Knowledge management in healthcare: towards ‘knowledge ...sraza/papers/IJMI01.pdf · a suite of...

International Journal of Medical Informatics 63 (2001) 5–18

Knowledge management in healthcare: towards‘knowledge-driven’ decision-support services

Syed Sibte Raza AbidiHealth Informatics Research Group, School of Computer Sciences, Uni�ersiti Sains Malaysia, 11800 Penang, Malaysia

Abstract

In this paper, we highlight the involvement of Knowledge Management in a healthcare enterprise. We argue thatthe ‘knowledge quotient’ of a healthcare enterprise can be enhanced by procuring diverse facets of knowledge fromthe seemingly placid healthcare data repositories, and subsequently operationalising the procured knowledge to derivea suite of Strategic Healthcare Decision-Support Services that can impact strategic decision-making, planning andmanagement of the healthcare enterprise. In this paper, we firstly present a reference Knowledge Managementenvironment—a Healthcare Enterprise Memory—with the functionality to acquire, share and operationalise thevarious modalities of healthcare knowledge. Next, we present the functional and architectural specification of aStrategic Healthcare Decision-Support Services Info-structure, which effectuates a synergy between knowledgeprocurement (vis-a-vis Data Mining) and knowledge operationalisation (vis-a-vis Knowledge Management) tech-niques to generate a suite of strategic knowledge-driven decision-support services. In conclusion, we argue that theproposed Healthcare Enterprise Memory is an attempt to rethink the possible sources of leverage to improvehealthcare delivery, hereby providing a valuable strategic planning and management resource to healthcare policymakers. © 2001 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Knowledge management; Data mining; Decision-support; Healthcare enterprise memory; Strategic decision-supportservices; Healthcare delivery info-structure

www.elsevier.com/locate/ijmedinf

1. Introduction: knowledge management inhealthcare

Knowledge Management (KM) in health-care can be regarded as the confluence offormal methodologies and techniques to facil-itate the creation, identification, acquisition,development, preservation, dissemination andfinally the utilisation of the various facets ofa healthcare enterprise’s knowledge assets [1–3].

The health care industry has evolved intoan extended enterprise—an enterprise that ispowered by sophisticated knowledge and in-formation resources. In today’s knowledge-theoretic healthcare enterprises, knowledge isdeemed as a ‘high value form of information’[4] which is central to the enterprise’s ‘capac-ity to act’ [5]. The field of knowledge man-agement provides the methodological andtechnological framework to: (a) pro-activelycapture both the experiential knowledge in-

1386-5056/01/$ - see front matter © 2001 Elsevier Science Ireland Ltd. All rights reserved.

PII: S1386 -5056 (01 )00167 -8

S.S.R. Abidi / International Journal of Medical Informatics 63 (2001) 5–186

trinsic to what are we doing vis-a-vishealthcare practice and delivery, and theempirical knowledge derived from the out-comes of what have we done; and (b) oper-ationalise healthcare knowledge to serve asa strategic decision-making resource, vis-a-vis an ensemble of business rules, trend pre-dicting insights, workflow analysis, analyticoutcomes, procedural guidelines and so on[6].

Healthcare enterprises can be regarded as‘data rich’ as they generate massiveamounts of data, such as electronic medicalrecords, clinical trial data, hospitals records,administrative reports, benchmarking find-ings and so on. But, in the same breath wecan say that healthcare enterprises are‘knowledge poor’ because the healthcaredata is rarely transformed into a strategicdecision-support resource. For that matter,with the emergence of technologies such asKM and Data Mining (DM) [7], there nowexist opportunities to facilitate the migra-tion of raw empirical data to the kind ofempirical knowledge that can provide awindow on the internal dynamics of thehealthcare enterprise. We argue that suchdata-derived knowledge can enable health-care managers and policy-makers to infer‘inherent’, yet invaluable, operative princi-ples/values/know-how/strategies pertinenttowards the improvement of the operationalefficacy of the said healthcare enterprise.

