ContentslistsavailableatScienceDirect
Health
Policy
jo u rn al h om ep a g e :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
Measuring
research
in
the
big
data
era:
The
evolution
of
performance
measurement
systems
in
the
Italian
teaching
hospitals
Frank
Horenberg
a,b,∗,
Daniel
Adrian
Lungu
a,b,
Sabina
Nuti
a,baHealthandManagementLaboratory(MeSLab),InstituteofManagementandDepartmentEMbeDS,ScuolaSuperioreSant’Anna,PiazzaMartiridella
Libertà,33,Pisa,Italy
bSant’AnnaSchoolofAdvancedStudies,HealthandManagementLaboratory(MeSLab),PiazzaMartiridellaLibertà,33,56127PisaPI,Italy
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received2July2019 Receivedinrevisedform 21September2020 Accepted4October2020 Keywords: Teachinghospitals Researchproductivity Performanceevaluation Impactfactor
Field-weightedcitationimpact
a
b
s
t
r
a
c
t
Background:Inthehealthcaresystem,TeachingHospitals(THs)notonlyprovidecare,butalsotrain health-careprofessionalsandcarryoutresearchactivities.ResearchisafundamentalpillarofTHs’missionand relevantforthehealthcaresystemmonitoredbyPerformanceEvaluationSystems.Researchactivities canbemeasuredusingcitationindexservicesandthispaperhighlightsdifferencesbetweentwo ser-vicesbasedonbibliometrics,describesopportunitiesandriskswhenperformanceindicatorsrelyon datacollected,controlledandvalidatedbyexternalservicesanddiscussesthepossibleimpactonhealth policyatasystemandproviderlevel.
Methods:Abibliometricanalysiswasdoneondatabetween2014−2016fromISIWebofScienceand Scopusof18.255physiciansworkingin26ItalianTHs.Quantitywasdefinedasthenumberofpublications andqualityasImpactFactororField-WeightedCitationImpact.
Results:Overall,41.233and66.409documentswereextractedfromrespectivelyISIWebofScienceand Scopus.Whilebenchmarkingresults,significantdifferencesinrankedpositionbothinmetricsemerged. Discussion:UtilizingsecondarydatasourcestomeasureresearchactivitiesofTHsallowsbenchmarking atan(inter)nationallevelandovercomingself-referment.Toutilizeindicatorsformultiplegovernance purposesatthesystemandproviderlevel,indicatorsneedtobeprofoundlyunderstood,require for-malizationsindata validation,internalanalysis andasharingprocess amonghealthprofessionals, managementandpolicymakers.
©2020TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Background
Performanceevaluationsystems(PESs)arecrucialfor
account-abilityandserveasafeedbackandguidancetoolforthemanagerial
leveloforganizations[1].PESsareusedtoevaluatehowwell
orga-nizationsaremanagedandtomeasurethevaluethatorganizations
delivertocustomersandotherstakeholders[2,3].Fromthe1980s,
PESshavebroadenedthekindofindicatorsmonitored,but
main-tainedthefocusonfinancialones[4,5].Alongside,developmentsin
Abbreviations:IRPES,Inter-Regional-PerformanceEvolutionSystem;LHA,local healthauthority;PES,performanceevaluationsystem;TH,teachinghospital;WoS, ISIWebofScience.
∗ Correspondingauthorat:Sant’AnnaSchoolofAdvancedStudies,Healthand ManagementLaboratory(MeSLab),PiazzaMartiridellaLibertà,33,56127PisaPI, Italy.
E-mailaddresses:horenbergfrank@gmail.com(F.Horenberg),
danieladrian.lungu@santannapisa.it(D.A.Lungu),sabina.nuti@santannapisa.it
(S.Nuti).
informationandcommunicationstechnology(ICT)facilitateddata
availability,completeness,andaccessibilityandtheevolutionof
theso-calledBigDataturnedusefultoenrichthePESinformation
[6–8].
StillnowPESsineconomicsectorsthatareprofit-orientedare
mainlyfocusedonmeasuresregardingprofitandrevenues,while
thisisnotthecaseinhealthcarewherethegoalistoproducevalue
forpatientsandthepopulation[9–11].Withinthehealthcare
sec-tornon-financialindicatorsarecrucialandPESs,mostlyinpublic
universalcoveragehealthcaresystemswhererevenuesarebased
onapercapitaquota,aredesignedandimplementedtobeable
tomeasureononesideoutcomes,qualityofcareandlife,identify
issues,andontheotherhandresourcesmadeavailablebysociety.
Inorderfor PESstobeeffectiveinpublicuniversalcoverage
healthcaresystem,itshouldbecharacterizedbythefollowing
ele-ments[12,13]:
• Multi-dimensionality:Indicatorsshouldincludemultiple
dimen-sions(process,qualityofcare,equity,etc.);
https://doi.org/10.1016/j.healthpol.2020.10.002
0168-8510/©2020TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4. 0/).
• Evidence-based:onresearchorclinicalpractice;
• Shareddesign:allstakeholders,andespeciallyhealth
profession-als,shouldbeinvolvedinthedesignandthefine-tuningprocess
ofthePESandtheindicators;
• Systematicbenchmarking:allowstoovercomeself-referentiality
andtomeasureavoidablevariationandspaceforimprovement;
• Transparentdisclosure:stimulatesdatapeer-reviewandmakes
professionalreputationleveragepossible;
• Timeliness:allowspolicymakerstomakedecisionspromptly.
Withthesepremises,PESs inhealthcare have beenevolving
overcomingtheorganizationalboundariesofsingleproviders[14].
Moreover,amongtheaboveelements,themostcrucialistorely
on benchmarking which facilitates and triggers organizational
improvementprocessestoincreaseeffectivenessbasedon
repu-tation[15].
Consideringtheserelevantfeaturesfor thehealthcaresector,
thispaperfocusesonthePESsadoptedbyteachinghospitals(THs)
withinthehealthcaresystem.Thesekindofinstitutions,evenif
theymaybenodifferentfromotherhospitalsintermsofquality
ofcare[16],fulfillaspecialroleinthehealthcaresystembecause
theirmissionisnotjustprovidingcare,butalsototrainhealthcare
professionalsandtocarryoutresearchactivities.
