rapid application development in small unit settings

RapidapplicationdevelopmentinsmallunitsettingslnlernaHninlJournalOrIMedicalELSEVIERInternationalJournalofBio-MedicalComputing40(1995)157-163ComputingRapidapplicationdevelopmentinsmallunitsettingsJasonC.H.YapMinistryofHenlih,Collegrof’MedicineBuilding,CollegeRoad,Singapore0316,SingaporeAbstractSmallunitswithspecialisedfunctionsoftenhavedifficultyindevelopinginformationsystemsbecauseoftheiruniqueanddynamicrequirements,thesmallnessoftheirapplicationsrelativetothelargeinter-organisationalnetworks,andthesheernumbersofsuchunits,especiallyingovernment.Personalexperiencewiththedifficultyofestablishingasuitableinformationsystemhasledtothedevelopmentofamethodologyforenduserstodeveloptheirownapplicationswithonlytheguidanceofinformationtechnologyprofessionals(ITPs)-hereentitled‘SmallUnitEnduserRapidApplicationDevelopment’orSERAD.Adatabasemanagementsystem(DBMS)isusedtomanagedataabout‘focalobjects’withstatusflagsandqueries/viewsatinterfacepointsthroughUDA(UserAction,DataEntry,ApplicationAction)cycles.Thesystemwhichpermitsdataentryalmostimmediately,notonlyeventualiycreatesafullapplication,butalsotrainsendusersintheprocess.ThemerhodologyleadsITPsandendusersthroughasequenceofexercisesinsystemsanalysis,tailoredtominimisethetimeandeffortinputofITPsdespitetherelativelackofITtrainingandconceptsinendusers,withaviewtorapidlycreatingdynamic,robustandsoundinformationsystems.Keywords:Smallunitsettings;Informationsystems;Rapidapplicationdevelopment1.IntroductionOurchaoticworldincreasesinspeedandcom-plexityeveryday.Astechnology,inparticularinformationtechnology(IT)transformstheworld[l],noorganisationcanignorethepotentialofIT.Leadingedgeorganisationsshowhowtheinnova-tiveuseofITleadtoefficientoperations,com-mercialeffectivenessandstrategiccompetitiveadvantage[2].Howeversmallunits(SUs)withincertainorganisations(especiallygovernment)areindangerofbeingleftbehind,ifendusers,withalbeitlimitedtrainingandresources,arenothelpedtoimplementsound1%.TheMinistryofHealth(MOH)inSingaporehasimplementedsomeexcellentnationalinforma-tionsystems(1s~).Nationwideapplicationsin-cludeMediNet(anationwideED1networkforhealthcareestablishments),PatientBillingSystem(forgovernmenthospitals)andCentralClaimsProcessingSystem(forthesubmissionofHospitalInpatientDischargeSummariestoMOHandMedisaveclaimsinthenationalpersonalhealthcaresavingssystem).ExecutiveISsandAI-as-sistedDecisionSupportSysfemsarealsobeingexplored.TherehasbeenexplicitconsiderationoftheroleofITinhealthcare[3].ThereishoweverpotentialfortheapplicationofITtoindividual0020-7101/95/$09.5001995ElsevierScienceIrelandLtd.AllrightsreservedSSDI0020-7101(95)01140-A158J.C.H.Yap1InternationalJournalofBiomedicalComputing40(1995)1.57-163specialisedunitswhicharestillmostlybasedonmanualprocessesoroffice-automation.Theworldisposingnewandgreaterchallenges.WhilethebasicaimoftheMinistryremainsalwaystoprovide,manageandregulatedeliveryofhealthcareatthenationallevel,therearethenewchal-lengesofincreasinghealthcarecostsandrapidevol-uationofmedicaltechnology[4].InasmallcountrywhereaccessibilitytoITisunquestioned,maximalutilisationofresourcesmeansthatend-usersmustimplementtheirownISswherenecessary.1.1SmallunitsandtheirinformationsystemsSmallunitswithlessthantwentypeopleoftenneedtheirownfunction-specificISsiftheyhaveclearlydefinedspecialisedportfolios.Whileperhapsinfrequentincommercialfirms,unitswithhighlyspecialisedfunctionsaboundingovernmentespe-ciallyatpolicysettinglevels.Forexample,oneunitinMOHlicensesprivatehealthcareestablishments(HCEs),anothertrainsanddeploysmedicalofficersthroughoutthepublicsector,purviewsnotsharedbyanyotherunits.Lowinterrelatedness[5]meansthatSUstendtostandalonephysicallyandhaveminimalrequirementsforelectronicdatainter-changeornetworking.SuchunitshavenoinhouseITPs.Staffareadministrativeand(non-IT)profes-sional,andcomputingexpertisevaryfromenthusi-asttoneophytetoterrified.ITtrainingtendstobelargelyincidentaltoworkexperience.Environ-mentsvary-somearestableandlongestablished,othershaverapidlyevolvingneedsasdirectionschangewitheachweek.Unitsdealingwiththecontrolofdrugsandmedicalpractitionersarefairlystable,butunitsdealingwithnewerconcerns(e.g.,licensingandaccreditationofHCEs)havetremen-dousworkloadsinjustlearningthesubject.1.2NeedsofsmallunitsDevelopmentandfirstdeliverymustfollowtheinitialspecificationsveryquickly.Thereisoftenverylittleleadtimeaspoliticalandorganisationalforcesrequireextremelyrapidresponses,ofteninweeksordays(andevenhours).Inadynamicenvironment,thesystemthatgoesupfirstispraisednotthebetterone.Inthecivilservice,themanualsystemisfamiliarandeasilyimplemented.Ittakesonlyamemowiththeappropriatepaperformsattached.Thisactuallyworkssimplybecausepeopleputinalotofeffortinbrute-forcepaper-shufflingandpen-and-paperanalysis.Flexibilityandspeedinmodificationisascriticalasitisdifficulttospecifyneedsinnewpurviews.Theendusers’relativelackofabilityalsomeansthatthemethodologymustproduceasystemrobustenoughnottosuccumbtothevagariesofendusers’actions,andsimpleenoughtobeachievableandyetnotinvitekeyboardadventuring.Securitydependsonthesubjectoftheunit’sattention.Inmanycases,SUsholdconfiden-tialinformation,andtheappropriatelevelofcareshouldbetakentoensuresecurity.NotehoweverthattheISdoesnotneedtobecompletelycompatiblewiththerestoftheorganisa-tion.TheSUworksseparatelyandtransactionsaremostlypaper-based.Electronicdataexchangecanbebatchedandtranslatedenmasseratherthanreal-time.Althoughdatastructureandspecifica-tionsshouldideallyfolloworganisationalnorms,itmaybereasonablenotto.Extensivecommunica-tionsorintegrationarenotneeded.Finally,thereislittleneedforsophisticatedhardwareorsoftware.Theendusersneedtostartwithwhattheyhaveandknow,andmaybedrivenoffbycomplexity.ThemovetoIT-basedapplicationsusingoff-the-shelfsoftwareisenoughtodauntendusers.Asuccessfulapplicationwouldprovideenoughincreasedpro-ductivitytoenticetheendusertolaterexplorethemoreesoteric.1.3TheplightofsmallunitsSUsareinadifficultposition.Theirenvironmentchangesrapidlyandtheymustadapt.Manualsystemshavereachedtheirlimitsbutthenecessaryexpertisetobuildcomputer-basedISsisnotavail-able.BeingSUs,thereisalsolowerorganisationalpriorityfortheirsystems.EvenifmanagementisconvincedthattheISisworththecosts,conven-tionalsystemsdevelopmentlife-cyclessimplytaketoolongandriskbeingovertakenbyeventse.g.,politicalandsocialchanges.SUsindynamicenvi-ronmentsrequireISs‘setupyesterdayandchangedtoday’.’’Whatisneededisawaytodevelop1%fromthemostelementaryofcomponents,i.e.,PCs,off-the-shelfPC-baseddatabasemanagementsystems(DBMSs),andthelike.Amethodologyp&posedf&rtheimplementationofISsinSUsettings.entitled‘SmallUnitEnduserRapidApplicationDevel--,opment’,willbedescribedbelow.J.C.H.Yap/InternationalJournalofBiomedicalComputing40(1995)157-1631592.Methodology2.1SourcesThismethodologydrawsonavarietyofsources.FromObjectOrientedSoftwareEngi-neering[6]cametheusecaseandthethree-foldcategorisationofobjectsintoentity,interfaceandcontrol.Thelattercontributesnotonlydi-rectlytothemethodology,butalsotothecon-ceptualseparationoftherolesofdata,enduserinterfaceandstatusflags.Client-servertechnol-ogyandrelationaldatabasesareassumed.TheUDAcycleistomyknowledgeoriginal.2.2BriefdescriptionThebasisoftheapplicationsdevelopmentmethodologyissimple.Workflowisbasicallylinearormulticollinear,withinputandoutputpoints.Pathwaysconnectnodeswhichareen-duserinterfacepoints.Ateachinterface,onlyrelevantdataitemsofinterestforafocalobjectareavailableandhandled.Focalobjectsappearanddisappearatinterfacepointsdependingonstatusflagswhicharethemselvesdataitems.Thestatusflagsthusserveasentranceandexit‘tick-ets’.Eachinterface(whichshowsaqueryset)hasalistoffocalobjectsforactionselectedbysatisfyingsomestatusconditions.Aftertherele-vantaction,theirexitiseffectedbysettingtherFig.1.PrinciplesofSERAD.(1)Enduser(withITPhelp)definesthescope,objectives(Objectives&Scope)andfocalobject(*),(2)Enduserdescribestheusecases(0).(3)Enduser(withITPhelp)assembles(-)usecasesintoworkflow.(4)Enduser(withITPhelp)reordersworkflowintoUDAcycles(UDA).(5)EnduserandITPdesignandsetuptablesviews,statusflags,andprocedures.(6)Enduserstartsuseofapplica-tionwithhelpofdatabasequerysoftware.(7)ITPandenduserdesignandimplementinterfacescreenandbatchproceduresasindividualsmallprogrammes,oneforeachinterface.(8)En-dusercontinuestouseapplicationasminiprogrammesareadded.ITPandendusermayconsolidateminiprogrammesintoonemenu-basedprogramme.statusflagtosomeothervalue,whichallowsthemtoappearatsomeotherinterface.Thusallthefocalobjectsaremovedthroughthenet-work.TheprinciplesareillustratedatFig.1.Thestepsdescribedareclassicsystemsdevelopmentsteps,exceptthatcarehasbeengiventomakingeachstepunderstandableandusablebytheav-erageenduserwithsomeguidancefromtheITPs.2.3AnalysisScopeAndObjectives:Theenduserdefinestheobjectivesoftheapplication.AsSUsgener-allyhavefairlylimitedandclearfunctions,thisshouldnotbetoodifficultintermsoftheactivi-tiestobeassistedbythecomputer-basedIS.OftentheISsupportsheterogeneousactivitieswithseeminglylittleinherentorder.Forexam-ple,thelicensingofHCEsincludessendingoutlettersandreports,telephoneenquiries,receivingapplicationforms,visitingestablishments,print-inglicences,etc.Eachactivityisclearlycon-nectedtoothersbuttheentireschemamaybedifficulttomap.Theconceptofthefocalobjectishelpfulhere.FocalObject:TheSUISisstructuredaroundanentityobjectwhichmaynotbeimmediatelyobvious.Forexample,theHCEmayseemtobethefocalobjectinlicensing,butitistheapplica-tionsforlicencesthattheunitdealswith.Whilethereareobject-oriented(00)methodologiesforanalysis[7],itisdifficulttoteach00concepts(inheritance,polymorphismetc)toendusersandtoexpectgooduse.However,theidentificationofjustoneobjectofmaininterest(i.e.,thefocalobject)issimplerandfeasible.UseCasesEndusersdescribetheirownwork.Individuallyorinabrainstormingsession,en-dusersimaginewhatactivitiesareconductedaroundthecentraldepositoryofdata(whethercomputer-basedorafilingcabinet).Eachsepa-rateactivityisinitiallyausecase.Eachpersonisanactor.Anactormaytakeonseveralroles,dependingontherangeofdutiesassigned.ThisislooselysimilartotheconceptofusecasesbyJacobson[6].160J.C.H.Yap/InternationalJournalof’BiomedicalComputing40(1995)157-163Foreachusecase,thegrouporindividualisaskedcertainquestionslike:Whodoesit?Whoelse?Whendoestheresponsibilityforactionmovetoeach?Whatistheexactactivity?Whatisthesequenceofsteps?Isthesequencecriticalorcansomestepsbedoneindependentlyofothers?Wheninthenaturallifeofthefocalobjectistheactivityundertaken?Whichotherusecasesareimmediatelybeforeandafter?Theexactdetailsofeachusecasearethusextracted.Theprocessisiterativewhichcompleteswhentherearenoaddi-tionalanswers.Theremaybesomerevisionoftheusecasesasthegranularityofthefirstpasslistingmaynothavebeenappropriate.Someusecasesmayneedtobesplitupandotherscombined.Workflow:Theusecasesdrawnupabovearestitchedtogethertoformtheworkflow,byonepersonorthewholegroup.Anefficientmethodwouldbeforonepersontostudyandsequencetheusecasesandthenpresentthecompletedworkflowfordebateandconcurrence.Thisstepisrelativelystraightforwardandendusersshouldbeable,withminimalguidance,tohandlethisontheirownsinceineffect,itisthemeredocumenta-tionofroutinework.UduCycle:TheUDAcyclecameoutofawhiteboardbaseddiscussionontheworkflowforaspecificunit.Insteadofdrawingarrowsinvari-ousdirectionstoconnectactivities,itwasnoticedthatmuchgreaterorderwasimposedontheworkflowiftheactivitiesinvolvedwereclassifiedas(a)UserActionoutsidetheapplication,e.g.,receivingapplicationsforlicenses(b)DataEntryintothedatabase,and(c)ApplicationActione.g.toautomaticallyprint,updateortakesomeotheractiononobjects.TheUDAcyclemakesitclearthatworkcanbeconceptualisedasrunningincycles.Theapplica-tiontakescareofthesecondandthirdphasesofthecyclewhiletheenduserswilldothefirstmanually.Thisnotonlyclarifieswhichpartsaremanualandwhicharecomputerised,andalsoneatlypointsoutalltheinterfacesandbatchprocedurestobeusedlater.InthemovementfromtheworkflowtotheUDAcyclechart,theremayagainberevisionofusecasesandworkflow,whichshouldbebackwardpropagated.Backtrackingandjumpingareallowedandarelb)JuniorthenIc)Juniorprintsdraftsreplyanreply.-2aPCDraftPrinted4,~DrafredDraftPrintedAmended2e)Junior2-JJunmreprintsamendsletteronreply93aAmendPC.Printed02fAmendedZb)Semoramends,2~)SeniorsignssendstileloJuniorletter,sendsfileto-2dJunior03d2d)Junmrreceivesfile,notescOmmentS42.AmendPrintedAmendPrintedDone3a)Juniorpusin3e)Juniormarks30Applicationdoerfile,sendsfiletocorrespondenceautomatedarchivalSenior93bor3cforarchival43fofobsoleteDonecorrespondence4ArchivedEND3b)Senioramends3clSeniorsiensFig.2.UDAcycleforhandlingaletter.indicatedby‘goto’statements.ThereshouldbenointermediaryenduseractionsbetweeneachDataEntrypointandApplicationAction.WhileApplicationActionsusuallyinvolveprintingofformsandreports,itcouldalsoinvolvetheupdat-ingofdatabases,backupofdataorevennoaction.DataEntryincludesastheminimumthechangeofstatusflagstomovethefocalobjecton.Fig.2showssomeUDAcyclesforthesimpleactionofwritingalettertobesignedbyone’sboss.2.4DesignWiththecompleteUDAcycle,theenduserisnowreadytodesignhisdatabase.ThisisagainrelativelysimpleandwellwithinthecapabilitiesofmostenduserswithsomeelementarycoursesonDBMSslikedBASEIV.Notethatskillsincreatingdatabasetablesarenotinquestion;thelackistheknowledgeindesigningtheISitself.Tables:TheenduseragainsimplyanswerssomequestionswhilelookingattheUDAchart.Allthedataitemsenumeratedforthefirstques-tionirrespectiveofthelocationofdataentrybecomedataitemsinthetableusedfortheappli-cation.Thedatatableisthendefinedwithindataadministrationguidelines.Questionsinclude:WhatdataaboutthefocalobjectdoIwanttocapture?WhatdatainadditiondoIwantduringJ.C.H.Yap/InternationalJournalofBiomedicalComputing40(1995)157-163161thedataentry,eithertocheckortomakesure1havethecorrectfocalobject?ArethereanyotherdataIwanttocapturewhichdonotbelongdirectlytothefocalobject?Areanyofthesedatatobecollectedforthefocaldataobjectconsistingoflistsofdata?TheITPexaminestheanswerstoconsiderthelevelofnormalisation.Thissplitapproachmakessurethatthedatabasetableforthefocalobjectremainsnormalisedasmuchaspossible(withouthavingtoteachenduserstheintricacies)whileallowingtheendusertodothebulkofthework.Thedecisionmustbemadewiththehelpoftheenduseronhowtocapturedatawhichdonotnormalisetothefocalobject’sdatabasetable.Iffew,theycanbeincludedasadditionalcolumns(deliberatelyacceptingincompletelynormalisa-tion)oraslistsinmemofields(iftheywouldnotbehandledasseparatedataitems)orinseparatetablestobelinkedthroughforeignkeys(difficultfortheaverageenduser,theITPmusthelp).MoreelaboraterelationaltablescanalsobedesignedbuttheITPnowonlydealswithdifficultpartsthattheendusercannothandlealone.Interfaceviews:Intheprocessofdefiningtheabove,theinterfaceviewshavealsobeendefined.ThedataitemsinvolvedwitheachDataEntrypoint(whethertobeenteredortolookat)arethedataitemswhichmakeupthetableviewforthatDataEntrypoint.Theadvantagehereisthateachenduserissofaronlyaskedtodoverysimpleexercises.EachquestiontheenduserisaskedisnotIT-orientedbutbusinessoriented.Forexample,heisaskedatthisDataEntrypoint,whatdatadoyouwanttoknow?Whatdatadoyouwanttoenter?Theseareeasyquestionsfortheenduser.ThereislittlescopeforerrorandtheITPoverseesthefinalproduct.ExpensiveITPtimeisnotneededtolearnanunfamiliarbusinessdomainastheendusercomesupwithadequateanalysesonhisown.Statusflags:Herethefirstdifficultpartarises.Thereisaneedtomanagethemovementofthefocalobjectthroughtheworkflow.TheITPhelpstheenduserthroughsomequestionsagain,suchas:AftereachDataEntrypoint,andApplicationAction,howdoesthestatusofthefocalobjectchange?Inthecaseofbacktrackingandjumps,howdothestatusofthefocalobjectschangeastheymove?ForeachDataEntrypointandAppli-cationAction,whatmustthestatusofthefocalobjectforittoappearintheDataEntryinterfaceorforsomeApplicationActiontobeperformedonit?Ashortlistofstatusesshouldbedrawnupthatlogicallyandclearlylabelseachusecase.Thesestatusesmayberepresentedbycodes.Intheexampleofthelettertoreplyabove,asamplesetof(expanded)codesareshowninthechart.Thestatusflagsallowthedatabasetablestobeseenasakindofundergroundunseencavern,anddatawellsupatdifferentplaces.Whichfocalobjectsappeardependsonthestatusflags.Ifthesystemprocessesobjectsthroughstages,eachchangemovestheobjectalittlefurtherdownstream.Supposethataletterisreceived(Fig.2).Iftheenduserabortsdraftingthereply,thestatusre-mainsatNullandthehalfdonereplyisstillavailablethenexttimesincethefirstinterfaceshowsallitemswithNullStatus.Aftercompletingthedraftreply,theenduserdoesnotneedtoprintitimmediately.Hecandraftanotherreply,andthenprintalloutstandingreplieslaterwithabatchprocedurewhichselectsallunprintedreplies(StatusDrafted),printthemandsetStatustoDraftPrinted.Shouldheselecttheprintprocedureagain,thelettersarenotreprinted.Otherproce-duresallowreprintingoflettersiftheneedarises.AmendmentsontheinitialdraftoronsubsequentamendedreplieswillbeavailableatthesecondDataEntryinterface.WhentheletterissentandthestatusissettoDone,theletterdisappearsfromtheinterface.Atperiodicintervals,thelet-terswithStatussettoDonewillbearchived.2.5FirstImplementationTablesandviews:ThecreationoftablesandviewsaredonewithconventionalDBMSs.AnSQLscripttodefineviewsisusefulasdocumenta-tionandfordynamicrecreationofviews.Someadditionaltablesmayberequiredtoensuredataintegrity.Forexample,atablewithallpermissiblestatusflagswouldallowalltablestocheckforvalidityofenteredstatuses.Aviewforalldataitemswouldalsobeuseful,especiallyforgeneralupdatestothedatabase.162J.C.H.Yap/InternationalJournalofBiomedicalComputing40(1995)157-163Controlmechanism:Thefirstimplementationofthecontrolmechanismtomovefocalobjectsthroughthenetworkisthesimplesttobuildbutthemostdangeroustouse.Thestatusflagismadeavailabletotheenduserintheviewattheveryendoftherow.Theenduserknowswhattoupdatefortherowandalsoupdatesthestatusflagtothenextvalue.Thatrowwillthenreappearintheviewwhichadmitsthenewstatusflag.Procedures:Intheabsenceofrigorousdatasecurityandintegritymechanisms,theprocedureandtheenduserbecomecritical.Thestepstobeperformedateachinterfacepointmustbeclearlylaidout.Theendusermustbemotivatedandtrainedtofollowtheprocedureclosely.Danger-ousasthissounds,withclearheadedendusersandwelllaidoutprocedures,itisreallyquitesafe.Asmuchoftheproceduresinvolvetheprintingoflettersandforms,thereportformatsmustbelaidoutalsobythisstage.Printingofhardcopiesisdoneenblocbasedonsimpleroutineswhichprintandthenupdatethestatusflags.2.6UsingtheapplicationZnitial:Theapplicationisnowreadyforitsfirstuse.ThemainaccessatthistimeisviatheDBMSwithappropriateview.Thefirstinterfacepoint,wherethefocalobjectisaddedtothesystemisusedtoenterthefirstdata.Theenduserstickstowrittenprocedures.Thisbarebonesapproachworksintheshorttermwhiletherestoftheapplicationisbeingdeveloped.Notethatthetimefromthecommencementofthedevelopmentpro-cesstothispointisveryshort.Ifdesperate(andsomeexperience,)theentireprocesstothispointcanbecompletedinoneday!Imagineatransac-tionprocessingapplicationthatacceptsthefirstdataentryinlessthan24hours!Subsequentimplementation:Thesystemgetstolookmorelikearealapplication.ThisiswheretheITPneedstoeithergivemoreattentionandhelptotheenduser,ortakeoverthedevelopmentofthefinishedinterfaces.Thelevelofactualprogrammingdonedependsonthesoftwareofchoice.MoresophisticatedprogrammeslikeSQL-Windowsrequiremoreprogrammingknowledgewhile‘nocode’softwarelikeBorland’sObjectVi-sionaremuchsimpler.Theapproachisstillpiece-meal,replacingeachdataqueryinterface(orview)withone‘miniprogramme’incorporatingthequery/viewrelevantbatchproceduresencodedinbuttons.Start-upoftheminiprogrammepre-sentsonewindowcontainingthecolumnsandrowsthattheenduserisalreadyfamiliarwithinthedirectDBMSaccesswiththeappropriateview,exceptthestatusflag.Buttonsareavailablefornavigation,abortingoftheprocedureandselectedsubprocedures(likeprintingofanop-tionalform)ifnecessary.The‘OK’buttonwillcontainalltheenduserproceduresincludingtheupdatingofthestatusflag.Inotherwords,theinterfacescreenallowstheendusertodoexactlywhattheydiddirectlywiththeDBMSateachdataentrypointincludingtheapplicationactions.EachinterfacescreenisthussequentiallybuiltandaddedtoaoverallaccessprogrammelikeWin-dowsProgramManager.Therearemanyadvantagesofthisrepetitivepiecemealapproach.Interfacesc