We contend that the operational efficacyof a healthcare enterprise can be signifi-cantly increased by (a) procuring diversefacets of empirical knowledge from theseemingly placid healthcare data reposito-ries, and (b) by operationalising the pro-cured empirical knowledge to derive a suiteof packaged, value-added Strategic Health-care Decision-Support Ser�ices (SHDS) thataim to impact strategic decision-making,

planning and management of the healthcareenterprise [8,9]. The vantage point of theaforementioned SHDS is that they providestrategic insights/recommendations/predic-tions/analysis to assist healthcare managers/policy-makers/analysts to device policies ormake strategic decisions or predict futureconsequences by taking into account the ac-tual outcomes/performance of the health-care enterprise’s current operative values—which may not necessarily be the same asthe espoused operative values.

To meet the above objectives, we proposeto design a KM-oriented info-structure,based on a novel approach that effectuatesa synergy between knowledge procurement(via DM) and knowledge operationalisation(via KM) techniques. The modus operandiof the proposed synergy is as follows: DMtechniques are used to ‘mine’ healthcaredata repositories to inductively derive deci-sion-quality healthcare knowledge, whereasKM techniques are subsequently used tooperationalise the inductively derivedhealthcare knowledge to yield a suite ofSHDS. The description of the functionaland architectural specification of such aKM-oriented info-structure is the theme ofthe work reported here. In this paper wewill present:

A reference KM info-structure—aHealthcare Enterprise Memory (HEM)—that purports the functionality to acquire,share and operationalise the various modal-ities of knowledge existent in a healthcareenterprise [2,10].

A demonstration SHDS Info-structurethat leverages existing healthcare knowl-edge/data bases to (a) derive decision-qual-ity knowledge from healthcare data and (b)generate and deliver a variety of SHDS [6].Here, we will discuss the end-user perspec-tive of the SHDS info-structure as opposedto technical details.

S.S.R. Abidi / International Journal of Medical Informatics 63 (2001) 5–18 7

The work reported here derives from ourpresent interest in the Malaysian Tele-Medicine initiative [11,12] under the auspicesof the Multimedia Super Corridor project.

2. Strategic healthcare decision-supportservices: an overview

The effective delivery of healthcare serviceshinges on the ability to deliver appropriate,proactive and value-added services to differ-ent client segments on a timely basis. Ingeneral, healthcare services need to be sys-tematically determined based on needs; pack-aged according to usage patterns,demographics and behavioural psychograph-ics; and delivered in a ubiquitous, proactiveand continuous manner. These mutually in-ter-related constraints are hard to formulate,let alone satisfy, using conventional strategicplanning techniques. Henceforth, for en-hanced healthcare services efficiency and in-formed strategic planning and management,there is an imminent need to model and

measure healthcare processes using definitivehealthcare process models that are induc-tively derived from the collected healthcaredata. We propose that the controlled simula-tions of data-derived healthcare process mod-els can be effectively used to derive‘knowledgeable’ (strategic) insights about theintrinsic behaviour of the healthcare enter-prise. The rationale of the approach is thatby understanding what worked—or didnot—healthcare practitioners can identify ar-eas for improvement and/or capitalise onpast successful methods.

SHDS can best be defined as a suite ofknowledge/data-driven, strategic, decision-support services derived from both healthcaredata and the health enterprise’s knowledgebases, with the objective to improve the deliv-ery of quality healthcare services (see Fig. 1for a typical SHDS environment). Methodo-logically speaking, SHDS cater for the migra-tion of data (the what), through a sequenceof sophisticated operations to information(the trend or behaviour) to knowledge (thewhy) to the ultimate level of wisdom (the

Fig. 1. An overview of a SHDS environment.


Table 1A synopsis of the proposed SHDS

necessary actions to be taken). The innova-tive meta-level of wisdom provides an addi-tional dimension to SHDS where, the datacollected, the information determined, theknowledge acquired can be used towards thesubtle shaping of the healthcare enterprise soas to acquire the highest degree of healthcarequality by either fine-tuning the existing en-terprise-wide practices or by defining newaction plans for policy making and healthplanning [9,12].

