Proceedings from Scandinavian Conference on Health ...15th Scandinavian Conference on Health...

i

Proceedings from The 15th Scandinavian Conference on Health

Informatics

2017

Kristiansand, Norway

August 29–30, 2017

Editors Santiago Martinez, Andrius Budrionis,

Ann Bygholm, Mariann Fossum, Gunnar Hartvigsen, Maria Hägglund,

Carl E Moe, Elin Thygesen, Vivian Vimarlund, and Kassaye Y Yigzaw

ii

Copyright The publishers will keep this document online on the Internet – or its possible replacement – from the date of publication barring exceptional circumstances.

The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/her own use and to use it unchanged for non-commercial research and educational purposes. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility. According to intellectual property law, the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.

For additional information about Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/

Linköping Electronic Conference Proceedings, No. 145ISBN: 978-91-7685-364-1ISSN: 1650-3686 eISSN: 1650-3740 URL: http://www.ep.liu.se/ecp/contents.asp?issue=145Linköping University Electronic Press Linköping, Sweden,

© The Authors, 2018

iii

Scientific Program Committee Chair: Santiago Martinez, Norway Andrius Budrionis, Norway Ann Bygholm, Denmark Mariann Fossum, Norway Gunnar Hartvigsen, Norway Maria Hägglund, Sweden Ole Hejlesen, Denmark Carl E Moe, Norway Elin Thygesen, Norway Vivian Vimarlund, Sweden Kassaye Y Yigzaw, Norway

Sponsors Svensk förening för medicinsk informatik Norwegian Centre for E-health Research Aalborg University University of Agder Linköping University Karolinska Institutet University of Tromsø – The Arctic University of Norway

iv

Table of Contents Articles New technology in Norwegian municipalities' health care services: national advices meets regional conditions Hilde G. Corneliussen and Marit Haugan Hove………………………………………………1 Mapping FHIR Resources to Ontology for DDI reasoning Raees Abbas, Islam Fathi Hussein Al Khaldi, Getinet Ayele and Jan Pettersen Nytun…………………………………………………………………………………………...8 EDMON - A Wireless Communication Platform for a Real-Time Infectious Disease Outbreak De- tection System Using Self-Recorded Data from People with Type 1 Diabetes Ashenafi Zebene Woldaregay, Eirik Årsand, Alain Giordanengo, David Albers, Lena Mamykina, Taxiarchis Botsis and Gunnar Hartvigsen……………………………………….14 Complex Medication Reconciliation in the Danish Medication System: Shared Medication Record for patients in transition of care across sectors Helene Tarp, Lili Worre Høpfner Jensen, Nikolaj Krabbe Jepsen, Mads Clausen, Nina Aagaard Madsen, Henrik Majkjær Marquart and Louise Pape-Haugaard………………….21 Using mobile sensors to expand recording of physical activity and increase motivation for prolonged data sharing in a population-based study André Henriksen, Gunnar Hartvigsen, Laila Arnesdatter Hopstock and Sameline Grimsgaard...............................................................................................................................28 Collective action in national e-health initiatives: findings from a cross-analysis of the Norwegian and Greek e-prescription initiatives Polyxeni Vassilakopoulou, Aleksandra Pesaljevic, Nicolas Marmaras and Margunn Aanestad....................................................................................................................................36 Systems integrating self-collected health data by patients into EHRs: a State-of-the-art review Alain Giordanengo, Meghan Bradway, Miroslav Muzny, Ashenafi Woldaregay, Gunnar Hartvigsen and Eirik Arsand.......................................................................................43 Infrastructure for the Learning Healthcare System: Centralized or Distributed? Andrius Budrionis and Johan Gustav Bellika.......................................................................... 50 A significant increase in the risk for exposure of health information in the United States: result from analysing the US data breach registry Johan Gustav Bellika, Alexandra Makhlysheva and Per Atle Bakkevoll.................................55 Innovative Simulation of Health Care Services in the Usability Laboratory: Experiences from the Model for Telecare Alarm Services-project Berglind Smaradottir, Santiago Martinez, Elin Thygesen, Elisabeth Holen-Rabbersvik, Torunn Vatnøy and Rune Fensli...............................................................................................60 Fictional Narratives for Clinical App Development Tanja Svarre and Tine Bieber Kirkegaard Lunn......................................................................66

v

Social Network Analysis and Tele Homecare Karsten Niss, Ole Hejlesen and Louise Bilenberg Pape-Haugaard.........................................72 Factors affecting seniors’ attitudes on the use of information technology Tor-Ivar Karlsen, Marit Bolstad Tveide and Rune Fensli........................................................78 The role of IT-service in future health care, can they be ignored? Karen Stendal and Janne Dugstad...........................................................................................82 Evaluation of a telemedical based care pathway for patients with COPD Elin Thygesen and Carl Erik Moe............................................................................................87 Sensor Technology for night surveillance: the experiences of next of kin Linda Iren Mihaila Hansen1, Mariann Fossum and Carl Erik Moe..........................................93

Paperfromthe15thScandinavianConferenceonHealthInformaticsSHI2017,Kristiansand,Norway,29-30August.ConferenceProceedingspublishedbyLinköpingUniversityElectronicPress.

©TheAuthor(s).

NewTechnologyinNorwegianMunicipalities'HealthCareServices:

ExperiencesinSmallRuralMunicipalities

HildeG.Corneliussen1andMaritHauganHove11WesternNorwayResearchInstitute,Sogndal,Norway

AbstractNewassistivetechnologyappearsaspartofthesolutionforacoming'crisis'inhealthcareservices.Thispaperanalyses experienceswith introducingwelfare technology in smallmunicipalities in the regionof SognandFjordane.Theempiricalmaterial isbasedonnine interviewswithprojectgroupsandhealthcare leaders inmunicipalities workingwithwelfare technology.Whilemany of the national advice and recommendationsaboutassistivetechnologyaregoodguidelines,thefindingssuggestthatcertainchallengesareincreasedbythesmallsizeofthemunicipalities.Thispointstotheimportanceofrecognisingdifferentneedscreatedbyavariationamongmunicipalities.

KeywordsWelfaretechnology,Assistivetechnology,Municipalityhealthcareservices

1 INTRODUCTION

Thereisawidespreadagreementthatwewillneedtodeliverhealthcareservicesinnewwaysinthefutureduetoanincreasinggapbetweenagrowingpopulationinneedofhealthcareservicesandthenumberof'warmhands'workingwithinmunicipalhealthcare.ItisprojectedbyStatisticsNorwaythatthe'careburden'inNorwaywilldoublewithinthenext25years.Alackofresourcesforsupportinginstitutionalcaremeansthatmorepeoplewillhavetobesupportedintheirownhomes.TheNorwegianreportInnovationinCare(NOU2011:11)suggeststhatassistivetechnology,or'welfaretechnology'whichitisoftencalledinScandinavia,willhelptomakemorepeopleabletoremainandlivesafelyintheirownhomeuntilanolderage.Atthesametime,welfaretechnologywillincreasequalityofhealthcareservicesforusersandrelativesand,accordingtothereport(NOU2011:11),itwillreduceorpostponetheneedforinstitutionbasedcare.Welfaretechnologyisfurtherdefinedinthereportastechnologicalassistancethatcontributetoincreasedsafety,socialparticipationandmobility,aswellasphysicalandculturalactivity.Itenablestheindividualtomanageeverydaylifedespiteofsickness,oldageormentaland/orphysicaldisability.Thegoalisthattechnologywillprovideanearlywarningaboutaccidentsorunwantedepisodes.Thiscanincreasequalityofcare,aswellassavetimeforhealthcarepersonnel.Seenfromtheperspectiveoftheusersorrelatives,suchtechnologymightincreasepersonalfreedom.Forinstance,usingaGPSoradigitalsystemforlocalisingapersoncanhavesignificantpositiveeffectsforpersonswithacognitivedisabilitythatliketogoforawalk.WithaGPSrelativesorhealthcarepersonnelcanbenotifiedandthewanderingusercanbefoundifshe

cannotfindherwaybackonherown(Øderud,2015).Wewillnotgointothediscussionofwhether'coldtechnology'or'warmhands'arebetterforhealthcareservicesortheusers.However,werecognizethatthereisastrongpoliticalpressureformunicipalitiestointroducewelfaretechnologyinNorway(Melting,2017;Melting,2015;Helsedirektoratet,2012).Similartrendsconcerningassistivetechnologyinhealthcareservicescanbeseeninmanywesterncountries.Althoughthisinvitestoexchangeofknowledgeandexperiencesacrossnationalborders,somechallengesinthisworkarenational.InNorway,thereisanationalprogrammeforwelfaretechnologyrunbytheNorwegianDirectorateforHealthincollaborationwiththeNorwegianAssociationofLocalandRegionalAuthorities(KS),providingadviceandguidelinesforintroducingwelfaretechnologyinmunicipalities'healthcareservices.Inthispaper,weanalyseanddiscussexperiencesfromaprojectaimingtoimplementwelfaretechnologyinaselectionofmunicipalitiesintheruralregionofSognandFjordane.ThemunicipalitiesintheregionSognandFjordanearemainlysmallregardingpopulation,butwithrelativelylargeareasandlongdistances.Thegeographyischallengingandmanyofthemunicipalitieshavescarcefinancialaswellashumanresources.Aswewillshowintheanalysis,someofthechallengesthatthemunicipalitiesmeetcanbefoundalsoinotherregions.However,someofourfindingssuggestthatthesizeofthemunicipalitiesalsoproducesomeoftheirchallenges–someofwhichhavenotbeenthoroughlyinvestigatedandclarifiedbythenationalprogramme.

1.1WelfaretechnologyinmunicipalitiesinSognandFjordane

ThisstudyanalysesexperiencesfromtheprojectWelfareTechnologyintheMunicipalitiesinSognandFjordane,

15thScandinavianConferenceonHealthInformaticsSHI2017,Kristiansand,Norway,29-30August. 2

whichstartedin2014.Theprojecthassincethenbeenrunbyacross-sectoralworkinggroup,includingseveralmunicipalities,countycouncil,representativesfromresearchandeducationandregionalhealthauthorities.TheCountyGovernorfundedtheprojectwith50%,whilethemunicipalitieshaveprovidedtheremaining50%.In2014,noneofthe26municipalitiesintheregionhadstartedimplementingwelfaretechnology.Consequently,thefirstphaseofthisprojectfocusedoninvitingallmunicipalitiestoparticipateandencouragedthemtostarttheprocessofplacingwelfaretechnologyontheiragenda.Thesecondphaseoftheprojectwasrunningthrough2015.Atthattime,theNorwegianDirectorateforHealthhadsetadead-lineforimplementingwelfaretechnologyinmunicipalhealthservicesby2018andtherewasapparentlynotimetolose.Thisdeadlinehaslaterbeenpostponedseveraltimes.Projectphase2focusedonestablishingpracticalexperienceandknowledgebytestingspecificwelfaretechnologicalsolutionsinpilotmunicipalities.Thisincludedsafetysystemsforshelteredhousingandprivatehomes,withvarioussensorsandalarmsthatwarnhealthcareservicesorrelativeswhenunforeseenandunwantedeventshappen.Suchunwantedeventscouldbeapersonfalling(fall-detectiontechnology),leavingherhome(GPSormotion-detectionsensors)orleavingbedatnightwithoutreturning(sensorinbed).Othertechnologicalsolutionsthatweretestedincludeddigitalmedicinedispensersthatremindstheuserwithanalarmwhenitistimeforhermedicine.Also,ifthemedicineisnottaken,analarmcanwarnrelativesorhealthcarepersonnel.Onemunicipalityintroducedactivitytechnologyintheirpilot,representedbyanexercisebikecoupledwithavideoshowingrecordingsfromlocalvillageroadstoincreasemotivationforindoorexercise.Atthetimeofthethirdprojectphase,in2016,thenationalwelfaretechnologyprogrammehadstartedtopublishresultsandnationwideadviceformunicipalities.Thus,whenthepilotscontinued,nationaladvicewasmoredirectlyinvolvedthroughtheintroductionofaframework,createdbythenationalprogramme,forrealisingcostandqualitybenefitsfromusingtechnology.Thepilotmunicipalitiescontinuedtheiractivitieswhilereceivingtraininginthiscostandqualitybenefittool.Themunicipalitieswerethenencouragedtousethismethodintheirownwork,andtheworkwasevaluatedthroughinterviewsfivemonthslater.Theaimoftheregionalprojectwasnotonlytogainexperienceinsinglemunicipalities.Theexperiencesandknowledgegatheredthroughthesepilotswouldbecomethebasisofadviceformunicipalitiesintheregionandsimultaneouslycreatelocalandregionalspearheadmunicipalities.Thetechnologicalsolutionsthatwerepilotedwerenotnewtechnology,buttheywereunfamiliartothemunicipalitiesofthisregion.ManyotherprojectsacrossNorway,aswellasothercountries,havedocumentedpositiveeffectsofsuchtechnologies

(Melting,2017;Melting,2015).Theaimofthisarticleishowevernottoevaluatethetechnology,buttoanalysetheprocessesofintroducingtechnologywithinthemunicipalities.

