10:00 UTC on Wednesday 26 October 2022 - 90 mins
- FHIR Terminology Services and Resources discussion
- FHIR Terminology Binding discussion
Chat: public-snomedintl.slack.com # snomed-hl7-fhir (ask for invite!)
Note no recording for 2022-10-18 due to small group, off-agenda discussion.
|Notes & Actions
|Welcome and introductions
Recording, notes & attendance.
SI Business Meetings and Expo - 47 countries represented.
Hospitals on FHIR October 27
2nd X-eHealth Innovation Day (October 26th, 2022) link: https://www.x-ehealth.eu/2nd-xehealth-innovation-day/
FHIR North - October 26 & 27
Australia Connectathon - Late Nov
NZ Connectathon Mon 5 Dec Health Terminology Server (Ontoserver) Connectathon run by CSIRO followed by joint standards workshop - "Escape from burning Platforms with SNOMED on FHIR" with PJ and AK. Part of Digital Health Week (Rotorua)
Other Regular Meetings:
HL7 Group TSMG (meeting Wed PM ET every other week) - Terminology Service Management Group (HTA Thursday AM is now a subgroup of the TSMG). 2022-05-17 RH Joint session with Vocab at last business meeting. 2022-06-14 Group has approved minimum capabilities for terminology servers. Now looking at bigger/better HL7 Terminology Server
|ValueSet of all of SNOMED CT
What is the proper valueset URI from which to reference a SNOMED CT concept?
http://snomed.info/sct (ie across all editions and all versions, but this is the CodeSystem URI, not a ValueSet one)
Whatever we use, it should be populated in CodeSystem.valueset see https://www.hl7.org/fhir/codesystem.html
However, certainly in the Snowstorm implementation that ValueSet would only expand to the default edition & version loaded. That said, $validate-code does check across all concepts known to the server (in any extension).
|Round Table Introductions
Australia - 2022-04-13 CSIRO released next version of Ontoserver that implements some R5 features as pre-adoption, mixed with R4 eg typed values in concept maps. Also code properties without expansion. 2022-08-31 ML CSIRO taking over responsibility for production (including authoring) of SNOMEDCT-AU. Further announcements in the pipeline.
2022-09-25 ML Australia specific FHIR training package in development (not focussed on clinicians, rather focussed on domains, developers and Information Modellers) Ontoserver approaching R5 however level of demand current unclear.
Belgium - Marie-Alexandra Lambot Anne Nerenhausen multi-discipline data exchange programme for care-sets, FHIR objects codified in SNOMED CT (firstly allergy intollerance).
Estonia - Rutt Lindström Kerli Linna switching to FHIR Nationally, preferring international terminologies. No National TS yet, currently capturing requirements. 2022-09-25 KL Developing National Terminology Server, piloting
France - François Macary Phast 2022-09-25 FM Profiling FHIR based on R4 Nationally. Intending to join Snomed International as a member. FM coordinating translation with LP (Canada).
India NRC - Manisha MantriSayali Pophalkar Gaurav Vaishnav Using SNOMED with LOINC and ICD-10. FHIR recommended nationally as a standard. Looking to build FHIR compliant TS at National Level, especially for lookup and validate & expand. 2022-05-11 MM focused on insurance domain profile - just starting (requirement gathering, looking at HL7 profiles). ValuesSets for Patient, Practitioner, Organisation. 2022-07-06 Successffuly imported of multiple code system using HAPI FHIR Server (also pending the beta release of the Snowstorm modifications in this area - next few weeks).
2022-09-25 SP FHIR is being adapted as a standard for exchange in National Digital Health Mission of India called Ayushyaman Bharat Digital Health Mission (ABDM). We have FHIR implementation Guide for ABDM where we have created profiles for FHIR resources as per Indian needs. We are using multiple terminology standards like SNOMED CT, LOINC, and ICD depending on the context. We are also working towards having a nation specific term server who can support multiple terminology standards mentioned above. And we are having hands-on on Snowstorm for that.
Netherlands - Pim Volkert Chantal Schitmeijer Feikje Hielkema-Raadsveld Ronald Cornet - using FHIR based terminology server which they supply to various vendors free to use except for usual licence fees - is an instance of Ontoserver, but not intended for use in "Live" system, instead used as offline supplier of terminology. Attempting to lower barriers to adoption. (Sander Mertens now working for a startup creating EHR system for GPs). 2022-07-06 Hitting technical issues with termserver (suspected scrambled mounting points).