Typical SHDS may include: trend analysisof diseases/epidemics [13,14], treatment pat-terns, hospital admissions, drug patterns andso on, benchmarking and best-practices re-porting, outcomes measurement, what-if sce-nario analysis; comparing medical practiceswith medical business rules, market research,feedback routing to R&D institutions (e.g.drug effectiveness on outcomes of treatment);data analysis for healthcare financing, healthsurveillance and resource allocations [9].


Table 1 gives an abridged list of possibleSHDS. Principal beneficiaries of SHDS areenvisaged to be the Ministry of Health(MoH), pharmaceuticals, medical service or-ganisations, private health providers, univer-sities, and community health organisation.

3. The healthcare enterprise memory

The Healthcare Enterprise Memory(HEM) can be characterised as a conglom-erate KM architecture—comprising func-tionally independent computing systems—that provides the functionality to acquire,share, reuse and operationalise the variousmodalities of healthcare knowledge (e.g.tacit and explicit knowledge of healthcarepractitioners, healthcare related documents,data, processes, workflows, experiences andlessons learnt). The technical realisation ofHEM involves a confluence of data, infor-mation and knowledge management tech-nologies that in unison aim to operat-ionalise healthcare knowledge so as to re-alise a suite of SHDS. Typical services of-fered by HEM may support healthcareplanning and management, automatic dis-semination of knowledge, reuse of knowl-edge and experience, support of intelligentknowledge management services, timely pro-vision of knowledge and experience, trans-forming information to action, connectingand converting knowledge and above allhealthcare (meta) modelling [1,2] [15]. Onan operational note, HEM is a consequenceof the need to explicate the abstract seman-tics of healthcare knowledge. On a func-tional note, the HEM allows for thesynchronisation of the heterogeneous health-care knowledge resources to yield a com-mon goal— i.e. the realisation of a dynamic,progressive and pro-active healthcare enter-

prise. We briefly identify the four function-ally distinct layers of our proposed HEM(see Fig. 2):

1 Object Layer: Consists of varioushealthcare information and knowledgesources such as data-, document-, knowl-edge- and scenario-bases.

2 Knowledge Description Layer: The mainpurpose of this layer is to facilitate the ac-curate selection and the efficient access torelevant object-level healthcare knowledge ina given application situation. Medical on-tologies reside at this layer to (a) maintain astandard vocabulary to describe conceptsand relationships between entities that at-tempt to share knowledge [16,17] and (b)facilitate the incremental scaling-up ofhealthcare knowledge.

3 Application Layer: Models and executesvarious healthcare processes and tasks, re-alised in different ways ranging from dedi-cated programs to flexible query interfaces.

4 Ser�ices Layer: Provides specialisedhealthcare services through the use of vari-ous dedicated applications [8,9].

In this paper, we will focus on just onecomponent of the HEM, which is theSHDS info-structure (shown as the shadedarea towards the right side of Fig. 2).

3.1. The SHDS info-structure

We understand that the healthcare do-main is replete with many DM applicationsand solutions, but we contend that most ef-forts, by and large, are limited to specialisedand focused problems that are handled by apre-selected DM algorithm [18–21]. Anothernoticeable limitation with such efforts isthat the onus is on the user to define amodel of the data— the user is required toselect the input data source, the relevantdata attributes, the pertinent DM al-


gorithm(s) and finally the most appropriatedata visualisation format— in order to deriveactionable knowledge from the collecteddata. Indeed, this places too much demandon a novice user, who is merely interested inthe automated generation of ‘just-in-time’strategic decision-support knowledge. To ad-dress the above limitations, we have devel-oped a generic, modular and query-drivendecision-support SHDS-info-structure thatincorporates an ensemble of DM techniquesto ‘mine’ heterogeneous healthcare datarepositories to derive ‘just-in-time’ decision-quality (healthcare) knowledge in response tospecific user-requests.

