Guidelines for report out:
1. 1 paragraph summary: what was the track trying to achieve
2. 1 list of participants (with logos if you have time and energy)
3. 1 paragraph: notable achievements
4. 1-2 screenshots if relevant and interesting and/or links to further information about implementations/achievements
5. 1 paragraph: discovered issues / questions (if there are any)
6. 1 paragraph: now what?
The first goal of the CDS Hooks track was to facilitate CDS Clients and CDS Services with their implementations of the 1.0 specification.
The second goal was to work on defining updates for the next version of the spec. CDS Service providers at this connectathon were mostly focused on creating their first CDS Services, so the 1.1 exploration was primarily discussion among CDS Clients, along with GitHub activity which we hope to try out more concretely in San Antonio.
Cerner and Epic integrated with Alphora’s CDC Opioid Guidelines services.
Cerner and Epic integrated with Anthem/CareEvolution’s CDS service which reconciles externally available AllergyIntolerances and Conditions for the patient back into the CDS Client’s FHIR Server.
Telstra Health created a new CDS Service for recommending medications for patients with acute sinusitis, based on a variety of criteria including allergy to penicillin, creatinine level, age, weight, and whether they are taking Coumadin. They integrated this service with the public CDS Hooks sandbox.
CDS Hooks Specification pull request activity:
● Merged enhancement to add selectionBehavior option of 'any', allowing more than one suggestion to be accepted - https://github.com/cds-hooks/docs/pull/498
● Merged enhancement to add preference indication to suggestion, allowing one suggestion to be indicated as the preferred option among multiple choices - https://github.com/cds-hooks/docs/pull/497
● Discussion around adding a topic to a CDS card to indicate the nature of the guidance (e.g. Zika virus management) - https://github.com/cds-hooks/docs/pull/505
Discovered issues / questions
JKM Software and Telstra Health brought up other forms of decision support. CDS Hooks is focused on synchronous actions providing support to an active user. They have use-cases of:
● Subscribing to changes in data (e.g. new Observation) in addition to user events
● Providing clinical support either by
○ creating a task / messaging the user’s inbox so they can asynchronously respond next time they log in. Or,
○ Trigger a series of orders/tasks or creation of other FHIR resources. Or,
○ Have the CDS Client return a value (e.g. a rating, risk score) that the CDS Client can then act upon with its own logic.
Some of the above may involve:
● using FHIR subscriptions
● allowing a service to create FHIR resources or call clinical messaging APIs
● using the system actions extension
● other extensions for specific workflows
JKM Software and Telstra Health both asked to document use-cases for examination so we can move towards clarifying workflows and creating recommendations.
CDS Hooks PRs needing further discussion and trial implementations in San Antonio:
● Communicating Card topics - https://github.com/cds-hooks/docs/pull/505
● Acknowledgement Reasons - https://github.com/cds-hooks/docs/pull/513
● Feedback Service - https://github.com/cds-hooks/docs/wiki/Feedback-endpoint-for-CDS-Hooks-1.1
Alphora plans for San Antonio:
● Continue testing Opioid Guidelines Services
● Add suggestions for creating a ServiceRequest, such as urinalysis
● CDS Client ability to accept ServiceRequest. (Prior connectathons already used ServiceRequest with radiology orders in the context of PAMA scenarios.)
- Test and seek feedback from a broad set of Australian and international implementers on the usability of the proposed FHIR implementation guide.
- Increase awareness of the proposed implementation guide and allow vendors to gain early experience in implanting against it.
- Test the logic and capabilities within the FHIR server against multiple different implementations.
