International Journal of Medical Informatics (2005) 74, 886—898

Nursing constraint models for electronic healthrecords: A vision for domain knowledgegovernance

Evelyn Hovenga ∗, Sebastian Garde, Sam Heard

School of Information Systems, Health Informatics Research Group, Faculty of Informatics andCommunication, Central Queensland University, Rockhampton MC, Qld 4702, Australia

KEYWORDSNursing;Domain knowledgeontology;Electronic healthrecords;Terminologies;Knowledge governance;Standards

Summary Various forms of electronic health records (EHRs) are currently beingintroduced in several countries. Nurses are primary stakeholders and need to ensurethat their information and knowledge needs are being met by such systems informa-tion sharing between health care providers to enable them to improve the qualityand efficiency of health care service delivery for all subjects of care. The latestinternational EHR standards have adopted the openEHR approach of two-level mod-elling. The first level is a stable information model determining structure, while thesecond level consists of constraint models or ‘archetypes’ that reflect the specifica-tions or clinician rules for how clinical information needs to be represented to enableunambiguous data sharing. The current state of play in terms of international healthinformatics standards development activities is providing the nursing profession witha unique opportunity and challenge. Much work has been undertaken internationallyin the area of nursing terminologies and evidence-based practice. This paper arguesthat to make the most of these emerging technologies and EHRs we must now con-centrate on developing a process to identify, document, implement, manage andgovern our nursing domain knowledge as well as contribute to the development ofrelevant international standards. It is argued that one comprehensive nursing ter-minology, such as the ICNP or SNOMED CT is simply too complex and too difficultto maintain. As the openEHR archetype approach does not rely heavily on big stan-dardised terminologies, it offers more flexibility during standardisation of clinicalconcepts and it ensures open, future-proof electronic health records. We concludethat it is highly desirable for the nursing profession to adopt this openEHR approachas a means of documenting and governing the nursing profession’s domain knowl-edge. It is essential for the nursing profession to develop its domain knowledgeconstraint models (archetypes) collaboratively in an international context.© 2005 Elsevier Ireland Ltd. All rights reserved.

∗ Corresponding author.E-mail address: e.hovenga@cqu.edu.au (E. Hovenga).

URL: http://www.infocom.cqu.edu.au/hi.

1. Introduction

Successful management of information, knowl-edge, information systems and technology is

1386-5056/$ — see front matter © 2005 Elsevier Ireland Ltd. All rights reserved.doi:10.1016/j.ijmedinf.2005.07.013

A vision for domain knowledge governance 887

crucial in modern health care organisations toprovide competitive advantage, support clinicaldecision-making, patient management, financialmanagement, resource planning, resource alloca-tion, priority-setting, strategic management andto change organisational processes. Health careorganisations operate in a dynamic environment,and consequently, must be able to collect informa-tion or knowledge, such as evidence-based clinicalor best practice guidelines as required, communi-cate internally and externally, apply new or existingknowledge and process information so that man-agers and clinicians can make decisions quickly andeffectively. Typically, different systems have beendesigned to serve the different functions within ahealth care setting. The challenge is to design sys-tems that can serve a range of functions and/or tointegrate information systems such that semanticinteroperability is achieved. System integration canbe technologically difficult and tends to be costly[1], unless all systems comply with the same set ofmessaging/communication standards that:

• support medico-legal accountability and privacy;• enable fast information retrieval;

• explore how terminologies may be used todescribe the many nursing concepts that makeup these models (archetypes);

• highlight advantages of standardising nursingarchetypes and their link with the adoption ofevidence-based practice;

• show why the nursing profession needs to adoptdomain knowledge governance protocols.

2. Current state of play

Every day, new technologies become availableclaiming to offer solutions for a range of businessproblems. There are a number of technologies andinitiatives that are being adopted by the healthindustry, such as improvements to supply chainmanagement, evidence-based practice, use of theInternet to deliver health products, intranets andimplementation of electronic health records. Oneof these new technologies is the freely availableArchetype Editor downloadable from the openEHRwebpages (http://www.openEHR.org).

Governments as well as individual health careenterprises and professional organisations have ardc

p

•

•

•

Aamttgw

gtdrTiwato

• support unambiguous clinical information repre-sentation;

• ensure that the EHR contains meaningful andvalid information;

• enable key patient information sharing betweenindividual care providers;

• facilitate communication regardingrequest/instruction activation or completion inshared care environments;

• support the ability to extend the system to meetnew information requirements without having torebuild systems.

The openEHR approach of two-level modelling isable to meet these needs. The first level is a sta-ble information model determining structure, whilethe second level consists of constraint models or‘archetypes’ that reflect the specifications or clin-ician rules for how clinical information needs to berepresented to enable unambiguous data sharing.

The overall aim of this paper is to explore to whatextent nursing constraint models (archetypes), asdefined by the openEHR foundation, can improvesemantic system interoperability and enhance thebenefits to be obtained by the introduction of EHRs.Within the context of the current state of play, wewill:

• describe ontology-based knowledge domain con-straints models (archetypes) and how theserelate to EHRs;

esponsibility to create an environment that is con-ucive to the take up of many and varied new appli-ations of IT in the health industry.