New Zealand - Peter Jordan (HL7) National Terminology Server for NZ currently under development. Peter's own implementation "Terminz". FHIR Release 5 underway eg changes to Concept Map.
New Zealand - PJ Medicines Data Repository. National Terminology Service Mon 5 Dec Connectathon including "Escape from burning platforms"
Norway - Oskar Hurtig using several Snowstorm servers in combination with SNOMED International's Managed Service. Oskar's team in start-up phase looking at the introduction of SNOMED on FHIR for Norwegian customers. Existing environment involves many classification systems (eg HealthTerm) which they're trying to consolidate with FHIR. Developing own FHIR Resources. 2022-05-11 Currently running assessment of various FHIR servers including HAPI, Microsoft. Focused on VitalSigns.
2022-03-16: This past month is that we’ve done some internal training on FHIR to prepare our team and (softly) requested our leaders to help us create a national recommendation on the use of SNOMED on FHIR in Norway. The use of FHIR and SNOMEDCT are separate recommendations already. We’ve also identified a use case where we would like to create some VitalSign profiles with terminology bindings to SNOMEDCT and be able to locate test patient data using a SNOMEDCT code search. This is primarily a way to learn terminology bindings and ask better questions to this forum, as well as to some extent understand our terminology server needs.
Sweden - Daniel Karlsson, Anna Rossander No clear decision made yet on using FHIR with SNOMED but there are a number of projects underway eg with HL7 Sweden. Some competition in this space for Terminologies.
Swiss Annatina Foppa
United Kingdom - Andrew Perry(NHS) Various FHIR profiles in use for moving data between primary and secondary care incorporating SNOMED. National Terminology Server procured (not publicly available - constrained by copyright eg non UK access). Programme of implementations which use the National Server underway. See https://digital.nhs.uk/services/terminology-servers 2022-09-25 AP GP to GP transfer happening in V3. NHS Digitial being brought into the main body of the NHS. Medicine Connectathon happening in a couple of weeks. NHSx for interoperability also re-absorbed.
2022-04-13 JR working with MS Excel add in to call FHIR $lookups from cells. Issues with oauth in that environment.
|Snowstorm support for other CodeSystems
Support based on HAPI library support for generic CodeSystem imports (doesn't allow for loading maps or valuesets)
ML Question on motivation for adding support for other code systems to Snowstorm KK Was at the request of members. Also for a SI use case, we'll have access to the terms associated with published core mappings so we can be more expressive than just returning "Q.74" or whatever the code is.
2022-07-06 Question about specific vs general support. Specific code written for LOINC and ICD-10 / ICD-10CM to allow import in native format Existing HAPI support for generic code system format is also available. CodeSystems which declare "IS A" semantics allow parent child ie $subsumption testing to be done. DK asked if non-international ICD-10 maps have been loaded - they have not.
2022-08-31 KK Snowstorm-X fork of the SI master Snowstorm project (beta) allows for loading packages from Simplifhir (pre-downloaded).
|R5 changes suggested around specifying language using BCP-47
See Tuesday meeting item #6
Note that where a language reference set pulls from more than one language, the BCP-47 tag could potentially missing out the <lang> element and start directly with "x-". Options for referring to multiple languages: could be done either with a * (see https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Accept-Language#syntax ) or with a comma separated list to indicate a priority order (although this isn't supported by the FHIR parameter (unless the server went out of it's way to allow for this), or use the accept-language header). Use case is in the displayLanguage input parameter to a ValueSet $expand operation
2022-08-31 ML CSIRO attempting to get R5 compliant Ontoserver ready in time for HL7 Connectathon.
|Post Coordinated Expressions in FHIR
2022-09-25 ML Main focus of languages working group is around risk, particularly in the transformation from close to user form.
|Clinicians on FHIR
MAL suggests discussing FHIR Resources for clinicians (clinical terminologists).
16 Feb MAL More of a question around teaching resources to bring clinical terminologists on board. Existing information is vast, split between different websites and finding a step by step path through it is very difficult. Something that could be taken forward by HL7 although there is an intersection with SNOMED CT when it comes to developing ValueSets. PJ suggests contacting Director of Education Sadhana Alangar sadhana@HL7.org (US based) in HL7 re "FHIR for Clinicians" . Crossover from OpenEHR may be useful.
|Cardinality on Designation.use
2022-05-11 DK FHIR-36231 has been marked as resolved.
This group recommends using the FHIR "consumer" use where appropriate as this aids interoperability (as opposed to using some custom extension like DesignationUse). It might be helpful for all "consumer" type refsets to have a common parent so that we can detect them and map appropriately to use = consumer in FHIR (Linda Bird?)