Ronald Dendre - University of Queensland
Gowrie Siriman - Clinical Coder - Sydney Children's Hospital / Royal North Shore Hospital
Heath Frankle - Ocean
Oliver Krauss - University of Upper Austria
Marvin Malcom - Telstra Health
Brett Esler - Oridashi
Min Zaw Lwin - Collaborative
Blair Thompson - Collaborative
Nichol Hill - Collaborative
Shovan Roy - Collaborative
Andrew Patterson - Melbourne Geonomics
William Durnford - BP
Thomas Clarke - BP
Lisa Nelson - Gravity Project USA
Jim Steel - CSIRO
Kate Ebril - CSIRO
Mathew Cordell - Australian DIgital Health Agency
- Multiple parties succeeded in interoperating with our hub.
- Good discussions on the project objectives and how to look at aliment with similar projects (e.g. primary care, gravity, SDC).
- Some good feedback on where our approach worked well and some valuable feedback on where some further clarity is required.
- Good discussions on approaches to common form collection and local repository of information.
- Technical issue on one of the API response. The response doesn’t match specification.
- Technical issue on making
- Feedback received on the profiles using Observation FHIR resource and how the ontology for Observation outcome has been modeled.
- Feedback received on using CarePlan resource in the context of Child Health Check Assessment and Schedule.
- Suggestions received on how to extend the use case and introduce analytics on the data received in Child Data Hub
- Remeditate technical issues
- Further discussion on fhive modelling feedback
- Develop an overview of potential data capture flow and discuss as a team.
The Clinical Reasoning Track focused on continued testing and evaluation of decision support and quality measurement. This round we focused on testing CQL-based decision support integrating through CDS Hooks with Epic and Cerner environments, as well as discussions around next steps for the CQL specification, and coordination with the EBM-on-FHIR track on expressing EvidenceVariable examples.
Integration of CQL-based CDC Opioid Prescribing decision support with both Epic and Cerner (see the CDS Hooks track report for more)
Coordinated with EBM-on-FHIR to express 8 EvidenceVariable resources (see the EBM-on-FHIR track report for more)
* Fixed several issues with the CDS Hooks Service implementation
* Discussed approaches to supporting scope discovery based on decision support content
* Discussed approaches to recursive query support in CQL
* Started working on expression of a decision table for cancer screening registries as a PlanDefinition
* Started expressing QI Core profiles using FHIR Shorthand
* Target returning a ServiceRequest from Opioid Recommendations in the CDS Hooks Service
* Continue exploring potential use of FHIR Shorthand for QI Core profiles
ePrescribing Track (Active Script List)
This track aims to offer vendors across acute, primary and residential aged care an opportunity to implement the new paperless ePrescribing workflow end-to-end and simulate the workflow for prescribers, dispensers and patients in a paperless prescription world. It also aims to provide vendors who are not currently experienced in using FHIR and opportunity to work with vendors who are familiar or more experienced.
Integration to the new ePrescribing APIs is a key success factor in rolling out paperless prescriptions. ePrescribing changes to support paperless prescriptions impacts all vendors.
Danielle Bancroft (Fred IT Group) - Australia
Paul Barry (eRx Script Exchange) - Australia
Greg Pryde (RxOne) - New Zealand
Alberto Isidro (Cerner) - Australia
Ryan Davis (GuildLink) - Australia
William Durnford (Best Practice) - Australia
Craig Schnuriger (MedAdvisor) - Australia
Anjani Jaswal (Fred IT Group) - Australia
Manjeet Singh (eRx Script Exchange) - Australia
Joaan Vargheese (Fred IT Group) - Australia
Sarwar Erfan (Modeus) - Australia
Weyn Ong (MedAdvisor) - Australia
Brian Postlethwaite (Telstra Health ) - Australia
Marc Belej (All Scripts) - Australia
Andy Robb (InterSystems) - Australia
Jesus Nunez (Rosemary Health) - Australia
Danielle Friend (Epic) - United States
Lloyd McKenzie (Gevity) - Canada
● Fixed and updated IG issues.