Asmedicalknowledgecontinuouslyevolves,THsareatthe
cen-terofinnovationsinhealthcare withrespecttotreatments and
cures.Theyareinchargeofconductingresearchandaddingnew
knowledgetoscientific literature.For this reason, researchis a
fundamentalpillarof theTHs’missionand thereforea relevant
componenttobeincludedinPESs,notjustforthesingleprovider
butforthewholehealthcaresystem.TheabilityofTHstoperform
researchactivitiesguaranteesthemaroleofreferenceandguide
intheprocessesofimprovingcareatregionalornationallevel.The
morehealthcareprofessionalsknowhowtobeonthefrontiersof
clinicalresearch,themorelikelyitisthatcarewillbealignedwith
thebestandmostupdatedclinicalprotocolsbenefitingforpatients.
Itisinfactproventhatthebesthospitalsarethosewheremore
researchiscarriedoutandinturnhealthcareoutcomesimprove
[17,18]and patientsbenefitfromaccesstonewandinnovative
treatmentsthatwouldnotbeotherwiseavailable[19,20].
Further-more,researchactivitiesshouldguideprocessestoimprovequality
ofcaregenerationknowledgewhichleadstoupdatedandtrained
stafftoestablishteamsofexpertsandcentersofexcellence[23–26].
Measuringperformanceofresearchactivitiesthusbecomesa
relevanttopicforthewholehealthcaresystemandforeachTHthat
operatesinit.However,isanendeavorasresearchactivitiesresult
inbothintangible(e.g.knowledge,experience)andtangible
out-puts(e.g.scientificarticles,products)andaccuratemeasurements
dependonmanypreconditions[27,28].
Ameasurethatisfrequentlyusedisscholarlyoutput[27],by
countingthenumberofpublisheddocuments.Thisproxycanbe
accessedviareadilyavailabledatasourcesfrompublishers,
jour-nals,citationindexingservices,andothersecondarydatasources.
ItisasimplemeasurethatcanbedetectedinternallybyeachTH,but
whichalsohasanexternalvalue:thearticleshavebeenpublished
andthereforerecognizedbythescientificcommunityassignificant
contributionstotheevolutionofscience.
Quantifyingresearchusingthenumberofpublishedarticlescan
thereforebethefirststeptomeasureresearchperformancebut
doesnotprovideanyinformationaboutthevalueandimpactof
thesepublishedworks.Othermetricsarethereforeneededto
mea-surethequalityoftheseworkstoprovidecontextandtheimpact
withintheresearchcommunity[29].Inthisperspective,citations
canbeareferenceelement,sinceitisanindirectpositive
evalua-tionthattheworkhasbeenreadandtakenintoconsideration-and
thereforementioned-bycolleagues.
Inordertoprovidecontextandassesstheimpactofscholarly
output,aquantitativemethodcanbeemployed,namelya
biblio-graphicanalysis[30].Nowadays,variouscommonlyusedmetrics
canbeusedtoassess valuei.e. downloadsand views,citations,
impactfactor,h-indexandfield-weightedcitationimpact(FWCI),
altmetrics (storage,links, bookmarks, conversations) and many
others[31–34].Thesemetricscanbeutilizedbothonajournalor
individualresearcherlevel.Mostbibliometricsarecalculated,
man-aged,andtrackedusingcitationindexservicessuchasScopus,and
ISIWebofScience(WoS)andcanbeaccessedviatheInternet.
Thesedifferentcitationindexservicesareanaccesspointto
differentrepositorieswhichstoreandcategorizescholarlyoutput.
However,eachoftheseservicesdiffersintheircoverage,method
of tracking, and available metrics [35]. Since data provided by
theseservicesaremanagedandcontrolledbyexternalparties,new
opportunitiesandchallengesariseforPESswhenusedtoevaluate
andbenchmarkperformance.
• What are the main differences between the commonly used
scholarly metrics extracted from citation index services and
derivedusingbibliographicanalyses?
• What are the opportunities and risks when PESs and their
correspondinggovernancetools arerelying ondatacollected,
controlled,andvalidatedbyexternalsources?
• Whatistheimpactintermsofhealthpoliciesatasystemlevel?
Thispaper,therefore,describesthedifferencesbyfocusingon
thequantityandqualityofresearchperformanceofTHsthatresult
fromtheuseoftwodifferentscientificcitationindexingservices
ontheweb,ScopusandISIWebofScienceCoreCollection,by
per-formingabibliometricanalysisof26THsinItaly.Moreover,results
arecontextualizedbydiscussingthepossibleimplicationsonPESs
whenmeasuringresearchperformancethroughalternative
exter-nalcitationindexservices.
Thenextsectionconceptualizesdifferentmetricsusedto
eval-uatescientificproductionprovidedbybothScopusandWoSand
adescriptionoftheimplementationofresearchperformancein
aregionalperformanceevaluationsysteminItaly.Thethird
sec-tiondescribesthemethodologyandcomparingmetricsfromboth
services.Findingsarecontextualizedin thefinaldiscussionand
conclusionsections.
1.1. Measuringscientificperformance
Asearlyasin1927Grossetal.identifiedtheproblemof
dissem-inatingliteratureandin1955EugeneGarfieldproposedtoutilize
a citationindex forscientific literatureto eliminatethe
uncrit-icalcitationof fraudulent,incomplete,or obsoletedata [31,36].
Later,in 1961,asa founderof theInstitutefor Scientific
Infor-mation(ISI)GarfieldlaunchedtheScienceCitationIndexasatool
forresearchers,librariansandscholarstomanagethelarge
num-beroflibrarycollections.Overtime, thepurposeofthecitation
indexchanged,nowknownasimpactfactorwhichwasintended
todescribejournalimpactbasedonthenumberofcitations[37].
Althoughthismetricwasneverdesignedorintendedtobeusedas
anevaluationindicator,inpracticeitisoftenusedtoindicatethe
qualityofindividualscientificwork[38].Thescientificcommunity
hasoftenexpressedconcernsregardingthebiasedimpactfactorof
journalsderivingfromasymmetric-leftskewed-distributionof
papercitations[39,40]andcausingquitesomecontroversywithin
theresearchcommunity[38,41–43].
Althoughmanyalternativesevaluationmetricshavebeen
pro-posedbytheresearchcommunity,impactfactorremainsdominant
inusage.However,Larivièreetal.proposedasimpleandrobust
methodologytodeferthecitationdistributionsthatunderliethe
pro-posedmethodwasadoptedbytheJournalCitationReports(JCR)
in2018andseemstobeanearnestfirstattempttoaddressthe
concernsofthescientificcommunity[45].
