ContentslistsavailableatScienceDirect
Health
Policy
jo u rn al h om ep a ge :w w w . e l s e v i e r . c o m / l o c a t e / h e a l t h p o l
Managing
the
performance
of
general
practitioners
and
specialists
referral
networks:
A
system
for
evaluating
the
heart
failure
pathway
Sabina
Nuti
a,
Francesca
Ferré
a,∗,
Chiara
Seghieri
a,
Elisa
Foresi
a,
Therese
A.
Stukel
baLaboratorioManagementeSanità,IstitutodiManagement,DipartimentoEMbeDS,ScuolaSuperioreSant’AnnaofPisa,PiazzaMartiridellaLibertà,33,
56127,Pisa,Italy
bICES,UniversityofToronto,Canada
a
r
t
i
c
l
e
i
n
f
o
Articlehistory:Received1April2019 Receivedinrevisedform 16September2019 Accepted5November2019 Keywords:
Heartfailure Clinicalpaths
Referralnetworksofcare Performanceevaluation Integrationofcare
a
b
s
t
r
a
c
t
Highqualitychronicdiseasemanagementrequirescoordinatedcareacrossdifferenthealthcaresettings, involvingmultidisciplinaryteamsofprofessionals,andperformanceevaluationsystemsabletomeasure thiscare.Inter-organizationalperformanceshouldbemeasuredconsideringtheprofessional relation-shipsbetweengeneral practitioners(GPs)andspecialists,whoare usuallylinkedthroughinformal referralnetworks.
Theaimofthispaperistoidentifyandevaluatetheperformanceofnaturallyoccurringnetworksof GPsandhospital-basedspecialistsprovidingcareforcongestiveheartfailure(CHF)patientsinTuscany, Italy.Theanalysisfocusesontheidentificationandclassificationofnetworks,followingCHFpatients (n=15,841)throughprimarycareandinpatientcareusingadministrativedata,andontheassessment ofprocessandoutcomeindicatorsforCHFpatientsinthesereferralnetworks.
WedemonstratetheexistenceofinformallinksbetweenGPsandhospitalsbasedonpatternsofpatient flow.Thesenetworkswhicharenotgeographicallybasedvaryintheintensityofrelationshipsandquality ofcare.Suchreferralnetworksmayrepresentthemosteffectiveaccountabilitylevelforchronicdisease management,sincetheyencompassthemultiplecaresettingsexperiencedbypatients.Overall,an inte-gratedapproachtoevaluationandperformancemanagementthatconsidersthenaturallyoccurringlinks betweenprofessionalsworkingindifferentsettingsmayenablemoreefficient,integratedcareandquality improvements.
©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
Chronic conditions are becoming increasingly important for healthsystemsworldwide,andaccountforapproximately46%of theglobalburdenofdisease[1].Themanagementofthese condi-tionsposessignificantchallengesforpatients,healthprofessionals, andhealthsystemsmorebroadly.Congestiveheartfailure(CHF) is a chronic diseasewith highprevalence[2], affecting tens of millionsofpatientsworldwide.Itfrequentlyarisesasaresultof ischemicheartdisease(IHD),theworld’sleadingcauseofdeath in2010[3].TheprevalenceofCHFisexpectedtoincreasedueto higherlifeexpectancyandreductionsinacuteIHDmortality[2]. CHFisfrequentlydiagnosedinhospitalizedpatients.Theproper management of CHF involvesboth primary and specialist care, withco-managementofpatientstoensuretheimplementationof
∗ Correspondingauthor.
E-mailaddress:francesca.ferre@santannapisa.it(F.Ferré).
evidence-basedtherapy,effectivemanagement ofcomorbidities, andtimelyfollow-up[4,5].
Improving coordination of care for chronic disease patients reduceshospitalizationrates,increasesqualityoflifeforpatients and improveshealth systemsustainability[6].Multidisciplinary careteamsarerecommendedinCHFguidelines,withtheevidence demonstrating improvedoutcomes,alleviationof suffering,and betterexperienceforpatientsandtheirfamilies[7].Additionally, CHFpatientsmanagedinacoordinatedmannerbyprimarycare andspecialistphysicianshavehighersurvivalratesthanthose fol-lowedonlybyfamilyphysicians[8,9]aswellasreducedhospital readmissionrates[10].
However,primarycareandinpatientsettingstypicallyoperate insilos,wheregeneralpractitioners(GPs)andspecialistsworkin separatesystemswithpoorcommunication,informationand lim-itedsharedresponsibility[11,12].Indeed,performanceevaluation systems (PES) in healthcare favor measures at the organiza-tionallevel,bysettinggoalsandmonitoringperformanceresults, stressingdepartments’productivity(e.g.,volumesandcomplexity)
https://doi.org/10.1016/j.healthpol.2019.11.001
0168-8510/©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4. 0/).
[13,14]andcreatingcompetitionforresourceallocation.ThesePES limittheabilityofhealthcarestakeholderstoassessperformance throughtheperspectiveofcarepathways[15]andaccordingto thepublicvaluecreationparadigm-whichshouldbethereference paradigmofpublichealthcaresystems[16].Effectiveperformance evaluationsystemsarethereforeneededtomonitorintegratedcare pathwaystoenhancesystemaccountability,reduceunwarranted variation,andimprovequalityofcare.
Accurately measuring the performance of care pathways requires a focus on inter-organizational networks that allow coordinationbetweenhealthprofessionalsacrossorganizational boundaries.Thesenetworks,whichmayormaynotbeofficially constituted,shouldcareforpatientsalongallphasesofthecare pathway.Therelationshipsamongnetworkprovidersare charac-terizedbyinterdependence,complexity,andcontinuouschange. Thesefeatures,plustheabsenceofaclearhierarchy,makestheir assessmentproblematic[17].