Registered patient successfully
submitted medication order from prescribing system
retrieved patient MySL in patient app as medication request in patient APP
rendered as token
scanned and pulled down and dispensed by pharmacy system
Received token and added to dispense queue
pull down script and mark as dispense
remove script/replace with new repeat on MySL
Issues identified during interactions:
● Missing drug codes in Medication Request (added to support both AMT and PBS+MF)
● Use case for queue mgmt software to access MySL - Authentication needs investigation as cannot use site PKI
● Patient match operation needs review
● Need to ensure patient validation exist to avoid more than one patient being addressed
● Request consent for org has error in the xml example
● Profiles need to be reviewed with regard to minimum mandatory fields etc
● Request consent for org has error in the xml example
Identified certain workflow scenarios requiring further work in the next phase:
● Edit patient call
● Carer mode access
● Identifier for patient app to use (cannot/don't have IHI)
● Flag med in request as hidden
● Mapping of pharmacy site IDs from third party apps against PDS site IDs
● Patient App auth model to be reviewed
Evidence Based Medicine (EBM) on FHIR Track
Goal was to demonstrate the first useful clinical decision support tool using actual FHIR Evidence Resources
Brian S. Alper, EBSCO Clinical Decisions
Khalid Shahin, EBSCO Clinical Decisions
Muhammad Afzal, UC Lab, KHU/Sejong University
Bryn Rhodes, Database Consulting Group
Updated Evidence Resource in R5 FHIR Build to reflect CDS WG-approved changes
Created 67 Evidence Resources, 63 Group Resources, and 8 EvidenceVariable Resources to represent the evidence base for a thrombolytics for acute ischemic stroke decision support tool
Resolved multiple challenges to create these Resources including:
1) creating an ActivityDefinition Resources to codify complex dosing regimen
2) creating a CQL Expression to codify EvidenceVaraible.characteristic.valueExpression representing a set of values on a scoring tool
We started with published evidence reports (conveying the risk of being alive without disability 90 days after stroke onset, and the risk of having a fatal intracranial hemorrhage within 7 days, both with and without alteplase treatment, in patients presenting within 4.5 hours of onset of acute ischemic stroke) :
We converted selected evidence and statistics to computable expressions with 139 FHIR Resources:
How it could be used clinically:
Clinician fills in details to obtain ASTRAL score (individualizing risk prediction for the patient)
Once the ASTRAL Score is completed, the relevant Evidence Resources are accessed and a decision support tool could look like:
FHIR Shorthand Track
● FHIR Shorthand is a domain-specific language (DSL) designed for the job of profiling and IG creation
● It is a textual language
● FSH is concise, understandable, and aligned to user intentions
● Formal grammar (ANTLR4)
● Reference implementation compiler (SUSHI)
● Some users might find that the FSH representation is more understandable and agile
● FSH is ideal for collaborative development under source code control
● Supports meaningful version-to-version differentials, and merging and conflict resolutio
● Ease of refactoring through global search/replace operation
● These features allow FSH to scale and accelerate development in ways that other approaches cannot
OBJECTIVE OF CONNECTATHON
● Allow track participants to gain hands-on experience with FHIR Shorthand
● Try producing an Implementation Guide using Shorthand
● Provide feedback on the FHIR Shorthand specification and SUSHI tooling
● Log any issues with Shorthand and SUSHI
● Consider using Shorthand for your next IG project
● Mark Kramer, MITRE (Track Lead)
● Rick Geimer, Lantana
● Sarah Gaunt, Lantana
● David Hay
● Ward Weistra , Firely
● Nick George, Google
● Eric Haas
● Natasha Singh, Childrens Hospital of Philadephia
● Hugo Leroux, CSIRO
● Alejandro Metke, CSIRO
● Michael Lawley, CSIRO
● John Rhoads, Phillips Healthcare
● John Moerke, By Light
● Bob Milius , NMDP/CIBMTR
● Andres Schuler, HL7 Austria
● Reinhard Egelkraut, HL7 Austria
● Vince McCauley
● Held a tutorial to teach FSH
● All participants were able to learn and use the FHIR Shorthand and successfully create implementation guides
