Does anyone have experience connecting the openEHR and semantic web technologies?
An increasing amount of care organisation in the Netherlands have their clinical data available in openehr format.
The regulators require data in semantic web technologies:
A custom ontology for relevant information, RDF store and SPARQL queries for key performance indicators.
For example the number of staff, average time to receive treatment etc.
Currently they are mapping admin data from EHR’s using an ‘exchange profile’ to the ontology. That mapping is custom per EHR system (and requires fine tuning for each care organisation). And they are interested to move into clinical performance indicators. Where data may be available in a standardised way in openEHR.
Now those regulators have requested me to advise if openEHR could help with those challenges.
The key question I have is how to efficiently make data in openEHR available to perform the SPARQL query (e.g. time after admission till Advance care plan recording).
One of the scencarios is mapping to an ontology, exporting data and then running the SPARQL. A bit similar to EOS/OMOCL by @SevKohler.
Does anyone have experience with a similar project?
Hello Joost, in Spain several groups have worked on that regard. Both Jesualdo Fernandez Breis from Murcia University and Santiago Frid from Clinic Hospital from Barcelona have worked extensively with the OWL/Rdf representation of openEHR/ISO13606 data. https://www.researchgate.net/profile/Jesualdo-Fernandez-Breis https://www.researchgate.net/profile/Santiago-Frid
We have also experimented with this in the past, specifically with the enrichment of EHR data with semantic web sources (e.g. drug-to-drug interactions). We participated in a paper on the matter Augmented EHR: Enrichment of EHR with Contents from Semantic Web Sources
This implied the transformation of openEHR data into rdf, and launching of SPARQL queries to get data back into the data instance
I’m starting to work on a RDFa/HTML5 representation of an ISO 13606 represented care record, which is of course related. I have an early draft of the 13606-1 record model in OWL if that helps anyone.
Tom’s post shared by @linforest makes pretty good points.
I am not implying anything against the use of Semantic Web technologies, but it takes some(!) effort to get the balance right. The reason I got introduced to openEHR and archetypes was a EU research project that aimed annotation of clinical data with OWL in … 2001. I’d say 24 years should have been enough for an approach to get adopted, but then openEHR itself bloomed just recently, after … 25 years or so So who knows, this may be the spark!
Oh, btw, if my memory is correct Josh Mandel’s first attempt at SMART was based on RDF. It then became SMART on FHIR. So there’s a story there worth learning about.
(For those who’ll google his name, I’m not talking about the ex treasurer from Ohio, add the term Microsoft to your query)
Certainly for our use case, OWL based implication reasoning from the openEHR/13606 model end isn’t a primary goal (although the record component and archetype identifier classifiers might prove useful if you’re classifying in the SNOMED CT space).
We’re looking more at using HTML5+RDFa for property/attribute graphs that empower generic search and AI technologies. Unlike (say) CDA, the semantics of the record are properly inline with human readable content and can be traversed by both human and machine alike. Such a representation would be useful to records that originated outside openEHR, which is important in settings like the NHS.
So who knows, this may be the spark!