In 2004, a European e-health action plan wasublished [2] to enable:

better provision of information to patients onhow to get treatment in other member states;coordinating the evaluation and assessment ofnew health technology;making the most of new IT and better integratinge-health policies and activities across Europe.

s a consequence, health care consumers in Europere expected to have a far greater direct involve-ent in making decisions about their care and

o use information technologies to interact withheir providers. These changes are far-reaching andlobal; they are not just about technology, as theyill affect everyone, everywhere.Nurses constitute a major group of direct care-

ivers who between them need to determine whatests to conduct, therapies to prescribe, proce-ures to perform or to whom the patient should beeferred for a second opinion or additional care.he next level of decision-making for this group

nvolves a decision about when to deliver care,here it should be delivered and which materialnd labour resources should be used. This is wherehe treating health care professional interacts withther direct and indirect care-givers. Such inter-

888 E. Hovenga et al.

actions determine what information needs to becommunicated to whom and when. Direct care-givers also need information about the quality andoutcomes of the treatments they have delivered.Providing this type of information in formats thatare readily accessible, timely and understandable isincreasingly important in health care organisations.In Australia, a national framework for clinical infor-mation capture, storage, representation and use tounderpin electronic health information interchangeand to facilitate semantic interoperability of clin-ical information across the health system is underdevelopment [3].

2.1. Electronic health records (EHRs)

Electronic health records are considered to offergreat benefits to patients and health providers,as current paper-based records have serious defi-ciencies. Australia is in the process of establishingHealthConnect a joint federal and State govern-ment initiative that aims to improve the flow ofinformation across the Australian health sector viaa secure network for on-line retrieval by consumersand authorised providers [4]. The expected benefits

mented across Canada. This document will guidethe development of EHR solutions, help jurisdic-tions develop their own technical roadmaps atlower costs and is freely available from the Infowaywebsite [5].

Early 2004, President Bush announced that mostAmericans will have electronic health recordswithin the next 10 years [6]. This was based onan earlier Institute of Medicine report [7] that hadidentified demonstration systems, indicating thatelectronic medical records, computerised order-ing of prescriptions and other medical tests, clin-ical decision support tools and secure exchangeof authorised information, improve quality, reducemedical errors and prevent deaths. His plan isbeing supported by significant funding and includesthe adoption of health information standards, anincrease in investment for demonstration projectsas well as the provision of leadership via a sub-cabinet level position of a national health infor-mation technology coordinator to coordinate part-nerships and guide health informatics standardsdevelopment as well as via the many federallyfunded health services, such as veteran affairs. Thisproposed national health information network has

are detailed in Box 1.Canada is another world leader with its elec-

tronic health record solution blueprint (EHRSblueprint) [5] that describes the business and tech-nical architecture of EHR solutions to be imple-

Box 1: HealthConnect’s expected benefitsrealisation:

• rapid access to vital and accurate healthinformation;

• reduced duplication of services;• more time available for direct care;• greater portability of health records for an

increasingly mobile population;• more control for consumers over who can

access their health information;• more active participation by consumers in

decisions about their health care;• better quality information exchange

between health care providers for improveddiagnoses and better quality care;

• a more comprehensive picture of Aus-tralians’ health to promote advances in thediagnosis and treatment of illnesses and bet-ter targeted decisions about health care.

Source: HealthConnect, http://www.healthconnect.gov.au/about/index.htmaccessed December 2004.

been endorsed by thirteen major health and infor-mation technology organisations in an unprece-dented joint collaboration [8].

2.1.1. What are EHRs?There are many definitions as to what constitutes anelectronic health record. The ISO Technical Report(ISO-TR20514) [9] defines a basic generic EHR as‘a repository of information regarding the healthstatus of a subject of care (patient or consumer),in computer processable form’. This document alsoidentifies and describes sharable and non-shareableEHRs as well as integrated EHRs and EHR systems.It notes that:

‘‘The difference between a non-shareable EHR anda shareable EHR is analogous to the differencebetween a stand-alone desktop PC and a networkedPC, where the latter adds enormous benefits interms of locating, retrieving and exchanging infor-mation using the internet, an intranet, email, work-group collaboration tools, etc.’’

This 2004 ISO Technical Report is the most autho-rative document to describe the EHR as is has beendeveloped through international consensus. Box 2contains this report’s definition of shareable EHRstaking place at three different levels where thethird level is referred to as an ‘integrated careEHR’. This is defined [10] as:

A vision for domain knowledge governance 889

Box 2: Shareable EHR definitionThe sharing of EHR information can take

place at three different levels:

Level 1: between different clinical disciplinesor other users, all of whom may be usingthe same application, requiring different ororganisation of EHRs;Level 2: between different applications at asingle EHR node, i.e. at a particular locationwhere the EHR is stored and maintained;Level 3: across different EHR nodes, i.e.across different EHR locations and/or differ-ent EHR systems.

The shareable EHR used for Levels 1 and2 will contain mainly detailed informationrequired for patient care within a single loca-tion, and it will be created and maintained ona local EHR system as described in Clause 7.3.However, it will also usually contain at leastsome health summary information such as aproblem list, allergies, past medical history,family history, current medications, etc.

When level 3 sharing is achieved and theobject of the EHR is to support the integratedcare of patients across and between healthenterprises, it is called an Integrated Care EHR(ICEHR).

Source: ISO Technical Report (ISO-TR20514)Health Informatics, Electronic Health Record,Definition, scope, and context Clause 4.5.