Inter-organizationalperformancecanbemeasuredby consider-ingtheprofessionalrelationshipsbetweenGPsandspecialists,who areusuallylinkedthroughinformalreferralnetworkswhichcould bebasedonthecollectiveexperienceofworkingtogether.These informallinksbetweenGPsandhospitalhealthprofessionalsare describedusingpatternsofpatientflowwhichcanbeidentified throughlinked administrativehealthcaredatabases. Webelieve that these primaryand hospital-based specialist carenetworks mayrepresentaneffectivelevelofaccountabilityforhealth sys-temmanagement.Indeed, makingprofessionalsaccountablefor patientstheyhaveincommon[8]bymeasuringperformanceon theiroutcomesandonprocessindicatorscanenhance collabora-tionandintegrationacrossdifferentcaresettings.Ourworkuses thesemulti-specialistcarenetworksasunitsofperformance mea-surement.
2. Measuringandevaluatingintegrationofcareusing administrativedata
Previousstudieshaveidentifiedinformalmulti-specialistcare networksthroughthelinkageofhealthadministrativedata.Bynum etal.[18]developedamethodologytoassignU.S.Medicarepatients andtheambulatoryphysicianswhoservethemtoindividual hos-pitals. Otherworkshows how administrativedata canbeused tounderstand howcare is actually deliveredand organizedby groupsofprimarycaredoctors,andthustomeasurehow prac-titionersareinterconnectedthrough theircare ofpatients[19]. Landonetal.(2012)appliedsocialnetworkanalysistoU.S. Medi-caredatatodefineprofessionalnetworksbasedonpatientsharing among physicians,and alsoexaminedhow suchnetworksvary acrossgeographicareas[20].Stukeletal.(2013)identifiedinformal multispecialtyphysiciannetworksinOntario(Canada),bylinking GPsandspecialiststothehospitalswheremostoftheirpatients wereadmitted[21].
BuildingontheexperienceofOntario,thepresentstudyaimsat identifyingnaturallyoccurringlinksbetweenprimarycare physi-ciansandhospitals-basedspecialistsinvolvedinthemanagement ofCHFinTuscany(Italy)usinghealthadministrativedata. Addition-ally,theperformanceofeachprimaryandspecialistcarenetwork isevaluatedthroughasetofevidence-basedindicatorsforCHFcare thatincludebothprocessandoutcomemeasures.
3. Context
Italy’shealth-caresystemisaregionallybasedNationalHealth Service(NHS),whichprovidesuniversalcoveragelargelyfreeof chargeatthepointofdelivery.Tuscanyisalargeregionincentral Italy,withanon-competitivehealthsystemprovidingprevention
andprimarycareaswellashospitalservices.Theregionalhealth systeminTuscany(TRHS)isresponsibleforthehealthof3.7 mil-lioninhabitants(6.2%oftheItalianpopulation[22])andcomprises threelocalhealthauthoritieswithabout40communityhospitals, fourteachinghospitals,onemono-specialistcardiaccentre,and 34healthdistricts.Thelocalhealthauthoritiesprovidepreventive medicineandpublichealthservices,primarycare,andinpatient andoutpatientcare,whiletheteachinghospitalsfocusonacute careandprofessionaltraining.Primarycarephysiciansarefunded onacapitationbasisandtheiractivityiscoordinatedwiththe ser-vicesprovidedbythehealthdistrict.IntheTRHS,about2,650GPs providefamilymedicineservicesandactasgatekeeperstohigher levelsofcare[23].Multipleprimarycarephysiciansandspecialist networksformacrosstheTRHS,sinceGPscanreferpatientstoany specialist(inpatientandoutpatient),andpatientsarefreetoseek carefromanyprovider.
Recently,theTRHSdevelopedanewformofprimarycare pro-fessionalintegrationatthelocallevel(territory)byadoptingthe TerritorialFunctionalAggregations(AggregazioniFunzionalie Ter-ritoriali-AFTs).AFTsarecompulsorynetworksofGPsexpectedto applyclinicalgovernanceprinciplestocontinuouslyimprovethe qualityofservicesandsafeguardhighstandardsofcare[24]. Cur-rently116AFTs(2017)havebeenestablishedthroughouttheTRHS. Onaverage,eachAFTservesapopulationofabout30,000patients by23GPs[23].OurresearchanticipatesthatthesenetworksofGPs willshowinter-networkvariabilityinoutcomesandprocessesfor thecareofCHFpatients.
4. Materialsandmethods
Adopting the methods of Stukel [21], the identification of primaryandspecialistcarenetworkswasbasedona retrospec-tiveanalysisusingdifferentsourcesofroutinelycollectedhealth administrativedatafromtheTuscanregionfor2014-2016. Specif-ically, theindividuallevel healthdatabases usedinthepresent study include:(i)hospital inpatientdata wheredata arecoded usingthe9threvision,ICD9-CM;(ii)emergencycaredata;(iii)
regis-teredpersonsdatabasewhichincludesdataonallpersonsenrolled intheTuscanhealthcaresystemincludingdateofbirth,dateof death,andassignedprimarycarephysician.Thedifferent adminis-trativedatabaseswerelinkedatindividual(patient)levelthrougha uniqueidentifier.Additionaladministrativedatawereusedto mea-sureperformanceindicators,namelyoutpatientdrugprescription data,usingtheAnatomicalTherapeuticChemical(ATC) classifica-tionsystemandoutpatientcaredata.