2 LITERATURE

ThereareseveralstudiesanalysingthechallengesofimplementationofwelfaretechnologyinNorwegianmunicipalities(Ørjasæter,2016;Dugstadetal.,2015;Nordtugetal.,2015;Dischetal.,2015;Grutetal.,2013).Thesemostlyfocusontheeffectsofthetechnologies,aswellastheexperiencesofhealthcareworkersregardingimplementation.Dugstadetal.(2015)identifyseveralbarriersregardingimplementationofwelfaretechnologyinvariousmunicipalitiesineasternandsouthernNorway(Drugstadetal.,2015).Someofthesearepresentedasuniversalchallenges,thatprobablycouldbefoundinmostmunicipalities.Examplesofsuchbarriersareknowledgeoftechnologyingeneralaswellasthespecificwelfaretechnologicalsolutionsandcommunicationbetweentechnologistsandhealthprofessionals.Nordtuget.al(2015)studiesthreedistrictsinmid-Norwayimplementingsensortechnologyintheirhealthcareservices.Theyfoundhighmotivationamongthehealthprofessionals,aswellasaneedforextraresourcestofocusoncross-sectionalcooperation.Grutetal.(2013)foundthat'tobesuccessful,theadoptionofwelfaretechnologyneedstobeanchoredatseveraldifferentlevelswithinthemunicipalorganization'andthatcommunicationbetweenallpartiesiscrucialinsuchprocesses.Oneoftheearlywelfaretechnologyprojectsconcludedthatwelfaretechnologycouldbeperceivedasathreattotheexistingcorporateculture,structuresandvaluesofmunicipalhealthcare(DetMidtnorskevelferdsteknologiprosjektet2014).Thisprojectalsoprovidedalistofadviceformunicipalitiesthatwereinterestedinwelfaretechnology,whichbecamepartoftheinspirationfortheworkintheregionofSognandFjordane.Smallmunicipalitiesinsomewhatruralareasareincludedinsomeofthestudiesmentionedabove.However,nonehavecasesfromtheregionofSognandFjordane,andnoneareidentifyingspecificchallengesconnectedtosizeofthemunicipality.Inthisarticle,wewillpointtoexperiencesfromSognandFjordanethatsuggestthatalsosizeofmunicipalitiesisimportantintheprocessofimplementingwelfaretechnology.

3 METHODSANDTHEORETICALFRAME

Theempiricalmaterialincludesatotalofnineinterviewswithwelfaretechnologyprojectgroupsorrepresentativesfromdifferentmunicipalities'healthcareservicesintheregion.Eightoftheinterviewswerewithsinglemunicipalities,whileonewasagroupconversationwithseveralmunicipalities.Theinterviewslasted


between30and90minutes,andhadbetweenoneandnineparticipants.Theseweremainlytheleaderofhealthcareservicesinthemunicipalityandpeoplefromtheprojectgroup,oftenincludingmiddleleaderswithresponsibilityforhomecareservices.Otherparticipantspresentinoneortwointerviews,wererepresentativesofITservices,representativesofuserorganisationsand,inonecase,apolitician.Theinterviewswerecompletedduring2016and2017.Theresearchersresponsiblefortheinterviewswerepartofthecross-sectoralworkinggroupthatmetanddiscussedthelocalprojectsseveraltimesintheprojectperiod.Theresearcherswerenotinvolvedinthemunicipalities'localprojectgroups.Attheendofphase2andphase3theresearchersvisitedthemunicipalitiesormettheprojectgroupsinvideomeetings,tointerviewthemabouttheirexperiencesintheworkofintroducingwelfaretechnologicalsolutionsaswellastheirexperiencewithusingthecostandqualitybenefittool.Twodifferentinterviewguideswereused,bothdesignedtoevaluatetheprocessofimplementingwelfaretechnology,howeveronealsofocusingonthecostbenefittool.TheinterviewguidesbuiltonfindingsfromsimilaranalysisinotherprojectsinNorwegianmunicipalities,thusaimingtocaptureexperiencesrelatedtoaspectsthathadbeenfoundimportantinotherprojects.Theinterviewguidesalsoreflectedthetheoreticalframeworkfromtechnologystudies,seeingtechnologynotonlyasaphysicalobject,butalsoincludingknowledge,skills,symbolsandroutines.Thetheoryofdomesticationemphasisesthatwhentechnologyisimplementedinanewcontext,thiscontextalreadyhasitsownroutines,normsandvaluesthatthenewtechnologyhastoadjustto,simultaneouslyasintroductionofnewtechnologychangesthecontext(Silverstoneetal.,1997).Thus,technologyhasan'interpretativeflexibility'fordifferentusergroups(Bijkeretal.,1993).Ouraimwastoexplorehowtheprocessofimplementingwelfaretechnologywasperceivedandexperiencedbythosewhowereclosetothisprocessinthemunicipalities,whilerecognisingthecomplexityandnegotiationsgoingoninsuchprocesses.Weweretworesearcherspresentduringtheinterviews,withoneaskingquestionsandtheotherinchargeofmakingextensiveinterviewnotestocapturethedialoguesintheinterviews.TheextensivenoteswerelateranalysedwithGroundedTheoryMethod,meaningthatwecodedthetexttodiscoverpatternsinthematerial(Charmaz,2006).Belowwewillgothroughsomeoftheexperiencesthatstandoutasparticularlynoteworthyforthemunicipalitiesweinterviewedinthisregion.

4 RESULTS

4.1AbundanceofwillandlackofknowledgeWefoundthatintroducingwelfaretechnologywaswelcomedbothbypoliticalandadministrativeleadership

inthemunicipalitiesthatactedaspilotsintheproject.Thiswasnotsurprising.Infact,theyhadbeenrecruitedaspilotsexactlybecausetheyhadalreadyputtechnologyontheagenda,makingthemready,ornearlyready,toimplementtechnologicalsolutions.Simultaneouslyaswefoundpoliticalwilltosupportwelfaretechnologyinitiatives,wealsofoundanequallylargelackofknowledgeaboutthesame.Inoneofthemunicipalities,themessagefrompoliticianswasthattheysupportedwelfaretechnologyregardlessofwhatkind.Inanothermunicipality,acquiringandinstallingwelfaretechnologyhadbeenpartofapublicprocurementinwhichthetechnologywasnotspecified,thusplacingthechoiceoftechnologyoutsidethehealthcaredepartmentofthemunicipality.Forathirdmunicipality,apoliticaldecisiontocloseshieldedhousingforpersonswithdementiawasfollowedbyarequirementthattechnologyshouldsolvethesituation,withnofurtherspecification.Thepoliticalwillwasnotonlyaccompaniedbyalackofknowledge,butalsoalackofpracticalsupport.Tooneoftheprojectgroups,itfeltasifthesupportfrompoliticiansandadministrationwas'almosttoobig'–theywere'cheeredon',butdidnotreceiveadequateresources,timeorfundingtoseeprojectsthrough.Thiscausedfrustrationamongthosewhohadbeenappointedasresponsiblefordrivingthelocalprojectforward.Itwasdifficultforthemtoachievetheirgoalsandtheyhadtodoprojectwork,likepreparingreports,intheirleisuretime.

4.2ProjectleadershipandprojectgroupsAllthemunicipalitiesinthisstudyhadestablishedaclearprojectleadership–inmostofthemthiswastheheadofhealthcareservices,occasionallydelegatedtomiddleleaders.Theyhadalsoestablishedprojectgroups,althoughithadbeenachallengeforsomeofthemunicipalitiestofindparticipantsfortheproject,thusthesizeoftheprojectgroupvariedbetweenmunicipalities.Projectleaderstogetherwithprojectgroupswerethevitaldrivingforcesintheworkofimplementingwelfaretechnologyinallthemunicipalities.Oneofthelargestchallengesthattheprojectleadersreported,wasthenewcompetencerequirementstheymetwhenstartingtoworkwithwelfaretechnology.Mostoftheseleadershadaprofessionalbackgroundinhealthcare,thus,technologyrepresentedanewcompetenceforthem:'NowIhavetolearnthingsIshouldn'treallyknow',oneofthemsaid.Anotherprojectleaderexplainedhowshehadtoactasa'Jackofalltrades',doingthingsshe'doesnotknow',andwhenpeopleasked,shecouldnotgivesatisfyinganswers.Anotherchallengeforleadersandtheprojectgroupswastobuildmotivationamongotheremployees.Motivatinghealthcareworkerstostarthandlingnewtechnologywasnotasimpletask.Onlyoneofthemunicipalitieshadreservedtimeforinvolvingemployees.Intheothermunicipalities,thenewtaskshadtobeintegratedinan


alreadyfullypackedworkdayschedule,makingthesmallestchallengesdifficulttoovercome.Wefoundthatinsomeofthemunicipalitieswithaleader-drivenprocess,theemployeeswerenotinvolvedfromthestartoftheproject.Theresultwasaweakmotivationandalackof'ownership'amongtheemployees.Thisinturnmadetheleadersevenmoreimportanttokeeptheprojectgoing.Thus,increasingtheleader-drivencharacteroftheproject,inonemunicipalityevenresultinginthetechnologynotbeingusedaccordingtotheplan.

4.3TechnologycompetenceChallengestoputtogetheraprojectgroupwaspartlyreferredtoasalackofpeople,andpartlyasalackoftherightpeople.Lackofpeoplereferredtothesizeofthemunicipalities:beingsmallalsomeantfewhandstosharethework.Thelackofrightpeoplehadtodowiththenewcompetencerequiredtodealwithtechnology.Wesawthateventechnologythatisapparentlysimpletohandlerequiresbasictechnologicalcompetence,suchaspluggingtherightcableintotherightsocket.Lackoftechnologycompetenceamongemployeeswhowereresponsibleforthedailycontactwiththetechnologywasachallengeinseveralmunicipalities.Someofthemclaimedthatemployeeswerehesitantordidnotfeelsafebeingintroducedtonewtechnology.Rhetorically,thechallengeoffindinghealthcarepersonnelwithasatisfactoryabilitytohandletechnologywasseveraltimesreferredtoasachallengetiedtogenderandage,andinsomemunicipalitiestheyleftthehandlingoftechnologytotheyoungerpersonnel.Somealsosuggestedthatscepticismtowardsnewtechnologyhadtodowithgender:the'womansyndrome',orthe'whatifeverythinggoeswrongsyndrome'.Thislackoftechnologycompetenceamonghealthcareprofessionalswasfurthercomplicatedinsomeofthemunicipalitiesthatalsohadalackofgeneraltechnologicalcompetenceinthemunicipalities–alackofITpeople.Thiswasaresultofinter-municipalcompaniesandagreementsaboutcollaborationaroundICTsupportforthesmallmunicipalitiesintheregion.Inaddition,severalofthemunicipalitiesemphasisedthattheyneededapersonwhocould'translate'betweenhealthcareandtechnology:a'healthcare-ICTperson'whoknewenoughabouttechnologybutatthesametimeunderstoodtheneedswithinhealthcareservices.