4. The SHDS info-structure: architecturaldetails

The key to the generic functionality of ourSHDS info-structure is the implementation ofa wide range of pre-packaged, yet user cus-tomisable, query-driven DM solutions ormodules, each with pre-defined functionality.The DM modules are designed to interactwith users to acquire the specification of theproblem and in return provide decision-sup-port services. Technically, the use of Dis-tributed Object Technologies— i.e. middleware technologies such as CORBA, ActiveXand JavaBeans [22]—allow users to (a) read-

Fig. 2. The healthcare enterprise memory model.


Fig. 3. An overview of the multi-tier functional architecture of the SHDS info-structure.

ily select the most pertinent DM module (interms of functionality) for the problem athand; (b) present the problem specification(as per the internal specification language) tothe DM module; and (c) customise the pre-packaged DM module as per the problemspecification to generate the desired SHDS.

The stratified design of the SHDS info-structure is based on the following princi-ples: (i) a user-friendly and cognitivelytransparent top-level user interface for thespecification of a decision-support serviceand the visualisation of the DM outcome;(ii) a library of functionally diverse DMmodules that cover the entire spectrum ofdecision-support services deemed relevant tothe healthcare domain; (iii) a service cus-tomisation facility to generate specialised;

and (iv) a rich source of healthcare datavis-a-vis health databases and data ware-houses.

Conceptually, the SHDS info-structure hasa multi-tier architecture that effectuates aconfluence of KM and DM techniques(shown in Fig. 3 and further elaborated inFig. 4). If we look at Fig. 3, the lowest layeris the Content Layer which comprises thecontent sources—healthcare databases, datawarehouses and knowledge bases. The secondlayer— the Knowledge Description Layer—provides both an abstraction and specifica-tion of the data and knowledge resources interms of metadata and ontologies, respec-tively. The third layer— the ApplicationLayer— is the core layer which consists oftwo engines, one geared towards DM tasks


whilst the other handles KM tasks. Finally,the top layer— the Ser�ices Layer—serves asthe user-interface, allowing users to select andspecify the SHDS required and to receive theresults/findings/conclusions/recommenda-tions. The multi-tiered architecture presented

here offers several key advantages. Firstly,data issues such as data cleansing, integrationand consolidation are taken care of at thesystem level. Secondly, the use of APIs toconnect the various inter-layer componentsallows (a) seamless message and data passing;

Fig. 4. The architecture of the SHDS info-structure.


(b) on-line and interactive selection of DMfunctions and customisation of modules,whilst maintaining the constraints and pro-tocol requirements at each level. We nowbriefly describe the functional characteristicsof the various components of our SHDSinfo-structure. Fig. 4 gives the architectureof the SHDS info-structure.

4.1. SHDS modules

An all-encompassing solution that meetsthe diverse decision-support needs of ahealthcare enterprise is achieved by the im-plementation of a number of individual,self-contained, specialised DM modules,each designed to provide a particularSHDS. For instance, three SHDS (1)analysing the trends in hospital admission(2) analysing the trends in treatment pat-terns and (3) analysing the trends outcomesof treatment, will be implemented as threeindividual SHDS-modules. The SHDS mod-ules are implemented as an integration of‘standard task-codes’ stored in libraries spe-cific to tasks ranging from data collection toDM algorithms. In technical terms, eachSHDS is written as a script in a high-leveldeclarative DM language (comparable toSQL). The script specifies the followingcharacteristics of the SHDS: (a) the minableview— the data resources to be mined, morespecifically the part(s) of the database(s) tobe mined; (b) the attributes of the data tobe used, together with their significance val-ues and relational information, in any; (c)the DM algorithm(s) to be used; (d) thetype of patterns/rules to be mined; (e) theproperties that the chosen pattern shouldsatisfy; (f) the constraints imposed to reflectthe user’s intentions and the peculiarities ofthe SHDS; and (g) the data visualisationmethod to be used to display the DM re-sults.