DatawereanonymizedattheRegionalHealthInformation Sys-temOfficewhereeachpatientwasassignedanencryptedunique identifier.Thestudywascarriedoutincompliancewiththe Ital-ianlawonprivacy,andapprovalbyanEthicsCommitteewasnot required.
The datamanagement andtheanalyseswere runusingSAS version9.4(SASInstitute).
4.1. Cohortselectionandprimaryandspecialistcarenetworks identification
ThemultispecialtyphysiciannetworksforCHFcarewere iden-tifiedusingatwo-stepprocess:usinghealthadministrativedata, acohortofpatientswithcardiacconditionswasselectedandeach patientassignedtohis/herGP(phase1);eachGPwasthenassigned to the hospital where most of his/her patients were admitted foranycardiaccondition(referencehospital)(phase2).Current administrativedatadonotallowtheidentificationofthehospital specialistwhovisitedtheCHFpatientinhospitalthereforelinkages betweenproviderswerecreatedbetweentheGPandthehospital.
Thecohortofcardiacpatientsinphase1wasidentifiedby select-ingallresidentsaged18–100yearswhohadatleastoneplannedor unplannedhospitalizationforheartdisease,arrhythmiaorcardiac decompensationinanypublichospitalinTuscanyinthethree-year periodfromJanuary1st2014toDecember31st2016(the
diagnos-ticcodesfortheidentificationofCHFpatientsarereportedinthe OnlineAppendix).Inphase1weincludedallpatientshospitalized foranycardiac conditionincludingCHF,sincewe hypothesised that linkages betweentheGP and thehospital-based specialist wouldholdforsimilarclinicalconditions(i.e.,thereference spe-cialistwouldbethesameforallcardiomyopathies).Thisconsiders agreaternumberofcardiacconditionsallowingamoreaccurate identificationofGPsandhospital-basednetworks.
Thestudycohortincluded51,760patientsand67,674 hospital-izationsforcardiacconditions.Thepatientswerethenassignedto theirGPs.PatientswhochangedGPduringtheobservationperiod wereassigned totheGPwhoprovided mostof thecareinthe observationperiod. For each GP,wecalculated thedistribution ofadmissionsofhis/herpatientstoallTuscanhospitalsoverthe 3-yearperiod.Basedonthisdistribution,GPswereassignedtoa ref-erencehospital-thehospitalwherethemajorityoftheirpatients wereadmitted(phase2).Throughthismethodweidentified2,881 linkagesofGPtohospitalwhichcomprised51,760patientswith cardiacconditionsassistedby2,881GPswho,inturn,wereassigned to41referencehospitals.Foreachpair,weclassifiedthe“strength” oftheprofessionallinkage(GPtohospital)asstrong,moderateand weak.Stronglinkageswerethoseforwhichatleast60%ofthe patientsof aGPwereadmittedtothesamehospital; moderate linkageswerethosewhere40%–60%wereadmittedtothesame hospital;andweaklinkageswerethosewherelessthan40%ofthe patientswerehospitalizedtothesamehospital,indicatingthatGPs tendedtoreferhis/herCHFpatientstodifferenthospitals(phase3). Allanalyseswereperformedusingthestrongestlinkage net-worksonly,sincewehypothesizedthattheyconstitutethehighest hospitaladmissionloyaltytoallowforoptimalperformance bench-marking and continuity of care. Out of the 2,881 linkages, we identified2,062strongGP-hospitallinkages.
TheperformanceevaluationwasrestrictedtoCHFpatients (car-diacpatientshavingat leastone hospitalizationfor CHF inthe 3-yearperiod)whowereadmittedforCHFtothereferencehospital. Fromthese,weobtained1,965strongGP-hospitallinkageswhich included15,841CHFpatientslinkedto1,965GPsand38reference hospitals.Webasedtheperformanceanalysisonthesenetworks. The diagnostic codes for theidentification of CHF patientsare reportedintheOnlineAppendix.
Finally,weranavalidationtesttoassesshowself-containedthe networkswere.WemeasuredthefrequencyofCHFpatientsthat hadatleastonecardiacoutpatientvisit,oneechocardiographic ser-vice,oroneelectrocardiogramdeliveredinthereferencehospitals inthe3-yearobservationperiod(January1st,2014toDecember
31st,2016).Wefoundthatonaverage,in74 %of thenetworks
thereferencehospitalforinpatientserviceswasalsothereference hospitalforoutpatientservices.
4.2. Performancemeasuresregardingheartfailurecareprocess andoutcome
From yearly performance reports of the Tuscan Perfor-manceEvaluationSystem(PESfreelyaccessibleonlineathttp:// performance.santannapisa.it/pes/toscana)thatmeasureandassess multiplehealthcareperformancedimensionsattheproviderand AFTlevels,weselectedevidence-basedindicatorsforCHF.These indicatorshavebeenmeasuredandevaluatedinindividualcare set-tingsandsharedwithpractitionersandpolicy-makerstosupport performanceimprovementandalignmentwiththestrategicgoals ofthehealthcaresystem[25].Publicreportingofbenchmarked
performance,togetherwithclinicalinvolvementindevelopingthe rulesandcriteriaofperformanceindicators,enabletheeffective engagementofprofessionalsindiscussionandfeedbackabout per-formanceandoutcomes.
