An economic evaluation of perioperative adverse events associated with spinal surgery

2012OutstandingPaper:Value

Aneconomicevaluationofperioperativeadverseeventsassociated

withspinalsurgery

ErikK.Hellsten,BAa,b,MichelleA.Hanbidge,BEScc,AspasiaN.Manos,BSca,

StephenJ.Lewis,MD,FRCSCd,e,EricM.Massicotte,MD,FRCSCe,f,MichaelG.Fehlings,MD,PhD,FRCSCe,f,PeterC.Coyte,PhDa,

Y.RajaRampersaud,MD,FRCSCd,e,*

InstituteofHealthPolicy,Management,andEvaluation,UniversityofToronto,HealthSciencesBuilding,155CollegeSt,Suite425,

Toronto,ONM5T3M6,Canada

bHealthQualityOntario,130BloorStWest,10thfloor,Toronto,ONM5S1N5,Canada

cDepartmentofElectricalandComputerEngineering,UniversityofToronto,10King’sCollegeRd,RoomSFB0,Toronto,ONM5S3G4,CanadadDivisionofOrthopaedicSurgery,DepartmentofSurgery,UniversityofToronto,100CollegeSt,Room302,Toronto,ONM5G1L5,CanadaeUniversityHealthNetwork,KrembilNeuroscienceCentre,SpineProgram,TorontoWesternHospital,399BathurstSt,Toronto,ONM5T2S8,Canada

fDivisionofNeurosurgery,DepartmentofSurgery,TorontoWesternHospital,399BathurstSt,Toronto,ONM5T2S8,Canada

Received2February2012;revised27November2012;accepted8January2013

aAbstract

BACKGROUNDCONTEXT:Besidestheirclinicalimpact,theeconomicimpactofhealthcare–relatedadverseevents(AEs)issignificant.AlthoughanumberofstudieshaveattemptedtoestimatetheeconomicimpactofAEs,fewhavedirectlylinkedcoststoclinician-reportedeventseverity.PURPOSE:Toestimatetheeconomicimpactintermsoftheincrementalcostandlengthofstay(LOS),attributabletodifferentseveritygradesofAEsthatoccurredduringperioperativespinalsurgery.

STUDYDESIGN:Healtheconomicevaluationofdatafromaprospectiveobservationalstudyfromtheperspectiveofanacademichospital.

PATIENTSAMPLE:Consecutivepatientsatasingle,tertiary-quaternarycareinstitutionwhohaveundergoneinpatientspinalsurgery.

OUTCOMEMEASURES:ThecostandLOSimpactswithrespecttotheseverityoftheAEs.METHODS:Weanalyzed4yearsofpatientdischargesbetweenJanuary1,2007andDecember31,2010.TheSpineAdverseEventsSeverityinstrumentwascompletedbythesurgicalteamatdischarge.ClinicalimpactsoftheAEsweregradedasI(requiresno/minimaltreatment),II(re-quirestreatmentandisnotlikelytocauselong-term[O6months]sequelae),III(requirestreatmentandismostlikelytocauselong-termsequelae),andIV(death).Atotalof1,815recordswerelinkedwiththepatient-levelcostinginformation.WematchedeachAEcasewithfourcontrolcasesbasedontheirpropensityscorefortheriskofexperiencinganAE,regressedagainstcasecharacteristics.WeestimatedanincrementalcostandLOSforeachseveritygradebycalculatingthedifferencesinmeansacrosscasesandcontrols.Weconductedasensitivityanalysisbyestimatingthealternatemodelsusinggeneralizedlinearmodel(GLM)regressionwithagammaloglink.

RESULTS:Adverseeventswerereportedin316(17.4%)cases,with126ofthesepatients(40.2%)experiencingmultipleevents.Theincrementalcost/LOSforeachseveritygradeareasfollows:

FDAdrug/devicestatus:Notapplicable.

Authordisclosures:EKH:Nothingtodisclose.MAH:Nothingtodis-close.ANM:Nothingtodisclose.SJL:Consulting:Medtronic(C);Speak-ing/TeachingArrangements:Stryker(E);Trips/Travel:Baxter(B);FellowshipSupport:Medtronic,Synthes,DePuy(Groupsupport,paiddi-rectlytoinstitution).EMM:ConsultingfeeorHonorarium:AOSpineNorthAmerica(A);Other:AOSpineNorthAmerica(D,paiddirectlytoinstitution);Consulting:WatermarkResearchPartners(none);ScientificAdvisoryBoard/OtherOffice:CanadianNeuroscienceFederationSociety(none).MGF:Royalties:Depuy(D);Consulting:Covidien(B);1529-9430/$-seefrontmatterÓ2013ElsevierInc.Allrightsreserved.http://dx.doi.org/10.1016/j.spinee.2013.01.003

FellowshipSupport:Medtronic,Synthes,DePuy(Groupsupport,paiddi-rectlytoinstitution).PCC:Nothingtodisclose.YRR:Consulting:Med-tronic(E).

ThedisclosurekeycanbefoundontheTableofContentsandatwww.TheSpineJournalOnline.com.

TheauthorsMHandAMcontributedequally.

*Correspondingauthor.TorontoWesternHospital,UniversityHealthNetwork,399BathurstSt,EastWing,1-441,Toronto,Ontario,CanadaM5T-2S8.Tel.:(416)603-5399;fax:(416)603-3437.

E-mailaddress:raja.rampersaud@uhn.on.ca(Y.R.Rampersaud)

E.K.Hellstenetal./TheSpineJournal13(2013)44–5345

I5$4,224(p5.0351)/3.63days(p5.0001);II5$23,500(p!.0001)/14.03days(p!.0001);III5$147,285(p5.0036)/74.50days(p5.0018);andIV5$121,366(p5.0323)/46.44days(p5.0036).Thetotalcostinmillions/LOS(days)associatedwitheachgradeoverthe4-yearstudyperiodareasfollows:I5$0.66million/569.9days;II5$2.96million/1,767.8days;III5$4.27mil-lion/2,160.5days;andIV5$0.49million/185.8days.Oursensitivityanalysisproducedcomparableoverallresultsusingalternatemodelingtechniques.Overall,AEscontributedanestimated$8.38million(16.0%ofthetotalcostsforallpatientsinthesample)inincrementalcostsand4,684ad-ditionalbeddaysoverthe4-yearstudyperiod.