The design of SHDS as modules predi-cates the provision of specifying externalconstraints, over and above the designatedfunctionality of a SHDS, in order to cus-tomise the inherent processing of a SHDSmodule according to the specific user de-mands. This is akin to constraint-basedDM, as reported in the DM literature [23].At a high-level, the constraints implementedwithin the SHDS modules are divided intosix broad categories [23]:

1 Knowledge Constraints specify the typeof healthcare knowledge to be mined. Forinstance, association rules, classifications,clusters, sequential patterns, concept de-scriptions and so on. This is the primaryconstraint as its definition determines thenature of subsequent constraints.

2 Data Constraints specify the set ofdata— the origin of the databases, the rele-vant features and relationships between dataitems—relevant to the selected SHDS. Suchconstraints are usually represented as SQLqueries to databases.

3 Dimension/Le�el Constraints specify thedimensions or levels of data to be examinedin a database. Such constraints are usuallyapplicable in concert with multidimensionaldatabases.

4 Rule Constraints specify certain rules,pertaining to the patterns being discovered,that need to be observed whilst mining thedata. Such constraints have a controllingand filtering behaviour.

5 Interestingness Constraints specify thedegree of usefulness or interestingness(statistical point of view) of the knowledgebeing discovered. Typically, if the knowl-edge being is deemed less interesting vis-a-vis the requirements of the SHDS then it isignored.

6 Qualitati�e Constraints specify measuresto examine the validity and correctness ofthe derived knowledge item.


Fig. 5. (a) The generation of a new SHDS-module by combining the data elements from multiple existingSHDS-modules. (b) The generation of a new SHDS-module in response to a request from a healthcare agency.

4.2. Strategic healthcare decision-supportser�ices menu

SHDS-Menu is the GUI application forthe specification of a SHDS via two modes:(1) The user may choose a pre-defined SHDS,and (2) The user may ‘design’ a specificSHDS.

4.3. Strategic healthcare decision-supportser�ices workshop

The SHDS-Workshop is an application en-vironment that supports: (1) the storage of allSHDS modules in the SHDS Library; and (2)the generation of demand-specific, newSHDS modules by the Module CustomisationWorkbench. The SHDS-Library stores a suiteof SHDS modules, with the functionality toadd new SHDS modules without disturbingthe internal dynamics of the available SHDS-modules. The Module Customisation Work-bench caters for the ‘dynamic’ generation ofspecialised SHDS modules as per user re-quests by: (a) Mixing existing SHDS-modulesin some principled manner to realise a ‘hy-brid’ SHDS-module (shown in Fig. 5a). Inthis case, either entire modules are synthe-

sised or else specific elements of each moduleare synthesised (note there exist certain con-straints to the synthesis of modules); and (b)Customisation of existing SHDS to deriveuser-specific services (shown in Fig. 5b).

4.4. Strategic healthcare decision-supportser�ices broker

The SHDS-Broker is the system to dynam-ically ‘mix and match’ multiple existingSHDS-modules to deliver innovative, ‘need-of-the-hour’ services customised according tothe user requirements. Architecturally, theSHDS-Broker comprises a number of inde-pendent ‘slots’, where each slot can be filledby a SHDS-module. The eventual SHDS gen-erated by the SHDS-Broker derives from thesystematic amalgamation of the multipleSHDS-modules within the different slots (seeFig. 4). Design issues addressed here are: (i)scheduling protocols for each module to ac-cess the services data and services knowledgelayers; (ii) the ‘hooks’ to assimilate the mod-ules in the slots; (iii) the sequence in whichthe modules will be visited by the processingengine; (iv) the compilation of the analysis byeach module to generate a global report.


4.5. Ser�ices-data layer

The Services-Data layer is responsible fordelivering the required data from the databases and data warehouse(s) to the SHDS-Broker. Information Broker [24] is a dataaccess application to collect and collate rele-vant information from the multiple and het-erogeneous healthcare data repositories.