4.4Findingtechnologyforusers,orusersfortechnologyOtherstudieshavefoundthatmunicipalitieshavehaddifferentpathwaysintotheirworkwithwelfaretechnology,fromstartingwithatechnology,auserneed,oraserviceinneedofinnovation(Grutetal.,2013).AsimilarvariationisfoundamongmunicipalitiesinSognandFjordane.Onemunicipalitywasapproachedbyaproducer;onewasforcedtosolveauserneed;onechoseatechnologicalsolutionthatoneoftheinvolvedhealthcarepersonnelhadlearntaboutit;andfinally,onemunicipalityhadleftthechoiceoftechnologytothe

entrepreneurbuildingtheirhealthcareinstitutionandnewshelteredhomes.Whatallthesecaseshaveincommonisthedifficultiesinestablishingsufficientknowledgeaboutdifferenttechnologicalsolutionsduetolimitedpersonnelresourcesandlimitedtechnologicalcompetenceamongthehealthcareleaders.Wesawthatoneresultofthiswasreluctancetomakeanychoices,butrathertowaitformoreconcreteadvicefromhealthauthorities.Thiswaythemunicipalitieswouldnothavetodoalltheexplorativegroundworkthemselves.Anotherresultwefound,wasatendencytofeel'stuckwith'thetechnologicalsolutionstheyhadalreadychosen,ortothetechnologythatappearedtobewithinreachoftheirpracticalandeconomicalresources.Thus,whenamunicipalityalreadyhadacquiredatechnologicalartefact,theirnextchallengewastofindauserforthattechnology,turningthehealthcareauthorities'adviceto'startwiththeusers'upsidedown.

4.5WhensizeisachallengeThemunicipalitiesdidnotonlyexperiencechallengesinestablishingadequateknowledgeaboutthetechnologicallandscape,butalsoaboutthetechnologicalsolutionstheyhadchosen.Onemunicipalityexperiencedthiswhentestingdigitalmedicinedispensers.Thedispensersnotifiedtheuserwhenitwastimetotaketheirmedicine,andcouldsendatextmessagetorelativesorhealthcareservicesiftheuserdidnotrespondtothisnotification.However,theysoondiscoveredthatthedigitalplatformthatsendsmessagesfromthedispenserrequiredalargernumberofdispensersinusethanthetwothismunicipalityhadacquired.Thesmall-scaleoperationsrepresentedachallengeforthemunicipalitiesalsoinanotherway.Havingalimitedsetoftechnologicalartefactsorsolutionsinuse,alsomeantthatittooklongertimetobuildexperienceamongemployeesresponsibleforusingthetechnology,particularlywhenthiswasnotpartoftheireverydayroutine.Lackofknowledgeabouthowtooperatethetechnologicalsolutionscouldresultinthetechnologynotbeingused:'Standingthereandfumblingwithsomethingtechnicalthatwecan'tmakeworkisdifficult–thenwe'dratherdoittheoldway',oneofthehomecareworkersexplained.Althoughmostofthesetechnologysolutionsarenotnew,itisstillachallengetomakedifferenttechnologicalsystemscommunicatewitheachother.Threedifferentdigitalhealthcaresystemsareinuseinthemunicipalities,andsomeofthewelfaretechnologysolutionscommunicatewithoneofthese,butrarelywithall.Thepilotmunicipalitiessawthemselvesas'toosmall'tonegotiatewiththeproducerstomakethemdevelopthetechnologytofitthemunicipality'sneeds.Themunicipalitiesfounditparticularlyfrustratingwhenitappearedthattechnologicalsolutionsrecommendedbythenationalprogrammewerenotfullydevelopedandreadytobeimplemented,likeoneofthemillustratedbysaying:


'itrepresentslargecostsforasmallmunicipalitybudgetandit'sperceivedasariskwithmunicipalmoneywhenthesolutionthatisrecommendedisnotproperlytestedorcommunicatingwithestablishedprofessionalprograms.Wearenotcertainthatthesereallyarethefutureofgoodsolutions.'

4.6IdentifyingbenefitsfromusingtechnologyThepilotmunicipalitiesdidnotonlytesttechnology,theywerealsoinvitedtoacoursetolearnandlatertestacostandqualitybenefittooldevelopedbythenationalprogramme.Atthecourse,largesheetsofpaperwithtablestofillinwerepresentedfortheparticipants.Somewrotedirectlyonthepaper,othersusedpost-itnotes.Theystartedwiththeuser'sneedsandexploredhowthesecouldbe(better)metwithtechnologicalsolutions,andfurther,whichchangesthathadtobemadetoachievetheidentifiedbenefits.Themunicipalitiesagreedthatthiswasagoodmethod.Theysuggestedthatitincreasesconsciousnessofthepre-andpost-technologyscenarios,makingitpossibletoseebenefits.Thus,theysawthisasatoolfordocumentingprogressforpoliticiansandtodemonstrateforscepticalcolleaguesthatusingtechnologycouldbenefitusersbyincreasingservicequalityaswell.However,wealsofoundaunanimousagreementthatthetoolwastoocomplexfortheirneedsandresources,andnoneofthemunicipalitiesusedthemethodinasystematicway.'Itistoocumbersomewithallthosesheetsofpaper.Thatwasabitdeterrent',representativesfromonemunicipalityexplained.Inoneoftheothermunicipalitiestheywereplanningtousethemethod,howevertheyalsoperceiveditastoocomplexandtimeconsumingcomparedtothesizeoftheirwelfaretechnologyproject.Usingacomplexmethodlikethatisprobablyfineinlargeprojectsinvolvinglargesumsofmoneyandmanyusers,theysuggested.However,theycouldnotjustifyusingthesamecomplexmethodinvolvingmanyemployeesfortheirlimiteduseoftechnologyforahandfulofusers.Onlyoneofthemunicipalitiesweinterviewedclaimedtousethemethod,however,theyhadalsoadjusteditandsimplifiedthetooltotheresourcesandcompetencesfoundwithinthemunicipality.Ontheothersideofthescalewefoundthatoneofthemunicipalitieshadmoreorlessgivenuptheirwelfaretechnologyproject,explainedbytheinsecurityitrepresentedforthemasasmallmunicipalitywithlimitedresources,toinvestintechnologythatwasstillnotfullydevelopedandadjustedforthemunicipalities'needs.Abetterstrategyforasmallmunicipalitywouldbetowait,theysuggested,andletthelargermunicipalitiesdotheworkwithadjustingandadaptationofwelfaretechnology.

5 DISCUSSION

Summarizingthefindings,wesawahighlevelofwill(orwish)tousetechnology,butalsoalackofknowledgeandalackofpracticalsupportinthepilotmunicipalities

resultinginlimitedfundingandpersonnelresources.Wesawleader-drivenprocesses,withalackoftechnologycompetenceresultinginhealthcareleadersdoingthingsthey'shouldn'treallyknow',furthercomplicatedbyITpersonnelpostedininter-municipalICTcompaniesratherthaninthemunicipalityorganisation.Theimportanceofgettingthesefactorsrighthavebeenpointedoutinotherprojects(Dugstadetal.,2015;DischandJohnsen,2015;Grutetal.,2013;DetMidtnorskevelferdsteknologiprosjektet,2014).Agoodprojectstructure,dedicatedtime,funding,acombinationofhealthcareandtechnologycompetenceetc.,areasimportantforlargeasforsmallmunicipalities.However,ourfindingssuggestthat,firstly,theeffectsofthesechallengesaresometimesescalatedinsmallmunicipalities–theextremeversionseeninonemunicipalitythatdecidedtoputonholdalltheiractivitywithwelfaretechnology.Secondly,findingsolutionsarealsocomplicatedbysize,likethechallengeofincludingpersonnelwithtechnologycompetenceinasmallmunicipalitythatreliesontheinter-municipalcollaborationratherthaninternalITpersonnel.Whileotherprojectshavereportedonchallengesofcommunicationbetweentechnologistsandhealthcarepersonnel(Dugstadetal.,2015)thechallengeinthiscasewasratheralackoftechnologiststocommunicatewithinthefirstplace.Eventhoughthemunicipalitiesoperatedinsmallscalesandexperienceswithtechnologywerelimitedintheseprojects,thenewexperiencesincreasedtheirknowledgeabouttechnology,andmoreimportantly,alsotheirattitudetowardstechnology.Themunicipalitiesexpressedanincreasedwilltousetechnologyandallpilotmunicipalitiesengagedindiscussionsofhowtofurtherdevelopthis.'Don'ttest,juststart',wastheadvicegivenbythehealthcareauthoritieswhentheyvisitedoneoftheregionalconferencesaboutwelfaretechnologyinSognandFjordane.Ourfindingsdohoweversuggestthatsmall-scaletestingandtryingouttechnologicalsolutions,includingtheorganisationalaspects,intheirownspeedandscope,wasimportantforthesmallmunicipalities.Eventhoughthetechnologyitselfisnotnewanymore,itisstillnotanaturalpartofthegeneralcompetenceinthemunicipality-basedhealthcaresector.Likefindingsinothermunicipalities(MoeandNilsen,2015)alsointhisregionwefoundstaffthatwerereluctantorscepticalofincludingtechnologyasapartofthehealthcaresolution.However,staffwiththeabilitytousethetechnologicalsolutionsthatarebeingintroducediscrucialtoachievebenefitsfromthetechnology.Thechallengesoftechnologycompetenceseemtobemorecriticalinsmallermunicipalitiesduetothereasonsmentionedabove,whereaccesstotechnologycompetenceappearstobeoneoftheweakestpointsforsomeofthesmallpilotmunicipalities.Thissuggeststheimportanceofsmall-scalepilotactivity,asthatprovidedthemwithapossibilitytoestablishexperienceandtobuildknowledgethattheyneedinthefurtherprocessofimplementingwelfaretechnology.


Otherchallengesrelatedtomunicipalitysizearefoundinthetechnology.Someofthetechnologicalsolutionsonthemarketrequireoperationsonalargerscalethanthesmallmunicipalitiesinthisregionneeded.Learningfromlargermunicipalities(Bjørkquist,2015)aswellasadvicefromthenationalwelfaretechnologyprogramme(Melting,2015;Melting,2017)hadnotpreparedthemfortheirsmall-scaleoperationbeingaproblem.Thesmallmunicipalitieswithlimitedresourcesalsoappearedtoreactnegativelytowhattheyperceivedastechnologicalsolutionsthatwerenotreadyforimplementation,inparticularwhen'authorised'throughadvicefromthenationalprogrammeforwelfaretechnology(Melting,2015;Melting,2017).Investingintechnologythatisnotthe'right'solutionrepresentsahighriskforthem.Arecurringquestionfromthehealthcaremanagerswas:'Howcanweknowthatthetechnologywechooseandspendourfinancialresourceson,arenotoutdatedinayearortwo?'Theadviceandtoolscomingoutofthenationalprogrammeareindeedimportantresourcesformunicipalitiesintheirworkwithwelfaretechnology.Manyofthechallengeswefoundappeartoreflectmunicipalitysize.However,thenationalprogrammehasnotfullyrecognisedthattheprerequisitesforintroducingwelfaretechnologyinthehealthcareservicesinsmallandruralmunicipalitiesmightbedifferentfromtheprerequisitesinlargercity-basedunits.MunicipalitiesinSognandFjordanearemotivatedtoworkwithwelfaretechnology,buttheyalsoneedtoscaleandorganisetheworktofitwithintheirownlimitedresourcesandabilities.

