concept reference terms in 1.9

2 views
Skip to first unread message

Stanisic, Lauren R.

unread,
Oct 30, 2012, 3:39:48 PM10/30/12
to implem...@openmrs.org
Hey folks,

For those who have upgraded to 1.9 and are now using the new concept features, I have a question regarding concept reference terms and your general workflow.  In the demo that Andy did during a University call (http://connect.iu.edu/p3mg1bi7j6y/?launcher=false&fcsContent=true&pbMode=normal), I believe the concept reference terms had to each be added individually in order to map a local concept to an external source.  

Is this how groups using this feature are managing their concept reference terms, by manually creating each reference term in advance of setting up a mapping?

thanks for any info!
-lauren  

 

Burke Mamlin

unread,
Oct 30, 2012, 4:50:38 PM10/30/12
to implem...@openmrs.org
In general, it might be easier to import reference terminologies (or chunks of them -- e.g., all SNOMED-CT diagnoses) prior to doing the mapping rather than manually building the external vocabulary references as you create mappings.  There's no harm to having reference terms that are not mapped to local concepts.

-Burke

--
OpenMRS Implementers: http://go.openmrs.org/implementers
Post: implem...@openmrs.org
Unsubscribe: implementers...@openmrs.org
*** OpenMRS 2012 Implementers Meeting: http://2012.openmrs.org/ ***
 
 

Andrew Kanter

unread,
Oct 30, 2012, 8:28:24 PM10/30/12
to implem...@openmrs.org
Many reference terminologies require licensing. The complete maps included with the 1.9 database from CIEL/MVP include the reference codes, but not descriptions. It would be possible to have us provide a version which includes the reference terms if you do have a license (eg., SNOMED CT). It makes sense for someone to centrally manage all of these terminologies and maps (such as what CIEL does for the OpenMRS community).

Andy
 
--------------------
Andrew S. Kanter, MD MPH

Asst. Prof. of Clinical Biomedical Informatics and Clinical Epidemiology
Columbia University
Email: andrew...@dbmi.columbia.edu
Mobile: +1 (646) 469-2421
Office: +1 (212) 305-4842
Skype: akanter-ippnw
Yahoo: andy_kanter


From: Burke Mamlin <bu...@openmrs.org>
To: implem...@openmrs.org
Sent: Tuesday, October 30, 2012 4:50 PM
Subject: Re: concept reference terms in 1.9

Ellen Ball

unread,
Oct 31, 2012, 3:05:22 PM10/31/12
to implem...@openmrs.org, lsta...@regenstrief.org
The addition of 85 radiology order concepts along with corresponding LOINC codes were added with a mysql script (see attached).    A similar script could be written to only add concept reference terms.

Ellen Ball
Partners In Health
CreateConcepts.sql

Jeremy Keiper

unread,
Nov 1, 2012, 3:49:21 PM11/1/12
to implem...@openmrs.org, Stanisic, Lauren R.
Does anyone maintain Metadata Sharing Module packages of complete concept sets?


Jeremy Keiper
OpenMRS Core Developer
AMPATH / IU-Kenya Support


--

Darius Jazayeri

unread,
Nov 1, 2012, 4:01:09 PM11/1/12
to implem...@openmrs.org, Stanisic, Lauren R.
Hi Jeremy,

Just to be clear, that's the sql script that Ellen used to populate our central concept server, based on a spreadsheet of radiology orderables designed by our clinicians. We actually push this to our Mirebalais development server(s) using Metadata Sharing.

-Darius

Manage your OpenMRS subscriptions at https://id.openmrs.org/
 
 

Stanisic, Lauren R.

unread,
Nov 2, 2012, 2:01:07 PM11/2/12
to Ellen Ball, implem...@openmrs.org
Thanks, guys.  For Ampath (like PIH), already have a concept dictionary that has been developed over several years.  When we upgrade to 1.9, we will begin using the reference terms.  However, for reasons of efficiency as well as wanting to avoid errors, it'd be nice avoid adding each reference term manually.  Like Burke said, we'll probably import reference terminologies rather than build them manually.

I really don't know a lot about this... Where would be a good place to get those imports?  Directly from the sources (eg LOINC), themselves? 

thx,
-lauren

From: Ellen Ball [elln...@gmail.com]
Sent: Wednesday, October 31, 2012 3:05 PM
To: implem...@openmrs.org
Cc: Stanisic, Lauren R.