Theindicators,calculatedfromadministrativedatabases,are: • Medicationadherence:%ofpatientswithtwoormore
prescrip-tionsofbeta-blockerswithinoneyearoftheindexhospitalization (C11a.1.4indicatorcodeonthePESwebplatform).Indeed,recent researchrevealedthebenefitsoftreatingheartfailurepatients withlong-termbeta-blockertherapyespeciallyinpatientswhere thereisaIArecommendation[26].%ofpatientswithtwoormore prescriptionsofACEinhibitorsorARBswithinoneyearofthe indexhospitalization(C11a.1.3indicatorcodeonthePESweb platform);and%ofpatientswithtwoormoreprescriptionsof anti-aldosteronewithinoneyearoftheindexhospitalization. • Outpatientfollow-upduringoneyearpost-index
hospitaliza-tion:%ofpatientsseenbyacardiologistatleastoncewithinone yearoftheindexhospitalization.Evidencereportingthat30-day follow-upafterdischargeisassociatedwithlowerriskof1-year mortality[27]andearlyphysicianfollow-up(within7days)can furtherreducethisriskand30-dayreadmissionrate[28,29].% ofpatientshavingatleastoneechocardiogramwithinoneyear oftheindexhospitalization;%ofpatientswithatleastone mea-surementofB-typenatriureticpeptide(BNP)andN-terminalpro b-typenatriureticpeptide(NT-proBNP)withinoneyearofthe indexhospitalization;and%ofpatientswithatleastone crea-tinine,sodiumandpotassiumlevel measuredwithinoneyear oftheindexhospitalization(C11a.1.2aandC11a.1.2bindicator codesonthePESwebplatform).
• Outcomes:unplannedreadmissionswithin30and180daysafter theindexhospitalizationexcludingpatientwhodiedinhospital; 30and180daymortality(includinghospitalmortality)afterthe indexhospitalization.Theindexhospitalizationwasidentifiedby randomlyselectingonehospitalizationduringthestudyperiod [30].
Outcomeswereindirectlyadjustedbyage,sexandElixhauser indexthroughamultiplelogisticregression.Tocomputethe Elix-hauserindex,weadoptedtheComorbiditySoftwareVersion3.3 developedaspartoftheHealthcareCostandUtilizationProject (HCUP)bytheAgencyforHealthcareResearchandQuality(2008) [31].
4.3. Performancemeasurement
Weconductedatwo-stepanalysis.First,wemeasuredthe per-formanceindicatorsatthenetworklevelfor2017.Followingthe methodologyoftheTuscanPES[32]foreach indicator,the per-formance of the networks was benchmarked using five-colour evaluationbandsdefinedonthebasisofthepercentiledistribution, wherered(poorperformance)representsthelowestquintileofthe distributionanddarkgreen(excellentperformance)thehighest, sincetherearenointernationalornationalstandardsortargets againstwhichtomeasureperformance.
Secondly, toprovidean effectivegraphical representationof shiftingthefocusfromasingleorganization’sperspectivestothe performanceofnetworksasawhole,wedisplayedtheperformance indicatorsalongthemainphasesofthecarepathway.This illustra-tionfollowsthemethodinNutietal.[15]usinga5-bandevaluation. Thisrepresentationdisplaysthepathways’ performanceatboth primarycareandhospitalnetworkandatAFTlevels,andallows simplerecognitionofthestrengthsandweaknessesofperformance inthedifferentpathwayphases.TheAFTrepresentationenablesthe identificationofintra-AFTvariationintheperformanceofdifferent carenetworks.
Fig.1.Percentageofpatientswithtwoormoreprescriptionsofbeta-blockerswithinoneyearoftheindexhospitalization.
Fig.2.Percentageofpatientswithatleastonecardiacvisitatoneyearfromtheindexhospitalization.
5. Results
ResultsshowthepresenceofinformalstronglinksbetweenGPs andhospitalbasedonexistingpatternsofpatientflow,whichare notconstrainedgeographically.Inparticular,outof1,965strongGP -hospitallinkages,whichincluded15,841patientswithCHF,linked to1,965GPsand38referencehospitals,wedefined38networks tobeconsideredfortheperformanceevaluationanalysis.About 49%ofCHFpatientsaremale,withanaverageageof81(range, 20–100years).73.5%ofpatientshadatleastonecomorbidityas measuredbytheElixhauserindex(seetheOnlineAppendixforthe dataofeachnetwork).Significantvariabilityincarepracticeand performanceexistamongthenetworksinTuscany(Figs.1,3and2 intheOnlineAppendix).
Onaverage63%ofTuscanypatientshavetwoormore prescrip-tionsforbeta-blockers12monthsposthospitalization(Fig.1).High performingnetworksarethosewithmorethan69%ofpatients beingprescribedbeta-blockers(green).
Fig.2showsthepercentageofpatientswithatleastone car-diac visit within one year of the index hospitalization (range,
28.8%–87.5%).Theregionalmeanis47.9%.Thereishighvariability acrossnetworks.
When looking at outcome indicators, 30-day adjusted mor-talityshowssignificantvariabilityamongthe38networks,from 3.6%–16.8%withaverageregional30-dayadjustedmortalityof11.6 %(Fig.3).
Fig. 4 shows the performance of the care delivered to CHF patientsbythestrong-linkingprofessionalnetworkswith69GPs distributed in16 AFTstoamono-specialty cardiaccentrein an urbanareaofTuscany.Outcomesforthesepatients(n=184)are verygoodwithastatisticallysignificantlower30-daymortality comparedtoaverage30-daymortalityinTuscany.Adherenceto drugtherapyisverygood(e.g.80%ofCHFpatientshavetwoor moreprescriptionsforbeta-blockers12monthspost hospitaliza-tion)aswellasgoodperformanceintheoutpatientfollow-ups,with forinstance,73%ofpatientshavingatleastonecardiacvisitwithin oneyearoftheindexhospitalizationbutwith41%ofpatientswith atleastoneechocardiogram.