2022-08-03 Final recommendation
2022-09-25 RH New location for terminology specific information (Terminology Services Management Group), this is advantageous because the content is no longer subject to FHIR Ballot cycle. See https://terminology.hl7.org/ updates via UTG (Unified Terminology Guidance, but TSMG will take Jira tickets directly)
|HL7 FHIR Update
FHIR Release Schedule : R5 will be trial use Q1 2023. New normative content will be added Q1 2026.
Minimum capabilities https://jira.hl7.org/browse/FHIR-24698 (aimed at R5)
2022-05-11 RH Date for R5 ballot still open
OH question about ConceptMap - maturity level? RH Current updates first need to make it officially into R5 (which is expected to happen at ballot). ML the changes are being driven by increasing maturity, but of course are then 'new' RH option for 'parts' of a resource to be called 'trial use'. Dependent mainly on reports of production level implementations.
|ICD 10 Maps in FHIR
Feikje Hielkema-Raadsveld suggested that the"group" element could be used to transmit additional information.
PJ: StructureMap, although not technically a 'terminology' resource and generally used for model mapping, is a possible solution for complex maps.
|Recording suspected & negated diagnoses
JR: Current UK use case is: primary care physicians do not want to record the available code for "Long COVID" because (a) they may not have any proof that the patient ever had COVID, because they had it before testing was available and (b) even in those patients who tested positive for COVID there is no test to prove that their current chronic symptoms are caused by that. So the proposal is to give them a "Suspected Long COVID" code.
MAL: Belgium use the confirmation code on the resource rather than a possible SNOMED CT code.
But that raises interesting questions about what happens when somebody sends a FHIR resource with code="suspected Long COVID" and verificationStatus= confirmed. Or refuted.
Historically, before the information models got around to adding their own support for contingent diagnoses, there have been codes for a small range of "suspected condition" only within the terminology. There are about 400 in SNOMED today. Of these, almost all are NEVER used in UK Primary Care. The significant exception is "Suspected urinary tract infection" plus a handful of others. Which also raises the question of whether clinicians really want, or should be allowed, to enter contingent diagnoses at all. In the case of the "Suspected Long COVID" code, for example, would it not be better to send what you know to be true e.g. "Post viral fatigue syndrome"?
Peter G. Williamsverification status is an example of an HL7 valueset (with a required binding) where some authorities may wish to use SNOMED CT codes instead. Belgium have created an extension to allow them to do this. Is there a way forward with the specification that would leave to a more interoperable outcome FYI Rob HausamPJ suggested that the binding strength be loosened eg to extensible or preferred (would go through the Patient Care Workgroup). Suggestion that any ticket raised is shared with this group and also that country HL7 Affiliate Chairs are involved (interest from at least 4 countries here). FMc states issue that existing valueset is unsatisfactory for clinicians because"suspected" is missing.
MAL provided: 410605003 | Confirmed present (qualifier value) | 410592001 | Probably present (qualifier value) | 415684004 | Suspected (qualifier value) | 410593006 | Probably NOT present (qualifier value) | 410594000 | Definitely NOT present (qualifier value) | 261665006 | Unknown (qualifier value) | 723510000 | Entered in error (qualifier value) |
PW: Given that this is a required binding to a codeable concept, could we add in the SNOMED CT code as a coding to the existing HL7 Code. This would force an equivalence that may not be generally agreed with.
JR: Brings up use case of differential diagnosis although this is supported in the HL7 ValueSet.
MAL: There are SCT ways to qualify a diagnosis, see << 106229004 |Qualifier for type of diagnosis (qualifier value)| could be added as an additional field.
SS: Differential Diagnosis is often used in Germany as a secondary option from the most likely diagnosis. MAL suggested DD was used more often when there was a set of more equally likely diagnoses. JR: UK doctors use the order of the DD list ie most likely first. RC: This was the point of SNOMED CT was that the logical definition of concepts allowed for the detection of equivalence.
SS: Also to note that some languages like German would express suspected asthma as a single word, which needs to be computable.
PW: Preference here, in general, would be to use the fields in the information model (avoids excessive pre-coordination in FHIR or forcing others to use PostCoordination) and then constrain profiles to remove the SNOMED codes where things like verification are included in the concept to avoid ambiguity or worse, contradiction.
JR: The AdverseEventActuality valueset looks to be a broadly similar idea for the AdverseEvent resource, but is also different.
