10:00 AEDT
Meeting 35
Attendees
Andrew McIntyre (Co-Chair)
Michael Legg (Co-Chair)
Liam Barnes
Gil Carter - Voronoi
Jakub Sielewicz
Nick Ferris
Madison Black
Dalisay Giffard
Cindy Chiu Chong
Lawrie Sim
Loma Avery
Appologies
Gil Carter - Voronoi engaged by RANZCR to perform landscape scan. Preferred terminology set, recommendations to ADHA.
Nick updated from RANZCR workshop from Friday:
- desire not disrupt workflow of referrers by introducing new terminology (names of procedures) - don't make them learn new terms
- survey of what was is in use. terminology sets.
- majority so far of replies - MBS 13/22 (doesn't fully cover what is done)
- locally hand made catalogues 9/22
- 5/22 state health department catalogues mostly in Qld
- 4/22 claimed SNOMED
- 4/22 claimed LOINC
- ACR (American College of Radiology) system
- what would motivate their choice
- 17 billing
- integrated with other systems.
- 12 legacy reasons
- EMR of the hospital they do the work for
- what they would like in a selected catalog
- fit for purpose
- easy to use
- Perceptions
- comprehensive
- fit for purpose
- Discussion about challenges in adopting a catalogue
- mapping
- maintenance
- Candidates
- RADLEX (developed in 1990s, after review of SCT, harmonised with LOINC in last 5 years. Structured Ontologies. Large vendors are claiming to use it. Not widely used in AU. Not in practice wide use in US)
- SNOMED
- Mapping between the two is lacking artifacts
- mapping to MBS is useful
- mapping to and from snomed is supported by tools
- need to be manageable by small practice through to large enterprises
- local Australian version control would be important
Interested to hear ideas on selection criteria
Most terms in radiology are organised by modality and anatomic region of interest.
Both HL7v2 and FHIR both use Table 0074 and a coded terminology for test ordered.
Protocols
Michael Legg: suggested to establish principles of terminology - refer to SPIA for what has been done previously in pathology. Determining synonyms frequency. Agree on information model before get too carried away with terminology. Pre and post coordination becomes important.
preferred term unambiguous but most acceptable to largest number of users
NCTS
SCT was chosen for pathology. would be nice if radiology aligned as they may have reason to share codes
How can terminology be used after it is received. can be influential in choice. Once terminology is chosen the hard work begins.
RADLEX: Radiology vendor support
SCT: NCTS supportable (Australian version)
How close disciplines in diagnostics work together?
Andrew McIntyre: SNOMED has fairly official language for primary terms but also supports synonyms
collect all interface terms, abbreviations,
smaller set of common codes. code that tells you what it is.
A neuro surgeon will order a specific sub test of CT Head. Whereas a GP will order a CT Head.
A hierarchy is useful to reason from more specific to more generic. Walk back up hierarchy to
Sub specialised versions of the same thing.
Preferred term goes on the order form, but users search with interface terms. Find it via familiar term.
For decision support need to be able to reason on an order code.
How you can use. Search for all Cat scan for a patient. A patient with 2000 individual results, being able to show all cat scans that is very useful (at report code level). Show me the cadiothoracic ratios for patient and graph over time (search report content).
RADLEX radiology specific terminology - unusable outside of radiology specialty and vendors.
Gil: interested in how iEMR.
Adopting patient common identifier. IHI preferred? What would
Lawrie replied the amount of effort and time that is put in to maintain catalogues across 3 RIS and one iEMR. Enormous effort → getting all terms. looking at term frequency of items in the catalogue. Synchronizing terms across. A smaller catalogue of orderables, eg. CT Head can be ordered but that expands into a list of 20 other sub procedures that that the radiographer can choose appropriate for the condition of the patient. Emphasized that there is an
Code and plain text
Short code and long code (descriptor)
Interface uses code or short code.
Qld Catalogue: Modality, Body part, Laterality. Reviewed in light of SCT, alignable with SCT, minor gaps. Clinical notes - some orderers provide large amount of detail, others provide none to very little
Peter Scott: Systems such as Best Practice choose term via modality and then region, but also free text.
Outstanding Meeting actions:
- Next Meeting to be scheduled for 25th May 2021. 10:00-11:30 AEST (Sydney Time).
Attachments:
Comments:
|
Intro to Radlex here :
|
|
The playbook extract presumably is not the underlying ontology? The pdf from the link suggests it is a 'portion' thereof. Does that mean its an information space like view of terms (ie stuff for humans to read) or is it making the claim to sufficiently define and canonise concepts. Assuming Radlex proper is a (mathematical) ontology key quality metrics that could be applied can be sourced from efforts such as OQuaRE, which builds on the ISO/IEC 25000:2005 Software Quality standard. This would be getting really empirical and needs resources including the source termset as we discussed today. This url contains links to the Radlex owl file which has some metrics already which is great: https://bioportal.bioontology.org/ontologies/RadLex_OWL , and as its available in Web ontology language (owl) format that clarifies in the affirmative, that it is a mathematical ontology. This infers that a reasoner or semantically aware query language could make appeals to it as the first order predicate logic would/should hold true. Mapping a Radlex concept to a SNOMED concept would be possible, as long as folk understand what is really going on with an ontology. Ontologies formally define concepts in terms of its parent- child IS-A relationships eg a 'CT abdo' is a 'CT scan' and its lateral defining relationships, the constrained set of which in SNOMED for a procedure is viewable here. In SNOMED, 169070004 | Computed tomography of abdomen (procedure) | Is_A 303678006 | Computed tomography of regions (procedure) | and Is_A 60654006 |Diagnostic radiography of abdomen (procedure)| ; and its lateral defining relationships { 260686004 |Method (attribute)| = 278110001 |Radiographic imaging - action (qualifier value)|, 405813007 |Procedure site - Direct (attribute)| = 818982008 |Structure of abdominopelvic cross-sectional segment of trunk (body structure)| ). So that is how it is built in SNOMED and all that is 'under the hood' for the apparently simple concept code 169070004 In Radlex looking at https://bioportal.bioontology.org/ontologies/RADLEX/?p=classes&conceptid=http%3A%2F%2Fradlex.org%2FRID%2FRID10321 it appears to be a postcoordinated system. is that right? In any case the attributes look more specific to DI so what I am trying to say is a 'CT abdo' concept means similar but yet different things in the two different systems, and they both would be true. |
|
Yes, the RadLex-LOINCPlaybook is a "core" set derived from the RadLex Playbook, a larger set of procedure terms, which is in turn a subset of the RadLex ontology : for background, examples and further links |