Anadditionalfocusisontheprocessandoutcomeperformance ofCHFpatientscaredforbyasingleAFT.TheresultsofFig.5
high-Fig.3.30dayadjustedmortalityfromtheindexhospitalization.
Fig.4.AnexampleoftheCHFpathwaydisplayingtheperformanceofamono-specialtycardiaccentre(Network28)linkedwith69GPsdistributedin16AFT.
lighttheheterogeneityofreferralbehaviourtospecialistswithin agroupof16GPsworkinginthesameAFT;62.5%haveas refer-encenetworkthenumber10,31.25%Networknumber28andthe remaining6.25%Networknumber20.WecanseethatNetwork 28hasbetterperformancecomparedtoNetwork20andNetwork 10onallprocessandoutcomeindicators,withtheexceptionof 180-dayreadmission.
6. Discussion
Thestudyhighlights theexistenceofinformallinks between GPsand hospitalsbasedonpatternsofpatientflow.These net-workswhich arenot geographicallybasedvaryin theintensity ofrelationshipsandqualityofcare.Suchinformalnetworksmay representthemosteffectiveaccountabilitylevelforchronic dis-easemanagement,sincetheyencompassthemultiplecaresettings experiencedbypatientsandovercomesomeofthelimitationsof formalintegratedcaremodels.
Integratedandcoordinatedcareaimsatbridgingtheboundaries betweenprofessions,providersandinstitutions,andovercoming existing organizational and funding silos. However, the chal-lengesaremultiple,andrequirestrategiesatthemicro,mesoand macrolevels[33,34].Numerousexamplesofmicroandmesolevel approacheshavebeenproposed,suchas,focusingonclinical inte-grationorprofessionalintegration)asmodelsforintegration[35]. However,fewmacrolevelactionintegratedprogramshavebeen putinplace[33].
Froma healthsystemperspective,severalfeaturesshouldbe redesignedtoincentivizeandoptimizeintegration:funding, reg-ulatorymechanisms,managementsystemssuchasperformance evaluationtools,andfinancialandhumanresourcesmanagement systems.Insomehealthcare systems,includingtheItalian,
pro-fessionalslackcommonlydefinedobjectives,informationsystems capable of following the patient across different care settings, andmechanismsofjointaccountabilitytomonitoroutcomesand appropriatenessofcare[11].Someexamplesofhealthgovernance builtaroundteamsofprofessionalsaretheAccountableCare Orga-nizations(ACOs)forMedicarebeneficiariesdevelopedintheUnited States[36],adaptedelsewhereascaregroupsforchronicdisease patients(TheNetherlands)[37].
In some settings with market, quasi-market or social secu-ritysystems, modelsof caresuchas ACOshavebeenproposed. Theseareintendedtoshiftthepricingsystembasedonvolume towardsmechanismstocapturethecontributionofhealth pro-fessionals working together to deliver outcomes and value. In BeveridgesystemswherebothfundingbyDRGforhospital activ-itiesandcapitationfundingforgeneralmedicinearestillfocused oncaresettingsandnotonpatientcare,newperformancetools areneededtoimprovethecontinuityandoutcomesofcare[38]. Intheory,settingsofcaredefinetheboundariesofprofessionals’ actions.However,inreality,thedeliveryofmedicineisshapedby theindependentstatusofprofessionals,andreflectstheirrelative autonomy.
Healthcareprofessionalsarenotrandomlyinvolvedinthecare pathway, but are linked to one another through long-standing relationshipsoftrust,creatinginformalreferralnetworkslinking multi-specialistgroupsofphysicians.Thisleadstothecreationof informalprofessionalnetworksacrosssettings.InthecaseofCHF, patientsareusuallyreferredtothespecialistbytheGP,whotends toestablishconnectionswithateamofspecialistswithwhomthey createaninformal ¨professionalnetwork¨.Improvingthe dynam-icsofinter-professional interactionsisakey issueforachieving importantorganizationaloutcomes,includingthediffusionofbest practice,routines[39]andinnovation[40,41].Thisimpliesthat
per-Fig.5.AnexampleoftheCHFpathwaydisplayingthenetworkperformanceofoneAFTwith16GPswhohavestrongreferrallinkswiththreedifferentreferralnetworks (GraphA–Network10;GraphB-Network20;GraphC-Network28).
formancemeasuresshouldvaluehorizontalrelationshipsbetween healthcareorganizationsandprofessionals,and mitigate profes-sionalandorganizationalbarrierstonetworking[17,42–45].From this perspective, it is necessary to identifywhich professionals (providers)havebeeninvolvedin thecarepathwayand canbe consideredco-responsiblefortheoutcomes.
Inthislight,ourworkprovidesadata-drivenapproachtofillthe “responsibilitygap”amonghealthcareprofessionals,indeed mea-suringinformalnetworkscanidentifytheprofessionalsinvolved along thecarepathway,and mayenhancecollaboration. More-over,PEStrackingtheresultsachievedbymultispecialtyphysician networkscanimproveperformanceforchronicdiseasepatients through strong primary care, coordinated and integrated care amongGPs,specialists,hospitals,engagementofinterdisciplinary healthprofessionals[46]andfocusonefficiency[47].