● Validated the FSH concept as something that is useful and needed by the FHIR community
● Several participants expressed their intention to using FHIR Shorthand on current and future IG projects
● Google and Lantana expressed interest in FHIR Shorthand to augment existing FHIR toolsets
● Rick Geimer developed a translator that converts StructureDefinitions to FSH
● Rob McClure inspected our ValueSet implementation and did not tell us we were doing it wrong
FHIR Shorthand Zulip channel (#shorthand)
FHIR Shorthand Documentation and Tutorial
FHIR Shorthand Documentation Issue Reports
SUSHI Code Repository
SUSHI Issue Reports
HL7 Confluence Site
See http://hl7.org/concalls for details
● About 10 issues (both bugs and suggestions) were filed in our SUSHI git repository
● Error messages were appreciated, but red color on black was not appreciated
● Version compatibility between FHIR, FSH, and SUSHI
● Users want to refer to to local packages in package.json
● FSH needs a way to designate which instances are examples, as opposed to conformance resources
● FSH needs some better documentation in some places and more examples
● A high priority will be to improve the relationship between FSH and the IG Publisher
● One of these tasks will be to put generated resources into multiple folders
● Provide syntax highlighting and look ahead
International Patient Summary Track
The primary focus for the IPS track this time was finishing preparations for the upcoming publication of the IPS FHIR IG.
We had productive conversations around issues with global representation of patient names around requirements for use of given, family and full ‘text’ descriptions of names in the IPS profile of the FHIR Patient resource.
○ Reached out on Zulip to solicit further comments on this issue and how to resolve it for optimum global interoperability.
We identified the desirability to remove the additional constraints on Patient.address subparts below the Patient.address element itself.
We also investigated the proper licensing text to use for the SNOMED CT Global Patient Set and how to define the GPS-specific content of the SNOMED-based value sets and how to represent them in the IG.
JSON-LD/RDF Implementation Track
We tried out some JSON-LD 1.1 contexts that we are now generating in the FHIR build process that, in combination with some supporting tools, allow us to transform vanilla FHIR JSON to the current FHIR R4 RDF standard. We also tried tooling that converts in the opposite direction: from FHIR-”like” RDF back into FHIR JSON. The JSON-LD Contexts that are generated also hold the promise of providing a “semantic anchor” for the FHIR model, providing formal documentation from other communities.
We also demonstrated samples and tooling that prototyped how one might maintain an active FHIR triple-store back end, as well as how one might use RDFa-like markup on clinical documents.
We had one participant, Damian Brede, who was relatively new to FHIR.
This ended up being largely a one-person work session, and we (I) used the time to address some issues in the Python Shape Expressions (ShEx) processor as well as to develop an alternative to BNode “reification” using the GraphDB triple-store -- something that should be useful in the long run
In general, while there is a relatively large community from the biomedical and translational research community that is anxious to consume (and produce!) FHIR using RDF, they did not come to this event. However, we plan to do some webinars to share our progress with them.
LOINC - InVitro Diagnostic (LIVD) Mapping Track
Validate the representation for non-quantitative IVD tests' result values mappings to LOINC and/or SNOMED towards the next, second version of LIVD Implementation Guide.
Hans Buitendijk (Cerner) - USA
Rob Hausam (Hausam Consulting) - USA
Yatish Gupta (Google) - Canada
Ralf Herzog (Roche) - Switzerland
Detected an issue in the implementation guide, regarding the context of the capability-extension of the device definition.
- Fixed the Implementation Guide (to be able to validate the examples again)
- Discussions regarding Result Value Mapping which will result in an improved profile (discussions around how to link from value sets to code system and Concept Map)
The Patient track continues to provide a useful way to introduce the FHIR specification while providing hands on experience with the RESTful API.