Subject: Re: concept reference terms in 1.9

Jeremy Keiper

unread,
Nov 2, 2012, 4:05:53 PM11/2/12
to implem...@openmrs.org, Ellen Ball
Ah, I see ... so there is no need to worry about Sync Module picking up on changes to the database on your concept development server, and that is why it is okay to use SQL to import them. I believe AMPATH will be in a similar situation.

My question about "concept sets available as MDS packages" might have been confusing.  Here is what I meant to ask, and it is somewhat redundant because I think I know the answer: do PIH, MVP, etc maintain MDS packages of entire concept groupings, those that all belong to one source (e.g. SNOMED) or those that are specific to a given field (e.g. Radiology)?  I know that individual concepts are available in MDS packages from a few websites, and I do not know if OCC was meant for this purpose (or if it is even in use). 


Jeremy Keiper
OpenMRS Core Developer
AMPATH / IU-Kenya Support


--

Ellen Ball

unread,
Nov 3, 2012, 4:04:54 PM11/3/12
to implem...@openmrs.org, Ellen Ball
Dear Jeremy -

You know too many of our tricks.  

Although PIH uses sync on our ZL/Haiti and IMB/Rwanda servers, our "concept" server is not synched.  This gives us much more flexibility to update the concept dictionary and "clean up" concepts without having to deal with sync.  When we want to share concepts with other sites (Haiti, Rwanda, Lesotho), we use MDS.

As for your other question about "maintain MDS packages of entire concept groupings", we don't do that on purpose.  Although it would be nice if MDS allowed users to select a mapping and package up everything with that association (ie. SNOMED, LOINC or org.openmrs.module.mdrtb mapping), that's not available now.   For convenience, we have created sets of concepts where each member has the same mapping (ie.  "MDR-TB Core Concepts" is a set of concepts with the org.openmrs.module.mdrtb map; All concepts in the "Radiology order" set have a LOINC mapping).  The PIH dictionary doesn't come close to the magnitude of concepts in MVP/CIEL with ICD10, SNOMED or RxNorm maps.  If we need one of these, we export from MVP/CIEL and import into our dictionary with MDS.

Let me know if this doesn't address your question,

Ellen Ball
Partners In Health

Jeremy Keiper

unread,
Nov 5, 2012, 8:52:51 AM11/5/12
to implem...@openmrs.org, Ellen Ball
Thanks, Ellen!

This may be off-topic, or only relevant to the last few emails.  I presume you have a MDS package for your entire concept dictionary.  Can you share your experience with upgrading to OpenMRS 1.9 and what you did to clean up your dictionary so it was compatible with MDS?  If you didn't have to, that's good!  Thanks.



Jeremy Keiper
OpenMRS Core Developer
AMPATH / IU-Kenya Support


Ellen Ball

unread,
Nov 5, 2012, 4:37:11 PM11/5/12
to implem...@openmrs.org, Ellen Ball
There's a pretty good wiki page about concept validation:  https://wiki.openmrs.org/display/docs/Concept+Validation.  It covers many of the painful areas for our OpenMRS 1.6 to 1.9 migration:
  1. Duplicates concept names are not allowed for the same locale.
  2. We use mappings (ICD10, PIH, mdrtb).  During upgrade, concept_map_type.name is set by default to "NARROWER-THAN".  This isn't true for PIH, mdrtb and others.  You can make this happen by setting the concept_map.comment field with the appropriate setting (ie.  SAME-AS).  If you can't find it, remind me and I'll find the setting. 
  3. Validation does not allow null concept_name.name or concept_description.description.
Validation will tell you about all the problems.  Fixing all the problems with a 6000 concept dictionary is a pain.

Ellen

Andrew Kanter

unread,
Nov 5, 2012, 5:29:10 PM11/5/12
to implem...@openmrs.org
Thanks, Ellen...

The CIEL dictionary does have a lot of comments in the mapping table to help set the map type during the migration. We are adding more and more of these as we get into more sophisticated modeling adding new procedure comments, etc. For example, the new radiology codes have laterality either explicitly added as a reference map (HAS-LATERALITY) or implicitly by a specific SNOMED code.

Is it really true about descriptions not being null? We have many concepts which don't have a definition.... (probably should, but don't).