ThemainstrengthsofourapproachlieintheselectionofCHF patientsbasedontheprimaryandhospitalcarenetworkandthe identificationofreferralnetworksforreportingqualityand per-formanceencompassingmultiplecaresettings.Suchprimarycare -hospitalnetworksshowedhighaccuracywhenassessedagainst outpatientappointments,confirmingthestrengthoftheidentified networks.Indeed,wefoundthatonaverage,in74%ofthenetworks thereferencehospitalforinpatientserviceswasalsothereference hospitalforoutpatientservices.
Our approach stressestheimportance ofpublicreportingof benchmarkedperformancebyencouragingcompetitionforhigh performance driven by professional reputation [48] and also emphasizes theimportance of clinical engagement to create a learningenvironmentina communityofpracticewhere discus-sion and feedback about performance are conducted. Thus, for effective use of the performance evaluation, provision of feed-backanddiscussionwithprofessionalsis essential[49].Finally, the pathway performance we propose highlights the contribu-tionofeachprovider/professionalindeliveringcareduringeach pathway phase,stressingjointresponsibility intheoverall care pathwayperformance[15].Byadoptingthepathwayperspective, attentionisdirectedtowardthepatient,embracingthevalue cre-ationparadigmwherebyperformancesystemsforcoordinatedcare shouldaimtoincludesystematicassessmentofthepatient experi-ence,thelevelofparticipationinshareddecision-makingbetween patientsandproviders,andself-managementinitiatives[50].
Futureresearchcouldapplythismethodtootherchronic condi-tions(e.g.,complexandmultiplechronicconditions)togeneralize theresults.Moreover,themethodshouldassessandcomparethe utilization of resources among networks.Beyond the PES, new approachesforresourceallocation(e.g.,yearlybudgeting)basedon theneedsofcohortsofchronicdiseasepatientsshouldbe devel-opedatthenetworklevel.
Thefindingshave somelimitations becauseonly asubgroup oftheCHFpopulationwasassessed;indeed,only“strong” link-agesbetweenGPsandreferencehospitalareconsidered.However, strong linkages are where the locus of responsibilities among settings can be definitively established. Additionally, current administrativedatadoesnotallowtheidentificationofthehospital specialistwhovisitedtheCHFpatientinthereferencehospital,so linkagesof“professionalstoprofessionals”arenotfeasible.Itwill beimportanttoaddthisdimensioninthefuture,aseach profes-sionalplaysakeyroleinthecarepathwayandmayhelpreduce hospitalreadmissionsandincreasetherapeuticcompliance. 7. Conclusions
Howcanperformancemeasurementsystemsholdprimarycare physiciansandspecialistsaccountableforthechronicpatientsthey arejointlyresponsiblefor?Ourcaseprovidessupportingevidence thatan integratedapproach that considersthenaturally occur-ringlinksbetweenprofessionalsworkingindifferentsettingsmay representthemosteffectivelevelofaccountabilityforquality eval-uationofthecareofchronicpatients,byencompassingmultiple caresettingsandthereforecontributingtoefficient,integratedcare andqualityimprovement.
The Tuscan experience in assessing the performance of the heart failure care pathway through the primary and specialist carenetworksrepresentsanexampleofaninter-organizational performanceassessmentsystem,fosteringcollaborativepractices, networkingandsharedresponsibilitybetweenprofessionals, espe-cially in themanagement of carepathwaysfor chronicdisease patients.
Authorcontributions
SabinaNuti:conceptualization,review&editing
FrancescaFerrè:validation,writingoriginaldraft,review& edit-ing
ChiaraSeghieri:methodology,datacuration,validation,review &editing
ElisaForesi:datacuration,formalanalysis ThereseStukel:methodology,review&editing Funding
Theresearchleadingtotheseresultshasreceivedfundingfrom RegioneToscana undergrant agreementNET-2016-02363853-4 (ProjectCARE-NETS)BandodellaRicercafinalizzata2016, Minis-terodellaSalute.
DeclarationofCompetingInterest
Theauthorsdeclarethattheyhavenoconflictsofinterest. Acknowledgements
This study was supported by the regional administration -Direzione Dirittidi cittadinanzaecoesionesociale- ofRegione Toscana. We thank Professor Michele Emdin, Scuola Superiore Sant’Anna, for his clinical competence in interpreting the data on heart failure and the profitable discussions throughout the researchproject.Theauthorsareparticularlygratefultothework ofGiusepped’Orio,LaboratorioManagementeSanità,fortheearly adaptationoftheOntariomultispecialtyphysiciannetworkmodel toTRHSandhisconstanthelpduringdataprocessing.
AppendixA. Supplementarydata
Supplementarymaterialrelatedtothisarticlecanbefound,in theonlineversion,atdoi:https://doi.org/10.1016/j.healthpol.2019. 11.001.
References
[1]WorldHealthOrganisation[Internet]Globalstatusreporton noncommu-nicable diseases 2014; 2014. Available from: https://www.who.int/nmh/ publications/ncd-status-report-2014/en/.
[2]Mosterd A, Hoes AW. Clinical epidemiology of heart failure. Heart 2007;93:1137–46.
[3]MurrayCJ,VosT,LozanoR,AlE.Disability-adjustedlifeyears(DALYs)for291 diseasesandinjuriesin21regions,1990-2010:asystematicanalysisforthe GlobalBurdenofDiseaseStudy2010.TheLancet2012;380:2197–223. [4]FaxonDP,SchwammLH,PasternakRC,PetersonED,McNeilBJ,BufalinoV,
YancyCW,BrassLM,BakerDW,BonowRO,SmahaL.Improvingqualityof carethroughdiseasemanagement.Principlesandrecommendationsfromthe AmericanHeartAssociation’sexpertpanelondiseasemanagement.Circulation 2004;109:2651–4.
