OHDSI Home | Forums | Wiki | Github

Adopting SNOMED as target terminology (non SNOMED member strategy)

(Vojtech Huser) #1

In light of recent smoking concept discussion (SNOMED vs. LOINC): For countries that are members of IHTSDO (USA, UK), it is easier to adopt SNOMED. No need to pay for license (there is national license).

For current non-members, SNOMED license requirement may be an obstacle. (e.g., France, Germany, Italy, China, Japan, Brazil, Mexico).

I understand there is no magic solution here, but I wanted to bring up this difficult subject for at least some discussion. There are some existing CDM datasets from non-member SNOMED countries.

list of SNOMED members: (IHTSDO plans to change its name in 2016, I think)

(J. Marc Overhage) #2

Of course LOINC does not require any license fees anywhere in the world so
are you suggesting that LOINC would be the preferred option?

(colin e.) #3

For the sake of someone who is comfortable with SNOMED, but not so much LOINC: What are the primary differences between SNOMED and LOINC?

(Christian Reich) #4


Lots of small differences, but both organizations are working on harmonizing wtih respect to each other. One main difference is that LOINC tends to be very detailed and comprehensive, while SNOMED is trying to grasp the large picture. Also, LOINC is constructed as fixed question (measurement, test, etc.) - answer (result, possible response) value pairs, while SNOMED is not working in that strict model. But we could use either one for the smoking problem.

Somebody needs to propose the solution. It’s Public Domain: If you miss something it is your job. I know Rimma did something in her PCORI project. Let’s see whether she wants to propose.

(J. Marc Overhage) #5

they are complimentary in many ways (through formal international
agreement) with LOINC providing detailed laboratory and clinical
observation codes which is an area where SNOMED is “thin”. see loinc.org
for details

(Philip Quinlan) #6

I am new to the community and still trying to grasp all the concepts. The issue of SNOMED licensing has already been raised when I have discussed using OMOP, how reliant is OMOP on a SNOMED license being in place, if someone in Germany wanted to use OMOP, how would they proceed?

(Christian Reich) #7


You would have to obtain a SNOMED license yourself. We are currently talking to SNOMED about an OHDSI license, but that is not in place, yet.

(Philip Quinlan) #8

And does that prevent use of OMOP?

(Christian Reich) #9

Without a license? Pretty much yes. Most of the Conditions, many Procedures and some other domains are in SNOMED.

(Andrew Williams) #10

In case there is any lack of urgency about this: you’ll be as unhappy as I was to learn that it has led to a very significant opportunity loss, but one that might be reversed if addressed in time.
The CD2H Dream challenges will use a version of OMOP that substitutes ICD9-10 for SNOMED. https://ctsa.ncats.nih.gov/cd2h/innovate-with-us/dream-challenges/
The licencing is the reason. As a result, a largish set of academic medical centers will start developing solutions build processes around an incompatible version of OMOP. Not good.

There might still be time to get this reversed if it could be resolved quickly.

(Christian Reich) #11


Can’t find where they are making a decision about coding like that. But assuming you are right: they are beating the sack and meaning the donkey.

If this were US-based the ICD-9/10-CM is a natural choice, since a large majority of the data are coded in that system due to the CMS mandate. There are tons of issues for analytics (lack of hierarchies, crazy concepts, main purpose is for justifying payment instead of recording conditions), and we know folks ignore them bravely, but it is the workhorse today, unless places like IMO start opening up. SNOMED licensing is not the problem in the US, it is fully paid up.

If it were meant as an international initiative the SNOMED license is an issue. However, the solution of reverting to ICD9/10 will quickly fall apart, even though people keep glossing over the problem by calling it ICD-9 and ICD-10, when in fact they are massive expansions and slight modifications of the original thing in every country. As a consequence, you have more inaccuracies than with the SNOMED mapping we are doing.

What needs to happen?

(Andrew Williams) #12

I sent an email to folks involved at CD2H introducing you and the topic of discussion.
Let me know if you don’t hear back from them or if there are other ways I can help.