‘‘A repository of information regarding the healthstatus of a subject of care in computer processableform, stored and transmitted securely and acces-sible by multiple authorised users. It has a stan-dardised or commonly agreed logical informationmodel which is independent of EHR systems. Its pri-mary purpose is the support of continuing, efficientand quality integrated health care, and it containsinformation which is retrospective, concurrent andprospective’’.

For an EHR to realise any of the above benefits,it is essential that a detailed set of user and tech-nical requirements be agreed upon. Another 2004ISO technical specification detailing these [11] hasrecently been adopted. In addition, there needs tobe national agreement on an EHR architecture and amodel of the generic features of an EHR. Agreementon these issues is necessary to enable the desiredusage and the sharing and exchanging of electronichealth records. This needs to be independent of the

technology used to implement the EHR system. TheEHR architecture should also be unconstrained bycurrent organisational structures, as all health careorganisations need to adopt the same standards fora true EHR to exist.

The requirements for a truly global EHR shouldensure that it can be used, shared and exchangedbetween clinicians of all disciplines, across all sec-tors of health, different countries and differentmodels of health care delivery. It should also sup-port secondary uses, such as research, epidemi-ology, population health, health administration,financing and health service planning. Finally, theEHR should facilitate the evolution of existing sys-tems as well as the construction of new systems.

Much standards development work is contin-uing internationally with considerable Australianinput to enable effective electronic communica-tion as required for widespread EHRs adoption. Inparticular, the European standard, EN-ISO 13606-1 Electronic Health Record Communication, origi-nally a four-part standard is now being enhancedto become a five-part standard by adopting theopenEHR two-level modelling approach [12]—–part1: reference model is almost ready for an interna-ttaim

dfiteetopreiaabtt[ctmemct

ional ballot; part 2: archetype model is expectedo be available some time during 2005. This isrguably the single most important standard requir-ng international consensus, as it defines the funda-ental EHR infrastructure requirements.The adoption of ‘archetypes’ or knowledge

omain-based constraint models is essential, as thisacilitates EHRs to accommodate changing med-cal and health service delivery practices overime. This represents an innovation in knowledgengineering as it facilitates knowledge domainxperts (clinicians) to electronically document con-ent models, such as blood pressure, an ECG resultr discharge summary, describe clinical workflowrocesses, such as a patient’s assessment, incorpo-ate clinical or best practice guidelines or a knowl-dge domain ontology. This is achieved by separat-ng knowledge domains from runtime informationnd design time models that collectively make upn information system. Archetypes are the buildinglocks of the health knowledge environment. Adop-ion of the EN-ISO 13606-1 standard ensures the sys-em is able to manage this. The openEHR approach13] focuses on the semantic interoperability ofomplete EHRs or EHR extracts and is the basis forhe new European standard. Current clinical infor-ation systems directly incorporate clinical knowl-

dge concepts within the software and databasesaking these systems very costly to maintain, as

hanging knowledge will eventually make such sys-ems obsolete. Whereas through the adoption of

890 E. Hovenga et al.

archetypes, all clinical and other domain specificknowledge, that frequently changes, now residesoutside the software in archetypes that can beexpressed and shared using the archetype defini-tion language (ADL). Their link to a particular datamodel, such as the one defined by EN-ISO 13606,provides the basis for querying information. Thedevelopment of gateways will make it possible toaccess and use information from legacy systems[14].

2.1.2. The openEHR approachThe aim of openEHR is to enable the developmentof open specifications and software for EHR sys-tems. It is based on the results of the EuropeanUnion’s GEHR-Project. GEHR is an acronym for GoodEuropean Health Record, respectively, later GoodElectronic Health Record. Following GEHR, severalprojects extended and refined its results (e.g. thesynapses and SynEx projects). All these projectsinfluenced the openEHR architecture and the pio-neering of a two-level modelling approach for EHRs[13]. An overview of this approach is given in Fig. 1.

The first level is the reference informationmodel, which is pared down to the minimum to

A blood pressure archetype, for example, repre-sents a description of all the information a clinicianmight want or has to report about a blood pres-sure measurement. Basically, one archetype modelsor represents one clinical or other domain spe-cific concept. They are specifications for groups ofdata that are discreet, highly related and clinicallymeaningful. Archetypes define the business rules(constraints) for valid values and they use terminol-ogy to identify components within the archetype.They are necessary for value-added EHR applica-tions, such as intelligent decision support and careplanning and are technology independent.

Through the use of freely available archetypetools, e.g. the Archetype Editor, clinical groupsare empowered to control the way that EHRs arebuilt up, using designed structures to express therequired clinical data and assuring that all nec-essary constraints on the values of record com-ponents are observed. This ensures that all datain an EHR system are valid at two levels, due todata conformance with an information model andthe knowledge domain via agreed concept defini-tions. Design principles of openEHR are describedin more detail in [15], but the key innovation of

roation(e.g

support the medico-legal requirements and recordmanagement functions. This ensures that clinicianscan always send information to another providerand receive information, which they can read —thus ensuring data interoperability. The secondlevel involves the openEHR archetype methodol-ogy — a way of sharing evolving clinical informa-tion, so that it can be processed by the receivingprovider — thus ensuring semantic interoperability.