FHIR Connectathon 23 in Sydney highlights:
● 15 participants from 7 organizations and coming from 4 countries
● 3 FHIR servers tested -- in addition to publicly available test servers
● 4 FHIR clients used for testing -- in addition to multiple participants using Postman
● successful Touchstone testing with 2 FHIR clients and 2 FHIR servers
● many learning opportunities and discoveries made along the way
● participants also branched out and found success on many other tracks as well
Primary Care Track (Practice-to-Practice Record Transfer)
The aim of the track was to validate the Australian Primary Care Practice-to-Practice Implementation Guide, to assist vendors with implementing it, and to explore the next steps, in particular the inclusion of social factors and how they can be represented in the record.
○ Jim Steel, CSIRO (track chair)
○ Lisa Nelson, Max MD
○ Christol Green, Anthem
○ Thomas Clarke, Best Practice
○ Heath Frankel, Ocean Health Systems
○ Frequent visits from Nicholl Hill, NSW Health
We had very productive discussions with other groups, particularly child health and the Gravity project, about alignment of clinical resource representations and use of document sections. We fell a bit short of exchanging resources, but we had productive discussions about prioritising which sections for vendors to concentrate on.
For social factors/determinants, we found some overlap between Australian instruments’ requirements for accessing social information in the clinical record, in particular in the Australian MBS 7-series of assessments, and the work being done by the Gravity project, but we see a real need to gather up the various sources and uses of this data in other assessments and tools, and how they are used downstream in clinical and analytics workflows. Doing that should help us to work out how to represent them both in resources and in terminology.
● What was the track trying to achieve
This track focused on testing the modeling and information exchange mechanisms documented in the SDOH-CC IG for Use Case 1
● List of participants (with logos if you have time and energy)
○ Lisa Nelson
○ Christol Green Anthem
● Notable achievements
○ Good discussion to confirm use of the QuestionnaireResponse for the “model of use” screening tools that have standardized (and when possible validated) questions and answers, and the use of an Observation profiled for a simple assertion for a Clinical Finding where the code comes from the SNOMED CT Clinical Findings hierarchy and the value is present or absent.
● 1-2 screenshots if relevant and interesting and/or links to further information about implementations/achievements
● 1 paragraph: discovered issues / questions (if there are any)
○ Implementers agree it would be a good start to agree on the modeling for SDOH observations.
○ Seems like there is pretty strong consensus aroudn the use of Observation with a code from an extensible SNOMED CT Value Set.
○ Also seems to be agreement that some Observations might be in multiple places, for example Ethnicity
● 1 paragraph: now what?
○ We need to be clearer about how to help Connectathon Participants to be ready to interoperate using SDC, CDex and SDOH-CC task transactions
○ Our prep sessions need to help implementers identify the role their system will play and how to get ready.
○ We need to sort out how to represent the Screening Interpretation Observation separate from the Clinical Assessment Observatation. The Screening Interpretation will be a searchable resource associated with the Questionnaire which could support, for example, a particular Screening that had a certain interpretation result.
Structured Data Capture
Brian Postlethwaite (Telstra Health - Australia)
Michele Mottoni (Care Evolution - USA)
Paul Lynch (SU National Library of Medicine NLM - USA)
Buminda Nawagamuwa (Telstra Health - Australia)
Steven Edge (Genie Solutions - Australia)
Yana Beda (Beda Software - Russia)
Ilya Beda (Beda Software - Russia)
Alexander Bartschke (BIH - Germany)
Lloyd MacKenzie (Gevity - Canada)
Vladimir Smirnov (Medlinx, Russia)
3 rendering engines, 6 data providers, 2 $populate implemantations
The example populate demonstration was shown and tested by several attendees
Michele is attempting to import a FHIR Questionnaire definition into his internal structures
Yana/Ilya are attempting to export their internal form structure to the FHIR Questionnaire definition, and produced a demo app: http://ui.hl7.beda.software/
Managed to use the Populate demo page to use a questionnaire definition from TelstraHealth, populate code from Beda software, data extracted from Oridashi, and rendered by Telstra Health! (now that's some interoperability)
Brian discussed with the Primary Care track getting their questionnaire definitions going, also spoke with Lisa Nelson regarding using the Telstra Health populate demonstration for their Social Determinants of Health guide and how to leverage existing content.