2022-04-13 Suggestions for a way forward here? PW Something to raise with Rob Hausamfor the HL7 Vocab working group (also mention the transversal nature of this VS - MAL). JR List of options eg profile out the codes that leave us open to contradiction. Guidance on how to interpret various combinations eg a confirmed diagnosis of a SNOMED code that indicates "suspected". DK Refers to the previous work done that resulted in orange mappings Free SNOMED CT set for FHIR MAL: Also discussed on the EAG call at the Business Meeting. Also inconsistency of these "certainty" fields in various FHIR Resources. Also query around level of certainty when several potential diagnoses are presented. ML: Suggestion that several condition objects would be linked in this case. PW Can I say: "This group recommends that profiles should be used to remove these "contextual" SCT codes from ValueSets which could contradict other fields (AP 'Semantic Clashes') in the FHIR Resource." ? DK We may already have implementations that are using something like 722545003 |Suspected rabies (situation)| and has decision support systems that would look for this. ML "The verification status pertains to the condition, itself, not to any specific condition attribute" (ALL this part of the spec was thought to be ambiguous. Does this mean just the condition code or the whole Condition resource ie are we verifying all aspects of the resource? ML Adding in a SNOMED code as an additional coding in a CodeableConcept has a little wiggle room due to "differences in granularity". TB Provisional (or "Working Diagnosis") is often used when a disorder perhaps takes a long time to fully confirm eg Motor Neuron Disease and in fact a treatment plan may already by underway. So it's further down the line than "suspected". JR Suspected may also be used when a condition has cleared up and it is no longer to test/prove that that is what it was. MAL notes that the Refuted status is valuable in terms of saving time and money. General agreement to avoid the situation hierarchy (No known allergy) and instead pass the allergy condition with the refuted flag. (Note: psychological feature for humans to fail to hear negation!). MAL Suggestion to tag the EAG on this Jim Case also with a view to being clearer about the definitions of these concepts (eg suspected diagnosis) in SCT to either align or clearly differentiate with the FHIR VS. Discussion about "Family History Of" and complications around roles vs sex see HL7 Gender Harmony Project. (see "In the Land of Invented Languages" - ML)
Interest here was specifically on FHIR Education for Clinicians.
Peter Jordan to discuss with Sadhana, still pending.
MAL would like to compile background information for Clinicians to approach FHIR (with SNOMED CT) and the rules around what you can and can't do with FHIR Resources such as the cost / pros / cons of creating FHIR Extensions.
See also Hospitals on FHIR: https://www.linkedin.com/pulse/hospitals-fhir-launch-event-catherine-e-chronaki/
MAL: More complex examples fully expressed with medical specifics ie real life problems including business rules.
Belgium looking at specifics of substance use / misuse / abuse. Alcohol has some representation in FHIR but not other substances. SNOMED CT coverage considered inconsistent (ie a line gets drawn where an observation becomes a disorder), would be keen to work in tandem so that any changes in the SCT hierarchy have an intended destination in a FHIR resource. PJ suggested solution use FHIR Resource as building blocks. MAL would like to see guidance so that whatever is put together can be commonly shared.
Peter G. Williamsto bring up internally that the Content Team is under represented in the working group.
DK Linking observations to diagnoses then CarePlans is a common problem and links between them may be spotty and open to interpretation. A generic discussion about linking resources would be useful. Secondly lifestyle factors (Social Determinants of Care). Both a content and terminology binding issue. Some modeling needed in SCT but may not be scaleable. If not, information model support would be needed.
MAL Non-substance addictions (also habits) not well represented in SNOMED CT
Copied from 2022-03-22 Agenda: https://jira.hl7.org/browse/FHIR-29821 "Updated cardinality of designation.use" Expected to be part of R5.
2022-02-22 Add HL7 Jira ticket about designation use context and designation.additionalUse. For discussion on HL7 Vocab call March 3.
2022-03-08 See also https://jira.hl7.org/browse/FHIR-36231 "Add designation context to designation.additionalUse"
Feedback from this group: relying on the order of elements in the additionalUse array could be tightened up. If a new element is being proposed, perhaps we could add some structure/hierarchy so that the relationship between refset and acceptability could be more explicit.
DK Compromise using BCP-47 standard seems hopeful. designation.additionalUse being added with 1..* but DK/ML/RH thought that this would be confusing (more than one way to do things)
DK Suggests implementing BCP-47 in Snowstorm.
PW Has documenting examples of these on TODO List.
FHR: NL have patient friendly text definitions - how is that going to work?