4.6. Strategic healthcare decision-supportser�ices �isualiser

The SHDS-Visualiser implements a com-bination of SHDS (result) visualisation for-mats, such as graphs, tables, maps, abstractmaps and visualisation hypercubes.

4.7. SHDS deli�ery

Two intuitive, easy-to-use, visual applica-tions are designed to deliver the SHDS: (1)On the Internet and (2) Direct Links toSHDS info-structure, which will be In-surance and Pharmaceutical companies, hos-pitals and govt. agencies.

5. Predictive modelling of bacteria-antibioticsensitivity and resistivity patterns: anexemplar strategic healthcaredecision-support service

Here we present an exemplar SHDS im-plemented as an individual module in theSHDS info-structure. The purpose here isnot to give the technical details of theSHDS, rather to highlight its end-user per-spective— i.e. its origination, functionalityand decision-support information.

Nature of Ser�ice: The scope of theSHDS is to provide effective infectious-dis-ease epidemic risk management [13,14]. Thedecision-support knowledge provided by this

SHDS is characterised as the predicted fu-ture effectiveness of candidate antibiotics to-wards a bacteria. Effective infectious-diseaseepidemic risk management services canbenefit from this knowledge by increasingthe usage of effective antibiotics during theprojected time period, whilst at the sametime avoiding the usage of the ‘ineffective’antibiotics.

Data Collection: The bacterial sensitivity/resistivity data was provided by UniversitiSains Malaysia Hospital located in KotaBaharu, Malaysia. This data-set compiledover six years (1993–9198) comprises dataon the sensitivity/resistivity of 89 organisms,for which 36 different antibiotics wereprescribed.

Implementation of the SHDS: A Back-Propagation Neural Network (BPNN) wasused to perform the core DM task— theBPNN was simply taught historical bacterialsensitivity/resistivity data of the time-seriesand the learnt BPNN is used to predict fu-ture bacterial-antibiotic sensitivity. In func-tional terms, this SHDS module requiresthree past sensitivity/resistivity Bacteria-An-tibiotic (BA) values (i= −3,−2,−1) andthe present (i=0) to generate three futuresensitivity/resistivity BA values (i=1,2,3).

End-User Functionality of the SHDS : Thisparticular SHDS is made accessible tohealthcare practitioners via a WWW inter-face. The workflow is as follows: (i) the userneeds to specify the bacterial organism, thelist of possible interacting antibiotics whoseeffectiveness is being examined, the natureof the forecast profile to be generated, andthe predictive time frame (as shown in Fig.6a); (ii) the selected SHDS performs thenecessary DM activities on the availabledata; and finally (iii) the forecasted resultsare displayed on a dynamically generatedWeb-page (as shown in Fig. 6b).


6. Features of the SHDS info-structure

Functionally, the proposed SHDS info-structure not only meets the specific demandsfor SHDS, but extends further to add valueby way of supporting a number of attractivefeatures, such as:

Data Collation: Comprehensive data analy-sis by collating data from multiple datarepositories to form a virtual, seamless andcontinuous data source.

Suite of Strategic Decision-Support Ser-�ices: A wide range of SHDS, categorisedinto four prominent classes: (1) TreatmentManagement; (2) Pharmaceutical Require-ments; (3) Healthcare Planning and Servicesand (4) Best Practices Benchmarking.

User-Specific Decision-Support Ser�ice Re-quests: To cater for usage preferences andtemporal constraints, we allow users the flexi-bility to customise existing SHDS modules tomeet user-specific decision-support requests.

Addition of New Decision-Support Ser�ices:Dynamic addition of new SHDS-modules tothe existing SHDS info-structure is possiblewithout disturbing the existing SHDS-modules.

Multiple Data & Knowledge Analysis Tech-niques: A confluence of functionally diversedata analysis methodologies realise the gener-ation of a variety of decision-support ser-vices, such as data mining, statistical analysis,symbolic rule extraction, time-series forecast-ing and benchmarking.