6 CONCLUSION

Ourstudyillustratesthatsizemattersforthemunicipalities'experiences,attitudesandpossibilitiesinthefirstphaseofimplementingwelfaretechnology.Onthenegativeside,themostcriticalfactorappearstobelackoftechnologycompetence,notonlyamonghealthcareworkers,butingeneralinthemunicipalities.ArecurringthemeintheNorwegiandiscourseisthecomplexityofwelfaretechnologyandtheneedtofocusonhumanfactorsratherthanthetechnologicalobject(NOU2011:11;Helsedirektoratet,2012;CorneliussenandDyb,2017).Thereisnodoubtthatimplementingwelfaretechnologyinvolvesmanyotheraspectsthanthephysicalartefactitself.Still,thepilotmunicipalitiesillustratethatthetechnologyitselfcreatesprerequisitesandchallengesthatcannotbemeasuredinpercentages.Althoughnotknowingwhichcabletoconnectisperhapsnotthemostintricatechallengetosolve,suchsmallissuescanalsopreventtheuseoftechnologyinpractice.Onthepositiveside,however,beingsmallisnotonlyadisadvantage.Thesmallmunicipalitieshavearatherlargedegreeofflexibilitywhentheycanworkattheirownspeedandwithinthescopethatisachievablewithintheirlimitedresources.

Althoughweshouldnotgeneralisefromtheinterviewsinthisproject,theyillustratethatmunicipalitiesaredifferentandhavedifferentaccesstoeconomicresources,aswellaspeoplewithnecessarymotivationandcompetences.Someofthesedifferencesarenottrivial,andsizeappearstobeonefactorthatneedstoberecognizedasmakingadifferenceinmunicipalities'effortstoimplementwelfaretechnology.Moreresearchisneededinthisfield,tolearnmoreaboutthesituation,andtoproduceadviceaboutimplementationofwelfaretechnologythatcanreflectthedifferencesbetweenmunicipalities.

7 ACKNOWLEDGMENTS

WewanttothankcollaboratorsintheprojectWelfareTechnologyintheMunicipalitiesofSognandFjordane.TheprojectwasfundedbytheCountyGovernorofS&Fj.

8 REFERENCES

[1] NOU2011:11,Innovasjoniomsorg.[2] Øderud,T.,Grut,L.andAketun,S.2015,Samspill–

GPSiOslo-PiloteringavTrygghetspakke3.[3] Melting,J.B.2017,Andregevinstrealiseringsrapport

medanbefalinger.Nasjonaltvelferdstekno-logiprogram,Helsedirektoratet.

[4] Melting,J.B.andFrantzen,L.2015,Førstegevinst-realiseringsrapportmedanbefalinger.Nasjonaltvelferdsteknologiprogram,Helsedirektoratet.

[5] Helsedirektoratet2012,Velferdsteknologi.Fagrap-portomimplementeringavvelferdsteknologiidekommunalehelse-ogomsorgstjenestene2013-2030.

[6] Ørjasæter,N.-O.andKistorp,K.M.2016,Velferds-teknologiiSentrum:InnføringavvelferdsteknologiisentrumsbydeleneiOslo-Enkartleggingaveffekten.

[7] Dugstad,J.,Nilsen,E.R.,Gullslett,M.K.,Eide,T.andEide,H.2015,Implementeringavvelferdsteknologiihelse-ogomsorgstjenester:opplæringsbehovogutformingavnyetjenester–ensluttrapport.

[8] Nordtug,B.,Aasan,H.M.andMyren,G.E.S.2015,Implementeringavvelferdsteknologi.Enkvalitativstudie:hvilkennytteoghvilkeutfordringererfareransatteikommunalhelsetjeneste?Senterforomsorgsforskning,rapportserie–nr.1/2015.

[9] Disch,P.G.andJohnsen,H.2015,PrioriteringogkompetanseibrukavvelferdsteknologiikommuneneiTelemark,VestfoldogBuskerud.Senterforomsorgs-forskningSør/HøgskoleniTelemark.

[10] Grut,L.,Reitan,J.,Hem,K.-G.,Ausen,D.,Bøthun,S.,Hagen,K.,Svagård,I.S.andVabø,M.2013,Veikartforinnovasjonavvelferdsteknologi.Rapport:SINTEF,NOVA&KS.


[11] DetMidtnorskevelferdsteknologiprosjektet2014,Sluttrapport.

[12] Silverstone,R.,Hirsch,E.andMorley,D.(1997(1992)).Informationandcommunicationtechnolo-giesandthemoraleconomyofthehousehold.InR.SilverstoneandE.Hirsch,eds.,ConsumingTech-nologies:MediaandInformationinDomesticSpaces.LondonandNewYork:Routledge.15-31.

[13] Bijker,W.E.,Huges,T.P.andPinch,T.,eds.,(1993(1987)).TheSocialConstructionofTechnologicalSystems.Cambridge,Mass.:TheMITPress.

[14] Charmaz,K.2006,Constructinggroundedtheory:Apracticalguidethroughqualitativeresearch.SagePublicationsLtd,London.

[15] Moe,C.E.andNilsen,G.S.2015,Velferdsteknologi,tilgledeellerbesvær?NorskkonferansefororganisasjonersbrukavIT.Vol.23.

[16] Bjørkquist,C.2015,ErfaringerfrapiloteringavvelferdsteknologiiSarpsborgkommune:Mobiltrygg-hetsalarm,digitalmedisindispenserogKOLS-moni-torering.HøgskoleniØstfold:Rapport2015:4.

[17] Corneliussen,H.G.andDyb,K.2017,Omteknologiensomikkefikkværeteknologi–diskurseromvelferdsteknologi.InJ.R.Andersen,E.Bjørhus-dal,J.G.NesseandT.Årethun,eds.,Immateriellkapital.Fjordantologien2017.Universitetsforlaget.165-181.


©TheAuthor(s).

MappingFHIRResourcestoOntologyforDDIreasoningRaeesAbbas1,IslamFathiHusseinAlKhaldi1,GetinetAyele1andJanPettersenNytun1

1DepartmentofInformationandCommunicationTechnology,UniversityofAgder,Grimstad,Norway

AbstractFastHealthcareInteroperabilityResources(FHIR)specificationsareusedtoexchangeclinicalandhealthrelatedinformationbetweendifferentsystems.Thereisunfinishedon-goingworktorepresentFHIRresourcesusingSemanticWeb technology to support semantic interoperability. This same technology would then also fitapplicationsdoingreasoning.WeutilizeandcustomizetheFHIRunofficialdraftontologyfordoingdrug-druginteractionsreasoning.Wegiveoneusecaseofsuchreasoningbasedonfamilyhistory;thiskindofreasoningmay extend the capabilities given by Forskrivnings- og ekspedisjonsstøtte (FEST) alone.We achieve this bysettingupaFHIRserverandmakingaFHIRclientthatstoredrugandpatientinformationtotheserver;wethenlaterretrievesomeofthisinformation,translateditintoWebOntologyLanguage(OWL)basedontology,dodrug-druginteraction(DDI)reasoningexposingpotentialhealthrisks.KeywordsFHIROntology,DDIreasoning,SemanticWeb,SPARQL

1 INTRODUCTIONElectronichealthrecords(EHR)capturehealthinformationabout patients and their medication; standards andprotocols allow such records to be sharedbetween different stakeholders such as hospitals, labs,pharmaceutical companies, etc.OneofthesestandardsistheFHIRspecifications,builtonHealth Level 7 (HL7) which is a widely accepted set ofprotocols and standards used to exchange clinical andhealth related information between different systems(Benson,2016).

SomeconsiderHL7V2andV3themostrelevantexchangestandards in healthcare (Oemig and Snelick, 2017). Ourpurpose istodoreasoningaboutDrug-DrugInteractions(DDI)basedonpatientinformationtogetherwithavailableinformationaboutdrugsutilizingFHIR.Thereasoningcanhelp doctors making decisions, and pharmaceuticalcompanies in manufacturing drugs. OWL (Bechhofer,2009)beingbasedondescriptionlogicsupportsreasoningandisourchosenontologylanguage.Manyoftheexistingmedical ontologies are expressed in OWL, e.g.,SystematizedNomenclatureofMedicine -ClinicalTerms(SNOMED CT) (Whetzel et al., 2011), this supports ourchoice.

Combining data from different sources may allowenhanced reasoning to derive useful information, e.g.,detect drug-drug interactions and potential healthproblems, which again enables early beneficialinterventions.Toachievethisvariousmedicalinformationsystems, implemented on different platforms, need toexchangehealthcaredatainastandardizedway.

Thispaperdescribesanimportantsteptowardsreachingourgoal.Inshort,wedemonstratehowaFHIRclientcanstoredrugandpatientinformationonaFHIRserver;howtoretrievethisinformationandtranslateditintoanOWLbasedontology.Howtoperformdrug-drugreasoningonthisontologyandtransfertheresultsofthereasoningbacktotheFHIRserver.Theontologyweuseforreasoning isthe FHIR unofficial draft ontology (Anthony, 2016); wecustomize and extend this ontology for our needs.Thecontributionofthispaperisthedemonstrationofhowsystemcomponents(EHR,Druginformation,ontologyforreasoning) may be tied together with help of the FHIRstandardtosupportDDI.

Rest of the paper is organized by different sections asfollows:SectionIIdescribesthebackground,andliteraturereviewofrelatedconceptsandwork.SectionIIIdescribesthe solution architecture and a detailed description ofsolutionmethodsandresults.InSectionIV,wediscusstheusabilityofthisworkandprovidesomeoptionsforfuturework.

2 MATERIALSANDMETHODS

2.1 BackgroundandStateoftheArtWhenapatienttakestwoormoredrugsatthesametime,DDI’s canoccur.This interactionmaycreateunexpectedside effects. The situation becomes evenmore complexwhenthesideeffectsoftwodrugstogetherresultsinnewsymptomsfalselyindicatingsomepotentiallynewdisease(leadingtotheintroductionofnewdrugs).Insomecases,themedicalhistoryofthefamilyofapatientmaycontaininformationthathelpsreasoning.Oneoftheapproachesfor representing the domain of Drug-Drug Interactions(DDI) is medical ontologies which provide a vital

The15thScandinavianConferenceonHealthInformaticsSHI2017,Kristiansand,29-30August2017 9

integration of knowledge base and drugs data. Medicalontologies are oftendefined inOWLwhereas reasonerssuch as Fact++ and HermiT (Glimm et al., 2014) can beused to infer Drug-Drug interactions. SNOMED CT is anexample of a medical ontology defined in OWL wheremedical concepts like "organism" and "substances" arerepresentedasclassesinahierarchicalstructure.Anotherclasslike"chemical"hasasubclass"Inor-ganicchemical".Propertiessuchas"hasactiveingredient"representingtherelationsandrulesbetweenindividualswhicharecreatedunder each class. Reasoners apply the rules against theknowledgebaseanddeterminetheactionthatshouldbecarriedoutincaseaDDIisinferred.Another form of reasoning on OWL ontologies may beperformedwithhelpof rule languages liketheSemanticWebRuleLanguage(SWRL)(Horrocksetal.,2004).OWLstatementsmaybeserializedbyusingtheResourceDescription Framework (RDF)where the statements arerepresentedastriplesformingRDFgraphs(W3C,2014).Another key technology of the Semantic Web is theSPARQL Protocol and RDF Query Language (SPARQL)(W3C, 2013), which allows complex queries to retrievedatafromanOWLontology.TheDrug-Drug InteractionsOntology (DINTO)defined inOWLdescribesandcategorizesDDIsandall thepossiblemechanismsthatcanleadtothem; itdemonstratesthatthecombinationofdrug-relatedfactsinDINTOandSWRLrules can be used to infer DDIs and their differentmechanismson a large scale (Herrero Zazo et al., 2014)(Herrero-Zazo et al., 2015).Yoshikawa et al. (2004) developed Drug InteractionOntology(DIO)withafocusofinferencingapossibledrug-drug interactionbasedon themoleculesof thedrugs. Itchecks all possible biomolecular interactions betweendrugsthatmayleadtounexpectedsideeffect(s).Itinfersnot only drug-drug interaction, but also checks theindividual differences in drug response or geneticsusceptibility of drugs, and optimization of balance ofefficacyandsafetyofnewdrugs.“Foundedin1987,HealthLevelSevenInternational(HL7)is a not-for-profit,ANSI-accredited standardsdevelopingorganization dedicated to providing a comprehensiveframework and related standards for the exchange,integration, sharing, and retrieval of electronic healthinformation that supports clinical practice and themanagement, delivery and evaluation of healthservices.”(HL7,2017).HL7hasaffiliatesinallcontinents.EuropeanHL7affiliatescomprises21countriesincludingNorway.FHIR(Benson,2016)isasetofHL7standardsintroducedbyHL7asDraftStandard for Trial Use (DSTU) which includes variousfeaturesfromdifferentversionsofHL7suchasHL7v2,HL7v3andCDA.FHIR resources are modular components that can becombinedtobuildasolutionformanyproblemsexistingin the real world of clinics and can be used to solveadministrative problems as well. According to FHIR

standards,resourcesarethecommonbuildingblocksforexchanging documents and information related tohealthcare information (Benson, 2016).Figure1showsanexampleofaFHIRresourceintheformof aUMLDiagram; theMedication resource is primarilyusedfortheidentificationanddefinitionofamedication(FHIR-HL7,2017).Asshowninthefigure1,itincludestheingredients and the packaging for a medication;communicating a medication would come as a bundlecontaining an instance of this pattern in one of thesupportedserializationformats(e.g.,JSON).