Anotherwell-knownproblemwithimpactfactoristhe
skew-nessinspecificresearchfields.Forexample,Narinetal.reported
thatresearchinbiochemistry andmolecularbiologywerecited
aboutfivetimes asoftenaspharmacy articles[46].In orderto
correctforthisphenomenon,theDutchpublisherElsevier
imple-mented their own metric, namely the field-weighted citation
impactmetric. TheFWCIshowshowthenumberofcitationsof
asinglepapercompareswiththeaveragenumberofcitationsby
similarpublicationsindexedinScopus[47]resolvingtheissueof
differentresearchbehavioracrossdisciplines.
AnotherwellknowmetricgrouparetheAltmetricswhichuse
multipledatasourcessuchassocialmedia,numberofreadingsand
downloadstoassesstheimpactofthepaperbothinsideandoutside
thescientificcommunity.Theabilitytomeasureimpactofscientific
workoutsidethescientificcommunityisavaluabletrait.However,
theactualuseofAltmetricsneedstobefurtherconceptualizedto
becomeametriconitsownwhilestilllackingacleardefinition,an
ever-evolvingframework,lowdatatransparency,andorigin[48].
Evaluatingthelargebodyofavailablebibliography,itbecomes
clearthatitisnotpossibletomeasurescientificperformanceby
simplyusingonemeasureandwhileothermeasuressuchasthe
h-indexandg-indexareincreasinglyusedtoevaluateresearchers’
performance,usingdifferentmetrics,emphasizingboth
productiv-ity,qualityandcontext,insideandoutsidetheresearchcommunity
isimperative.
1.2. EvaluatingresearchperformanceinItaly
The Italian National Healthcare System follows a Beveridge
model,mainlyfinancedthroughgeneraltaxationandbasedonthe
principleofuniversalcoverage.Resourcesarecollectedatanational
levelandallocatedtothetwentyregionsonage-adjustedpercapita
basis.Theresponsibilityfortheorganizationandprovisionofcare
hasbeendecentralizedat a regionallevel, and regions allocate
resourcestoLocalHealthAuthorities(LHAs)whoareresponsible
forthedeliveryofallhealthcareservicesintheirgeographicalarea,
directlythroughpublicprovidersoraccreditedprivatehospitals.
THsareautonomousbodiesfromtheLHA,canbepublicorprivate,
andareusuallymanagedjointlybytheregionaladministrationand
auniversity.Thissharedresponsibilityinmanaginghasarelevant
impactontheirorganizationalcultureastheycanbeconsidered
doubleprofessionalbureaucracies[49].Withintheregional
health-caresystem,theyplayarelevantrolebecausetheyoverseetraining
offuturehealthprofessionalsandbecausetheyareinchargeof
leadinginnovationprocessesbasedontheresearchactivitiesthat
theycarryout.
Starting from2005, theManagementand Health Lab ofthe
Scuola Superiore Sant’Anna has developed a multidimensional
healthcareperformanceevaluationsystem(PES),initiallyadopted
by Tuscany’s regional administration. Over time, the PES was
adoptedbyanincreasingnumberofregionsandin2008the
Net-workofItalianRegionswasformed.TheNetworkexpandedand
nowadaystwelveregionshaveadoptedthesame
Inter-Regional-Performance Evolutionsystem (IRPES) [13] tobenchmark their
performanceusingmorethanthreehundredsharedindicators.In
2014,theNetworkofTHswasfounded,aimedatbenchmarkingTHs
performanceusingaPEStotakeintoaccountthespecific
charac-teristicsandmissionofTHswithintheregionalhealthcaresystem
consideringaroundsixtyindicators[16,50,51].
Reportingontheseindicatorsareconsideredanimportant
man-agementtoolbyallTHsandtheItaliangovernmentandregionsuse
themforseveralissuesasmonitoringandassessingperformance,
allocatefinancialresourcesforresearch,andalsotoevaluatethe
Table1
IndicatorsincludedintheIRPESregardingresearchevaluationofteachinghospitals.
Number Descriptionofindicator
1 Averageimpactfactorperphysician 2 Averagenumberofpublicationsperphysician 3 Percentageofpublicationswithanaverageimpact
factorhigherthanthebenchmarkspecialtyimpact factorreportedinISI
4 Percentageofpublicationswithamedianimpact factorhigherthanthebenchmarkspecialtyimpact factorreportedinISI
5 Medianimpactfactorperspecialty 6 Medianimpactfactorvariationperspecialty
GeneralManagers’performance[52].Giventhatresearchisoneof
thethreepillarsofTHs’mission,withinthesesixtyindicatorssome
arefocusedonresearchperformance.
Table1providesanoverviewoftheindicatorsincludedinthe
IRPESwhichareusedtoevaluateresearchactivities.
2. Methods
The bibliometricanalysis canbe performedusing two
well-knownpubliclyaccessiblecitationindexingservices,namely,ISI
webofknowledgeandScopus.Metricsaboutthescholarlyoutput
canbeextractedfromtheserepositoriesbysimplyprovidingauthor
firstname,lastname,andoptionallytheiraffiliatedorganization.
Thepossibilityofusingthesesearchengines,externaltothe
internal detection systems, has always been perceived as an
opportunitytohave ¨certain ¨andvalidateddata,afundamental
char-acteristictoguaranteethestrength,rigorandreputationofthePES
itself.
However,eventhesesystemsshowsomecriticalities.For
exam-ple,whenan authoris affiliatedwithmultipleorganizationsor
duplicatenamesareaffiliatedwiththesameorganizations
man-ualcorrectionisrequired.Dataextraction,articlede-duplication
of(co-)authors,andvalidationweredoneintwodifferent
man-ners for both repositories including all scientific documents in
thesedatabaseswhichhavebeenpublishedbetween2014−2016.A
detailedprotocolofthedataextractioncanbefoundas
supplemen-tarymaterial;AppendixA(inSupplementarymaterial)–Extraction
ofdata.THswereresponsibleforprovidingnamesofresearchers
affiliatedwiththeirorganization.
2.1. DataextractionWoS
DocumentspublishedinISIWebofSciencebetween2014–2016
were extracted and validated by an external affiliated party,
ResearchValueSRL,inMay2018.Datawassentto
correspond-ingauthorsforinternalvalidationoncompletenessusingrandom
samplingmethods.Allavailablemetricswereextractedfrom
ISI-WoS,including;title,discipline,documenttype,affiliatedauthors,
ISSN,ISBN,year,edition,pagenumbers,subjectcategories,DOI,
PubMedidentificationnumber,numberofcitationsandjournal’s
impactfactor.