Fig. 1 Overview of the openEHR two-level modelling appinstance of the underlying information model. The informawhich are instances of a knowledge description language

the openEHR architecture is that it separates recordkeeping concerns from clinical data collection usingarchetypes [16] and thus enabling patient-centred,longitudinal, comprehensive and prospective EHRs.Experiences in two field trials supported by the Aus-tralian General Practice Computing Group (GPCG)showed that the approach does work [17] and itis now the basis of a major trial of data sharingbetween hospitals and primary care in Australia

ch for EHRs [3]. The information stored in the EHR is anin the EHR is constrained by archetypes (and templates),. archetype definition language).

A vision for domain knowledge governance 891

within the framework of HealthConnect, the Aus-tralian initiative for a national health informationnetwork and used in further projects [3,16]. Firstresults are promising.

2.1.3. Achieving semantic interoperabilityGreater standardisation of health information andknowledge needs to be achieved if we are to makebetter use of the data and knowledge collectedby health care organisations. In health informat-ics, the adoption of a standard way to representboth the meaning of information (a ‘normative’reference or terminology) and the known place orcontext enable clinicians to share information ina form that computers can understand. Many ter-minologies have been developed to suit a varietyof purposes. Their effectiveness relates to a bal-ance between the number of terms (size) and thespecificity or level of detail. No matter how manyterms and categories are offered some conceptswill rightly belong to more than one category. Asa consequence, provision is made for either cross-referencing or by adopting more than one hierarchyfor some single concepts within a terminology sothat data can be retrieved consistently. Mappingbaa

dsatmmps

witmnftlttt(f

hisi

this difficulty is via the adoption of standard struc-tured messages. This is most successful betweentwo systems but very complicated when multi-ple systems need to be connected. It works whenthere are limited and pre-determined communi-cation requirements. HL7, an international healthinformation standards development organisation,has identified over 400 different coding systemscurrently associated with its messaging standards.

A major infrastructure project on data defi-nitions and standards is being managed by theCanadian Institute for Health Information (CIHI).Canada has invested in 17 major projects since 2002towards achieving system interoperability, includ-ing provider and client registries, an enterprisemaster person index, a national electronic claimsstandard, drug information and diagnostic imagingsystems [19].

Data dictionaries are one tool used to pro-mote the standardisation of data. The AustralianHealth Ministers’ Advisory Council has endorsedthe National Health Data Dictionary (NHDD) as therepository of data standards, including clinical dataspecifications to support electronic health recordsdevelopment in Australia. The NHDD standardisesttmmb(c(iaA(

2tAdstgdTvnttetace

ased on complex rules between the terminologynd the reporting classification is a critical step inchieving consistent reporting.

Coding and classification systems help to stan-ardise the collection of health information to suitpecific purposes, such as resource allocation. Theyim to define specific concepts within confined con-exts. Understanding context is critical to the com-unication of meaning or semantic value. ‘‘Anyeaningful exchange of utterances depends on therior existence of an agreed set of semantic andyntactic rules’’ [18].

Data types are technical specifications of theay different types of data are entered and handled

n information systems. Data types are fundamen-al building blocks of computer software, electronicessaging and the EHR. A small team of experts are

ow engaged in identifying Australian requirementsor health data types. There is a realistic expecta-ion that the Australian health data types, beingargely an amalgam of the HL7 and CEN TC251 dataypes with modifications to overcome implementa-ion difficulties, will have a significant influence onhe emerging ISO standard for health data typesMacIsaac et al. 2004, unpublished working paperor HL7 Australia).

Current clinical information systems tend not toave adopted standard data models or data types,ndeed these tend to be vendor specific, thus dataharing between many systems is difficult if notmpossible to implement. One way of overcoming

he meaning and representation of data used inhe communication and analysis of health infor-ation by stakeholder consensus. This, as well asany national minimum data sets, are managedy the Australian Institute of Health and WelfareAIHW) who are re-engineering existing metadataontent to develop the Metadata Online RegistryMETeOR) [20]. A business case for a National Clin-cal Terminology is under development as part of

2005 national work program managed by theustralian National e-Health Transition AuthorityNeHTA) [21].

.1.4. What is the relationship betweenerminologies and archetypes?rchetypes are agreed models of clinical or otheromain specific concepts and need to adopt atandard set of terms for each archetype. Theseerms can come from any number of terminolo-ies. Archetypes specify groups of data that areiscreet, highly related and clinically meaningful.hey define the business rules (constraints) for validalues and they use terminology to identify compo-ents within an archetype. They enable informa-ion to be specified in a far more complex formhan is possible in message structures, they canvolve over time yet remain standardised. In addi-ion, there is no need to only use one terminologys the term that best describes the relevant con-ept is adopted within the archetype by formallyxpressing the combination (context) of data. It

892 E. Hovenga et al.

is also possible where appropriate to define smallunstructured term sets within the archetype itself.This enables the addition of, for example, ‘patientposition’, such as sitting, standing, etc. to a bloodpressure measurement.

Archetypes are formal information specifica-tions written in ADL allowing clinicians (informationusers) to author archetypes so that these specifi-cations can become agreed standards. Archetypes‘model’ domain knowledge and enable machinelevel semantic interoperability as they describe richinformation structures by indicating:

• how the information is to be expressed;• what is optional and what is mandatory;• what is a sensible value for each data element;• any other rules that need to be expressed.

Another advantage of using archetypes is that com-puters can, at the time of data entry, validate dataand alert the user to any error as archetypes controldata quality through their rules. Sharing of infor-mation requires the adoption of standard terminol-ogy and data structures. Archetypes offer a genericapproach to structuring health information for alluses where information is shared.