CONCLUSIONS:Inthissurgicalspinecohort,AEsaccountedfor16%ofthetotalcostofin-hospitalcare.HigherseverityAEswereprogressivelymorecostlyonaper-casebasis;however,themorefrequentlowerseverityevents(ie,GradeIandII)alsohadasubstantialaggregatecost(43%).TheseresultssuggestthatastrongbusinesscaseexistsforpatientsafetystrategiesfocusednotonlyonsevereAEsbutalsoonthereductionoflowerseverityeventsthatmaybemoreame-nabletopreventionefforts.Ó2013ElsevierInc.Allrightsreserved.

Keywords:

Health-Economicevaluation;Spinesurgery;Adverseevents

Introduction

Anemergingbranchofthehealthservicesliteraturefo-cusesontheeconomicevaluationofadverseevents(AEs)resultingfromhospitalcare.Large-scalestudiessuchastheInstituteofMedicine’sToErrisHuman[1]andZhanandMiller[2]havesucceededinattractingtheattentionofadministrators,policy-makers,andcliniciansalikethroughquantifyingthestaggeringeconomicimpactofmedicalerrorsonhealthsystemcostsandresources.Akeythrustofthisworkistheimperativetobuildabusinesscaseforpatientsafetyefforts;althoughethicalandprofes-sionaldutiescompelallhealthcareproviderstodoevery-thingintheirpowertopreventmedicalinjuries,economicrealitydictatesthatinvestmentstoimprovepatientsafetyaremorelikelytobefundedinareaswhereacompellingcaseforthefinancialreturnoninvestmentexists[3].Con-sequentially,itisimportanttobetterunderstandtheepide-miologyandtheeconomicconsequencesofAEstotargetimprovementstrategiesthatwillbothimprovethepatients’healthandreduceavoidablehealthsystemcosts.

Fortheaforementionedreasons,theassessmentandre-ductionofAEsarecurrentlyasignificantfocusofmanysurgicalsocietiesandinstitutions.Spinalsurgeryhastradi-tionallybeenperceivedasahighrisk;however,there-portedAErateassociatedwithspinalsurgeryisquitevariableandmostoftendependentonthenatureofthedi-agnosis,patientfactors,andthetypeand/ormagnitudeofthesurgery[4–8].Furthermore,therateofAEsismostaf-fectedbytheworkingdefinitionofanAEandthenatureinwhichitiscaptured(ie,prospectivelyvs.retrospectively)[9–12].Consequently,apples-to-applescomparisonofdifferentAEstudiesisdifficult.Furthermore,mostlarge-scalestudiesusingadministrativedatabaseslacktheclini-calgranularityrequiredtoapplyfindingstoaspecificpatientorsubgroup[2,13].Alternatively,mostclinicalstudiesonlyfocusononeparticulardiagnosticgroupand,thus,cannotbeappliedtothebroaderpopulationofspinesurgery[14,15].Althoughresearcheffortshaveproduced

validatedtoolsforthereporting,classification,andseveritygradingofperioperativeAEsoccurringduringthespinalsurgery[11,12],thesesystemsarejuststartingtobeutilizedinmultiplecentersorregistries[9],andtheyhavenotyetbeenlinkedwiththeanalysesonthecostimpactsofsuchevents.Classenetal.[16]andParadisetal.[17]havedem-onstratedthatsuchanalysiswaspossibleusingvoluntarilyreportedAEsdatafromabroaderrangeofpatienttypes.Acombinationofthesemethodsisrequiredtoenablemulti-centerdatacollectionthatisrelevanttoboththecliniciansandthepolicymakersengagedinimprovingpatientsafety.Weproposethatresearchtobetterunderstandandeval-uatetheeconomicimpactsrelatedtotheparticulartypesofAEsisessentialtoinformthetargetedqualityimprovementinitiativesthatwillyieldthegreatestimpactinbothreduc-inghospitalcostsandimprovingpatientsafetyandout-comes.Theincreasingrateanddiversityofspinalsurgerypresentsignificantopportunitiesforapplyingevidence-basedandcost-effectiveapproachestowardrealizingim-provedpatientsafetyandagreatervalueforthescarcehealthcaredollars[18].Thisarticlemakesanewcontribu-tiontothisfieldofstudybydeterminingtheper-caseandoverallin-hospitalincrementalcostsandthelengthofstay(LOS)impactofperioperativeAEsofvaryingclinicalse-verityinacomprehensivespinalsurgerypopulation.

MaterialsandmethodsStudydesign

Dataforthisstudywereobtainedfromanongoingpro-spectivestudyatalargeacademichealthsciencecenterlo-catedinToronto,Ontario.Thestudyincluded4yearsofspinalsurgerydischargesfromJanuary1,2007toDecem-ber31,2010,withprocedurescompletedbysixfellowship-trainedacademicspinesurgeons.AdverseeventswerecapturedusingthelocallydevelopedandvalidatedSpineAdverseEventsSeverity(SAVESv.1)[11]instrument

46E.K.Hellstenetal./TheSpineJournal13(2013)44–53

Context

Thefinancialimpactofadverseevents(AEs)followingspinalsurgerycanbesubstantial.Usingcarefullycol-lectedprospectivedata,theauthorsexaminedtheactualcostsinvolved.

Contribution

Overa4-yearperiod,thecostsduetoAEsfollowingspi-nalsurgeryatthissingleacademicinstitutionwereanestimated$8.38Mandaccountedfor16%oftotalcostsofinpatientcare.CostsofAEsincreasedinlinewithse-verityoftheevent.

