’s why there is only one way to represent this data correctly: to map it UP to the ancestor (common parental concept for both conditions):

N93.9 45592158 Abnormal uterine and vaginal bleeding, unspecified
Maps to 4169954 Genitourinary tract hemorrhage

I48 45596206 Atrial fibrillation and flutter
Maps to 4068155 17366009 Atrial arrhythmia
(But not 4108832 195080001 Atrial fibrillation and flutter - according to SNOMED hierarchy, these are simultaneous conditions)

453.8 44834756 Acute venous embolism and thrombosis of other specified veins
Maps to 4028367 Acute disease of cardiovascular system
(But not 318775 234049002 Venous embolism only)

The same idea with "condition OR condition" (Accidental puncture or laceration during a procedure, not elsewhere classified).

We can’t map to the only one condition in this case. That’s why we map it UP to the common parental concept.

I63.5 45605808 Cerebral infarction due to unspecified occlusion or stenosis of cerebral arteries
Maps to 443454 Cerebral infarction
(but not ‘35610085 cerebral infarction due to stenosis of cerebral artery’ only)

3). Some suggested concepts look like correct according to the name, but, unfortunately, they are incorrect due to specific reasons:
– SNOMED hierarchy:
For example, you propose to map
‘Z89.6 1576257 Acquired absence of leg above knee’
to
‘4203427 308685008 Amputated leg’,
but if you look at the ancestor of this concept, you will see:
‘4100593 299644008 Postprocedural finding of lower limb’.
That indicates that a leg was amputated during a procedure, but concerning the ICD concept name, this fact is UNKNOWN. Therefore, accidental traumatic amputation can take place too. That’s why we will correct current ICD10CM mapping to
‘37109843 723722000 Absence of lower limb’ + ‘42537748 737303004 Acquired absence of organ’ (muscles, bones and vessels of this body part are considered to be organ’s combination)
– rules for exclusion in ICD guides:
For instance,
R63.0 35211405 Anorexia (ICD10CM)
783.0 44832712 Anorexia (ICD9CM)

Can’t be mapped to 436675 Anorexia nervosa, because of:
Coding Notes for R63.0
Anorexia nervosa - instead, use code F50.0
Loss of appetite of nonorganic origin - instead, use code F50.8 https://icd.codes/icd10cm/R630

783.0 Excludes
anorexia nervosa (307.1)
loss of appetite of nonorganic origin (307.59) http://www.icd9data.com/2012/Volume1/780-799/780-789/783/783.0.htm

Please, look at the attached file to see more comments!suggested_ICD10_mapping_comments.xlsx (17.3 KB)


#17

Thanks for the detailed explanation! I see I need to understand more about the ancestor concepts. Is there a good document to read about this and how to trace up the hierarchy? Am I understanding that R74.8 would end up getting mapped to two SNOMED codes 4046263 and 4135493? I was under the impression that these mappings were one-to-one.


(Polina Talapova) #18

You are always welcome! :slight_smile:

There you can get useful information about ancestor concepts: http://www.ohdsi.org/web/wiki/doku.php?id=documentation:cdm:concept_ancestor

And you are right! R74.8 would be mapped to two SNOMED codes 4046263 and 4135493 simultaneously.

Read more: http://www.ohdsi.org/web/wiki/doku.php?id=documentation:vocabulary:mapping&s[]=maps


t