Alexander was able to create some new questionnaires with the LHC Forms editor, and render them with the Telstra Health Renderer
Thomas Clarke from Best Practice was able to Substitute his data server for the Oridashi one and successfully provide Practitioner and Patient data, and is hoping to be able to provide DocumentReference data tomorrow.
Testing the pre-populate with alternate data providers: Oridashi (works), Health Intersections (works), Pyro (get works, however no batch search support), DXC (get works, batch support CORS preflight issue in browser)
Some of the lesser known extensions in SDC were discovered, and some clarifications added
Not all servers support the batch search capability - encouraged its adoption
Swapping $populate servers went well
Using the FHIRpath based populate seems pretty easy to understand and adopt
Extracting from entered QR into resources hasn’t been tested extensively as yet, and more gaps in this space
There are still some rendering/interactions gaps in the spec, and will get these logged
Further refine the SDC guide, and complete the STU1 publication
Sun 2:30 - 3:30 NLM FHIR Tools Overview ( https://lhcforms.nlm.nih.gov/ ) (Paul Lynch)
Telstra Health Tech Overview (Brian Postlethwaite)
General SDC discussions
Sun 3:30 - 4:30 Querying answer lists as a property of a selected answer
(Paul Lynch) http://rxterms.nlm.nih.gov
How to select content to be available as answers (e.g. read the list of conditions from the patient and permit the user to select them) - Brian Postlethwaite - using this extension?
Monday - we should do a breakout session to discuss the gaps that have been identified in the spec from the days experimentation
The primary goal was to review the changes that have happened with R4 and see if there is additional feedback that we need into the specification. Initial discussion during the start of the connectathon led the group to choose the R5 build version (rather than the Argonaut branch with more updates including renaming Topic to SubscriptionTopic: http://build.fhir.org/branches/argonaut-subscription ).
● HealthOne NZ
● Telstra Health
● Cerner: We hosted a server that supported the admission Topic (initially) and added a labs Topic on the second day. We were able to confirm notifications to several external rest hook clients. Encountered and corrected bug with our meta lastUpdated on Topic, was unable get get bug for lab notifications fixed during time.
● Apple: Were able to write subscriptions and get ping notifications back.
● Grahame: Was able to catch up on spec and provide feedback for Topic filters (See now what)
Grahame brought up some concerns with the way filters work with criteria today. Ongoing discussion right now on zulip: https://chat.fhir.org/#narrow/stream/179229-subscriptions/topic/SubscriptionTopic
The argonaut branch needs to finish applying some pending changes (eg: SubscriptionStatus) so that we can use those updates/get more feedback in the next connectathon.
Terminology Services Track
● Good track attendance, with considerable representation and contribution from Australia
● Breakout on SNOMED on FHIR terminology services
○ URI standards for use with SNOMED CT in FHIR
○ Specification of SNOMED CT “language” (or context) refsets as a parameter to value set $expand to specify the context-specific “preferred” terms to be returned by the expansion (e.g., for specific languages, “patient friendly” terms, etc.)
■ SNOMED on FHIR group will define the new parameter in the SNOMED on FHIR IG
● Breakout on issues and the recent completed and proposed changes to the ConceptMap resource
○ Overview of the proposed and completed changes for R5 currently in the CI build
○ Final codes to be used for “broader” and “narrower” relationships
● Terminology server testing
○ Discussion with Epic on how to begin testing the new terminology service they are developing.
● SNOMED Int. accomplishments
○ Developed a prioritised list of "Must Haves" for a FHIR Terminology Server implementation
○ Uploaded the IPS as a ValueSet to the development Snowstorm FHIR server and shared examples of specifying (with ECL!), creating and accessing that with the IPS track.
○ Answered a number of questions on SNOMED CT, SNOMED International and our open source tooling.
Multiple questions answered on a variety of terminology content and terminology service topics with multiple groups
Track Proposal Template