while thinking about terminology server requirements for openEHR systems
I would like to ask all openEHR implementers about experiences of
different solutions. Are there any experiences of using openEHR systems
with e.g. the FHIR terminology services, CTS2, Ocean TQL, homebrew, etc?
What are the use cases when the terminology servers are used (e.g.
design time, data entry, querying, etc.)? What are the "terminological
queries" that are used/needed (e.g. subsumption testing, subset
membership, subset expansion, etc.)?
I’ll listen to this discussion with interest. I expect that the answer will be: same functional needs as already covered by FHIR terminology services, but there’s some additional information features that are needed to enable seamless integration.
At Linköping University we did a demonstrator in 2012 using a homebrew REST interface to an expression repository based on the SNOMED CT query language at the time. The demonstrator showed querying over EHR content including both AQL and the SNOMED CT query language. The terminology server per default did expansion of results of the SNOMED CT queries, i.e. it returned a set of SCTID:s+expression id:s. The aim of this experiment was to show that some very complex quality indicators could be expressed as queries on a structured health record.
I say, source of inspiration, because not everybody needs editing capacity, most use-cases just want to query.
And the swagger/openapi is not optimal, there are some errors in the data-models on technical level, but these are very few.
When you look at it, and leave out all the branch-things (which are for editing and versioning), then you have a decent interface for a SNOMED service.
An then, it is also very obvious (afterwards) and it reflects good thinking in its simplicity.
As you may know, there are mappings for SNOMED and LOINC, ICDxx and other terminologies (also local), and others, on the way, or already finished, so this interface can also used for these mappings which gives in this way a route to query other terminologies.
Very timely questions for us at Operon - we are working with
terminology service providers to provide tight integration with
openEHR services.
We definitely need to be able to expand valuesets in the context of
an openEHR query
e.g Give me any patients with a problem/diagnosis of diabetes mellitus
type 2 or children. i.e is_a relationships
We need to be able to channel this via AQL or other opeNEHR query languages.
We need to be able to manage and query over termsets/resultsets.
I'm not so sure about the need for more complex queries over other
relationships / SNOMED concept model right now.
We definitely need some sort of vendor-agnostic API against which the
majority of terminlogy service calls can be run. The FHIR service
looks 'good enough' to me although I don't pretend to be an expert.
Ian
Dr Ian McNicoll
mobile +44 (0)775 209 7859
office +44 (0)1536 414994
skype: ianmcnicoll
email: ian@freshehr.com
twitter: @ianmcnicoll
Co-Chair, openEHR Foundation ian.mcnicoll@openehr.org
Director, freshEHR Clinical Informatics Ltd.
Director, HANDIHealth CIC
Hon. Senior Research Associate, CHIME, UCL
unfortunately that didn’t happen, at least until now.
Technically, the system used a relational database to persist SNOMED CT and expression versions over time and an OWL ontology of SNOMED CT and expressions and the two were kept in sync. Expansion was just a query of the the transitive closure table.
We have been working on this from several perspectives. We have been
mostly focused in providing terminology binding support for
archetypes, and first efforts were focused in providing term and
subset bindings via external terminology services (such as ITServer)
Lately we have implemented a Snomed expression syntax engine (for both
parsing and executing) called SNQuery (http://snquery.veratech.es) and
we are using it instead, as provides more flexibility.
Having Snomed expression syntax queries as value bindings in
archetypes allows to easily create queries for data validation ("this
code should be an allergy" or "the text of this coded text should be
one of the synonyms of the code") and data transformation (to have
conditional data transformation functions such as "if the diagnosis is
a cancer diagnosis then A else B"). We mostly use subset membership
queries, but I assume that subsumption testing will be handy in
archetype specialization definition (subsets in the specialized
archetypes must be a subset of the ones in the parent archetype).
I think that subsumption testing in archetypes is not feasible when the archetypes are not covered by SNOMED.
This is because SNOMED subsumption testing is only valid inside the SNOMED definitions.
The system of specialized archetypes and parent archetypes (without the use of SNOMED) is a parallel system in its own semantic world, which cannot be mixed with the SNOMED semantics, so there will be no automatic subsumption testing.
OpenEHR will need to implement SNOMED if you want to do what Ian proposed as an example:/
"e.g Give me any patients with a problem/diagnosis of diabetes mellitus type 2 or children. i.e is_a relationships"/
When you want to use advantages of SNOMED, you must not mix up a parallel world to SNOMED with SNOMED.