Implication

Beyondsavingthepatientsandtheirfamiliestheemo-tionalandphysicaldistressofcomplications,thepreven-tionofAEsmakessolidfinancialsense.Morerecently,thesetypesofdataarebeingusedtosetpolicy.Partoftheintendedshiftistotiesurgeonsandinstitutionstotheriskofadverseoutcomesandremoveevenindirectfinancialbenefitsfromincreasedsurgicalcomplications.Whilefinancialdisincentivesmaydecreasethepotentialrewardassociatedwithahigh-volume,low-qualityprac-tice,thereareotherunintendedconsequencesthatshouldbeconsidered.Thesickerandsocioeconomicallydistressedpatientsinthefuturemayfinditevenhardertoreceiveneededsurgicalinterventions.

—TheEditors(seethefollowingsection).Theclinician-drivenvoluntaryreportingSAVES-AEformwascompletedbythesurgicalteamatdischargeonconsecutivepatientsundergoingawidevarietyofemergent,urgent,andelectiveinpatientspinalsurgeries.Unlessdictatedbyasignificantactivemedicalcomorbidityorsocialcircumstances,mostproceduressuchasdiscectomyanddecompressionareperformedonanout-patientbasis.TheseproceduresareassociatedwithaverylowAErateand,thus,wereexcludedfromthisstudy.Patientsample

Informationwasexaminedfor1,860inpatientspinalcases.Eachpatientrecordincludedinformationonpatient’sage,sex,admissionanddischargedates,admittingdiagno-sis,AmericanSocietyofAnesthesiology(ASA)surgicalriskclass,numberofsurgicallevels,andinformationonwhethertheprocedurewasarevisionorafusion.Oftheinitialdataset,weexcluded31casesthatweremissingadmission,dis-charge,anddiagnosisdetails,threecasesmissinganAEgrade,sevencaseswithnoclassscore,threelowcost(!$500)outliers,andonecase(C2quadriplegic)withanLOSexceeding3years.Ourfinaldatasetcomprised1,815casesevenlydistributedacrossthe4yearsoftheanalysis.

Costingdata

Inadditiontothisinformation,eachrecordincludespatient-basedmicrocostingdatacollectedthroughtheOn-tarioCaseCostingInitiative(OCCI;www.o.com),astandardized,auditedprovincialdatabasethatincludes40reportinghospitalsacrosstheprovinceofOntario.AllhospitalssubmittingdatatoOCCIuseaconsistentmethod-ologythatinvolvestheallocationofhospital-specificcoststounitsofstaffactivity,supplies,andequipmentsandastep-downprocessforallocatingindirectandoverheadcosts.Thetotalcostrecordedforeachcaseincludesdirectcosts,suchaslabor,operatingroom(OR)hours,surgicalimplants,disposables,recoveryroomhours,wardnursingandalliedhealthcare,inpatientdiagnosticimaging,andlabcharges.Italsoincludedtheindirectcosts,suchasbur-densrelatedtoadministrationandfacilities.Allcostswereconsideredfromtheperspectiveofthehospitalanddonotincludefee-for-servicephysicianbillings,whicharereim-burseddirectlybytheOntarioMinistryofHealthandLong-TermCarethroughaseparatefundingmechanismandnotincludedinthehospitalbudgets.Costsforepisodesdischargedinyearsbeforethestudydate(ie,2006–2010)wereinflatedto2011CanadiandollarvaluesusingtheCa-nadianConsumerPriceIndexbasketforhealthcareservices.

SAVESsystemdata

Ofthe1,815patientsstudied,316(17.4%)experiencedatleastoneAEthatwastrackedintheSAVESsystem.SpineAdverseEventsSeverityisaclinician-inputed,voluntaryreportinginstrumentusedfortheidentificationandseverityclassificationofAEsatourinstitutionsince2002[11,19].EacheventrecordedinSAVESincludesconsistentlycategorizeddetailsonthenatureofthecom-plicationandaseveritygradeusinganumericalscalefromItoIVbasedontheclinicalconsequencesoftheevent.AnAEisdefinedasanyeventthatiscausedbythemedicalorsurgicalmanagementandnotbytheunder-lyingdiseaseprocessorinjury,whichleadstopatientharmorrequiresadditionalmonitoringortreatment[11].Atdischarge,anyeventmeetingthisdefinitionisthengradedforclinicalseveritybythespinalteam.Ad-verseeventsweregradedasI(resultsinnoorminimaltreatment)(eg,urinarytractinfection[UTI]),II(requirestreatmentbutisnotlikelyassociatedwithlong-term[#6months]sequelae)(eg,acutepostoperativewoundinfec-tion),III(requirestreatmentandismostlikelyassociatedwithlong-term[O6months]sequelae)(eg,functionallysignificantnerverootinjuryormyocardialinfarctionwithabnormalechocardiogram),andIV(eventresultingindeath).

SpineAdverseEventsSeverity(availableat:http://links.lww.com/BRS/A411)hasbeenvalidatedinapreviousstudyinwhichrecordedentrieswerecomparedwithchart

E.K.Hellstenetal./TheSpineJournal13(2013)44–5347

abstractsfor200randomlyselectedpatients.SpineAdverseEventsSeveritycorrespondedsubstantiallytothehospitalchartabstractionfornumber(70%)andtype(75%)ofAEs.However,theSAVESprocessconsistentlycapturedagreatnumber(15%–40%)ofAEs,especiallysurgicalAEs.Interobserverreliabilitywasassessedinconsecutiveoperativecasesofonespinesurgeonduringa1-yearperiodusingthreeraters(staffsurgeon,fellowandresident,oranursepractitioner).Interobserverreliabilitywasassessedasnearperfectagreement(kappa50.8)forthegradeofAEseverity[11].Economicevaluation