Fig. 2 Ontology: reality is conceptualised by a personwho then decides how best to represent this concept toreflect their perceived reality. Source: AS 5021-200X, TheLanguage of Health Concept Representation, 2004.

the same domain concepts. Such an ontology facili-tates unambiguous communication of complex anddetailed concepts. A domain ontology cataloguesvarious aspects known to exist within the domainof interest from a specific point of view to suita defined purpose. An ontological domain knowl-edge model shows how key concepts relate to oneanother and is usually based on a philosophical ortheoretical foundation. Such knowledge domainsform the basis of rules for decision support in clini-cal practice as well as for all other areas of decision-making. The term ‘ontology’ in the field of healthconcept representation describes the representa-tion of concepts and the reality they attempt todescribe (Fig. 2). For example, terminologies, clas-sifications, knowledge bases, nomenclatures, etc.are all examples of attempts to represent an ontol-ogy [25]. Beale [26] has identified five ontologi-cal levels, foundational (data types), data shar-ing where minimal ontological commitment enablesdata exchange, invariant domain concepts, such as‘observation’ or ‘protocol’, representing the baseontological commitments of a domain, variant re-usable atomic domain concepts, such as ‘apgarscore’, ‘blood pressure measurement’ and variantln

udbiatcsd

2.2. Ontological domain knowledge andarchetypes

Every health professional has command of a bodyof knowledge that needs constant updating. Pro-fessional organisational intellect is said to operateat four levels that increase in value. This intel-lect relates to: cognitive knowledge (know-what),advanced skills (know-how), systems understand-ing (know-why) and self-motivated creativity (care-why), where the last level is needed for people toadopt change [22]. Knowledge represents intellec-tual assets that are of primary importance in thesafe provision of quality health care. The dimen-sions of knowledge generation and its managementhave huge implications for the health industry.Knowledge management requires the adoption ofa set of processes that enable the creation, cap-ture, organisation, sharing, dissemination and useof knowledge, including the use of data miningtools.

It is argued by some [23,24] that the adoptionof an ontological framework facilitates the build-ing of better and more interoperable informationsystems. An archetype ontology built on a domain-based concept model, needs to be identified toenable the mapping of individual archetypes to ageneric knowledge framework. This enables theiruse alongside other knowledge resources sharing

ocal context specific concepts, such as ‘asthmaote’ or ‘ante-natal exam’.

Ideally, an ontological approach is used to writep archetypes such that they accurately reflectomain knowledge. Such knowledge needs to beased on evidence or best practice although it cannclude ‘tacit’ knowledge acquired by experts andgreed to by consensus. A good example of this ishe work undertaken in the European-based Wise-are Project [27], which we have used to demon-trate how such expert knowledge can be used toevelop archetypes.

A vision for domain knowledge governance 893

2.2.1. Information specifications for archetypedevelopmentOne main idea of the archetype approach is toempower domain experts (nurses) to create andchange the knowledge inherent in archetypes. TheopenEHR Archetype Editor was built as a toolto support the domain experts’ development ofarchetypes by providing them with an easy-to-use graphical user interface making the abstractopenEHR archetype model tangible. While the Edi-tor is quite intuitive, domain experts should havea basic understanding of archetypes and openEHR’slogical building blocks of an EHR (folders, compo-sitions, sections, entries, clusters, elements anddata values) to build useful archetypes. Basically,an archetype defines:

• a whole, distinct, domain-level concept;• constraints on instances of the openEHR informa-

tion model, which express a valid structure (i.e.composition, cardinality);

• constraints on instances of a reference model,which express valid types and values.

An archetype can be a specialisation of another

elements. It might be relevant for oncology nursesto document if the saliva is watery, thick or ropyor absent [27]. It is sometimes more intuitive forclinicians not to specialise archetypes. An alterna-tive is for the saliva-element to be included in thegeneral archetype, but made optional. Completedarchetypes can then be grouped together usingopenEHR templates. A template could be equiva-lent with a computer screen form, for example.Templates can also further restrict archetypes, thatis, make the saliva-element mandatory for use ina special oncology clinic. As an archetype has thesame meaning no matter where it appears in theEHR, there is no need to standardise templates.

2.3. Evidence-based practice andarchetypes

Evidence-based clinical practice is playing anincreasingly important role as health servicesdemand high quality and cost effective health caredelivery. Nurses are significant practicing clinicalstakeholders. Consequently, it is imperative thatnurses along with all other health professionals,identify what data sets are required to reflect evi-ddcshoabitiipnadaraetna

iob[

•

more general archetype. An EHR entry, whichhas been archetyped will have the same mean-ing no matter where it appears in the EHR. Whilearchetypes can be developed for various EHR build-ing blocks, the most common archetypes probablyare those for EHR entries. Entries are specialisedinto observations (e.g. the patient’s blood pres-sure), evaluations (e.g. the diagnosis) and instruc-tions (e.g. a medication order).

Commonly, archetypes have state and protocoldata; state data are required for safe interpretationof the archetype’s concept (e.g. patient is sittingor standing when measuring blood pressure). Pro-tocol data are not required for safe interpretationof the concept (e.g. the device used to measureweight and its last calibration date), but may bevaluable, e.g. for quality assurance purposes. Whilestate data must be displayed, protocol data are usu-ally not displayed to the clinician, but available incase it is needed.

Fig. 3 shows an example of how a nursingarchetype can be modelled using the ArchetypeEditor. Fig. 4 shows an extract from the actual ADL-representation of the corresponding archetype. Thedesign of this Editor guides the user to adopt anontological approach to model each domain con-cept. If we have consensus amongst all nursing spe-cialities than this shows the general oral assessmentrelevant for all nursing specialties, then the generaloral assessment could be modified to suit variousnursing specialities, by, for example, adding further

ence of their practices. Definition of minimumata sets for treatment practices, diseases and pro-edures allow the reporting of key indicators toupport outcome research and the evaluation ofealth and nursing services. Widespread adoptionf standard nursing archetypes will improve ourbility to undertake practice evaluation from whichest practice guidelines can be developed. Indeed,t will ultimately be possible to undertake ‘vir-ual’ or randomised clinical trials (RCTs) using exist-ng databases, real world demographics, includ-ng severity of illness, from which homogeneousatient cohorts are selected to enable evaluation ofursing care outcomes. Once standard archetypesre widely used, it will be possible to access largeatabases and manage the many confounding vari-bles and reliably evaluate nursing practice in theeal world. This effectiveness research, also knowns clinical practice improvement (CPI) [28], is,specially, beneficial for the nursing profession ashe results facilitate the improvement of routineursing care and it is far less costly to use as itccesses operational data.

Evidence-based clinical practice requires thedentification of data or information that representr reflect what is occurring. Current informaticsuilding blocks for evidence-based practice include29]:

standardised terminologies, such as the ICD-10-AM, SNOMED CT, ICNP and many others;

894 E. Hovenga et al.

Fig. 3 Example for modelling a nursing archetype using the openEHR Archetype Editor. This archetype models ageneral oral assessment (modified from the Oral Assessment Guide of the Wisecare Project).

• digital sources of evidence, for example, auto-mated care plans documenting actual care pro-vided;

• technical, messaging and terminology stan-dards that facilitate health care data exchangebetween information systems;

• informatics processes that support the acquisi-tion and application of evidence to a specificclinical situation;

• informatics competencies by care providersenabling the best possible data collection,retrieval and use of data from information sys-tems;

• adoption of appropriate computing and telecom-munication technologies.

Archetypes incorporate most of these, so that inthe future, we only need to concentrate on identi-fying which archetypes need to be standardised to

enable practice evaluation and the generation ofnew knowledge.

To enable the collection of evidence of bestpractice, it is also essential that we use uniquepatient identification, unique provider identi-fication and resolve issues of ownership andaccess/custodianship of electronic health records.Standard data and data structures are required ona per episode, per provider or per patient (casebasis over any time period incorporating practicedata provided by a number of providers and organ-isations. Such every day practice data can then becompared and evaluated in terms of patient out-comes from which best practice guidelines basedon evidence can be established. It is important thatnurses and all other clinicians are equipped with theskills to interpret this information and that healthcare organisations invest in information systemsthat can synthesise and present this information in

A vision for domain knowledge governance 895

Fig. 4 Extract from the general oral care observation archetype represented using the archetype definition language(ADL) as generated by the Archetype Editor.

meaningful ways. Health service managers need tocreate an environment that facilitates innovationand problem-solving and thus stimulates the gener-ation of new knowledge.

3. Nursing constraint models(archetypes)

Nurses as a profession need to take charge ofmanaging their domain knowledge. We need todevelop a framework for managing archetypes andfor identifying which need to be standardised,which are nursing domain specific. Nurses need towork in multidisciplinary teams to ensure that thosearchetypes that meet many clinical user needs ade-quately meet nursing information needs. We needto minimise redundancy.

We suggest that the current full draft interna-tional ISO standard-integration of a reference ter-minology model [30] for nursing be used as thebasis for such a framework as this provides the con-

ceptual structures as represented in a referenceterminology model. This standard consists of refer-ence terminology models for nursing diagnosis andnursing actions and reflects attempts at harmoni-sation with evolving terminology and informationmodel standards outside the domain of nursing. Itmay be that we should concentrate on develop-ing sets of archetypes to accommodate these twokey areas of nursing practice. However, it may bethat an archetype for an oral assessment, for exam-ple, is equally applicable to knowledge domainsother than nursing? The development of a struc-tural framework for archetype development andmanagement is expected to require the adoptionof principles not unlike those used for the develop-ment of a classification system or terminology. Thisis an area for further exploration.

We argue in accordance with Rector [31] thata comprehensive nursing terminology might sim-ply be too complex and too difficult to maintain.As the openEHR archetype approach does not relyon big standardised terminologies but on micro-

896 E. Hovenga et al.

vocabularies [32], it would offer more flexibil-ity during standardisation of clinical concepts andovercome the shortcomings of terminology-focusedapproaches. The degree of semantic system inter-operability is highly dependant on the quality ofthe terminology used. Our experiences confirm thatterminology harmonisation, maintenance and gen-eral governance are key factors for success in thisarea. The greater flexibility during standardisationprocesses offered by the openEHR archetyping is ofgreat value here. Furthermore, Garde et al. [33]concluded that openEHR can support multi-centreresearch based on routine clinical data and thusfacilitate the generation of new knowledge throughcontinuous practice evaluation.