2.2. DataextractionScopus
Scholarlyoutputproductionbetween2014−2016inScopuswas
extractedutilizing internallywrittenscriptsusingElsevier’sAPI
developers’program.Thescriptwasdividedintotwomain
func-tionalities,MatchandExtract,andwasexecutedinDecember2018.
ThefirstpartqueriestheScopusdatabaseauthornamestoobtain
auniqueidentificationcodeusedinallElsevier’sproductssuch
asScopusandScival.Onlywhenauniquematchwasfoundbased
Table2
NumberofpublisheddocumentsinbothScopusandWoSdatabasewiththeirrespectivedifferenceandchangeinrankingwhenbenchmarkedperTH.
Teachinghospital Numberof physicians Numberof documents Scopus Numberof documentsWoS Differencein documentsn(%) Differencein positionwhen benchmarked AOPadova 854 5854 3579 2275(38,9%) 1 AOUBologna 868 4605 2727 1878(40,8%) 1 AOUCareggi 1026 4343 2684 1659(38,2%) 1 S.Raffaele-MI 526 4140 3583 557(13,5%) -3 AOUVerona 876 3910 2247 1663(42,5%) 1 Fondaz.IRCCSCaGranda 780 3818 2470 1348(35,3%) -1 IRCCSS.Martino 862 3334 2223 1111(33,3%) 0 AOUPisana 952 3295 2222 1073(32,6%) 0
P.O.SpedaliCiviliBrescia 1039 3184 1766 1418(44,5%) 0
Ist.Clin.Humanitas-Rozzano 679 2928 1305 1623(55,4%) 4
AOUPol.Bari 895 2878 1708 1170(40,7%) -1
IRCCSPoliclinicoSanMatteo 571 2304 1485 819(35,5%) -1
AOUParma 696 2208 1374 834(37,8%) -1
AOUOsp.Riun.Ancona 758 2049 1184 865(42,2%) 3
Osp.S.Gerardo-Monza 735 1961 871 1090(55,6%) 5
AOUSenese 573 1958 1309 649(33,1%) -3
AOPerugia 576 1882 1304 578(30,7%) -2
AOUModena 489 1850 1199 651(35,2%) -2
Osp.L.Sacco-Milano 559 1623 859 764(47,1%) 2
ASUIUdine 745 1573 1017 556(35,3%) -1
AOUFerrara 523 1542 1020 522(33,9%) -3
Osp.S.Paolo-Milano 736 1441 742 699(48,5%) 1
Osp.diCircoloeFond.Macchi 589 1274 801 473(37,1%) -1
ASUITrieste 539 1165 640 525(45,1%) 0
OO.RR.Foggia 412 811 570 241(29,7%) 0
AOTerni 397 479 344 135(28,2%) 0
Total 18.255 66.409 41.233 25.176(37,9%)
searchresultedinmultiplepossibleauthorsamanualvalidation
wasdonebytheauthors,selecting,ormergingtheresearcher
pro-file(s).
Thesecondpart,extractspublishedworkfromScopusandScival.
AllavailablemetricswereextractedfromScopus,including;title,
DOI,ISSN,Journalname,typeofpublication,coverdata,number
ofcitations,affiliationorganization.AsScopusdoesnotallowto
trackanyvaluemetrics,theFWCIperauthorusingtheElsevier
identificationnumberwasextractedfromScival.
Adetailed description ofthe fullscript can befoundin the
supplementarymaterial;AppendixA(inSupplementarymaterial)–
Extractionofdata.FullscriptdetailsusedtoobtaindatafromScopus
andstatisticalprocedurescanberequestedviathecorresponding
author.StatisticalanalysiswasperformedusingRversion3.5.2.
3. Results
Afterextractingthescholarlyoutputofall26THs,atotal of
66.409and41.233documentsareincludedforanalysisfromScopus
andrespectivelyWoSfromatotalof18.255authors.Descriptive
statistics about the THs can be found in Appendix B (in
Sup-plementarymaterial)–DetailsTeachingHospitals.Documentsare
categorizedasarticles(69,2%Scopus;75,3%Wos),reviews(13,6
%Scopus;14,29%WoS),Letters(7,1%Scopus;9,4%Wos),
editori-als(1,89%Scopus),book(chapters)(2,97%Scopus)orother(5,2%
Scopus;1,0%WoS).Theprecedingtwocategoriesareonlyindexed
inScopus.Table2andFig.1comparethetotalnumberof
docu-mentsperTHinWoSandScopuspublishedbetween2014−2016.
Table3showsanoverviewofthetotalnumberofdocumentsper
THsbetweenWoSandScopusexcludingbook(chapters)and
edi-torialswhicharenotindexedinWoStoprovideamoreaccurate
comparison.
AWilcoxonsigned ranktest wasperformedtocomparethe
differenceofindexeddocumentsinbothdatabases,indicating a
significantdifference(p<0.005)inthedocumentsindexedin
Sco-pus(M=2.060,SD=1.122)andWoS(M=1.267,SD=863).When
rankingTHsbasedonthescholarlyoutputasshowninTable2,
almostallorganizationsarebenchmarkedatadifferentposition.
Onaverage,instituteschangetwopositionseitherpositiveor
neg-ative.Thebiggestpositivechangeintherankingwhenlookingat
ScopusistheTeachinghospital“AOSanGerardodiMonza”moving
fromthe20thpositionthe15thposition.
AlthoughthequalitymetricextractedfromScopusandWoS
can-notbe directlycompared witheach othersince WoSmeasures
impactfactorandScopusmeasuresFWCI,investigatingthe
qual-ityofthepublisheddocumentsshowsadifferenceinrankingwhen
benchmarked.AdetailedoverviewcanbefoundinTable4,showing
bothimpactfactor andFWCIoftheinstitutesandtheir
respec-tiverankingwhenbenchmarked.Onaverage,instituteschangefive
positionseitherpositiveornegative.Thebiggestpositivechangein
therankingcanbeseenwith“AOSanGerardodiMonza”moving
fromthe19thpositionthe3rdposition.However,some
organiza-tionsalsomovedownintheranking.AOUCareggiisplacedon18th
positionwhenrankingtheorganizationwithFWCIbutisranked7th
whenbenchmarkingwithimpactfactor.Noneoftheorganizations
remainatthesamepositionwhencomparingthebenchmarkon
ImpactfactororFWCI.
Finally,Fig.2showstherelationshipbetweenqualityand
quan-tity between the published works. Calculating the Spearman’s
rho shows a low but positive correlation between the quality
(FWCI)ofproduceddocumentsandthenumberofdocumentsper
researcher.