Figure1.FHIRMedicationresource(FHIR-HL7,2017)

The main challenge for healthcare standards is thecontinuous need to addmore fields and options to thespecificationwhich increases the costandcomplexityofimplementations.FHIRhasdefinedaframeworkinsuchaway that it becomeseasyenough toextend the currentresources and add custom definition (Kasthurirathne etal., 2015). As FHIR mainly focuses on implementation,therearemanylibrariesavailablefordevelopmentwithnorestrictionsonspecification(Kasthurirathneetal.,2015).There are also research efforts in the direction ofrepresentingFHIRdatatypesintermsofOWLconstructswhich enhances semantic interoperability of FHIRresources.One key changeofHL7 FHIRRelease3 is thedefinition of an RDF format, and how FHIR relates toLinkedData(Grieve,2017).An unofficial HL7 FHIR ontology draftwas presented on11thOctober2016bytheITSgroup(Anthony,2016).Thisunofficialdraftisaresultoftheongoingworkanditsefforttowards describing all the FHIR resources using OWLformat.TheyusedthesamenameofFHIRresources fortheirOWLclass.Contentthatdescribearesourceisgivenas OWL object properties. And, some of the content isrepresented usingOWL data properties. AnOWL objectproperty describes links going from one individual toanother,e.g.,thedescriptionofaspecificpatient(instanceof class Patient) linked by an object property calledprescribedtoaspecificdrug(instanceofclassDrug).OWLdata properties describe links from individuals to datavalues like text strings and numbers, e.g., representingname and year of birth.


There are several HAPI-FHIR FHIR RESTful serversavailable;HAPIFHIRisdevelopedbytheUniversityHealthNetwork(UHN)group.ItisaJavaimplementationofFHIRresources,and it isanopensourceRESTfulserverwhichgives opportunities for researchers and academicians touse it freely and provides java libraries for all resources(UHN, 2017).HAPI-FHIR test server has implemented all the FHIRresourcesandisbuiltfromnumerousmodulesoftheHAPIFHIRproject.Theserverhasuploadedsampledataforallresources. It is possible to read, edit, create, update,delete and validate these resources. The output of thereadoperationiseitherinJSONorXMLformats.RDFisstillnotdefinedforaccessingaRESTfulserver.There are several tools available forworkingwithOWL.Protégé is an open-source ontology editor used forcreating and editing OWL ontologies; it also supportsreasoning. Ithasaplug-inarchitecturewitharichsetofplug-ins,e.g.,SPARQL,SWRLandreasonerslikeFact++andHermiT(Musen,2015).JenaisacollectionofJavalibrariesused for creating Semantic web and Linked dataapplications(McBride,2002).Itprovidesaframeworkforinferencing,storage,querying(SPARQL)anditprovidesadatabase solution, called Fuseki (Apache, 2011), forhandlingontologies,e.g.,throughawebinterface.

3 SOLUTIONOVERVIEWFigure 2 shows the architecture of our solution. WedevelopedacustomizedFHIRontology,comprisedoftheresourceswhicharerelevanttoourdrug-drugreasoningexample. The part of our ontology concerning the FHIRresourcescoincideswithideaspresentedbyBeredimasetal. (2015); the part concerning DDI reasoning is partlyinspiredby(Herrero-Zazoetal.,2015)alltogetherwithourown contribution.

Figure2.Architectureoverview

In the ontology, we need to represent the informationgivenbyFHIRresources,andwedothis inaone-to-onestyle.WedonotrepresentalltheFHIRresourcesindetail,butusingasimilarapproach,wecanextendtheontology

for all the resources.Webuilt a FHIR client consisting of three Javamodules,whichcaneasilybemadeasthreeseparateapplications:•oneforinsertingpharmacologicaldataintotheFHIRserver.•oneforinsertingpatientdataintotheFHIRserver.•oneforretrievingpharmacologicalandpatientdata;mapping of these FHIR resource instances to thecorrespondingFHIRontologyclasses.

The server is locally hosted HAPI-FHIR server. We thenusedProtégéforreasoninganddisplayoftheresults; todothisintheJavaapplicationisalsostraightforward.Alsoa SPARQLendpoint, e.g., Fuseki, as shown in the figure,canbeusedforstoringtheontology.

3.1 FESTtoFHIRFEST is adatabaseused forprescription inNorway, andhave a huge collection of drug descriptions.Pharmaceutical data that we have used for therepresentation of drug-drug interactions is taken fromFEST (i.e., from the distributed FEST xml file). For thepurposeofdemonstratingadrug-drugreasoning,theFESTdataishandledmanuallybyidentifyingrelevantdrugsandimportingthemtotheFHIRserver(Legemiddelvek,2016).Adifferentproject,intheformofmasterthesiswithtitle“Ontological modelling of FEST with support for DDIreasoning”(Abbas,2017).

3.2 FHIRserverOf the FHIR resources, we mainly used Medication,Patient, Practitioner, DetectedIssues, andMedicationStatementre-sources.IntheFHIRserver,drugsare represented using the Medication resource. Weidentified 10 different medicines from FEST databasewhere some may have potential drug interaction witheachother,andcreatedFHIRresourcein-stances.Patientand Practitioner instances were created using thecorresponding FHIR resources. When it comes to theidentificationofamedicine,twodifferentcodingsystemswereused:AnatomicalTherapeuticChemical(ATC)Classi-ficationSystemandSNOMEDCT.

Figure3.SNOMEDCTandATCcodesexample

Aboveisanexampleofamedicine,whichisstoredinFHIRserver, identified by both using SNOMED CT andAnatomicalTherapeuticChemical(ATC)codes.


Infigure3above,systemvaluespecifiesthecodingsystemthat is being used and corresponding code displays theactual code that is used to describe the particularmedicine.Whiledisplayvaluedescribesthenameofthemedicine which is also specified by the coding systembeing used.DetectedIssue resource, forourusecase, isusedonlytorepresentdrug-drug interactions.Similartotheresourceinstances ofMedication, in the FHIR server, we storedresource instances of Practitioner, DetectedIssues,FamilyMemberHistory. In addition,MedicationStatementthatisusedtorecordthedrugsbeingusedbyapatient.

3.3 FromFHIRtoOWLPharmaceutical and patient data thatwas stored in thelocalFHIRserverwereretrievedandmappedtotheFHIRontology using JENA libraries. A part of the customizedFHIRontolo-gyhastheclassstructureshowninfigure4.

Figure4.UMLdiagramoftheFHIRresourceMedicationWechoosetheapproachtorepresenttheFHIRresources(ormorespecificallyFHIRresourcetypes)asclasses(e.g.,Figure 4 shows FHIR resource Medication in a UMLdiagram)usingsamenamingscheme,e.g.,Medication iscalledMedication.ThecontentsthatdescribeaparticularresourcearerepresentedasOWLobjectproperties.Someof the contents are also represented using OWL dataproperties. Figure 5 below shows some of the datapropertiesusedtodescribethedetailsoftheMedicationresource.

Figure 5. Data properties of the FHIR resource Medication Inthisway,wemappedthedetailsofalltheselectedre-sourcesfromFHIRservertoourontology.Figure6showsadetaileddescriptionoftheobjectpropertiesalongwiththeirpurpose.WhenitcomestothemappingofFHIRresourceinstances(e.g., description of an actual drug), we accessed theresourceinstancesfromtheFHIRserver,andaddedittothe ontology. E.g., a specific drug is then seen as anindividual of class Medication and inserted as such anindividualintotheontology.Figure7showstheexampleof mapping instance of Medication resource named“Efavirenz”.

Figure 6. Object properties

Figure 7. Medication instance mapping

Asthemedicationisidentifiedusingtwocodingschemes,DisplayvalueofcorrespondingdrugismappedtoMedica-


tion.Snomed.display property in the ontology and takesthe value as “Efavirenz 600 mg tablet“. Similarly, codevaluefromtheFHIRserver,incaseofSNOMED,ismappedto Medication.Snomed.code in the ontology. The samemappingprocedureisusedforATCcodemapping.

3.4 DDIReasoningSPARQLqueriesareusedtoretrieveandmanipulatedatafromtheontology,e.g.,wecancheckwhethertwodrugshaveaninteractionbythefollowingSPARQLquery:SELECT?XWHERE{?X:DetectedIssue.implicated:Fluvoxamine.?X :DetectedIssue.implicated :Rabeprazole. }

Figure 8. SPARQL query result

TheresultshowninFigure8abovecanalsobeverifiedbyrunning reasoner in the Protégé. Similarly, if we have acomplete database from the FHIR server, we can useSPARQLqueriesonthatexistingdatatogeneratespecificresults.Explicitaswellasimplicitresultscanbegeneratedusing SPARQL queries on the FHIR data, which was thepurposeofthispaper.Oneexampleofgeneratingimplicitresultsfromthedataispresentedbelow.We used the SPARQL CONSTRUCT query to generateimplicit triples about unknown DDI’s by examining theexisting data. For example, suppose that father of OleNordmann, having a heart disease, was prescribed acombination of drugs A and B; and his brother, havingsimilardisease,wasalsoprescribedsamecombinationofdrugs.Theirstatementhistoryexhibit thatbothof themsuffered fromsomesideeffects,possiblyaDDI.But thisDDIisnotexplicitlystatedinthedatabasenorthedrugsAandBweresupposedtoexhibitareaction.Ahypotheticaldiagnosecanbelinkedtosuchasituationthatgenesoftheconcerned persons may have a problem with thiscombinationofDrugs.Wecancheckthiswholescenarioby constructing a SPARQL query, and in response, OleNordmann can be prevented from taking the samecombinationasheispotentiallyinheritingthesamegenes.ApossiblewarningcanbegeneratedforthepractitionerresponsiblefortreatingOle.CONSTRUCT { ?X :hypotheticalDiagnose :UnknownDDI }

WHERE { ?X :Patient.hasFamilyMemberHistory ?a.

?a :FamilyMemberHistory.relationshipCodingCode "father";

:prescribed :DrugA;

:prescribed :DrugB;

:sideEffects :UnknownDDI.

?X :Patient.hasFamilyMemberHistory ?b;

?b :FamilyMemberHistory.relationshipCodingCode "brother";

:prescribed :DrugA;

:prescribed :DrugB;

:sideEffects :UnknownDDI.

}

Figure 9. CONSTRUCT query result

The implicit triplegeneratedbythequeryabove,canbeinsertedandstored intheontology.TheresultsaysthatPatient9952 individual, which is Ole Nordmann, maypossiblysufferfromthesameDDIashisfamilymembers,thus the practitioner should make further observationsregardingprescription.