EconomicevaluationsofAEstypicallyinvolveseveralkeycomponentsincludingAEincidencedata,costingand/orLOSdata,andstatisticalmodelstoestimatethere-lationshipbetweentheeventsandtheirimpactonresourceutilization.OurprimaryanalysismatchedAEcaseswithcontrolsonthebasisofapropensityscoretoestimatethemeanincrementalcostsandLOSassociatedwithAEs.Pairedttestswereusedforcomparingcontinuousout-comestoaccountforthematchednatureofthesample.Secondaryanalysesincludeddescriptivestatistics,chi-squaretests(fornominalvariables),ttests,andMann-WhitneyUtests(forskewedordinaldata).Sensitivityanalysiswasperformedbyusinganalternategeneralizedlinearmodel(GLM)regressionanalysistogeneratecompa-rableestimatesoftheincrementalcostsandtheLOS.AttributingeconomicoutcomestoAEspresentsanum-berofspecialchallengestotheuseofstandardstatisticalmethods.Hospitalrecordsamplesoftencontainarela-tivelysmallnumberofincidentcasescomparedwithlargenumbersofcontrolsandincludeadiversityofpatient-andfacility-specificcharacteristicsinfluencingthecostandLOS.Conventionalregressionanalysis

Table1

CharacteristicsofpatientswithandwithoutadverseeventsTotalcases51815

Meanage(standarddeviation)DiagnosiscategoriesTumorIntraduralDeformityTrauma

DegenerativeOtherASAclass12345

Revisions

Mean(standarddeviation),medianlengthofstayMean(standarddeviation),medianepisodecostASA,AmericanSocietyofAnesthesiology.

approachescontrollingforconfoundingvariablesareof-tensubjecttobiasandmodelmisspecificationinsuchsit-uations[2].Althoughoursamplehastheadvantageofcomprisingrecordsoriginatingfromacommonclinicalprogramatasinglehospitalfacility,evenwithinfacilities,spinalproceduresvarywidelyincomplexityfromasim-plediscectomytoamultilevelfusion,andthereisconsid-erableheterogeneityacrosspatients.Case-controlmatching

Toaddressthesechallenges,wedevelopedamodelus-ingmultivariablematchingofcaseswithcontrolsontheba-sisofpropensityscore.TomeasurethecausaleffectsassociatedwithAEs[20],weusedpropensityscorematch-ingtechniquessimilartothoseusedbyEncinosaandHel-linger[21]andWardleetal.[13].Weassignedallcasesandcontrolsapropensityscorethroughlogitregressiononeachpatient’sestimatedriskofexperiencinganAE,regressingagainstthepatientcharacteristicsdescribedinTable1.Riskadjustmentvariablesincludedage,collapsedthreedigitICD-10codefordiagnosisonadmission,fusionandrevisionstatuses,yearofdischarge,spinallevelsoperatedon,andtheASAclass.ASAclassisassignedfromGrade1(normalhealthy)toGrade5(moribundandnotexpectedtosurvivewithouttheoperation)[22].Forsurgicalpopula-tions,ASAclasshasbeenfoundtoprovideamoreaccuratecomplexityadjustmentthanCharlsoncomorbidityindex[23].

Usingthepropensityscore,eachofthe316AEcaseswasmatchedwithfourcontrolcasesbynearestneighbormatching(Stata12psmatch2routine;StataCorp,CollegeStation,TX,USA)withreplacement.Threehundredthir-teen(99.1%)ofthe316caseswerematchedonacaliperwidthoflessthan0.01withtheremainingthreecasesmatchedonawidthoflessthan0.015.

Adverseevent(n5316;17.4%)60.05(17.0)15.2%2.8%34.8%16.5%21.8%8.8%

(48)(9)(110)(52)(69)(28)

Noadverseevent(n51,499,82.6%)55.62(15.52)15.1%2.7%21.8%13.2%39.8%7.3%

(226)(41)(327)(198)(597)(110)

Significancep!.001p5.82p5.91p!.001p5.13p!.001p5.35p!.001

1.6%(5)22.2%(70)63.3%(200)13.0%(41)0.00%(0)14.6%(46)

23.88(36.43),14

,225.61(78,796.63),36,633.953.3%(50)35.0%(525)55.2%(828)6.0%(90)0.4%(6)14.6%(219)

7.67(15.982),5

23,465.43(32,575.58),18,282.52

p5.98p!.001p!.001

48E.K.Hellstenetal./TheSpineJournal13(2013)44–53

EstimatingincrementalcostandLOS

EachpatientrecordthatexperiencedanAEwasas-signedanindicatorvariableforthemostseveregradeofeventrecordedintheSAVEStool.Thesampleincludedasignificantproportion(124or39.4%)ofhospitalepisodeswheremultipleAEsweredocumented.Wereasonedthatinmostofthesecases,themostseveregradeofeventwaslikelytobethemostresponsibleforapatient’spotentialin-crementalcostandLOS.Althoughadditionaleventsoflessergradesmayalsooccurduringthesamehospitalepi-sode,thesearelikelytorunconcurrentlywiththemorese-vereeventsandhaveadiminishedadditiveeffectontheresourcesabovethoseattributedtothemajorevent.Bytak-ingthehighestseveritygradeforanypatientwithmultipleevents,thecostandincreasedLOSestimatesarealsorela-tivelyconservative.

Aftermatchingcasesandcontrols,wecalculatedthein-crementalcostandLOSasthedifferencesinthemeasuresbetweenacaseandthemeanofitsfourmatchedcontrols.Meanincrementalcost,LOS,andstandarderrorswerecal-culatedacrossallsets.WeperformedpairedtteststoassesswhetherthemeanincrementalcostandLOSforeachgradeweresignificantlydifferentfromzero.Sensitivityanalysis

Totestthesensitivityofourestimates,weusedanalter-nateestimationapproachusingmoreconventionalregres-sionanalysisoverthefullpatientsample.WeestimatedGLMswithagammaloglinktofitthenonzeroandrightward-skewednatureofthecostandLOSdata.Gener-alizedlinearmodelstendtoperformwellwhencomparedwithordinaryleastsquaresregressiononaloggeddepen-dentvariableandavoidretransformationissues[24].Incontrasttocase-matchingmethods,whichusedasubsetofcasesandmatchedcontrols,thegeneralizedlinearre-gressionmodelsincludedthefullpatientsample.TheGLMincludedallpatient-levelvariablesusedforpropen-sityscorematchinginthecase-controlmodel.