RH: Are there any remaining issues for the R5 deadline Aug 21?
ML: Extensions of the BCP-47 language tag specifications does not require any changes to FHIR specifications.
RH: This use of BCP-47 with SNOMED CT still needs to be documented.
Proposal to make a change proposal to the HL7 Using SNOMED CT with FHIR page, and particularly the section Display Terms for Specific Languages. https://www.hl7.org/fhir/snomedct.html#22.214.171.124.4
In this section we should describe:
There was an agreement in the group that the BCP-47 extensions should have the form <lang>-x-sctlang-<language reference set id>. Note that according to the BCP-47 specification, any string following the "x-" should be divided into at most 8 character sections divided by hyphens. https://tools.ietf.org/rfc/bcp/bcp47.txt
Further there was an agreement in the group that the list of aliases supported by ECL should be referenced, and that there should be one list consisting of aliases for any international or extension language reference set.
A link to a SNOMED-hosted page with more details about the use of languages in a SNOMED+FHIR context has been created (but not filled with content) and should be linked from the HL7 page above. FHIR-languages
The change proposal should be drafter on this page: Changes for R5 with a deadline for the Aug 9 SNOMED on FHIR call.
Next step is to add content to the FHIR-languages page.
|Allergy Resource Issues
Rob Hausamhas reviewed and document now in final draft (see previous agenda attachment)
Bruce presenting at Clinicians session Wednesday 12:30UTC
Annatina Foppahad issues with the 2 profiles in Switzerland (findings vs substances focus)
SS Notes that one could determine the substance from the finding. MAL in fact you could go both ways and having access to the entire range of substances. Further confusion caused by negation fields vs SNOMED CT concepts
FMc If using the Finding concept, the substance could be separately included. Would be keen to see preferred choice stated, or at least for each country to declare their implementation
DK Notes that IPS does not attempt to constrict the list.
|Metadata in FHIR
RC: Particular interest in the provenance of information. Interested in relationship between FHIR and FAIR (findable, accessible, interoperable and reusable) and metadata around patient eg why was patient included in study, potentially metadata for the dataset, rather than just an individual. FAIR See https://confluence.hl7.org/pages/viewpage.action?pageId=91991234 Findable, Accessible (conditions and ways rather than "open"), Interoperable & Reusable.
PJ Suggested looking at the bulk capabilities in FHIR (Zulip stream - https://chat.fhir.org/#narrow/stream/179250-bulk-data )
Ronald Cornetwill give updates at further meetings. What metadata are we interested in (question could be put to the MAG)
PWI Has interest in metadata from SNOMED as being progressed with annotations in the MAG. Also notes that the HAPI library has (had?) issues with aggressive caching of the Capabilities
TODO Ask Ronald about categories of Metadata. Done. NISO in US have specified this eg syntax.
PJ HL7 Vocab collaborating with OMOP on this (meeting weekly). RC OMOP well aligned with FAIR.
See also CDISC - Clinical Data Interchange Standards Consortium (focus on clinical trials cf OMOP using real world data)
2022-05-11 RC currently working on metrics to determine if datasets are 'fair' and detect gaps with a view to augmentation.
2022-09-25 RC Exchange of data may mean copy and pasting of data, rather than sharing of original source. So errors will be propagated or even re-introduced.
|NLP / Text Annotation
Most real world data is held in narrative. To train data, annotators are needed who should follow annotation guides to ensure consistency. New guidelines are needed which are informed by FHIR. Question from FMc about limitations of using German (eg availability of tooling).
DK Compound languages need to be able to "decompounded" for translation & analysis. Germany provides the majority of this work. Snowstorm does not do matching on compound words well (doesn't recognise same root)
2022-09-25 SS Early stages of annotation of medical records. Also looking at interface terminology
|Known FHIR Implementations / National Terminology Servers
|Definition of GPS
Request that SI make the GPS ValueSet available as an extensional definition.
SI are (I think - PWI) in agreement to do this.
16 Nov Update: "Search After" functionality will be available in Prod 1 Dec.
|SNOMED FHIR Implementation Guide
IG Documentation: http://build.fhir.org/ig/FHIR/ig-guidance/index.html
|Any Other Business
Make more time available for looking at the discussions in the previous 20:00 UTC call.
Terminology Binding Question: Observation (from 2019) "No need to to replace existing HL7 Interpretation Valueset with a SNOMED CT one as it is mature / metalled" - DK to report back on decision from X-eHealth project.
Potential Items for Discussion
|Notes & Actions
|Notes & Actions