4. Discussion
Thispaperdescribesthedifferencesinperformanceof26THsin
Italybyfocusingonthequantityandqualityoftheirresearch.The
goalofthispaperwastoperformabibliometricanalysisfocusingon
twocommonlyusedperformancemetrics,impactfactorandFWCI
usingScopusandISIWebofScienceCoreCollectiontoidentifymain
differencesandpotentialopportunitiesandchallengesforPESsas
Fig.1.NumberofpublisheddocumentsinbothScopusandWoSdatabasewiththeirrespectivedifferencewhenbenchmarkedperTH.
Table3
NumberofpublisheddocumentsinbothScopusandWoSdatabasewiththeirrespectivedifferenceandchangeinrankingwhenbenchmarkedperTH,excludingbook(chapters) andeditorials.
ISIWoS Scopus(articles&reviews) Scopus(articles,reviews&conferencepapers) Scopus(alldocuments) Numberofpublisheddocuments(2014−2016) 41.233 55.805 56.962 66.409
Difference(%) 14.572(26,1%) 15.729(27,6%) 25.176(37,9%)
Table4
QualityofpublisheddocumentsinbothScopusandWoSdatabasewiththeirrankingandrespectivechangeinrankingwhenbenchmarkedperTH.Qualityisdefinedasthe averageimpactfactororField-weightedcitationimpactofallauthorsaffiliatedtotheTH.
Teachinghospitalname ImpactFactor Field-WeightedCitationImpact RankingScopus RankingWoS Differenceinpositionwhenbenchmarked
AOUBologna 13,49 2,66 1 4 3
Ist.Clin.Humanitas-Rozzano 10,66 2,54 2 9 7
Osp.S.Gerardo-Monza 5,84 2,38 3 19 16
S.Raffaele-MI 35,70 2,14 4 1 -3
AOPerugia 10,57 2,07 5 10 5
AOPadova 18,03 1,97 6 2 -4
AOUPisana 8,99 1,93 7 12 5
AOUModena 10,19 1,91 8 11 3
IRCCSPoliclinicoSanMatteo 12,06 1,88 9 5 -4
Fondaz.IRCCSCaGranda 14,27 1,85 10 3 -7
P.O.SpedaliCiviliBrescia 7,41 1,84 11 15 4
AOUSenese 8,68 1,82 12 13 1
AOUVerona 10,96 1,79 13 6 -7
AOUOsp.Riun.Ancona 6,23 1,78 14 18 4
AOUFerrara 7,42 1,77 15 14 -1 ASUIUdine 5,40 1,77 16 21 5 IRCCSS.Martino 10,68 1,73 17 8 -9 AOUCareggi 10,69 1,62 18 7 -11 AOTerni 3,52 1,60 19 24 5 OO.RR.Foggia 4,38 1,59 20 23 3
Osp.diCircoloeFond.Macchi 5,03 1,57 21 22 1
Osp.L.Sacco-Milano 5,67 1,56 22 20 -2
AOUParma 6,99 1,54 23 17 -6
Osp.S.Paolo-Milano 3,43 1,52 24 26 2
AOUPol.Bari 7,00 1,48 25 16 -9
ASUITrieste 3,51 1,33 26 25 -1
4.1. Bibliometricanalysis
Extractingthescholarlyoutputshowedasignificant
discrep-ancybetweentheextracteddatafromthetworepositories.When
extractingthefullscholarlyoutputofthe18.255authorsinthe
sample,Scopus resultedin 37,9 %moredocuments (66.409Vs.
41.233).Toacertaindegree,thisdifferencecanbeexplained.First,
apartfromreviews,articles,conferencepapers,bookchapters,
Sco-pusalsoindexesbooksandeditorialsintheirdatabase.WoSdoes
notincludethesetwo categoriesintotheircorecollection,thus
explaining9,7%ofthevariation.Second,Scopusisknowntobe
moreextensiveintheircoverageincludingover71millionrecords
andcoveringover23,700peer-reviewedjournals[53]whileWoS
includesjustover20.000peer-reviewedjournals[54].Itis,
there-fore,possiblethatsomearticlesarenotindexedinbothdatabases.
Third,sinceauthorsaresearchedusingonlyname,surname,and
affiliationand itispossiblethatauthorsareindexeddifferently
inbothdatabasesresultinginamismatchwhenextracting
infor-mation.However,atthisstage,weareunabletoprovideanexact
quantificationofthisobservedvariation.
Whencomparingthequalityindicatorofbothdatabasesand
benchmarking THs based on impact factor and FWCI, none of
theorganizationsremainatthesameposition.Interestingly,data
Fig.2. Correlationbetweenquality(FWCI)andpublisheddocumentsperresearcherinScopus.
workspublishedbytheTHs.Haslam&Laham,hypothesizedthat
researchersinmoreprestigiousinstitutionsmayfollowastrategy
wherethefocusismoreonthequalityofpublishedpapersandless
onthequantity[55].Ourresultscontradictthishypothesisproving
apositiverelationshipbetweenquantityandqualityofpublished
worksindicatingthatTHspublishingmoredocumentsalso
pro-ducehigherqualitydocuments.However,itshouldbenotedthat
withthecurrentnumberofobservationstherelationshipisweak
andmightnotsustainwithanincreasedsamplesize.
Archambaultetal.,2009providesevidencethatindicatorsof
scholarlyproductionandcitationsatthecountrylevelarestable
and largely independent of the database reported and no
sig-nificantbibliographicdifferences betweenScopus and WoSare
found [35]. We were able to compare results on the
individ-ualresearcher,nowsuggestingthatasignificantdifferenceexists
betweenbothrepositories,rejectingthefindingsofArchambault.
However,Archambaultwasunabletoinvestigatethescholarly
out-putonanindividuallevelandfocusedonaninstitutionallevel.
4.2. Impactonperformanceevaluationsystems
Thepresentedfindingscanhaveimportantimplicationsforthe
currentuseofperformanceevaluationsystemsinthehealthcare
sector.
IntraditionalPESsdataaremeasured,calculatedandvalidated
bytheorganizationsthemselves,usingbenchmarkinginorderto
compareresultswithothers,ondifferentlevelssuchas
individu-als,departments,andorganizations[14].Usingsecondarybigdata
sourcesopensnewopportunitiestobenchmarkoutsidethe
orga-nizationalboundarieswithotherorganizationsonanationaland
internationallevel.