Allstatisticaltestswereconductedatthe0.05signifi-cancelevel.StatisticalanalyseswereperformedusingStata12(StataCorp,CollegeStation,TX,USA)andSPSS20(IBM,Chicago,IL,USA).

Results

Univariateanalysisofpatientsample

Tables1and2describethecharacteristicsofourstudy’spatientpopulation,comparingthesubgroupofpatientswhoexperiencedanAEwiththepatientswhodidnot.Com-paredwithcontrols,patientswhoexperiencedAEstendedtobeolder(mean60.1yearscomparedwithmean55.6years,p!.001).TherewasasignificantdifferenceintheASAscoresbetweenthetwogroups(p!.001).Patients

Table2

Incidenceofadverseeventsbypatientpopulationgroup

n

NoeventRateofanyeventASAclassASA515590.9%(50)9.1%(5)ASA5259588.2%(525)11.8%(70)ASA531,02880.5%(828)19.5%(200)ASA13168.7%(90)31.3%(41)ASA55

6100.0%(6)0%(0)DiagnosisgroupTumor27482.5%(226)17.5%(48)Intradural5082.0%(41)18.0%(9)Deformity43774.8%(327)25.2%(110)Trauma

25079.2%(198)20.8%(52)Degenerative666.6%(597)10.4%(69)Other

138

79.7%

(110)

20.3%

(28)

ASA,AmericanSocietyofAnesthesiology.

withAEshadahigherproportionofdeformity-relateddi-agnosesthanpatientswithout(34.8%comparedwith6%)andalowerproportionofdegenerativediagnoses(21.5%comparedwith39.9%).Otherdiagnosiscategorieshadsim-ilarproportionsacrossthetwogroups.Neithertherevisionstatusnortheyearofdischargeshowedsignificantdiffer-encesbetweenthetwosamples.

ThemeancostperdischargeandtheLOSweresignifi-cantlyhigherforthecaseswithAEsthanwithout($,225.61and23.9dayscomparedwith$23,465.43and7.7days,p!.001).Thetwolargestcostcomponentswithinthistotalcasecostwerenursing(meanof$26,181.59perdischargeforcaseswithAEscomparedwith$7,624.51forcaseswithout)andORcosts(meanof$15,263.72forcaseswithAEscomparedwith$11,521.68forcaseswithout).Notsurprisingly,caseswithAEsalsotendedtohavefarhigherIntensiveCareUnitcoststhancaseswithout(mean$7,663.68percasecomparedwith$979.73,p!.001).

Differencesinthetotalepisodecostsbetweenthetwogroupsalsotranslatedintodifferencesinthecostsperday;asaresultinincreasedLOS,themeancostperdayforcaseswithAEswassignificantlylowerthanforcaseswithoutAEs($2,229.92forAEcasescomparedwith$3,056.84,p!.001).Similartothecostsperepisode,thisdifferenceintotalperdiemcostscanbeexplainedthroughdecompositionofmajorcostdrivers:althoughtherewerenonsignificantdifferencesinnursingcostsperdaybetweenthetwogroups($1,134.51perdayforcaseswithAEscom-paredwith$995.90forcaseswithout,p!.106),ORcostsperdaywerefarhigherforcaseswithoutAEs($1,504.95comparedwith$661.41forcaseswithoutAEs,p!.001).ThismuchhigherintensityperdayinORresourceusecanbeattributedtothe‘‘frontloading’’ofthecostsofsur-gerywithinthedurationofthehospitalstay.AlthoughAEsresultedinincreasingpatients’LOSs,thebulkofresourceutilizationthroughoutthisincreasedstayfellonadditionalnursingcostsratherthanadditionalORusage.

Table2showsthebreakdownofAEsexperiencedbyourpatientpopulationintoGradesI,II,III,andIV.

E.K.Hellstenetal./TheSpineJournal13(2013)44–53

Table3

DistributionoftypesofAEsbyseveritygrade

Distributionofcomplicationtypesacrossseveritygrades

%oftotalAEs

TypeofcomplicationsUTI

Duraltear/leakUrinaryretentionCardiac

AlteredmentalstatusWoundinfectionRespiratoryPneumoniaGI

IntractablepainThromboembolicCutaneous

InstrumentationrelatedSystemicinfection

NeurologicdeteriorationAirwayOtherTotal

N(%)14160403728282721141414131210107661

(26.1)(11.0)(7.4)(6.7)(5.2)(5.2)(5.0)(3.9)(2.6)(2.6)(2.6)(2.4)(2.2)(1.8)(1.8)(1.3)(12.3)(100.0)

Grade1N(%)1345021616341152———2123269

(95.1)(83.3)(52.5)(16.2)(57.1)(10.7)(14.8)(4.8)(7.1)(35.7)(14.3)

Grade2N(%)5101923122019201171091283335191

(3.5)(16.7)(47.5)(62.2)(42.9)(71.4)(70.4)(95.2)(78.6)(50.0)(71.4)(69.2)(100.0)(80.0)(30.0)(42.9)(53.0)(35.3)

Grade3N(%)2——5—53—2224—253742(1.4)

49

Grade4N(%)———2(5.4)——1(3.7)—————————1(1.5)4(0.7)

(13.5)(17.9)(11.1)(14.3)(14.3)(14.3)(30.8)(20.0)(50.0)(42.9)(10.6)(7.8)

(20.0)(14.3)(34.8)(49.8)

AE,adverseevent;UTI,urinarytractinfection;GI,gastrointestinal.