Multiple Result Visualisation Methods:Multiple results visualisation formats, rang-ing from graphs to 3D hypercube maps toactive reports/documents can be selected bythe user.

7. Concluding remarks

For all practical purposes, modern health-care systems generate massive amounts of‘knowledge-rich’ healthcare data, but unfor-tunately this asset is not yet fully ‘cashed’ forimproving the management and delivery ofhealthcare services. In this paper, we havesuggested that the possible synergy of KMand DM techniques can provide opportuni-ties for the generation of strategic knowledge-driven decision-support services from theseemingly ‘mundane’ healthcare data. The

Fig. 6. (a) The main screen for the specification of the SHDS. (b) Forecast report for an exemplar bacteria-antibioticinteraction.


general idea is to leverage the healthcare en-terprise’s databases, data warehouses andknowledge bases to derive experientialknowledge from it, which can in turn be usedto optimise strategic decision-making andplanning. We have proposed a viable IT info-structure— the so-called SHDS info-struc-ture— that can facilitate the automatedtransformation of healthcare data to a suiteof heterogeneous strategic knowledge ser-vices. More importantly, the HEM providesan opportunity to migrate healthcare practicerules, primarily stated in texts, towards thegeneration of value-added, pro-active strate-gic services that may directly impact the be-haviour and efficacy of the healthcareenterprise as a whole. In conclusion, we em-phasise that the conception of the both theHEM and the SHDS info-structure has iden-tified opportunities to improve healthcaremanagement through the increased use KMtechnology. The HEM project is still underdevelopment, in particular the focus of theon-going work is the acquisition of tacitknowledge [25] and the creation of a varietyof medical knowledge bases.

Acknowledgements

The author gratefully acknowledges thecontributions of Prof. Zaharin Yusoff, MrCheah Yu-N and Mr Alwyn Goh who as-sisted in crystallising the ideas presented here.

References

[1] Y.-N. Cheah, S.S.R. Abidi, Evaluating the efficacyof knowledge management and organisationalmemory towards healthcare enterprise modelling,workshop on knowledge management and organi-zational memories at int. joint conference on ar-tificial intelligence IJCAI’99, Stockholm, 1999.

[2] Y.-N. Cheah, S.S.R. Abidi, Healthcare knowledgemanagement through building and operationalis-ing healthcare enterprise memory, in: P. Kokol, B.Zupan, J. Stare, M. Premik, R. Engelbrecht(Eds.), Medical informatics in Europe (MIE’99),IOS Press, Amsterdam, 1999.

[3] D. O’Leary, Knowledge management systems:converting and connecting, IEEE Intelligent Sys-tems 13 (3) (1998) 30–33.

[4] T. Davenport, D. DeLong, M. Beers, Successfulknowledge management projects, Sloan Manage-ment Review Winter (1998) 43–57.

[5] K. Sveiby, The new organizational wealth: manag-ing and measuring knowledge-based assets,Berrett-Koehler, San Francisco, 1997.

[6] S.S.R. Abidi, Y.-N. Cheah, A convergence ofknowledge management and data mining: towards‘knowledge-driven’ strategic services, 3rd Int.Conference on the practical applications of knowl-edge management (PAKeM’2000), Manchester,2000.

[7] U.M. Fayyad, G.P. Shapiro, P. Smyth, R. Uthu-rusamy (Eds.), Advances in knowledge discoveryand data mining, AAAI Press, California, 1996.

[8] S.S.R. Abidi, Z. Yusoff, Data driven healthcaremanagement: from a philosophy to an IT info-structure, in: Int. conference on multimedia andinformation technology. Kuala Lumpur, 1998.

[9] S.S.R. Abidi, Applying knowledge discovery inhealthcare: an info-structure for delivering ‘knowl-edge-driven’ strategic services, in: P. Kokol, B.Zupan, J. Stare, M. Premik, R. Engelbrecht(Eds.), Medical Informatics in Europe (MIE’99),IOS Press, Amsterdam, 1999.