ThisreducestheroleofeachsingleTHinthecollectionofdata
andreassurestheRegional orNationalHealthSystemaboutthe
reliabilityofthedataitself,asthereisareducedriskof
opportunis-ticdatamanipulation.Thebenchmarkingprocess,atafirstglance,
appearsmorerobust.
Servicestoconsultbibliographicinformationarepublicly
avail-ableandeasilyaccessibleviatheInternet.However,datainthese
systemsarenotmanaged,owned,andoftennotvalidatedbythe
organizationthemselves,butbyexternalpartiessuchasClarivate
Analyticsand Elsevierwhichhavepartly acommercialinterest.
Especiallysincenumerousstudieshaveprovidedevidenceabout
inaccurate information, falsification, and fabrication of data in
citationindexserviceswhichaffectandinfluencethe
bibliomet-ricmeasures[56–59]. Additionally,metricsmeasuringthesame
constructnamelyqualityoftendifferfromeachotherandareall
subjectedtotheirownadvantagesanddisadvantagesmaking
com-parisonschallenging.
Theseindicatorsareanimportantmanagement toolusedby
theItaliangovernment,theRegionsandtheTHs.Theyusethem
for monitoring and assessing performance, allocating financial
resources for research, and evaluating the General Managers’
performance.We want tounderlinethatchoosing oneofthese
databasesisnotsufficient norreliabletobaseimportanthealth
policydecisionsonwithoutincludingcontextualinformation.
Thedifferencesbetweenthetwometricsfoundintheresults,in
fact,highlighttheintrinsicweaknessofthesemetricswhich,tobe
effective,requireasignificantworktocriticallyassessthemeaning
usingcontextualinformation.Validationoftheoriginofthemetric
isakeystepintheageofBigDatabeforeassessingthemeaningof
themetricitself.Moreover,increasingthescopeofbenchmarked
organizationsprovidesnewinsightstopolicymakers,andcan
sup-port beneficialstrategies when using PESssuchas namingand
shaming[60]orrewardingorganizationsforhigherperformance
[61].Thisgoalcanonlybeachievedifdataarereliable.Thefact
thattheresearchindicatorsarebasedonsystemssuchasWoSand
Scopusdoesnotguaranteepersethepursuitofthiscondition.
Healthsystemsmustaccompanytheuseofthesemetricswith
acontinuoussharingprocesswithallthestakeholdersofthe
sys-temandfirstofallwiththeresearchersthemselves[62].Thissame
sharingprocessrepresentsthefirstmechanismtoaligneffortsand
commitmenttowardspursuingtheoverallmissionofthe
health-caresystemanditisthebasisoftherelationshipoftrustandesteem
thatallowstofeedandpromoteimprovementprocesses.
Finally, other several issues should be mentioned possibly
influencingthepresentedresults.First,althougharepresentative
samplesizeof 18.255authorswasused,authorswerenotable
tovalidateeachindividualresearcher.Namesofresearcherswere
providedbyallTHsintheIRPESnetwork,butauthorswerenotable
tovalidatetowhatextenttheseresearcherswereactivelyworking
fortheTHsorprovideanydescriptivestatisticsabouttheseauthors.
Second,datafromWoSwasextractedandprimarilyvalidatedbyan
externalcommercialparty,ResearchValueSRL.Duetocommercial
interests’authorswereunabletoassesstheextractionprocedure
tovalidateaccuracy.Authorswereabletovalidatetheextraction
fromScopusbyaccessingthedevelopers’platformfromElsevier.
Sincetheauthorswerenotabletocomparetheextractionaccuracy
itispossiblethatthedifferenceinthenumberofarticlesfrom
Sco-pusisattributedtoahigheraccuracywhenqueryingScopus.Third,
thescholarlyoutputfromWoSwasextractedinMay2018and
Sco-pus9monthslater.Theeffectonthenumberofpaperswouldbe
thisdelaysinceimpactfactorandFWCIrelyonthetotalnumberof
citations.Itis,therefore,possiblethattheFWCIispositivelyskewed
comparedwithIF.
Futureresearchshouldaddressthetopicsmentionedaboveby
aligningtheextractionmethodofWoSandScopusandperform
extractionsimultaneously.Additionally,byexpandingtheIRPES
moreTHscanbeincludedtoimprovegeneralizabilityonanational
level.Next,adetailedstudyshouldbeperformedtoanalyzeeach
documenttypeseparately,sincereviewshave,ingeneral,ahigher
impactthanmostotherdocumenttypessuchasarticles,letters,
notes[41].Finally,otherqualitymetricscanbeincludedintothe
analysistofurthercontextualizeFWCIwithotherqualityindicators
bylookingat,butnotlimitedto,cross-checkinggrants,
collabo-rationswithotherresearchinstitutes,andpercentage ofpapers
publishintop5percentilejournals.
5. Conclusion
Toourknowledge,nopriorresearchwasperformedto
iden-tifyandhighlightthedifferencesofresearchperformanceofTHs
withrespecttoquantityandqualitymetricsusingtheirpublished
workswhileincludingalargesampleofindividualphysicians.
Uti-lizingsecondary BigDatasourcesforperformance management
is,ontheonehand,usefulbecausetheyallowbenchmarkingat
anational andinternationallevel,but ontheotherhand,using
differentdatasources tomeasurethesameconstructofquality
andquantity,clearlyleadtodifferentresultswhenbenchmarked
againsteachother.
Researchactivitiesareanobjectivetobepursuedandispart
ofthemissionofboththeHealthcareSystemasawholeandthe
providerswhooperatewithintheSystem.Amongtheproviders,
inthefirstplacetherearetheTHs,withtheirtriple-foldmission
ofresearch,careandtraining.Followingtheirmission,THshavean
intrinsicmotivationtodeliverhighperformanceonallthreepillars.
Measuringtheperformanceoftheresearchactivitiesisessential
butcomplex.Web-basedtoolsallowtoensureabenchmarking
pro-cessondifferentlevelsandcanbeeffectivelyusedataHealthcare
Systemlevelfordifferentgovernancepurposessuchasplanning,
designingincentivesforresearch,andallocatingresources.
Web-basedtoolshaveweaknessesandrequireaformalinternaldataand
validationprocesstoavoidself-referral.Thiscanbeovercomeby
settingupatransparentprocesssharedamonghealth
profession-als,hospitalmanagementandpolicymakers,whichcancontribute
andinturnimproveresearchperformance.
Ethicsapproval
NotApplicable.