Onehundredtwenty-fourof315(39.4%)caseswithAEsexperiencedtwoormoreeventsduringthesameepisode.Table3showsthedistributionofthecomplicationtypesbytheirseveritygrades.Urinarytractinfections,duraltears,andurinaryretentioncomplicationswerethemostcommontypesofAEsandforthemostpartwerecatego-rizedaseitherGradeIorII.GradeIVAEsleadingtomortalitiesweremedicalinnatureandoccurredinrela-tivelylowfrequencies(n).Specifically,mortalitiesoc-curredinthreepatientswhopresentedwithmultilevelepiduralabscesses(sepsis,myocardialinfarctionwithcardiacarrest,andrenalfailure)andafourthpatienttreatedforaprimaryspinalcordtumor(respiratoryarrest).

Multivariateanalysis

ThelogitmodelappliedtoestimateapropensityscoreforeachcaseestablishedthesignificantpredictorsofAEstobegreaterpatientage(p!.001),sex(female)(p5.031),greaternumberofspinallevelsoperatedon(p!.001),fusion(p5.044),andanASAclassoffour

Table4

Grade1to4case-controlanalysis:incrementalcostperAEcaseAEgrade1234Total

No.ofcases157126294316

Incrementalcostpercase($CDN)4,224.823,500.57147,285.20121,366.1

Standarderror1,985.4313,885.235,239.632,057.66

(p5.037).ProtectivefactorsincludedanASAclassoftwo(p5.003)anddegenerativediagnoses(p!.001).

Tables4and5describethemeanincrementaldollarcostsandLOSestimatedforeachseveritygradeofAEbythematchedcase-controlmodel.Estimatesforbothoutcomesin-creaseincrementallywithhighergradesofAEsthroughGradeI($4,224.and3.6days),GradeII($23,500.57and14.0days),andGradeIII($147,285.20and76.50days)cases.GradeIVcasesareassociatedwithasmallerincremen-talcostandLOSimpactthanGradeIIIcases($121,366.10and46.44days).InadditiontotestingforthesignificanceofmeanincrementalcostandLOSforeachgradebeingsig-nificantlydifferentfromzeroatthe0.05level,ttestscon-firmedthatmeanattributablecostandLOSforeachincrementalgradearesignificantlygreaterthanthegradebe-lowthat.GradeIVcasesweretheexceptiontothisrule:nei-thercostnorLOSwassignificantlydifferentatthe0.05levelfromthatofGradeIIIcases.

ThetotalcostandLOSimpactsofeachgradeofeventaresubstantial.Alltogether,GradeIcasescontributedanaddi-tional$0.663millioninincrementalcostsand5,709.9addi-tionalbeddays,GradeIIcasescontributed$2.96millionand

95%Confidenceinterval299.59,15,794.22,53,4.04,19,344.36,

8,150.1931,206.91241,106.40223,387.9

pvalue(meanO0).0351!.0001.0036.0323

Totalincrementalcost($CDN)663,307.572,961,071.824,271,270.80485,4.408,381,114.59

AE,adverseevent.

50E.K.Hellstenetal./TheSpineJournal13(2013)44–53

Table5

Grade1to4case-controlanalysis:incrementalLOSperAEcaseAEgrade1234Total

No.ofcases157126294316

IncrementalLOSpercase(days)3.6314.0374.6.44

Standarderror0.91321.8420.88685.5747

95%Confidenceinterval1.821202,5.4323710.33601,17.7319531.06347,117.936528.695,.17855

pvalue(meanO0).0001!.0001.0018.0036

TotalincrementalLOSforstudyperiod(days)569.91,767.82,160.5185.84,684.0

LOS,lengthofstay;AE,adverseevent.

1,767.8beddays,GradeIIIcasescontributed$4.27millionincostsand2,160.5beddays,whereasGradeIVcasescontrib-utedanadditional$0.485millioninincrementalcostsand185.8incrementalbeddays.AcrossallrecordedAEsovertheentire4yearsofthestudy,therewasatotalincrementalcosttothehospitalof$8.38millionand4,684incrementalbeddays,orapproximately$2.10millionand1,171beddaysperannum.TheestimatedimpactsoftheAEsaccountedforapproximately16.0%ofthe$52.31milliontotalcostand24.6%ofthetotalbeddaysofallthespinesurgerypatientsinoursample.

Sensitivityanalysisresults

Thegeneralizedlinearregressionmodelsusedforsen-sitivityanalysisproducedcostandLOSestimatesinasim-ilarrangetothoseestimatedthroughmatchedcase-controldifferenceinmeans.AsimilarascendingtrendincostandLOSwasapparentacrossthehigherseveritygrades,dissi-patingasbeforewiththefourGradeIVcases.Asde-scribedinTable6,incrementalcostsattributabletoGradeI,II,III,andIVAEcaseswereestimatedat$6,370.56,$21,533.98,$91,827.17,and$59,252.85,re-spectively.Table7presentsincrementalLOSestimatesof4.2,12.9,50.9,and18.4beddaysforGradeI,II,III,andIVevents,respectively.

Incomparisonwiththematchedcase-controlanalysis,theGLMsestimatedahighercostandLOSimpactattribut-abletoGradeIevents,decreasedcostandLOSforGradeIIevents,andaverysignificantlyreducedcostandLOSforGradeIIIevents.Underthematchedcase-controlanalysis,GradeIIIeventsareestimatedtohavethegreatesttotalhos-pitalcostimpactofalltheseveritygrades,whereasunderGLM,GradeIIeventsmakeaslightlygreatercontributiontothetotalcost.TheGLMresultsinalowertotalcostfor

Table6

GLMgammaloglinkregression:incrementalcostperAEcaseAEgrade1234Total

No.ofcases157126294316

Incrementalcostpercase($CDN)6,370.5621,533.91,827.1759,252.85

Standarderror2,781.604,523.9023,880.0046,331

alleventscomparedwiththematchedcase-controlmodel($6.59millionvs.$8.03million).