[10] A. Abecker, A. Bernardi, K. Hinkelmann, O.Kuhn, M. Sintek, Towards a technology for orga-nizational memories, IEEE Intelligent Systems 13(3) (1998) 40–48.

[11] Government of Malaysia, CRFP telemedicineflagship application: lifetime health Plan. http://www.mdc.com.my, 1997.

[12] S.S.R. Abidi, A. Goh, Z. Yusoff, Telemedicineand medical informatics in the multimedia supercorridor: the Malaysian vision, in: B. Cesnik, A.McCray, J. Scherrer (Eds.), Ninth World Con-gress on Medical Informatics (MedInfo ’98), IOSPress, Amsterdam, 1998.

[13] S.S.R. Abidi, A. Goh, Applying knowledge dis-covery to predict infectious disease epidemics, in:H. Lee, H. Motoda (Eds.), Lecture Notes in Ar-tificial Intelligence 1531-PRICAI’98: Topics in Ar-tificial Intelligence, Springer Verlag, Berlin, 1998.


[14] S.S.R. Abidi, A. Goh, Neural network based fore-casting of bacteria-antibiotic interactions for infec-tious disease control, in: B. Cesnik, A. McCray, J.Scherrer (Eds.), Ninth World Congress on Medi-cal Informatics (MedInfo’98), IOS Press, Amster-dam, 1998.

[15] J. Euzenat, Corporate memory through coopera-tive creation of knowledge bases and hyper-docu-ments, tenth knowledge acquisition forknowledge-based systems workshop, Banff, Al-berta, 1996.

[16] B. Chandrasekaran, J.R. Josephson, What areontologies, and why do we need them?, IEEEIntelligent Systems 14 (1) (1999) 20–26.

[17] T.R. Gruber, Toward principles for the design ofontologies used for knowledge sharing, Int. work-shop on formal ontology, Italy, 1993.

[18] A. Babic, Knowledge discovery for advanced clin-ical data management and analysis, in: P. Kokol,B. Zupan, J. Stare, M. Premik, R. Engelbrecht(Eds.), Medical Informatics in Europe (MIE’99),IOS Press, Amsterdam, 1999.

[19] A. Kircher, J. Antonsson, A. Babic, H. Casimir-Ahn, Quantitative data analysis for exploring out-comes in cardiac surgery, in: P. Kokol, B. Zupan,J. Stare, M. Premik, R. Engelbrecht (Eds.), Medi-

cal Informatics in Europe (MIE’99), IOS Press,Amsterdam, 1999.

[20] B. Zupan, N. Lavrac, E. Keravnou (Eds.), Datamining techniques and applications in medicine,artificial intelligence in medicine, 1999, p. 15.

[21] N. Lavrac, E. Keravnou, B. Zupan (Eds.), Intelli-gent data analysis in medicine and pharmacology,Kluwer, Boston, 1997.

[22] M. Holena, A. Sochorova, J. Zvarova, Increasingthe diversity of medical data mining through dis-tributed object technology, in: P. Kokol, B. Zu-pan, J. Stare, M. Premik, R. Engelbrecht (Eds.),Medical informatics in Europe (MIE’99), IOSPress, Amsterdam, 1999.

[23] J. Han, L. Lakshmanan, R. Ng. Constraint-basedmultidimensional data Mining, IEEE computer,August 1999.

[24] S.S.R. Abidi, Internet information brokering: are-configurable database navigation, data filterand export system, Informatica 21 (1997) 185–191.

[25] Y.-N. Cheah, S.S.R. Abidi, A strategy for tacitknowledge acquisition and representation: a casefor scenarios, 3rd int. conference on the practicalapplications of knowledge management(PAKeM’2000), Manchester, 2000.

.

Knowledge management in healthcare: towards ‘knowledge ...sraza/papers/IJMI01.pdf · a suite of...

Documents

Transcript of Knowledge management in healthcare: towards ‘knowledge ...sraza/papers/IJMI01.pdf · a suite of...