Andy
 
--------------------
Andrew S. Kanter, MD MPH FACMI

Asst. Prof. of Clinical Biomedical Informatics and Clinical Epidemiology
Columbia University
Email: andrew...@dbmi.columbia.edu
Mobile: +1 (646) 469-2421
Office: +1 (212) 305-4842
Skype: akanter-ippnw
Yahoo: andy_kanter

From: Ellen Ball <elln...@gmail.com>
To: implem...@openmrs.org
Cc: Ellen Ball <elln...@gmail.com>
Sent: Monday, November 5, 2012 4:37 PM
-lauren
Hey folks,

For those who have upgraded to 1.9 and are now using the new concept features, I have a question regarding concept reference terms and your general workflow.  In the demo that Andy did during a University call (http://connect.iu.edu/p3mg1bi 7j6y/?launcher=false&fcsConten t=true&pbMode=normal), I believe the concept reference terms had to each be added individually in order to map a local concept to an external source.  

Is this how groups using this feature are managing their concept reference terms, by manually creating each reference term in advance of setting up a mapping?

thanks for any info!
-lauren  

 
Manage your OpenMRS subscriptions at https://id.openmrs.org/
 
 
--
OpenMRS Implementers: http://go.openmrs.org/ implementers
Post: implem...@openmrs.org
Unsubscribe: implementers...@ openmrs.org
Manage your OpenMRS subscriptions at https://id.openmrs.org/
 
 

Rafal Korytkowski

unread,
Nov 6, 2012, 9:10:07 AM11/6/12
to implementers
Concepts without descriptions are valid at the moment, but there's a ticket to "fix" this: https://tickets.openmrs.org/browse/Trunk-3616

Considering Andy's comment I think we should abandon the idea to require at least one description if there's a name in the same locale.


-Rafał

Ellen Ball

unread,
Nov 6, 2012, 11:30:33 AM11/6/12
to implem...@openmrs.org

My point about concept description -- concept description validation fails if the database has a row with a null value for the concept_description.description. 

But descriptions are not required for concepts.

Ellen

On Nov 6, 2012 8:10 AM, "Rafal Korytkowski" <ra...@openmrs.org> wrote:

Concepts without descriptions are valid at the moment, but there's a ticket to "fix" this: https://tickets.openmrs.org/browse/Trunk-3616

Considering Andy's comment I think we should abandon the idea to require at least one description if there's a name in the same locale.



-Rafał




On 5 November 2012 23:29, Andrew Kanter <andy_...@yahoo.com> wrote:
>
> Thanks, Ellen...
>

> T...

Unsubscr...

Burke Mamlin

unread,
Nov 8, 2012, 9:41:25 AM11/8/12
to implem...@openmrs.org
We intended on concept.description being required in order to promote/enforce best practice (every concept should have some indication of what it is and/or how it is intended to be used).  We don't prevent people from putting gibberish or just a copy of the name in the description to defeat the requirement, but the goal is to promote the use of well-described concepts.  With localization, the requirement would become at least one description in any locale for the concept.

If somewhere along the way we lost the requirement for at least one description (in any locale) per concept and are allowing concepts without any description, then I would be in favor in re-instituting the requirement.

-Burke

Rafal Korytkowski

unread,
Nov 8, 2012, 10:06:29 AM11/8/12
to implementers
So to alleviate future migrations I'd suggest that along with adding a requirement for a concept to have at least one description for a locale we will create a database changeset which will automatically copy a preferred name to a description if it is missing for all existing concepts. Is this acceptable?


-Rafał

Darius Jazayeri

unread,
Nov 8, 2012, 10:15:17 AM11/8/12
to implementers
I agree that we don't want to break otherwise-working installations.

One option would be to (re?)introduce the requirement for >= 1 description on each concept, but only after we've done the ticket that allows you to disable validation-on-save.

If we do auto-create descriptions, I'd rather do something like create a description of "No description" in the locale en.

-Darius

Burke Mamlin

unread,
Nov 8, 2012, 2:49:59 PM11/8/12
to implem...@openmrs.org
Sure.  I would probably use the opportunity to teach best practice – e.g., any concepts without any description in any locale would get a description like "Missing concept description for [PREFERRED NAME] (see http://go.openmrs.org/concept-description for best practices)."

Cheers,

-Burke

Andrew Kanter

unread,
Nov 8, 2012, 5:16:23 PM11/8/12
to implem...@openmrs.org
I think since these descriptions/definitions are shown as hover help within the application, I don't think I would do this. I would want to be careful to not make it look like the database is incomplete or non-standard or not-best practice....
 