[5]CowieMR,AnkerSD,ClelandJGF,FelkerMG,FilippatosG,JaarsmaT,etal. Improvingcareforpatientswithacuteheartfailure:before,duringandafter hospitalization.ESCHearFailure2015;1(January):110–45.
[6]Martínez-gonzálezNA,BerchtoldP,UllmanK,BusatoA,EggerM.Integrated careprogrammesforadultswithchronicconditions:ameta-review.The Inter-nationalJournalforQualityinHealthCare2014;26(5):561–70.
[7]RileyJP,MastersJ.Practicalmultidisciplinaryapproachestoheartfailure man-agementforimprovedpatientoutcome.EuropeanHeartJournalSupplements [Internet]2016;18(supplG).G43–52.Availablefrom:https://academic.oup. com/eurheartjsupp/article-lookup/doi/10.1093/eurheartj/suw046.
[8]EzekowitzJA,VanWalravenC,McAlisterFA,ArmstrongPW,KaulP.Impact ofspecialistfollow-upin outpatientswithcongestiveheartfailure.CMAJ 2005;172(2):189–94.
[9]LeeDS,StukelTA,AustinPC,AlterDA,SchullMJ,YouJJ,etal.Improvedoutcomes withearlycollaborativecareofambulatoryheartfailurepatientsdischarged fromtheemergencydepartment.Circulation2010;122:1806–14.
[10]FeltnerC,JonesCD,CenéCW,ZhengZ-J,SuetaCA,Coker-SchwimmerEJ,etal. Transitionalcareinterventionstopreventreadmissionsforpersonswithheart failure:asystematicreviewandmeta-analysis.AnnalsofInternalMedicine 2014;160(11):774–84.
[11]NutiS,FantiniMP,MuranteAM.Valutareipercorsiinsanità.Bologna:IlMulino; 2014.p.1–145.
[12]LeattP,PinkGH,GuerriereM.TowardsaCanadianmodelofintegrated health-care.HealthcarePapers2000;1(2):13–35.
[13]HeadBW,AlfordJ.Wickedproblems:implicationsforpublicpolicyand man-agement.Administration&Society2015;47(6):711–39.
[14]BouckaertG,HalliganJ,Availablefrom:Managingperformance:international comparisons[Internet].LondonandNewYork:Routledge,Taylor&Francis Group;2008.p.1–440 http://www.regscience.hu:88/record/421/files/DEMO-BOOK-2017-058.pdf?version=1.
[15]NutiS,NotoG,VolaF,VainieriM.Let’splaythepatientsmusic. Manage-mentDecision[Internet]2018,http://dx.doi.org/10.1108/MD-09-2017-0907. MD-09-2017-0907.Availablefrom:.
[16]GrayM.Valuebasedhealthcare.BMJ[Internet]2017;437:1–2,http://dx.doi. org/10.1136/bmj.j437.Availablefrom:.
[17]Van Der Meer-kooistra J, Scapens RW. The governance of lateral rela-tionsbetweenandwithinorganisations.ManagementAccountingResearch 2008;19(4):365–84.
[18]Bynum JPW, Bernal-Delgado E, Gottlieb D, Fisher E. Assigning ambula-tory patients and their physicians to hospitals: a method for obtaining population-based provider performance measurements. Health Services Research2007;42(1I):45–62.
[19]ManuelDG,LamK,MaatenS,Klein-GeltinkJ.Usingadministrativedatato measuretheextenttowhichpractitionersworktogether.OpenMedicine2011. [20]LandonBE,KeatingNL,BarnettML,OnnelaJP,PaulS,O’MalleyAJ,etal.Variation inpatient-sharingnetworksofphysiciansacrosstheUnitedStates.JAMA:The JournaloftheAmericanMedicalAssociation2012;308(3):265–73.
[21]StukelTA,GlazierRH,SchultzSE,GuanJ,ZagorskiBM,GozdyraP,etal. Multi-specialtyphysiciannetworksinOntario.OpenMedicine2013;7(2):1–16. [22]ISTAT,Availablefrom:Demografiaincifre[Internet].Demografiaincifre
-poplazioneresidente;2016http://demo.istat.it/.
[23]NutiS,VainieriM,BarsantiS,D’OrioG,ParentiA,VinciB.ILSistemaDi Valu-tazioneDellaPerformanceDelleAftToscaneReport2017;2018.
[24]SantosR,BarsantiS,SeghieriC.Payforperformanceinprimarycare–the useofadministrativedatabyhealtheconomists.In:Data-drivenpolicyimpact evaluation.Cham:Springer;2019.p.313–32.
[25]NutiS,VolaF,BoniniA,VainieriM.Makinggovernanceworkinthehealthcare sector:evidencefroma‘naturalexperiment’inItaly.HealthEconomics,Policy andLaw2016;11(1):17–38.
[26]PonikowskiP,PolandC,VoorsAA,GermanySDA,UkJGFC,UkAJSC,etal.ESC Guidelinesforthediagnosisandtreatmentofacuteandchronicheartfailure TheTaskForceforthediagnosisandtreatmentofacuteandchronicheart
fail-ureoftheEuropeanSocietyofCardiology(ESC)Developedwiththespecial contribution.EuropeanJournalofHeartFailure2016;18:891–975,2016. [27]AtzemaCL,AustinPC,YuB,SchullMJ,PharmdCAJ,IversNM,etal.Effectof
earlyphysicianfollow-uponmortalityandsubsequenthospitaladmissions afteremergencycareforheartfailure:aretrospectivecohortstudy.CMAJ 2018;190(50):1468–77.