This could lead to dangerous situations.
I talked about specializations because one of the basis of it is that
constraints in specialized archetypes must be included in the parent
constraints. e.g. parent archetype contains the "any allergy" subset
and the children archetype contains "all allergies caused by a drug",
and for that specialization to be correct, the second should be a
subset of the first one.
If you refer to label binding then I completely agree with you,
bindings to snomed labels are difficult or almost impossible to define
(I think that probably the only safe way of doing this is that the
country asks for an extension to snomed based on their national
archetypes and then binds the labels to the new codes)
That is what I mean with the parallel semantic world. In your archetypes, you create in this way a semantic world that has no trust-able semantic connection to the SNOMED semantic world. Maybe there is compatibility, but we can never be sure about it.
In fact, working this way is creating many semantic worlds (for every institution which creates archetypes sets), a situation which we are for twenty years very busy with, to avoid.
If we want to be sure that we are talking about the same thing, we need a common reference, and SNOMED is a good example for how a common reference can look like. I would not know a better one.
Why SNOMED is so suitable to be a good reference, is because SNOMED defines into many details, which are common ground for the member states.
We can be sure what SNOMED means when it talks about "allergy for drug", because we agree to the very extensive semantic world of which this statement is a part.
I’m just like Bert from the Netherlands, and it seems to have something to do with the locale nl that our browser sends to your server but that your server has not available?
This is my error message:
type Informe de Excepción
mensaje com.vaadin.server.ServiceException: java.util.MissingResourceException: Can’t find bundle for base name com.example.snomedquery.messages_nl, locale es_ES
descripción El servidor encontró un error interno que hizo que no pudiera rellenar este requerimiento.
excepción
javax.servlet.ServletException: com.vaadin.server.ServiceException: java.util.MissingResourceException: Can't find bundle for base name com.example.snomedquery.messages_nl, locale es_ES
com.vaadin.server.VaadinServlet.service(VaadinServlet.java:367)
javax.servlet.http.HttpServlet.service(HttpServlet.java:729)
org.apache.tomcat.websocket.server.WsFilter.doFilter(WsFilter.java:52)
causa raíz
com.vaadin.server.ServiceException: java.util.MissingResourceException: Can't find bundle for base name com.example.snomedquery.messages_nl, locale es_ES
com.vaadin.server.VaadinService.handleExceptionDuringRequest(VaadinService.java:1469)
com.vaadin.server.VaadinService.handleRequest(VaadinService.java:1426)
com.vaadin.server.VaadinServlet.service(VaadinServlet.java:365)
javax.servlet.http.HttpServlet.service(HttpServlet.java:729)
org.apache.tomcat.websocket.server.WsFilter.doFilter(WsFilter.java:52)
causa raíz
java.util.MissingResourceException: Can't find bundle for base name com.example.snomedquery.messages_nl, locale es_ES
java.util.ResourceBundle.throwMissingResourceException(ResourceBundle.java:1564)
java.util.ResourceBundle.getBundleImpl(ResourceBundle.java:1387)
java.util.ResourceBundle.getBundle(ResourceBundle.java:773)
com.example.snomedquery.Messages.getString(Messages.java:19)
com.example.snomedquery.StartView.OpcionesApp(StartView.java:588)
com.example.snomedquery.StartView.<init>(StartView.java:175)
com.example.snomedquery.SnomedqueryUI.init(SnomedqueryUI.java:65)
com.vaadin.ui.UI.doInit(UI.java:687)
com.vaadin.server.communication.UIInitHandler.getBrowserDetailsUI(UIInitHandler.java:214)
com.vaadin.server.communication.UIInitHandler.synchronizedHandleRequest(UIInitHandler.java:74)
com.vaadin.server.SynchronizedRequestHandler.handleRequest(SynchronizedRequestHandler.java:41)
com.vaadin.server.VaadinService.handleRequest(VaadinService.java:1414)
com.vaadin.server.VaadinServlet.service(VaadinServlet.java:365)
javax.servlet.http.HttpServlet.service(HttpServlet.java:729)
org.apache.tomcat.websocket.server.WsFilter.doFilter(WsFilter.java:52)
nota La traza completa de la causa de este error se encuentra en los archivos de diario de Apache Tomcat/8.0.26.
