Call for semantic health interoperability

A leading UK academic has called for a new push on the development of semantic electronic health records based on ‘clinical archetypes’ as the only viable route to deliver fully interoperable electronic health care records, writes Jon Hoeksma in Slovenia.

Dr Dipak Kalra of University College London said clinical archetypes, or “clinical shapes” need to be used in tandem with the advanced clinical standards terminology of SNOMED CT to develop fully interoperable EHRs. He described archetypes as a tool “for building clinical consensus in a consistent ways”.

Speaking at a e-health 2008 pre-conference EuroRec workshop on EHR accreditation Dr Kalra said there was a need for a sustained effort to develop a range of EHRs that can work together as a single seamless EHR

To the end-user he said there should be no apparent difference. Dr Kalra said the development of clinical archetypes was already being pursued by agencies, including the English NHS.

Dr Kalra said: “Many systems today can achieve semantic interoperability, but that cleverness is not sharable between systems.” The real problems, he said, arise when different EHR systems attempt to share clinical data. “This is the challenge of semantic interoperability.”

Dr Kalra set out three levels of semantic health interoperability. Level 1 is technical: “how data structures can be sent between applications and data can be imported and it knows where to put them and what they are”.

Level two he said was having correctly imported data a system then needs to know “how to exploit it and show it to users.”

Finally, at the third level comes full semantic interoperability, “when the user can’t tell the difference anymore” between data from his own local EHR or any other.

Dr Kalra said with full semantic health interoperability the objective should be that clinical data looks the same as clinician’s own data. Users can still view where data came from, but should be able to have as much confidence in the data as their own.

Dr Kalra stressed this goal remained some way off, but was essential to aim for to get away from the current situation in which data from different EHR systems was not semantically interoperable. “That is what we need to achieve. As a clinician I’m not going to trust a prescribing alert on data that is not interoperable or is well integrated.”

But to achieve this advanced level of semantic interoperability Dr Kalra said it will be necessary to have some standardisation of clinical practice. “If we are going to share clinical data as well as having key attributes of data, clinical practice has to start doing common things in common ways. If it doesn’t we will not achieve level three.”

The way to achieve this he argued is clinical archetypes. Dr Kalra said key standards now exist to support the development of archetypes, including ISO EN 13606-01 which is now a full international standard.

He said that clinical archetypes did not create complexity but provided a tool to accurately reflect the underlying complexity of clinical practice, potentially providing a more flexible and faithful way to enable interoperability than can currently be achieved by today’s messaging approaches such as HL7.

“We need to find way of managing semantic interoperability with complex next stage terminologies such as SNOMED CT,” said Dr Kalra. The use of clinical archetypes, he argued, offered a way data architectures can be used for care pathways, ensuring they are similarly formulated.

He said that an extract from one system to another, references archetypes and as a result knows how to better place data it receives in its own data architecture.

“If clinical practice can start to define in an intelligent way we have a hope of clinical records that are both human and machine readable,” concluded Dr Kalra.

In June the European Health Records Institute (EuroRec) will, in collaboration with the openEHR Foundation, publish a ‘state of the art’ report on where archetypes currently stand and make recommendations on what further work is needed.