Consenttoparticipate
Participating teaching hospitals in the network of
measur-ingperformanceprovidedauthorswiththeiraffiliatedemployed
researchers.Final analysis wasperformed onan organizational
levelandemployedresearcherswerenotinvolved,contactedor
analyzedonanindividuallevelinanywayduringthestudy.
Consentforpublication
Responsible region representatives have approved the final
results.
Availabilityofdataandmaterial
Thedatasets,scriptsoranyothersupplementarymaterialused
andanalyzedduringthecurrentstudyareavailablefromthe
corre-spondingauthoronreasonablerequest.DataobtainedfromSciVal®
database,ElsevierB.V.,http://www.scival.com
Funding
FH is working as a fellow in a project
(www.healthPros-h2020.eu)thathasreceivedfundingfromtheEuropeanUnion’s
Horizon2020researchandinnovationprogrammeundertheMarie
Skłodowska-CuriegrantagreementNo765141.Theoverallproject
ispartlyfinancedbyItalianregionswithintheIRPES.
Authors’contributions
StudyconceptionwascreatedbySN;studydesignwascreated
byFHandDAL.AcquisitionofdatawasperformedbyFH,DAL.
Anal-ysisandinterpretationofdatawasperformedbyFH.Draftingof
themanuscriptwasperformedbyFHandDAL.SNwasinvolved
incriticalrevisionsofthemanuscriptandcontributedinwriting
thebackground,discussionandconclusionparagraphs.Allauthors
havereadandapprovedthesubmittedmanuscript.
DeclarationofCompetingInterest
Theauthorsreportnodeclarationsofinterest.
Acknowledgments
Authorswouldliketothanktheparticipationoftheregional
networkinprovidinguswithinputfordatacollection.Thispaper
ispartofaproject(www.healthpros-h2020.eu)thathasreceived
fundingfromthe EuropeanUnion’sHorizon 2020researchand
innovationprogrammeundertheMarieSkłodowska-Curiegrant
agreementNo765141.
AppendixA. Supplementarydata
Supplementarymaterialrelatedtothisarticlecanbefound,in
theonlineversion,atdoi:https://doi.org/10.1016/j.healthpol.2020.
10.002.
References
[1]MarrB,GrayD.Strategicperformancemanagement.TaylorFrancis;2006,240 p.
[2]MoullinM.Performancemeasurementdefinitions:linkingperformance mea-surementandorganisationalexcellence.InternationalJournalofHealthCare QualityAssurance2007;20(3):181–3.
[3]MoullinM.Deliveringexcellenceinhealthandsocialcare:quality,excellence, andperformancemeasurement.OpenUniversityPress;2002.
[4]Bourne M, Mills J. Designing, implementing and updating performance measurementsystems.InternationalJournalofOperationsProduction& Man-agement2000;20(7):754–71.
[5]KaplanD,RobertS,NortonD.Thebalancedscorecard:translatingstrategyinto action.Boston:HarvardBusinessSchoolPress;1996.
[6]NeelyA. Theperformancemeasurementrevolution:why nowand what next? International Journal of Operations Production & Management 1999;19(2):205–28.
[7]KaleS,TamakuwalaH,VijayakumarV,YangL,RawalKshatriyaBS.Bigdatain healthcare:challengesandpromise.In:Smartinnovation,systemsand tech-nologies.Springer;2020.p.3–17.
[8]NgiamKY,KhorIW.Bigdataandmachinelearningalgorithmsforhealth-care delivery.LancetOncology2019;20:e262–73.LancetPublishingGroup.
[9]PorterME.Whatisvalueinhealthcare?NewEnglandJournalofMedicine [Internet]2010;363(26):2477–81,http://dx.doi.org/10.1056/NEJMp1011024. Dec23[cited2019Sep19];Availablefrom:.
[10]GrayM.Populationhealthcare:designingpopulation-basedsystems.Journal ofRoyalSocietyMedicine[Internet]2017;110(5):183–7,http://dx.doi.org/10. 1177/0141076817703028.May12[cited2019Jul3];Availablefrom:. [11]BerwickDM,NolanTW,WhittingtonJ.Thetripleaim:care,health,andcost.
HealthAffairs2008;27(3):759–69.
[12]NutiS,NotoG,VolaF,VainieriM.Let’splaythepatientsmusicsystemsin healthcare.ManagementDecision2018;56(10):2252–72.
[13]NutiS.Makinggovernanceworkinthehealthcaresector:evidencefroma ‘naturalexperiment’inItaly.PolicyLaw2016;11(February2015):17–38.
[14]BititciU,GarengoP,NudurupatiS.Performancemeasurement:challengesfor tomorrow.InternationalJournalofManagementReviews2012;14:305–27.
[15]BevanG.Reputationscount:whybenchmarkingperformanceisimproving healthcareacrosstheworld.HealEconomicsPolicyLaw2019;(14):141–61.
[16]NutiS,RuggieriT,PodettiS.Douniversityhospitalsperformbetterthan generalhospitals?AcomparativeanalysisamongItalianregions.BMJOpen 2016;6(011426).
[17]PrasadV,GoldsteinJA.USnewsandworldreportcancerhospitalrankings:do theyreflectmeasuresofresearchproductivity?PLoSOne2014;9(9):1–6.
[18]KrzyzanowskaMK,KaplanR,SullivanR.Howmayclinicalresearchimprove healthcare:outcomes?AnnalsOncology2011;22(Suppl.7):10–5.
[19]MajumdarSR,ChangWC,ArmstrongPW.Dotheinvestigativesitesthattake partinapositiveclinicaltrialtranslatethatevidenceintopractice?American JournalofMedicine2002;113(November(2)):140–5.
[20]KanavosP,SullivanR,LewisonG,SchurerW,EckhouseS,VlachopiotiZ.The roleoffundingandpoliciesoninnovationincancerdrugdevelopment. Ecan-cermedicalscience2010;Vol.4:1–139.
[23]JanniW,KiechleM,SommerH,RackB,GaugerK,HeinrigsM,etal.Study participationimprovestreatmentstrategiesandindividualpatientcarein par-ticipatingcenters.AnticancerResearch2006;26(September(5B)):3661–7.
[24]Birkmeyer JD,Stukel TA, SiewersAE, GoodneyPP, Wennberg DE, Lucas FL. Surgeonvolume and operative mortality in the United States. New EnglandJournalofMedicine[Internet]2003;349(November(22)):2117–27,
http://dx.doi.org/10.1056/NEJMsa035205 [cited 2020 May 9]; Available from:.