It is highly recommended for knowledge alliancesor clinical partnerships to be established forthe purpose of developing evidence-based nurs-ing archetypes. These may also be referred to as‘practice communities’ [34]. One good example isagain the process adopted in the European Wise-care Project where nurse experts agreed on datadefinitions, data collection and usage to improvenursing care and patient outcomes [27]. Knowledgerepresents intellectual assets that are of primary

profession that specialises in that area of practice.In medicine, the professional colleges are in chargeof their specific knowledge domains.

The Australian National Nursing OrganisationsGroup, now consisting of more than 50 organisa-tions, was established in the early 1990s. They meettwice annually to discuss national issues. One ofthese was the designation of nursing specialtiesand credentialing of specialty practice. In 1995,agreement was reached that there should be threecategories of registered nurse defined in Australia:the beginning RN, the experienced/advanced RNand the expert RN. The development of genericcompetency standards was undertaken. These werelater adapted and used by specialty nursing groupswho developed their specific competency standards[36]. These are now used as the basis for creden-tialing individual nurses for each specialty. Thus,each organisation governs their special knowledgedomain as expressed by means of competencies.

This type of work could also be adopted asthe means of differentiating between generic andspecialised nursing archetypes either nationally orinternationally. In addition, there is a need tohave a national and/or international entity that isatndrmsiiIItaatlabtost

4

Wpamc

importance in the safe provision of quality healthcare. The dimensions of knowledge generation andits management have huge implications for thehealth industry. Knowledge management requiresthe adoption of a set of processes that enable thecreation, capture, organisation, sharing, dissemi-nation and use of knowledge. Many of these ‘knowl-edge artifacts or repositories’, such as archetypesneed to be standardised, managed and maintainedby the relevant domain knowledge experts. Collec-tively, this will create the knowledge environmentrequired to effectively adopt EHRs.

3.1. Domain knowledge governance

Once we have professional consensus regardingbest practice based on evidence expressed in theform of archetypes, we need to consider howto manage, maintain, update and disseminatethese. We already have organisations, such as theCochrane collaboration, an international not-for-profit organisation, providing up-to-date informa-tion about the effects of health care. They also havea Cochrane library containing regularly updatedevidence-based healthcare databases [35]. It ishighly desirable to either use an existing or estab-lish a new organisation of this kind to managegeneric standard archetypes and to make theseavailable. However, there will be a number ofarchetypes that will reflect specific clinical knowl-edge domains that can only be managed by the

uthorised to work with knowledge domain expertso develop and manage standard archetypes. Weeed a formalised process to ensure that stan-ard archetypes are evidence-based, reflect cur-ent best practice, use the most appropriate ter-inologies, are updated as required, are easily dis-

eminated and made available to all who need themn a timely manner and to ensure that redundancys minimised. Organisations, such as the Australiannstitute of Health and Welfare or the Canadiannstitute for Health Information may well be inhe best position to do so, but it would require

change to their existing organisational missionnd operations. Such organisations are best placedo take care of generic archetypes but this stilleaves a gap of how to best manage specialisedrchetypes. The issue is about who takes responsi-ility for the domain knowledge content? We arguehat the knowledge domain experts need to taken this governance role. However, the issue ofcope and boundaries of practice will not be easyo resolve.

. Conclusion

e need to take note of changes currently takinglace, consider the long term implications, activateparadigm shift in thinking and be pro-active toake the most of the opportunities and challenges

onfronting us today. It is clear that internation-

A vision for domain knowledge governance 897

ally there is strong desire to develop and imple-ment health information systems that have seman-tic interoperability. There is a global trend towardsthe implementation of EHRs along with the desireto make the best possible use of available technolo-gies to improve patient safety and outcomes andcontain costs. We have presented the current stateof play regarding the most authoritative under-standing of EHRs. We have shown how the adoptionof the openEHR approach optimises the ability toachieve these overall objectives. Not only does thisrequire standards developers to work more closelywith domain knowledge experts, but it also requiresa paradigm shift regarding the methods adopted.We need to develop a framework or classificationsystem for archetypes. We need to establish whogoverns which knowledge domain. We need a pro-cess for engaging knowledge domain experts, foridentifying what archetypes, we need and to assistwith development. We need to identify individualontological knowledge domains and assess prioritiesfor archetype development. We should no longer beconcentrating on getting one standard nursing ter-minology but we should divert our efforts towardthe development of archetypes. Terminology needsabotl

R

http://www.amia.org/pdf/rfirelease011805.pdf, accessed13.06.05.

[9] ISO Technical Report (ISO-TR20514), Health Informatics,Electronic Health Record, Definition, Scope, and Context,Clause 4.4.

[10] ISO Technical Report (ISO-TR20514), Health Informatics,Electronic Health Record, Definition, Scope, and Context,Clause 4.6.2.

[11] ISO/TS18308, Health Informatics, Requirements for An EHRReference Architecture.

[12] CEN TC251-EN-ISO 13606-1, Electronic Health Record Com-munication.

[13] openEHR Foundation, http://www.openEHR.org, accessed13.06.05.

[14] T. Beale, Knowledge-Enabling the Enterprise withArchetypes: Lessons from Health, EVOLVE conference,Sydney, Australia, 18—20 August, 2003.

[15] T. Beale, A. Goodchild, S. Heard, EHR Design Principles,openEHR Foundation, 2001, http://titanium.dstc.edu.au/papers/ehr design principles.pdf, accessed 13.06.05.

[16] A. Goodchild, K. Gibson, L. Anderson, L. Bird, The Bris-bane Southside HealthConnect Trial, in: Preliminary ResultsHealth Informatics Conference (HIC), Brisbane, 2004.