[28]TungY,ChangG,ChangH,YuT.Relationshipbetweenearlyphysician follow-upand30-Dayreadmissionafteracutemyocardialinfarctionandheartfailure. PLoSOne2017;12(1):e0170061.
[29]HernandezAF,GreinerMA,FonarowGC,HeidenreichPA.Relationshipbetween earlyphysicianfollow-upand30-Dayreadmissionamongmedicare beneficia-rieshospitalizedforheartfailure.JAMA:TheJournaloftheAmericanMedical Association2010;303(17):1716–22.
[30]KrumholzHM,MerrillAR,SchoneEM,SchreinerGC,BradleyEH,WangY,etal. Patternsofhospitalperformanceinacutemyocardialinfarctionandheart fail-ure30-Daymortalityandreadmission.Circulation:CardiovascularQualityand Outcomes2009;2(5):407–13.
[31]AgencyforHealthcareResearchandQuality,Availablefrom:Healthcarecost andutilizationproject.ComorbiditySoftwareVersion3.3.[Internet];2008 http://www.hcup-us.ahrq.gov/toolssoftware/comorbidity/comorbidity.jsp. [32]NutiS,VainieriM.Strategiesandtoolstomanagevariationinregional
gover-nancesystems.In:StukelTA,JohnsonCCA,editors.Medicalpracticevariations. Springer;2014.p.1–40.
[33]AngusL,ValentijnPP.Frommicrotomacro:assessingimplementationof integratedcareinAustralia.AustralianJournalofPrimaryHealth[Internet] 2017;(Kodner2009):59–65.Availablefrom:http://www.publish.csiro.au/py/ PY17024.
[34]NolteEE,McKeeM.Caringforpeoplewithchronicconditions:ahealthsystem perspective.EuropeanObservatoryonHealthSystemsandPolicies [Inter-net]2008;XXI:259.Availablefrom:http://www.euro.who.int/ data/assets/ pdffile/0006/96468/E91878.pdf.
[35]ValentijnPP,IntegratedC,CareP.Understandingintegratedcare:a compre-hensiveconceptualframeworkbasedontheintegrativefunctionsofprimary care.InternationalJournalofIntegratedCare2013;13(March).
[36]BerwickD.MakinggoodonACOs’promise-thefinalruleforthemedicare sharedsavingsprogram.TheNewEnglandJournalofMedicine2011:1753–6. [37]StruijsJN,BaanCA.Integratingcarethroughbundledpayments—lessons
from the Netherlands. The New England Journal of Medicine [Inter-net]2011;364(11):990–1, http://dx.doi.org/10.1056/NEJMp1011849. Avail-ablefrom:.
[38]StruckmannV,QuentinW,BusseR,vanGinnekenE,Availablefrom:Howto strengthenfinancingmechanismstopromotecareforpeoplewith multimor-bidityinEurope?;2016http://www.euro.who.int/ data/assets/pdffile/0006/ 337587/PB24.pdf?ua=1.
[39]SchönD.Thereflectivepractitioner:howprofessionalsthinkinaction. Alder-shot,UK:Ashgate;1991.
[40]TortorielloM,KrackhardtD.Activatingcross-boundaryknowledge:therole ofSimmeliantiesinthegenerationofinnovationsactivatingcross-boundary knowledge:the roleof Simmelianties inthe generation ofinnovations. AcademyofManagementJournal2010;53(1):167–81.
[41]CarlilePR.Apragmaticviewofknowledgeandboundaries:boundaryobjects innewproductdevelopment.OrganizationScience2002;13(4):442–55. [42]KurunmäkiL,MillerP.Regulatoryhybrids:partnerships,budgetingand
mod-ernisinggovernment.ManagementAccountingResearch2011;22:220–41. [43]Berry A. Spanning traditional boundaries: organization and control of
embedded operations. Leadership & Organization Development Journal 1994;15(7):4–10.
[44]CuganesanS,JacobsK,LaceyD.Beyondnewpublicmanagement:does per-formance measurementdrivepublic valuein networks?In:Public value management,measurementandreporting(Studiesinpublicandnon-profit governance);2016.p.21–42.
[45]DekkerHC.Ontheboundariesbetweenintrafirmandinterfirmmanagement accountingresearch.ManagementAccountingResearch2016;31:86–99. [46]CrossonFJ,GutermanS,TaylorN,YoungR,TollenL.Howcanmedicarelead
deliverysystemreform?,Vol.71;2009.TheCommonwealthFund.
[47]RahmanF,GuanJ,GlazierRH,BrownA,BiermanAS,CroxfordR,etal. Asso-ciationbetweenqualitydomainsandhealthcarespendingacrossphysician networks.PLoSOne [Internet]2018;13(4):1–16,http://dx.doi.org/10.1371/ journal.pone.0195222.Availablefrom:.
[48]BevanG,EvansA,NutiS.ReputationsCount:whybenchmarkingperformance isimprovinghealthcareacrosstheworld.HealthEconomics,PolicyandLaw 2019;14(2):141–61.
[49]VainieriM,FerreF,GiacomelliG,NutiS.Explainingperformanceinhealthcare: howandwhentopmanagementcompetenciesmakethedifference.Health CareManagementReview2019;44(4):306–17.
[50]NutiS,DeRosisS,BoncianiM-,Murante AM.Rethinking healthcare per-formance evaluation systems towards the people-centredness approach: their pathways, their experience, their evaluation. Healthcare Papers 2018;17(2):56–64.