Semantic interoperability is absolutely compromised when for the same clinical concept variants of archetypes are created.
If justified for internal system development , the moment exchange with another system requires harmonizing on datapoint to datapoint level. I have done about 2000 in perinatology 800 in stroke care 1250 in youth care 100 in nursing oncology 20 in reuma, 400 in general nursing 250 in diabetes care 200 in GP care 100 in cardiology. In the past 13 years.
The inconsistencies for the same data element in the various domains are BIG, without clinical justifiable reasons.
That same situation exists if you have locally / vendor specific arechetypes .
You have produced a lot of archetypes based on clinical input, that in the end were, well considered, well discussed, co-ordinated, agreed, ad-hoc, collections of what clinicians expect in a local context, to see on a screen to look at, and be used for data capture.
It is a lot of experiences.
Questions:
Have you produced a lot of Template stuff?
Or,
have you produced semantic interoperability artefacts?
Hi William, I do not understand some things in your comment.
Do you mean with “same clinical concept variants of archetypes” that they have the same archetypeId? So different archetypes with the same id?
Do you mean with “harmonizing” mapping from datapoints in different archetypes (with a different archetypeId) to each other?
In my opinion, this kind of mapping must always be defined manually per new case, like you define manually where to put your datapoints in Nictiz-messages.
I think you can never let a machine guess and understand the context in which a datapoints exists.
I also cannot imagine a use case in which such (IMO) a risky mapping, defined by machine, would be appropriate.
I will be very pleased if you will clarify on this.
The same situation exists if you have local, regionally and nationally competing HL7 messages, CDAs, FHIR specs, CCDs, DICOM-SR, NCI BRIDG models, and all the rest. That the situation today. Archetype technology doesn’t solve the problem of organisations not collaborating, it just gives much better technical support and capability for building semantic specifications of content and creating software and UI from them. The sociological problem still needs work. That’s why identification, governance, and networks of cooperation are important. Still, we didn’t get nowhere. Now everyone seems to agree on how systolic BP should be represented - thomas
You are absolutely right. Yesterday I where in a meeting with two national stakeholders for data specification regarding patient with drug/alcohol problems. The problem we as a vendor and our customers and users face is that they ask for almost the same thing. And they can not agree. That’s when they ask if we, as a vendor, can map between the terms.
I said no. I said you have to agree about the meaning of your data. Then, maybe, we can do the mapping. Or even better: there will be no mapping and we can reuse data inside the EHR and between different organisations.
The base of semantic interoperability is in a firm core definition of what the data is about. Then the software can let magic things happen.
- thomas