[25]BirkmeyerJD,DimickJB,BirkmeyerNJO.Measuringthequalityofsurgicalcare: structure,process,oroutcomes?11Nocompetinginterestsdeclared.Journalof AmericanCollegeSurgery2004;198(April(4)):626–32.
[26]duBoisA,RochonJ,PfistererJ,HoskinsWJ.Variationsininstitutional infrastruc-ture,physicianspecializationandexperience,andoutcomeinovariancancer: asystematicreview.GynecologyOncology2009;112:422–36.
[27]AbramoG,D’AngeloCA.Howdoyoudefineandmeasureresearchproductivity? Scientometrics2014;101(2):1129–44.
[28]KreimanG,MaunsellJHR.Ninecriteriaforameasureofscientificoutput.Front ComputingNeuroscience2011;5(48):1–6.
[29]DewettT,DenisiA.Exploringscholarlyreputationit’smorethanjust produc-tivity.Scientometrics2004;60(2):249–72.
[30]Broadus RN. Toward a definition of “bibliometrics”. Scientometrics 1987;12(5–6):373–9.
[31]Gross EM. College libraries and chemical education. Science (80-) 1927;66(1713):385–9.
[32]SternbergRJ,StatesU.Journalofappliedresearchinmemoryandcognition evaluatingmeritamongscientists.JournalofAppliedResearchMemoir Cogni-tive2018;7(2):209–16.
[33]GrechV.Increasingimportanceofresearchmetrics:journalImpactFactorand h-indexH-index.InternationalUrogynecologicalAssociation2018;29:619–20.
[34]CockrielWM,McdonaldJB.Theinfluenceofdispersiononjournalimpact mea-sures.Scientometrics2018;116(1):609–22.
[35]ArchambaultE,CampbellD,GingrasY,LarivièreV.Comparingbibliometric statisticsobtainedfromthewebofscienceandScopus.JournalofAmerican SocietyInformationScienceTechnology2009;60(7):1320–6.
[36]GarfieldE.Citationindexesforscience.Science(80-)1955;6:31–2.
[37]GarfieldE.Thehistoryandmeaningofthejournalimpactfactor.Journalof AmericanMedicalAssociation2006;295(1):1–4.
[38]GarfieldE.Howcanimpactfactorsbeimproved?BMJ1996;313:411–3.
[39]MutzR,DanielH.Skewedcitationdistributionsandbiasfactors:solutions totwocoreproblemswiththejournalimpactfactor.JournalofInformatics 2012;6(2):169–76.
[40]SeglenP.Theskewnessofscience.JournalofAmericanSocietyInformation Science1992;43(9):628–38.
[41]VanLeeuwenT,MoedHF,ReedijkJ.CriticalcommentsonInstitutefor Scien-tificInformationimpactfactors:asampleofinorganicmolecularchemistry journals.JournalofInformationScience1999;25(6):489–98.
[42]SeglenPO.Whytheimpactfactorofjournalsshouldnotbeusedforevaluating research.BMJ1997;314:498–513.
[43]AlbertsB.Impactfactordistortions.Science(80-)2013;340(6134):787.
[44]LariviereV,KiermerVV,MacCallumCJ,McnuttM,PattersonM,PulvererB,etal. Asimpleproposalforthepublicationofjournalcitationdistributions.bioRxiv 2016:1–23.
[45]MinnickJ.The2018JCRishere!Clarivateanalytics;2018.
[46]NarinF, HamiltonSK.Bibliometricperformancemeasures.Scientometrics 1996;36(3):293–310.
[47]Researchmetricsguidebook.Elsevier;2018.
[48]HausteinS.Grandchallengesinaltmetrics:heterogeneity,dataqualityand dependencies.Scientometrics2016;108(1):413–23.
[49]MintzbergH,UpSaddleRiverStructureinFives;DesigninEffective Organiza-tions;1992.
[50]NutiS,D’OrioG,GambaMP.Ilsistemadivalutazionedellaperformancedei sistemisanitariregionali;IrisultatidelleAziendeOspedaliero-Universitariea confronto;2017.
[51]NutiS,RuggieriT.LavalutazionedellaperformancedelleAziende Ospedaliero-Universitarie.Finalità,metodierisultatiaconfronto2016:109.
[52]VainieriM,VolaF,GomezG,NutiS.Howtosetchallenginggoalsandconduct fairevaluationinregionalpublichealthsystems.InsightsfromValenciaand TuscanyRegions.HealthPolicy(NewYork)2016;120(11):1270–8.
[53]Aneyeonglobalresearch.Elsevier;2018.
[54]CarloniM.Webofsciencecorecollectiondescriptivedocument;2018.
[55]HaslamN,LahamSM.Quality,quantity,andimpactinacademicpublication. EuropeanJournalofSocietyPsychology2010;40:216–20.
[56]FranceschiniF,MaisanoD,MastrogiacomoL.Empiricalanalysisand classifica-tionofdatabaseerrorsinScopusandWebofScience.JournalofInformatics 2016;10(4):933–53.
[57]FranceschiniF,MaisanoD.Themuseumoferrors/horrorsinScopus.Journal ofInformatics2016;10(1):174–82.
[58]López-cózarED,Robinson-garcíaN,Torres-salinasD.Thegooglescholar exper-iment:howtoindexfalsepapersandmanipulatebibliometricindicators. JournalofAssociationInformationScience&Technology2014;65(3):446–54.
[59]Bartneck C,KokkelmansS.Detectingh-indexmanipulationthrough self-citationanalysis.Scientometrics2011;87:85–98.
[60]BevanG,WilsonD.Does‘namingandshaming’workforschoolsandhospitals? LessonsfromnaturalexperimentsfollowingdevolutioninEnglandandWales. PublicMoneyManagement[Internet]2013;33(July(4)):245–52,http://dx.doi. org/10.1080/09540962.2013.799801[cited2019Jul26];Availablefrom:. [61]VainieriM,LunguDA,NutiS.Insightsontheeffectivenessofrewardschemes
from10-yearlongitudinalcasestudiesin2Italianregions.International Jour-nalofHealPlanManagement2018;33(2):474–84.
[62]NutiS,BiniB,RuggieriTG,PiaggesiA,RicciL,GrilloRuggieriTG,etal.Bridging thegapbetweentheoryandpracticeinintegratedcare:thecaseofthediabetic footpathwayinTuscany.InternationalJournalofIntegrativeCare2016;16(May (2)):9.