4 DISCUSSIONANDCONCLUSIONSeveralclinicaltoolsprovidecomprehensivelistsofDDIs,often they lack the supporting scientific evidences anddifferenttoolsmaynotgiveconsistentresults(Tarietal.,2010).Oneofourgoalsistofindouttheimplicitdrug-druginteractions that are not explicitly stated or that areinconsistentinthecurrentlyavailabledrugcatalogs.Suchreasoningwould typically involve several ontologies anddatasources;inthiscontextstandardprotocolsneedstobe applied and ourwork success-fully demonstrate onewayofdoingthiswithhelpofFHIR.We have also demonstrated some beneficial DDIreasoningbasedonfamilyhistory.However,thereasoningpotentialseemshugeandshouldbefurtherinvestigated.Ourtaskmaybesimplified inthefuturesinceusingRDFwith the REST API is on the TODO list of HL7 (Anthony,2016).TheyhavealreadyspecifiedhoweachresourcecanberepresentedasasetofRDFtriplesrepresentedusingthe Turtle syntax. Our work fits well with this format.WhenFHIRserversgivetheRESTAPIsupportwecaneasilydirectly inserttheRDFGraphs intoourontologythatweuseforreasoning.

5 REFERENCES[1] Abbas, R. 2017,OntologicalModelling of FEST with

support for DDI Reasoning. MSc Information andcommunication technology, University of Agder,Norway.

[2] Anthony, M., Prud'hommeaux E. 2016, FHIR OWLOntology: Unofficial Draft [Online]. Available:http://w3c.github.io/hcls-fhir-rdf/spec/ontology.html[AccessedMay42017].

[3] Apache. 2011, Fuseki: serving RDF data over HTTP[Online]. Available:https://jena.apache.org/documentation/serving_data/[AccessedMay52017].

[4] Bechhofer, S. 2009, OWL: Web ontology language.EncyclopediaofDatabaseSystems.Springer.

[5] Benson,T.2016,PrinciplesofHealthInteroperability,Springer.


[6] Beredimas, N., Kilintzis, V., Chouvarda, I. &Maglaveras, N. 2015, A reusable ontology forprimitive and complex HL7 FHIR data types.EngineeringinMedicineandBiologySociety(EMBC),37th Annual International Conference of the IEEE,2015.IEEE,2547-2550.

[7] FHIR-HL7. 2017, Resource Medication - Content[Online]. Available:https://www.hl7.org/fhir/medication.html [AccessedMay52017].

[8] Glimm,B.,Horrocks,I.,Motik,B.,Stoilos,G.&Wang,Z. 2014, HermiT: an OWL 2 reasoner. Journal ofAutomatedReasoning,53,245-269.

[9] GRIEVE, G. 2017, FHIR Release 3 Posted [Online].Available: https://onfhir.hl7.org/2017/03/22/fhir-release-3-posted/[AccessedMay52017].

[10] Herrero Zazo, M., Hastings, J., Segura BedmarR, I.,Croset,S.,Martinez,P.&Steinbeck,C.2015-04-23.Anontology for drug-drug interactions. InternationalWorkshop on SemanticWebApplications and Toolsfor Life Sciences (SWAT4LS), 2013December, 11-122014Edinburgh,UK.CEUR,1-15.

[11] Herrero-Zazo, M., Segura-Bedmar, I., Hastings, J. &Martinez,P.2015,DINTO:usingOWLontologiesandSWRLrulestoinferdrug–druginteractionsandtheirmechanisms. Journal of chemical information andmodeling,55,1698-1707.

[12] HL7. 2017. About HL7 International [Online].Available: http://www.hl7.org/ [Accessed May 52017].

[13] Horrocks,I.,Patel-Schneider,P.F.,Boley,H.,Tabet,S.,GrosofF,B.&Dean,M.2004.SWRL:Asemanticwebrule language combining OWL and RuleML. W3CMembersubmission,21,79.

[14] Kasthurirathne,S.N.,Mamlin,B.,Kumara,H.,Grieve,G. & Biondich, P. 2015, Enabling betterinteroperabilityforhealthcare:lessonsindevelopinga standards based application programing interfacefor electronic medical record systems. Journal ofmedicalsystems,39,182.

[15] Legemiddelverk, S. 2016, Fest ImplementationGuidlines [Online]. Available:https://legemiddelverket.no/Documents/Andretemaer/FEST/Hvordan bruke FEST/ImplementationguideFESTv2.2.pdf[AccessedMay52017].

[16] Mcbride,B.2002,Jena:Asemanticwebtoolkit.IEEEInternetcomputing,6,55-59.

[17] Musen,M.A.2015,TheProtégéproject:Alookbackandalookforward.AImatters,1,4-12.

[18] Oemig, F. & Snelick, R. 2017, HealthcareInteroperability Standards Compliance Handbook:Conformance and Testing of Healthcare DataExchangeStandards,Springer.

[19] Tari, L.,Anwar, S., Liang, S.,Cai, J.&Baral,C.2010,Discoveringdrug–druginteractions:atext-miningandreasoning approach based on properties of drugmetabolism.Bioinformatics,26,i547-i553.

[20] UHN.2017,HAPI-FHIR.FHIRmadesimple. [Online].Available: http://jamesagnew.github.io/hapi-fhir/[AccessedMay52017].

[21] W3C.2013,SPARQLquerylanguageforRDF[Online].Available: https://www.w3.org/TR/rdf-sparql-query/[AccessedMay42017].

[22] W3C. 2014, RDF 1.1 concepts and abstract syntax[Online]. Available: https://www.w3.org/TR/rdf11-concepts/[AccessedMay52017].

[23] Whetzel,P.L.,Noy,N.F.,Shah,N.H.,Alexander,P.R.,Nyulas, C., Tudorache, T. & Musen, M. A. 2011,BioPortal: enhanced functionality via new Webservices from the National Center for BiomedicalOntology to access and use ontologies in softwareapplications.Nucleicacidsresearch,39,W541-W545.

[24] Yoshikawa, S., Satou, K. & Konagaya, A. 2004, Druginteractionontology (DIO) for inferencesofpossibledrug-druginteractions.Medinfo,11,454-458.

6 ADDRESSFORCORRESPONDENCEJanPettersenNytun,Departmentof ICT,UniversityofAgder,Norway,[email protected],address:Postboks509,4898Grimstad,Norway


©TheAuthor(s).

EDMON-AWirelessCommunicationPlatformforaReal-TimeInfectiousDiseaseOutbreakDetectionSystemUsingSelf-RecordedDatafromPeoplewithType1

Diabetes

AshenafiZebeneWoldaregay1,EirikÅrsand2,AlainGiordanengo1,2,DavidAlbers3,LenaMamykina3,TaxiarchisBotsis1andGunnarHartvigsen1,2

1DepartmentofComputerScience,UniversityofTromsø–TheArcticUniversityofNorway,Tromsø,Norway,[email protected]

2NorwegianCentreforE-healthResearch,UniversityHospitalofNorthNorway,Tromsø,Norway3ColumbiaUniversity,N.Y.,USA

AbstractTherelationbetweenan infection incidentandelevatedbloodglucose(BG) levelshasbeenknownfor longtime.PeoplewithdiabetesoftenexperiencevariableepisodesofelevatedBGlevelsuponinfectionsincident.Hence,weproposedanElectronicDiseaseSurveillanceMonitoringNetwork(EDMON)thatusesBGpatternandotherrelevantparameterstodetectinfecteddiabetesindividualsduringtheincubationperiod.Theprojectisanextensionoftheresultsachievedinthemobilediabetes(mDiabetes)fieldwithinourresearchteamforthelast15years.TheproposedEDMONsystemisakindofpublichealthsurveillance,whichuseseventsanalysisat individual levels (called micro events) to reach on a conclusion for uncovering events on the generalpopulations(calledmacroevents)basedonspatio-temporalclusterdetection.Itincorporatesself-managementmobileapps,sensors,wearables,andpointofcare(POC)devicestocollectreal-timehealthinformationfromindividualswithType1diabetes.Inthispaper,wewillpresenttheproposedEDMONsystemarchitecturealongwiththedesignrequirements,systemcomponents,communicationprotocolsandchallengesinvolvedherein.KeywordsType1Diabetes,WirelessCommunication,BGPatternDetection,InfectionDetection.

1 INTRODUCTIONDiabetesmellitusisachronicmetabolicdisorder,whichismostlycausedbyeitherfailureofpancreasβ-betacellstoproduceinsulinsecretion(Type1)orlackofbodyresponsetoinsulinaction(Type2)(IDF,2015). Accordingtorecentreports, there are approximately 450million adults withdiabetesworldwideandisprojectedtoraiseto642millionby2040(IDF,2015).Currently,thereisnocurefordiabetes;however, it can be prevented from creating furthercomplication with one’s proper self-management of thedisease. The advent of information and communicationtechnology (ICT) has much revolutionized self-management and made treatment of the disease a loteasier than before, which is mostly connected with theintroduction of mobile apps (mHealth), wearables andsensors, and POC devices that can provide individuallytailoredinformationforabetterinformeddecisionmaking(patient empowerment) (Walseth et al., 2005; Li et al.,2017; Issom et al., 2015; Botsis et al., 2009). Theseadvancementsinturnhavealsocreatedhugeaccumulationof the individual patient health data gathered on a dailybasis,which creates opportunities for further analysis ofthesedatatocapturerelevantinformationforbetterself-management and treatments (Béranger et al., 2016;Mohammadi, 2015). The introduction of big data, datamining and advanced analytic concepts have made thedetection of aberrant pattern, an outbreak signal, awaymorerelevantandeasierthanbefore(Vayenaetal.,2015).Inthisregard,useofpatientself-gathereddataforpublichealth surveillance purpose has now become more

apparent than ever; given the ubiquitous nature andwidespreaduseofmHealth apps,wearables and sensorsforself-managementpurpose(Walsethetal.,2005;Issometal.,2015).Likewise,theintroductionofcloudcomputingtechnologiessince the millennium has brought a significantimprovement in various healthcare delivery settings, ofwhichpublichealthsurveillanceisnotanexception(Swanetal.,2013;Council2017).Currently, most of the existing electronic diseasesurveillance systems rely on data sources (surveillanceindicatorsandevents) thatspanfromthe incidentof thefirst symptoms till physicians or laboratory confirmation,suchaswebbased(i.e.googlesearchengine)(Choietal.,2016),Over-the-counter(OTC)pharmacydrugssell(Pivetteetal.,2014),schoolabsenteeism(Lawpoolsri,etal.,2014)or work absenteeism (Paterson, Caddis, and Durrheim,2011),andothersleavingtheincubationperiodoutoftheirsystems.AccordingtotheCentresforDiseaseControlandPreventionreport(Holtetal.,2013)onindictorsforchronicdisease surveillance, more than 20 individual diabetesindicatormeasuresaregivenwhiletheuseofBGpatternsas surveillance indicators arenot indicated,which showsthecomplexityanduniquenessoftheproposedapproach.Botsisetal.(Botsisetal.,2012)presentedthemostnotableproofofconceptstudythatempiricallysupportstheuseofblood glucose pattern of diabetes individuals assurveillance event indicator. Moreover, blood glucosepattern has also been described as event indicators forsurveillancepurposeinotherrelatedliteratures(Botsisetal.,2012;Botsisetal.,2010;Lauritzenetal.,2011).


