Building Interoperability Into Medical Information

“The healthcare interoperability problem can be investigated in two categories: Interoperability of the healthcare messages exchanged and interoperability of electronic healthcare records [EHRs],” says Professor Asuman Dogac, Director of the Software Research & Development Center at the Middle East Technical University in Ankara, Turkey, and coordinator of the IST-funded ARTEMIS project.

Right now messaging interfaces, or interface engines, are used to exchange information among different healthcare information systems. Currently, the Health Level 7 (HL7) Version 2 Messaging Standard is the most widely implemented in healthcare.

Unfortunately, HL7 V2 compliance does not imply direct interoperability between healthcare systems, because V2 has no explicit information model. Instead it has rather vague definitions for many data fields and it contains many optional fields.

It offers great flexibility, but requires detailed bilateral agreements among healthcare systems to achieve interoperability. So Version 3 was developed, based on an object-oriented data model but that can't talk to V2.

EHRs suffer from similar problems. An EHR is digitally stored healthcare information about an individual’s lifetime that supports continuity of care, education and research, while ensuring confidentiality at all times.

A number of standardisation efforts are progressing to provide EHRs interoperability. However, an exchange of well-structured and machine-processable EHRs has not been achieved yet in practice.

Adding meaning
ARTEMIS provides the desperately required interoperability between medical information systems through semantically-enriched Web services that use defined meanings of individual pieces of input and output parameters. The team decided not to reinvent the wheel and used existing standards where they were relevant.

For example, an essential element in defining Semantic Web services is domain knowledge. Medical informatics is one of the few domains to have considerable domain knowledge developed in standards. HL7 categorises events and defines messages in healthcare based on service functionality, which reflects the business logic. ARTEMIS uses HL7 as a basis for defining both the service action semantics and the message semantics.

In the ARTEMIS project, their Message Exchange Framework provides the exchange of meaningful clinical information among healthcare institutes through semantic mediation. The framework involves first maps the source ontology onto the target message ontology with the help of a tool which produces a mapping definition. This mapping definition is then used to automatically transform the source ontology message instances into target message instances. Through a prototype implementation, ARTEMIS demonstrates how to mediate between HL7 Version 2 and HL7 Version 3 messages. However, the framework proposed is generic enough to mediate between any incompatible healthcare standards currently in use.

For EHRs, ARTEMIS uses constraint rules for specific clinical concepts, called ‘archetypes’, rather than distinct entities in the reference information model. Archetype-based interoperability discovers existing archetypes based on their semantics. It annotates their archetypes with ontologies, an established list of definitions, and then composes templates from archetypes and retrieves corresponding data from the underlying medical information systems. ARTEMIS used electronic business XML (ebXML) registry semantic constructs to annotate, store, discover and retrieve archetypes.

Continuity of care with EHR
One crucial aspect in ARTEMIS is to find and retrieve clinical information about a particular patient from different healthcare organisations where concrete sources are unknown. To complicate matters, in most countries there are no unique person identifiers that would be valid for the whole lifetime of an individual and used by all parties in healthcare and for all episodes of care. On the contrary, in many cases several identifiers for a patient exist even within a single organisation.

Consequently a protocol is needed that allows the identification of patients by means of non-unique patient-related attributes. ARTEMIS developed a ‘Patient Identification Process’ (PIP) Protocol’. PIP provides a solution for a common problem in the healthcare sector that is likely to become very important with the increasing mobility of the workforce in Europe: locating and accessing prior clinical records for continuity of care.

Healthcare information systems operate within a strict regulatory framework that ensures the protection of personal data and outlines the conditions where processing is allowed. The ARTEMIS project responded to these conditions by providing comprehensive security and privacy protection mechanisms.

Currently, the healthcare industry’s Integrating the Healthcare Enterprise (IHE) initiative proposes the Retrieve Information for Display (RID) integration profile to allow users to retrieve and display patient-related documents on systems other than the document keeping systems. Although the RID profile is well suited for use in a single hospital or within a trust of hospitals that belong to a single Patient Identifier Domain, it is not designed for cross-boundary access on information stored in different hospitals.

ARTEMIS developed middleware infrastructure that extends the IHE RID protocol for cross-enterprise search and access to patient-related clinical information, even if no Master Patient Index is available, and without modifications to existing information source actors. Applied to the ARTEMIS infrastructure, the RID Information Source and Display actors may be located in different institutions using different Patient ID domains and different sets of demographic data. Within the ARTEMIS network, clinical records can be located using the Patient Identification Protocol which can also be combined with the IHE Cross-enterprise Document Sharing (XDS) Integration Profile.

At the Healthcare Information and Management Systems Society (HIMSS) annual conference in San Diego, February 2006, "we will run a demonstrator that will realise a scenario where, after an accident, a patient is admitted to the most appropriate hospital from the ambulance,” says Dr Dogac. “The patient will be admitted before the ambulance arrives at the hospital, via a mobile device. The hospital admissions service will then automatically seek out any relevant healthcare records of the patient in the ARTEMIS P2P network, and presents them to the doctor, although the hospitals discovered may not be using interoperable standards with each other. This is a considerable improvement over current systems."

Looking ahead
In the meantime, the team will continue to work on improvements in the system, enhancing the peer-to-peer networking, security, privacy and patient identification modules and integrating them to the ARTEMIS architecture.

Dr Dogac emphasised that the current aim of the project is to demonstrate the project results to the industry players on board. "We need to show more people what we are doing and publicise it more widely. We need to show them that it works."

That's part of an on-going process to promote their work at conferences like HIMMS. In the meantime, Dr Dogac will begin a new project, SAPHIRE, in the next couple of months.

"That project will seek to extend the functionality of ARTEMIS. It will expose medical sensor data as semantically-enriched Web services. It will process sensor output, the patient's medical history and clinical guidelines to help physicians with diagnoses and treatment," says Dr Dogac.

While ARTEMIS is focused on the hugely important step of accessing disparate patient data through a semantic interoperability platform, SAPHIRE uses this system to create new services that are currently unavailable. It could mean a paradigm shift in patient care.

But even without these new services, ARTEMIS already represents a major advance; it enables clinicians to capture the complete clinical history of a patient that may be spread out over a number of different institutes that do not interoperate.