Discussion

Initially,Brennanetal.[25]definedAEsasunintendedinjuriesorcomplicationscausedbythehealthcaremanage-mentratherthanbythepatients’underlyingconditionsthatleadtodeath,disability,orprolongedhospitalstays.Previ-ousstudieshaveestimatedAEstooccurin7.5%ofallhos-pitalizationsinCanada,withcloseto38%oftheseestimatedasbeingpotentiallypreventable[26].Inadditiontotheirhu-mancostinincreasedmorbidityandmortality,AEshaveasubstantialeconomicimpact:directhealthcarecostsofAEshavebeenestimatedtoconsumeonedollarineverysevenspent(ie,14%)onhospitalcare[27].Theresultsofourstudyareconsistentwiththesefindings.Ourpropensityscorecase-matchedanalysisdemonstratedthatAEsassoci-atedwithabroadspectrumofspinalsurgeryatatertiary-careacademichospitalwereresponsibleforanincrementalcostof8.4milliondollarsover4years,whichrepresents16%ofthetotaldirectcostspentonhospitalcareinourstudycohort.Thistranslatedtoanestimatedadditionalcostof$2.1milliondollarsperyearand,approximately,1,171additionalbeddaysperyearin1,815patients.Remarkably,thisincrementalcostwasattributabletoonly316patients(17.4%)whosufferedclinicallyrelevantAEs.Asexpected,theincrementalcostofanAEprogressivelyincreasedastheseverityofAEincreased.However,becauseofahigherfre-quency,thelowerseveritygradeAEs(I–II)resultedinmorethanhalfoftheaggregateadditionalcost(55.9%)andbeddays(59.7%).NotonlyaretheseAEsthemostcommon,theyalsorepresentthemostpotentiallymodifiableAEs(eg,lowerUTI,incidentaldurotomies,urinaryretention,

95%Confidenceinterval918.67,12,667.20,45,022.70,À31,5.00,

11,822.4030,400.70138,632.00150,060.00

pvalue(meanO0).022!.0001!.0001.201

Totalincrementalcostforstudyperiod($CDN)1,000,177.922,691,747.502,662,987.93237,011.406,591,924.75

GLM,generalizedlinearmodel;AE,adverseevent.

E.K.Hellstenetal./TheSpineJournal13(2013)44–53

Table7

GLMgammaloglinkregression:incrementalLOSperAEcaseAEgrade1234Total

No.ofcases157126294316

IncrementalLOSpercase(days)4.2312.9450.8518.37

Standarderror1.653.0718.1222

95%Confidenceinterval1.065,6.918,15.340,À24.728,

7.52918.96686.35361.457

pvalue(meanO0).009!.0001.005.404

51

TotalincrementalLOSforstudyperiod(days)674.61,617.81,474.573.53,840.4

GLM,generalizedlinearmodel;LOS,lengthofstay;AE,adverseevent.

andwoundinfection).Infact,approximately56%ofthein-dividualAEsthatoccurredinthiscohortwouldbeconsid-eredpotentiallypreventableand31%(26%withUTIand5%withsurgicalsitewoundinfections)areontheUSCen-tersforMedicareandMedicaidServices‘‘neverevents’’list[28].

Toourknowledge,thisistheonlystudytoassessspinesurgery–relatedAEsusingprospectivelycollected,patient-specificclinicalAEseveritycategorizationandpatient-specificcasecostingdata.Inarecentlypublishedstudy,Whitmoreetal[10]assessedthecosttopayersasso-ciatedwiththecomplicationsafterspinesurgery.Usingdatafromtheirprospectiveobservationalstudyof226cases,theauthorsestimatedthecostofAEsbyapplying2008Medicarereimbursementratesforthepaymentasso-ciatedwithallDiagnosisRelatedGroupandCurrentProce-duralTerminologycodesforeachcomplicationcaseandcomparedwiththosetothebaselinecostderivedfromthemodelwhenallcasesareincludedandfactorsthatinfluencethecostofcarearecontrolledbyregression.Theauthorsreportedthatthedifferenceinthemeancostofcaseswithcomplicationscomparedwiththosewithoutwas$13,518(confidenceinterval:$9,378,$17,657)USD,withmajorcomplicationscostingsignificantlymorethanminorcom-plications.AsnotedbyWhitmoreetal.,theuseofMedi-carereimbursementratestoestimatedirecthealthcarecostsdoesnotnecessarilyreflectthetruecosttothehospi-talandmayunderestimatetheimpactofcostsassociatedwithprolongedLOS.Incontrast,ourstudyusedpatient-specificclinicalandhospitalcostingdatatodeterminetheincrementalhospitalcosts.Thisresultedinanoverallaver-ageincrementalcostof$26,522CDN.AsrecentlynotedbyKimetal.[29],theUSin-hospitalcosts(usingMedicarerates)forlumbarspinaldecompressionordecompressionandfusionareapproximatelythreetofourtimesgreaterthanCanadiancostsforequivalentprocedures;thus,themeanincrementalcostsofAEsreportedbyWhitmoreetal.arelikelytobeanunderestimation.TheliteratureisverylimitedregardingstudiesthatassesstheimpactofAEsoncostsforspinesurgery[14,15,18,30–33].Typicallythesestudieshavefocusedonaspecificdiagnosticgrouporspinalprocedureand,thus,areoflimitedscopeinconsid-eringthediversityofsurgicalcomplexityandcostsassoci-atedwithspinalsurgery.Consequently,thesestudiesdonotprovidedatathatenableabetteroverallunderstandingoftheeconomicimpactofAEsassociatedwithspinesurgery.However,allstudies(regardlessofmedicaldiscipline)uni-versallyshowasignificantincreaseinhealthcarecostsre-latedtotheAEs.