TheproposedEDMONsystemisareal-timeearlydiseaseoutbreakdetectionsystemthatusesself-recordedhealthdatafrompeoplewithType1diabetes.Itisakindofpublichealth surveillance, which uses micro events analysis(detecting infection induced elevated BG pattern on anindividual level) to reach on a conclusion foruncoveringmacroeventsonthegeneralpopulationsbasedonspatio-temporal cluster detection. The system will analyse theindividual’s blood glucose levels in real time, an onlinecontext,tolookforaberrantpatterns;variableepisodeofelevated blood glucose levels as a result of metabolicinstability due to infection incident (Woldaregay et al.,2016;Årsandetal.,2005;Rayfieldetal.,1982).Therefore,thesurveillancecasedefinitionencompassestheinfectioninduceddeviatedpatternoftheindividual’sBGdynamics.Inaddition,patternsofothersupportingparameterssuchas insulin injections, physical activity, and dietaryinformationalongwithphysiologicalparameterslikebodytemperature, blood pressure, and others will also beincluded. Of course, it is not only infections that couldcause variable episode of elevated blood glucose levels,and factors such as stress could also result in somehowsimilar pattern.As a result, theplan is to incorporate allcontributing variables known to the patients in thesurveillancecasedefinitionandanalysissoastosuppresstheeffectsof theseconfrontingvariables.Moreover, thespatio-temporal nature of the EDMON’s system issupposed to alleviate these challenges; given that theprobability of having sufficient number of people to bestressedataspecific locationandspecifictimeinterval isprobably low to trigger the necessary threshold ascompared to an infection incident. This characteristic ishighly dependent on the contagious nature and itsprogressive prevalence of an infection after any initialincident. EDMON will use techniques from big dataanalytics, social media, mobile computing and a novelhealthmonitoringsystems.Ifsuccessful,EDMONwillpavethe way for the next generation disease surveillanceapproaches. In this paper, wewill present the proposedEDMON’s system architecture along with the designrequirements, system components, communicationprotocolsandthechallengeinvolvedherein.

2 BACKGROUNDANDRELATEDWORKS

2.1 WirelesscommunicationplatformsCurrently, the rapid development of informationcommunicationandtechnology(ICT)andInternetofThings(IoT) have created opportunities for a quantified-self,whichaimstoempowerpatient’sdecisionmakingbasedondocumentingtheirownhealthcondition.This inturnhascreated a rapid pace on the integration, communicationanduseofwearables,sensors,POCdevicesandotherbodyareanetworkforphysiologicalmonitoringandotherhealthrelatedpurposes(Swan,2013;Béranger,2016).Diabetesisnot an exception in this case, experiencing a rapidadvancement in its field. In this regard, differentcommunication systems, protocols and standards forvarious purpose such as intelligent diabetes monitoring,remote diabetes surveillance, remote diabetes

management, tele-management and tele-monitoring,follow-up systems, data analysis, personalized andcustomized feedback and decision making haveincreasinglybeenstudiedandpresentedintheliteratures,e.g.(Huzooree,Khedo,andJoonas, 2017; Liao et al., 2004; Mougiakakou et al., 2010; Mougiakakou et al., 2005; Martinez et al., 2011; Al-Taee et al., 2015; Chang et al., 2016), but none of the recent studies have considereddetectionof infection incidents indiabetespeopleastheunderlyingpurposes.For example, Huzooree et al. (Huzooree, Khedo, andJoonas, 2017)developedacommunicationplatformforawireless body area network for remote diabetes patientmonitoring and analysis. The system integratesphysiological data from the body area network into astandalone mobile app, which sends these data into aremote server for further analysis and monitoring(Huzooree, Khedo, and Joonas, 2017). Moreover,Mougiakakouetal.(Mougiakakouetal.,2010)developeda communication platform for an intelligent remotediabetes monitoring, management, follow up andtreatmentsforType1diabetespatients.Thesystemusesstateofthearttechnologiesandstandardsandconsistedoftwounits;apatientunitandpatientmanagementunit(Mougiakakou et al., 2010). Mougiakakou et al.(Mougiakakouetal.,2005)alsodevelopedatelemedicinesystem that provides tele-monitoring and tele-management services for type 1 diabetes individuals.Besides, Liao et al. (Liao et al., 2004) developed acommunicationplatformforremotediabetessurveillance,wherethediabetesindividualismonitoredfromhomebyremote healthcare givers. The system promotes theindividual with diabetes to measure and update his/herstatusathome,whichiscommunicatedtotheirhealthcaregivers. Inaddition,Martinezetal. (Martinezetal.,2011)developedasystemthatprovidesaremotemonitoringoftheindividualdiabetespatient’smetabolicprofilesthroughApplication Hosting Device (AHD), which manages thesensor platform and allows sending IHE-PDC messagescompliant with Continua Health Alliance at aWAN levelalongwithamobileapplication.Thesystemincorporatesthree modules;1073 adaptation, Data Access API andSensorManagementmodule.Thepatientisabletoregisterphysical activity, food intake (menuandCHOquantities),blood pressure, weight and glycaemia measurementmanually,medicationintake(i.e.insulinandotherdrugs),andspecialevents(i.e.stressatwork,holidayandbirthdayparty), which are integrated into a diary application(Martinezetal.,2011).Furthermore,Al-Taeeetal.(Al-Taeeet al., 2015) presented a platform to support self-managementthroughremotecollectionandmonitoringofself-gathereddataandprovisionofpersonalizedfeedbackonthesmartphonebasedonInternetofThings(IoT).Basedonthecurrentandhistoricalself-gathereddata,thesystemenablesreal-timeclinicalinteractionandtailoredfeedbackto the individual needs (Al-Taee et al., 2015). Likewise,Chang et al. (Chang et al., 2016) developed a contextaware, interactive cloud basedmHealth system that canprovide a real time, two way communication between

15thScandinavianConferenceonHealthInformaticsSHI2017,Kristiansand,Norway,29-30August.

16

diabetes patients and caregivers by using Internet ofThingstechnology.

2.2 StateoftheartForthepast15years,ourresearchteamhasbeenworkingon the patient unit and created and developed theDiabetesDiary,whichisnowavailableinbothgoogleplay(Android)andappstore(Apple)(Nse,2017).Currently,ourteamisworkingtowardsatailoredversionofthediabetesdiary with more data integration, patterns analysis andmonitoring options. The tailored version will includemeasurementslikebloodpressure,heartrate,bodyweightandtemperatureinadditiontobloodglucose,carbsintake,physical activity, and insulin injection. Moreover, afeasibilitystudytowardstheuseofPOCdeviceshasbeenconducted(Botsisetal.,2009).Thestudyconcludedthatdeviceslikewhitebloodcellcount(XBCanalyser)seemstobe problematic due to usability issues and the coast isregarded as the main bottleneck (Botsis et al., 2009).Therefore,theplanistorequestmeasurementsfromthesePOCdevicesonlywhenitisnecessaryandappropriate.

There have been some research activities regarding theinfectiondetectionsystemusingbloodglucoselevelsasapotential indicator.Forexample,Årsandetal. (Årsandetal., 2005) presented an approach for developing anepidemic disease detection using blood glucose (EDDG)systembasedonbloodglucosemeasurements.Thepaperdescribesthesystemcomponentsincludingthenecessaryequipment,datastructuresanddatarepositoryalongwiththe proposed detection mechanisms. Furthermore, anumber of studies regarding the outbreak detectioncomputingalgorithmhavebeenconducted.Forexample,Woldaregay et al. (Woldaregay et al., 2016) havedevelopedan infectiondetectionalgorithmbasedonthecontinuousglucosemonitoring(CGM)readings.However,thestudyhasconsideredonlybloodglucosepatternsastheinput to the system.Moreover, other similar studies like(Granberg et al., 2007) have tried to detect infectioninduced blood glucose deviation. Even though thesestudieshaveshowntheproofofconcept,theyhavecertainlimitations, i.e. the number of input parameters, realinfectionBGdataandsamplesize.Therefore,theplanistoincludemore input parameters, a real infection BG dataandlargersamplesizetodevelopamorerobustapproachforthecomputingalgorithm.

3 EDMONDESIGNREQUIREMENTSDiabetes self-management mobile applications (mHealthapps),sensorsandwearables,includingbothinvasiveandnon-invasive, and other POC devices should collect thepatient’sbloodglucoselevel,insulintherapy,dietaryintake(carbs), physical activity, and physiological informationsuchasbodytemperatureandbloodpressure.Someotherideal and optional physiological parameters like whitebloodcellcount,CRPtest,heartrate,respiration,oxygensaturation, and stress level should be recorded and sentupon request from EDMON system. The measuredparametersshouldbeintegratedintoastandalonemobileapp,i.e.personalhealthrecordapplication,whichactsasa

gatewayforthedatatobetransferredtoaprivatecloud(remote server) in a real-time scenario. Therefore, dataqualityisthedeterminantfactorforsuccessfulprocessing,computation and interpretation of those health data(Huzooree,Khedo,and Joonas, 2017).Asa result, all therecordedkeydiabetesandphysiologicalparametersshouldbetransferredsecurelyandappropriatelythrougheitheramobile infrastructureoraprivatenetworkandshouldbesafely stored in the cloud (remote server) in a real-timeenvironment. Any possible failure thatmay arise due tonetwork coverage, sensors andwearables failure, lackofsignal strength, transmission reliability, and delay, couldlead to an unpredictable effect on the accuracy of thedetectionsystem,andalsoonthequalityandreliabilityofpatient tracking (Huzooree, Khedo, and Joonas, 2017;Sachidanandaetal.,2010).Hence,ensuringthequalityofinformation (QoI) attributes suchasaccuracy, timeliness,completeness, relevancy,and reliability (Sachidanandaetal.,2010;Zahedietal.,2008),alongwithsystemusability(ease of use) are key design requirements for theacceptanceoftheproposedEDMONsystem.

4 EDMONARCHITECTUREThe EDMON architecture consists of a patient unit,computing unit, and end users, as shown in Figure 1(Mougiakakouetal.,2010).Thepatientunitisresponsiblefor collecting the necessary parameters into the user’ssmartphone.Thecomputingunitwillanalysetheincomingdataforaberrantpatternsontheindividualaswellasonthe cluster level. The end users (desktop, laptop orsmartphone version of EDMON’s application) could bephysicians, patients, family and relatives, or the generalpublicoranyconcernedhospitalsorpublichealthauthoritythatshouldhaveaccesstotheoutbreakinformationfromthesystem.

4.1 CommunicationarchitectureandprotocolsEDMON is a three-tier architecture that incorporates;sensorandwearable tier,mobile computing tier, remoteserver (cloud) tier, as shown in Figure 2. This kind ofarchitecturemightbepronetoadegradeddataaccuracyduetoremotesitecomputationsasaresultoftransmissionand other errors. However, we prefer to minimize thepowerconsumption,andsavethememorystorageissuesincurredintheparticipants’smartphones.

Figure1.EDMONarchitecture.

In EDMON, the invasive and non-invasive sensors andwearables,andPOCdevices,recordthedataautomatically


andsendthesereadingstothesmartphoneapp(DiabetesDiary)usingexistingcommunicationprotocolsthatensuresecurity,robustnessandprivacy,i.e.BluetoothandZigBee(Huzooree,Khedo,andJoonas, 2017).Insomecases,whenthereisnosuchautomaticfacilitytheusermightbeaskedtorecordthedatamanually.Thesmartphoneappactsasagatewaynode,whichintegratesthedatafromthesensorsandwearablesnodesandforwardsittotheaccesspoint.Securecommunicationprotocolssuchas IEEE802.11/Wi-Fi/GPRScouldbeusedasacommunicationmediumwiththeaccesspoint(Huzooree,Khedo,andJoonas, 2017;RafeandHajvali,2014).Thecommunicationbetweentheaccesspointandtheremote(server)couldbeimplementedviaaProtectedNetwork,connectedviaanindependentsecureIP-network,i.e.theNorwegianHealthNetwork,toenablesecure electronic communication between the accesspointsandtheremotecomputingcentre.

Figure2.ThethreetireofEDMONArchitecture.

4.2 SensorandwearabletierThefirsttierisasensorandwearabletierthatincorporatesself-management apps, POC devices, wearables andsensors for collecting the diabetes key parameters andothernecessaryphysiologicalparametersoftheindividualdiabetes patients, as shown in Figure 2, where the dataentry is either automatic ormanual. The automatic dataentrywilluse thedeviceAPI, i.e.Bluetooth,whereas themanual data entry requires users to manually input likedietary information, i.e. menu and amount of CHOquantities and others. This node includes circadian cyclemeasurements of the individual’s physiologicalparameters, diabetes key parameters and other point ofcare test devices. The diabetes key parameters includedietary intake, insulin injections, CGM readings, physicalactivity and other measurements. The physiologicalparameters group includes body temperature, bloodpressure, oxygen saturation, respiration rate and stresslevelmeasurements.ThePOCmeasurementsincorporateswhite blood cell count, CRP test and other necessaryquantities. However, the frequency of these readings,physiological parameters and POC measurements, aredeterminedbytheEDMONsystemuponnecessityexceptthekeydiabetesparameters,whicharethedefault inputtothesystem.