--------------------
Andrew S. Kanter, MD MPH FACMI

Asst. Prof. of Clinical Biomedical Informatics and Clinical Epidemiology
Columbia University
Email: andrew...@dbmi.columbia.edu
Mobile: +1 (646) 469-2421
Office: +1 (212) 305-4842
Skype: akanter-ippnw
Yahoo: andy_kanter
From: Burke Mamlin <bu...@openmrs.org>
To: implem...@openmrs.org
Sent: Thursday, November 8, 2012 2:49 PM

Subject: Re: concept reference terms in 1.9

Darius Jazayeri

unread,
Nov 8, 2012, 5:23:56 PM11/8/12
to implementers
That's why I suggested "No description". It's clear, and it won't look bad.

-Darius

Ellen Ball

unread,
Nov 8, 2012, 10:16:43 PM11/8/12
to implem...@openmrs.org
I don't like the enforcement of descriptions -- especially if there's nothing useful there.   Do we really want to provide a description to yes, no, misc, negative, aspirin, ambulance, chest xray, cattle owned by patients family?  Does it provide anything useful to have "no description"?  Burke seems to suggest that future implementations will force descriptions for each locale.  For the PIH dictionary, we often have 3 locales (English, French and Haitian Kreyol).  It's actually quite a bit of work to come up with a one good description, let alone something in another language.

Can't we just provide guidelines that descriptions are recommended?  Or use a global property to enforce this rather than make it a requirement?

Sincerely,

Ellen Ball
Partners In Health
"Beyond the mountains, more mountains"  -- Haitian proverb

Michael Seaton

unread,
Nov 8, 2012, 10:41:02 PM11/8/12
to implem...@openmrs.org
I agree with Ellen.  Would love to see this be recommended but not a hard validation requirement, particularly given past challenges around validation.

Mike

Burke Mamlin

unread,
Nov 9, 2012, 10:58:56 AM11/9/12
to implem...@openmrs.org
Ellen,

My proposal was to promote best practice of at least one description in any locale; I did not mean to imply that we would require a description in every locale.

While "yes" & "no" may be the exception, they're not difficult to describe and nearly every other term can be misinterpreted.
  • yes – "Used to indicate an affirmative response."
  • no – "Used to indicate a negative response."
  • misc – Is this an abbreviation for miscalculate, miscarriage, miscellaneous, mischief, misconception, misconduct, or miscreant?  It may be obvious to people maintaining the dictionary, but not everyone using it.
  • aspirin – there are 224 forms of aspirin in RxNorm.  Does this refer to the active ingredient?  The pill?  A solution?
  • chest xray – is this portable, PA & lateral, or both?
  • cattle owned by patients family – is this cattle currently owned or ever owned?  is this whether or not cattle are owned… or is it the number of cattle owned?  does it apply only to immediately family or to extended family?
There is something harder than making up a clear description of a concept; that is, dealing with data that have been collected under differing interpretations of the concept.

Personally, I wouldn't accept a concept in a dictionary that didn't have a description; however, if Ellen Ball, queen of dictionary management is advocating for giving into entropy, then who am I to protest.  If we are not going to require any description for concepts, could we at least make the concept management page promote descriptions (i.e., explain that it's best practice for all concepts to have a description and provide a go.openmrs.org link to supporting documentation)?

-Burke

Stanisic, Lauren R.

unread,
Jan 18, 2013, 11:04:33 AM1/18/13
to implem...@openmrs.org
Hey Burke, and everyone, 

ReL "In general, it might be easier to import reference terminologies (or chunks of them -- e.g., all SNOMED-CT diagnoses) prior to doing the mapping rather than manually building the external vocabulary references as you create mappings"

Do you know where we get those reference terminologies to import?  From the source itself?  

I'm curious, for those implementations which have not doing mappings when creating concepts, do you plan on going back and mapping the currently unmapped items in your dictionary? 

-lauren

From: rafal.ko...@gmail.com [rafal.ko...@gmail.com] on behalf of Rafal Korytkowski [ra...@openmrs.org]
Sent: Tuesday, November 06, 2012 9:10 AM
To: implementers

Burke Mamlin

unread,
Jan 18, 2013, 12:50:23 PM1/18/13
to implem...@openmrs.org
The easiest path is to adopt the MVP-CIEL dictionary and benefit from other people doing the mapping & benefitting from efforts like those allowing people to subscribe to changes in that dictionary… or you can always go to the source and do the mappings yourself.  There are resources like the Maternal Concept Lab you could use to find what other people are doing and benefit from their mappings (perhaps for local concepts beyond what MVP-CEL provides).  The ultimate vision was for the OpenMRS Concept Co-operative (OCC), which has been an idea for many years with a fair amount of coding code, but hasn't yet been fully realized.

-Burke
Reply all
Reply to author
Forward
0 new messages