[17] L. Bird, A. Goodchild, Z. Tun, Experiences with a two-levelmodelling approach to electronic health records, J. Res.Pract. Inf. Technol. 35 (2) (2003).

[18] ISO TR 9007, Information Processing Systems—–Concepts andTerminology for the Conceptual Schema and the Informa-tion Base, 1987.

[19] Canadian Institute for Health Information (CIHI), http://

[

[

[

[

[

[

[

[

[

[

[

[

nd adoption, as well as the adoption of evidence-ased practice principles will then become partf that effort. Nurses are well placed to leadhis, as we have strong international professionaliaisons.

eferences

[1] K. Laudon, J. Laudon, Management Information Systems:Managing the Digital Firm, eighth ed., Pearson PrenticeHall, 2004.

[2] European Commission-eHealth, http://europa.eu.int/information society/eeurope/2005/all about/ehealth/index en.htm, accessed 13.06.05.

[3] CIP, Clinical Information Program, a NeHTA Project, 2004,http://cip.healthbase.info/, accessed 13.06.05.

[4] HealthConnect, A Health Information Network for All Aus-tralians, http://www.healthconnect.gov.au/about/index.htm, accessed 13.06.05.

[5] Canada Health Infoway, http://www.infoway-inforoute.ca/home.php?lang=en, accessed 13.06.05.

[6] G. Bush, Transforming Health Care: The President’s HealthInformation Technology Plan, 2004, http://www.whitehouse.gov/infocus/technology/economicpolicy200404/chap3.html, accessed 13.06.05.

[7] J.M. Corrigan, A. Greiner, S.M. Erickson (Eds.), Fos-tering Rapid Advances in Health Care: Learning fromSystem Demonstrations, The National Academy Press,Washington, DC, 2002, http://www.nap.edu/catalog/10565.html?onpi newsdoc1192002, accessed 13.06.05.

[8] AMIA Media Release, Leading Health Care and Infor-mation Technology Groups Endorse Common Frameworkfor Health Information Exchange to Support Improve-ments in Health and Healthcare, 18 January 2005,

secure.cihi.ca/cihiweb/home e.html, accessed 13.06.05.20] Australian Institute of Health and Welfare, The Knowledge

Base v.2, 2004, http://www.aihw.gov.au/knowledgebase/redev advice.html, accessed January 2005.

21] National e-Health Transition Authority (NeHTA), http://www.ahic.org.au/nehta, accessed January 2005.

22] J.B. Quinn, P. Anderson, S. Finkelstein, Managing profes-sional intellect, in: Harvard Business Review on Knowl-edge Management, A Harvard Business Review Paperback,Boston, 1998.

23] D.M. Pisanelli, A. Gangemi, If ontology is the solution,what is the problem? in: D.M. Pisanelli (Ed.), Ontologiesin Medicine, IOS Press, Amsterdam, 2004.

24] A. Kumar, B. Smith, D.M. Pisanelli, A. Gangemi, M. Ste-fanelli, An ontological framework for the implementationof clinical guidelines in health care organisations, in: D.M.Pisanelli (Ed.), Ontologies in Medicine, IOS Press, Amster-dam, 2004.

25] Australian Standard AS 5021-200X, The Language of HealthConcept Representation, 2004.

26] T Beale, Ontological principles of information modeling forfuture proof systems, in: Presentation at the CEN TC251Meeting May, Berlin, 2005.

27] W. Sermeus, N. Kearney, J. Kinnunen, L. Goossens, M. Miller(Eds.), Wisecare: Workflow Information Systems for Euro-pean Nursing Care, vol. 73, IOS Press, Amsterdam, 2004(Stud. Health Technol. Inf.).

28] S.D. Horn, Quality, clinical practice improvement, and theepisode of care Manag. Care Q. 9 (3) (2001) 10.

29] E. Hovenga, S. Lloyd, Working with information, in: Harris etal. (Eds.), Managing Health Services, Concepts and Practice,MacLennan & Petty, Sydney, 2002.

30] ISO/FDIS 18104, Health Informatics, Integration of a Refer-ence Terminology Model for Nursing, 2003.

31] A.L. Rector, Terminology and concept representation lan-guages: where are we? Artif. Intell. Med. 15 (1) (1999)1—4.

898 E. Hovenga et al.

[32] Health Level 7, HL7 EHR System Functional Model, AMajor Development Towards Consensus on ElectronicHealth Record System Functionality, A White Paper,2004, http://www.hl7.org/ehr/downloads/dstu/EHR-SWhitePaper.zip, accessed 13.06.05.

[33] S. Garde, P. Knaup, T. Schuler, E. Hovenga, Can openEHRarchetypes empower multi-centre clinical research? in:Paper Accepted for Presentation at MIE: 2005, Geneva,2005.

[34] E. Wenger, R. McDermott, W.M. Snyder, Cultivating Commu-nities of Practice, Harvard Business School Press, Boston,2002.

[35] Cochrane Collaboration, http://www.cochrane.org/index0.htm, accessed 13.06.05.

[36] National Nursing Organisations, Glossary of terms: Cri-teria for Specialties in Nursing—–Principles of Creden-tialling for Nurses, 2004, http://www.anf.org.au/nno/pdf/NNO Glossary of Terms.pdf, accessed 13.06.05.