Figure3.ComponentsofEDMONdatacollector’snode.

4.3 Mobilecomputingtier(GatewayNode)The second tier is a mobile computing tier, which is astandalonemobileappthat integrates thereadingof theindividual’skeydiabetesandphysiologicalparameters.Themobileappactsasasinkformeasurementsthatcomefromthe diabetes individual’s sensors, wearables and POCdevices.ItisbuiltonthetopoftheexistingdiabetesmobileApp-Diabetes Diary, which is developed by Norwegianscientists at theNorwegianCentre for E-healthResearch(NSE),as shown inFigure4 (bothEnglishandNorwegianversion).Thistirealsoactsasagatewaythatforwardstherecordedparameterstotheaccesspoint,asshowinFigure2. Currently, the tailored version of the diabetes diarysupports measurements like blood glucose, insulin,physical activity, carbohydrate, calories, weight andmedications(Årsandetal.,2016).Therefore,theplanistoaddmoremonitoringoptionincludingmorephysiologicalparameters to enhance the accuracy of the infectiondetectioncapabilitybasedontheself-recordeddata.

Figure4.DiabetesDiary-TailoredVersion.

4.4 Remoteserver(Cloudcomputing)tierThe third tier is a remote server (cloud computing) tier,which incorporates a repository and computing centresthatwillcarryoutthedatastorageandcomputationtasksrespectively. This tier also fetches and provide theoutbreakdetectionresultsfortheresponsiblebodies,i.e.,public health officials, physicians, patients, relatives andgeneralpublicaudiences.

15thScandinavianConferenceonHealthInformaticsSHI2017,Kristiansand,Norway,29-30August.

18

DataRepositoryTherepositoryisusedtostoretheincomingdatafromtheparticipatingType1diabetes individuals. Theuser’sdataarestoredinadatastructurestampedwithauserID,timeandgeographicallocationcontainingthekeydiabetesandphysiologicalparameters(Årsandetal.,2005).Thesizeofthedatarepositorywilldependonthesizeoftheareathesystemcovers.

ComputationalServiceThecomputationalserveristhemostcrucialpart(heart)ofthe EDMON system. The server carries out an intelligentdata mining and advanced data analytics concepts touncover both the micro and macro events. It performsvarious computation at individual and cluster levelsincluding BG profiling, analysing, reporting anddisseminatinginformation.BGprofilinginvolvesmodellinga personalized health model, which keeps track of theindividual’s blood glucose dynamics and predicts theupcomingbloodglucosevaluesdependingonasetofinputparameters, such as previous BG values, dietary intake,amountofphysicalactivity,amountofinsulininjection,andothers. The analysis will carry out a comparison of theindividualpredictedBGandactualBGvaluessotolookforany statistically significant aberrant patterns. Moreover,theaggregationalanalysiswilllookforamaximumnumberof micro events based on spatio-temporal clusterdetection. The reporting and dissemination part of thecomputationrespectivelywillorganizetheinformationinauser-friendly format (tables, graphs, and maps) anddistribute this information, such as the spatial andtemporaldistributionofthediseaseoutbreakonamapofthe region, the degree of severity and others, to theaudienceviaEDMONwebpageorapplication.

5 DISCUSSIONAdvancedsystemsandfunctionalitieslikeEDMONcouldbea breakthrough in digital disease detection (DDD), andpublichealthsurveillanceandmightalsohaveasignificantimprovementfordiabetesself-management.Theproposedwirelesscommunicationplatformwillusethestateoftheartcommunicationstandardsandprotocols,databaseandserver technologies. However, given the sensitivity ofhealth data there are challenges that need specialattentions such as user privacy/security, quality ofinformation and standardization issues, geographicallocation estimation and user mobility, and useracceptance.Healthrelatedpersonaldataareverysensitiveandneedtobe treated confidentially throughout the system’s dataflow.Inthisregard,inadditiontotherecommendedthreetier architecture, it is necessary to look for robustapproachestoensuretheuserprivacyandsecurityduringdatacollectionandtransmissionasthisishighlycriticalforsuccessfuldesignandacceptanceoftheproposedEDMONsystem.Forexample,privacypreservingmechanismssuchasde-identification (Office forCivil Rigths, 2012;Uzuner,Luo,andSzolovits,2007),which involvesremovalofuserdirect identifier, could be one possible options. Foraccurate data analysis, quality of data is the most

determinantfactorsincecorrupted,heterogeneous(duetomultiplesensors,wearablesandPOCdevices),missing,anddelayed data could result in unpredictable performancedegradation. Therefore, it is necessary to look for anadvanced data quality control and pre-processingalgorithm, which might pre-process the incurredheterogeneity, and check and request a retransmissionupon corruption, delay, and missing data (Huzooree,Khedo, and Joonas, 2017). User acceptance is also animportant factor that should be considered and tackledsincepeoplemightnotbewillingtoadoptanewsystemformultiplereasonssuchaslackoftrust,lackofmotivation,ifthe system hinders mobility, and lack of perceivedusefulnessandeaseofuse(Huzooree,Khedo,andJoonas, 2017).Therefore,itisnecessarytolookforapproachestobuy users trust and enhance their motivation andperception.Usermobility also could create challenges intermsofgeographicallocationestimationandtransmissionpower. However, a real time geographical locationestimationtechniquesrelyingonthesignalsentfromtheuser through GPS or Wi-Fi positioning data and energyaware communication protocols could be an option(Niewiadomska-Szynkiewicz, 2013; Gautam and Gautam,2009).

6 CONCLUSIONEDMON is a real-time early disease outbreak detectionsystem that uses self-recorded health data from peoplewithtype1diabetes.ItmainlyexploitsthepresenceofanelevatedBGlevelsuponinfectionincidents.Therefore,thesurveillance case definition will be formulated entirelybased the individual’s patternofBGdynamics.However,patterns of other supporting parameters such as insulin,physical activity, and diet along with physiologicalparameters like body temperature, blood pressure andotherswillalsobeincluded.Ifsuccessful,EDMONwillpavethe way for the next generation disease surveillanceapproaches. We presented the proposed EDMONarchitectures along with its persistent challenges thatneedstobesolved.Webelievesuchkindofsystemmightbenefit other similar systems, i.e. diabetes patientmonitoring, decision support and other patientempowerment system, and most importantly provokefurther thought in the challenging field of real timeelectronicdiseasesurveillancesystems.

REFERENCES[1] (Idf),I.D.F.2015,IDFdiabetesatlas-7thedition

[Online]. Available:http://www.diabetesatlas.org/.

[2] Walseth,O.A.,Årsand,E.,Sund,T.&Skipenes,E.2005, Wireless Transfer of Sensor Data intoElectronic Health Records. Connecting MedicalInformaticsandBio-Informatics.IOSPress.

[3] Li, X., Dunn, J., Salins, D., Zhou, G., Zhou, W.,Schussler-Fiorenza Rose, S. M., Perelman, D.,Colbert,E.,Runge,R.,Rego,S.,Sonecha,R.,Datta,S., Mclaughlin, T. & Snyder, M. P. 2017 DigitalHealth: Tracking Physiomes and Activity Using


Wearable Biosensors Reveals Useful Health-RelatedInformation.PLoSBiol,15,e2001402.

[4] Issom, D.-Z., Woldaregay, A. Z., Chomutare, T.,Bradway, M., Årsand, E. & Hartvigsen, G. 2015,Mobile applications for people with diabetespublished between 2010 and 2015. DiabetesManagement,5,539-550.

[5] Botsis,T.,Walderhaug,S.,Dias,A.,VanVuurden,K., Bellika, J.G.&Hartvigsen,G. 2009, Point-of-caredevicesforhealthyconsumers-afeasibilitystudy.JTelemedTelecare,15,419-420.

[6] Béranger, J. 2016, Introduction. Big Data andEthics.Elsevier,xi-xxxvi.

[7] Mohammadi, D. 2015, Turning big data intopersonaliseddiabetescare.TheLancetDiabetes&Endocrinology,3,935-936.

[8] Vayena, E., Salathe, M., Madoff, L. C. &Brownstein, J. S. 2015, Ethical challenges of bigdata in public health. PLoS Comput Biol, 11,e1003904.

[9] Swan,M.2013,TheQuantifiedSelf:FundamentalDisruption in Big Data Science and BiologicalDiscovery.BigData,1,85-99.

[10] Council,C.S.C.2017,ImpactofCloudComputingonHealthcareVersion2.0.

[11] Choi, J.,Cho,Y.,Shim,E.&Woo,H.2016,Web-basedinfectiousdiseasesurveillancesystemsandpublic health perspectives: a systematic review.BMCPublicHealth,16,1238.

[12] Pivette,M.,Mueller,J.E.,Crepey,P.&Bar-Hen,A.2014, Drug sales data analysis for outbreakdetection of infectious diseases: a systematicliteraturereview.BMCInfectDis,14,604.

[13] Lawpoolsri, S.,Khamsiriwatchara,A., Liulark,W.,Taweeseneepitch, K., Sangvichean, A.,Thongprarong, W., Kaewkungwal, J. &Singhasivanon, P. 2014, Real-timemonitoring ofschool absenteeism to enhance diseasesurveillance: apilot studyof amobile electronicreportingsystem.JMIRmHealthuHealth,2,e22.

[14] Paterson,B.,Caddis,R.&Durrheim,D.2011,Useof workplace absenteeism surveillance data foroutbreak detection. Emerg Infect Dis, 17, 1963-1964.

[15] Holt,J.B.,Huston,S.L.,Heidari,K.,Schwartz,R.,Gollmar,C.W.,Tran,A.,Bryan,L.,Liu,Y.&Croft,J. B. 2015, Indicators for Chronic DiseaseSurveillance—UnitedStates,2013.CDC.

[16] Botsis, T., Lai, A. M., Palmas, W., Starren, J. B.,Hartvigsen, G. & Hripcsak, G. 2012, Proof ofconcept for the role of glycemic control in theearly detection of infections in diabetics.HealthInformaticsJournal,18,26-35.

[17] T.Botsis, O.Hejlesen, J.G.Bellika & G.Hartvigsen2007,Bloodglucoselevelsasanindicatorfortheearly detection of infections in type-1 diabetics.AdvancesinDiseaseSurveillance.

[18] Botsis, T. & Hartvigsen, G. 2010, Exploring newdirectionsindiseasesurveillanceforpeoplewith

diabetes: lessons learned and future plans.StudHealthTechnolInform,160,466-470.

[19] Lauritzen, J. N., Arsand, E., Van Vuurden, K.,Bellika, J. G., Hejlesen, O. K. & Hartvig-Sen, G.2011, Towards a mobile solution for predictingillness inType1DiabetesMellitus:DevelopmentofapredictionmodelfordetectingriskofillnessinType1Diabetespriortosymptomonset.1-5.

[20] 20. Woldaregay, A. Z., Van Vuurden, K.,Årsand, E., Botsis, T.&Hartvigsen, G. ElectronicDiseaseSurveillanceSystemBasedonInputfromPeople with Diabetes: An Early OutbreakDetection Mechanism. Proceedings from The14th Scandinavian Conference on HealthInformatics2016,Gothenburg,Sweden,April6-72016,2016.LinköpingUniversityElectronicPress,23-27.

[21] 21. Årsand, E., Walseth, O., Andersson, N.,Fernando, R., Granberg, O., Bellika, J. &Hartvigsen,G.2005,Usingbloodglucosedataasan indicator for epidemic disease outbreaks.StudiesinHealthTechnologyandInformatics,116,217-222.

[22] 22. Rayfield, E. J.,Ault,M. J., Keusch,G.T.,Brothers,M. J.,Neche

Date post:	08-Feb-2021
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times