TheoutcomeofaneconomicevaluationofAEscanvarysignificantlydependingonavarietyoffactorsthatinclude,butarenotlimitedto,thespecificsourceandaccuracyofthepatient-leveldata;thetypeandaccuracyofthecostingdata;and,andthetypeofanalysisused.Inaddition,numer-ousfactorssuchaspatientage,comorbidity,primarydiag-nosis,andsurgicalcomplexityhavebeenshowntobeindependentlyassociatedwithahigherlikelihoodofdevel-opinganAEfromspinesurgery[4–8].Thesesamefactorshavealsobeenshowntosignificantlyinfluencecostand,therefore,representsignificantconfoundersthatneedtobeconsidered.Consequently,cautionmustbeusedwheninterpretingtheresultofcostingevaluationsassessingsur-gicalAEs.

Inthisregard,ourstudyhasseveralstrengths.First,wehadthebenefitofdataderivedfromavalidatedtoolwithspecificdefinitionsthatprospectivelyandcomprehensivelycapturesandreflectstheclinicalseverityofAEs;theSAVEStoolhaspreviouslybeenshowntohaveahighin-terobserverreliability(kappa50.8)[11]forthegradingofAEseverityandaclosematchwiththechartabstractionforthenumber(70%)andtype(75%)ofAEs[9–12].Fur-thermore,patient-specificdiagnosticandclinicaldataasso-ciatedwithahigherriskofdevelopinganAEhavebeencaptured.AsreportedbyWardleetal.[13],theuseofad-ministrativedatabaseshastheadvantageoflargecohortsbutthedisadvantageofpotentialcodingvariationsacrosshospitalsandthesignificantriskofdiagnosticandAEcod-ingerrors.Wardleetal.[13]notethattheimpactofcodingvariationonAEcostestimatescanbesubstantial.Second,wehaveusedpatient-specificcasecostingdata(seelimita-tionsalso)thatprovidesamoreaccuratereflectionofthetruecosttothehospitalratherthanestimatingcostsfromcharges[3,13,16,17,34,35].Third,ouruseofpropensityscorematchingandcase-controlmethodstoremovebiasandconfoundinginfluencesallowsforamoreaccuratees-timationoftheeconomicimpactsofcomplications,consis-tentwiththemorerecentacceptedpracticesinthebroaderliteraturearoundestimatingtheeffectofAEsonresourceutilization[3,13,16,17,34,35].Theuseofappropriatelyro-bustanalyticalmethodsisparticularlyimportantwhen

52E.K.Hellstenetal./TheSpineJournal13(2013)44–53

dealingwiththediagnosticallyandprocedurallyheteroge-neouspatientpopulationfoundinspinalsurgery;suchfac-torsbothcontributetotheriskofapatientexperiencinganAEandcontributetotheresultingcostandLOSimpactsobservedfromthecomplication.Fourth,weusedalterna-tivemodelingapproachestoperformasensitivityanalysisonourestimatesandfoundthattheresultsaregenerallyconsistentacrossthemethodsused,confirmingthatourpri-maryanalyticalmodelandresultsarerobust.Finally,ourstudyisunique,itcombinesclinicallyrelevant,patientsafetyincidentreportingwitheconomicevaluationthatprovidesimportantinformationregardingareaswhereamoretargeted,evidence-basedpatientsafetystrategycanbeconsideredataprogrammaticlevel(ie,spinesur-gery)andislikelytohavemaximumimpactinbothim-provingpatientsafetyandreducinghospitalcosts.Thiscombinedapproachcanproviderelevantinformationtocli-nicians,policy-makers,andadministratorsalike.

Ourstudydoeshaveseverallimitations.Wehavepre-sentedtheincrementalcostandLOSestimatesforeachse-veritygradeofAEratherthanidentifyingeconomicimpactsbythetypeofcomplication.ApromisingfutureareaofstudywillbetoinvestigatetheincrementalcostsandLOSassociatedwithspecificcomplications,suchasUTIs,duraltears,andwoundinfections.Intermsofthescopeofouranalysis,wewerelimitedbythehospital-specificfocusofourcasecostingdata;althoughOCCI-reportedcostingdataisgranularandrobust,itrepresentsonlythecostsaccruedtothehospitalanddoesnotincludethecostsofphysicianbillingsorpostdischargeservices,suchasoutpatientandambulatoryvisitsandhomecareandanyreadmissionstohospital.Intheir2008study,Enci-nosaandHellinger[21]foundthatthecostsassociatedwiththechainsofhealthsystemutilizationafteraninitialindexeventcanbesubstantial.Beyondthedirectcostsofhealthservicesalone,wedonotconsiderthebroadersocietalperspectiveinassessingtheimpactsofAEs.Indirectcostsassociatedwithseriouseventssuchastemporaryorperma-nentdisability,reducedproductivity,andlossoflifeandlimbmaybeconsiderable.Higherindirectcostsmaybeex-pectedtoaccruetomoreseriouseventsthatresultinpro-longedorpermanentconsequencesafterdischarge;thispresentsapromisingareaforthefuturestudy.

Inthepresentclimateofcostconscioushealthcare,strategiesaimatimprovingthequalityofhealthcaredeliv-eryand,thus,thevalueofeachhealthcaredollarspentarecritical.Thereisnoquestionthatimprovedpatientsafetyisofparamountimportancetoimprovingtheoverallqualityofhealthcare.Thefindingsofthepresentstudyfurtherstrengthentheeconomicargumentforsuchmeasures.OurfindingsalsohighlightthesignificanteconomicimpactoflowerseverityAEsthatareoftenconsideredminor(ie,thosethatarenotlikelytohavealong-term[O6months]clinicalimpact).ManyoftheseAEsfallintotherealmofso-called‘‘neverevents.’’Althoughfromaclinical

perspectivetheseeventsarecertainlymodifiable,evenun-derthebestcircumstancestheyareunlikelytobecom-pletelypreventable.However,ourfindingthatalargeshareoftheAEsthatoccurredinthisstudyarepotentiallymodifiableshouldmotivatethehospitalstobroadenthescopeofpatientsafetyinitiativestoincludetheselessclin-icallysignificantevents.Aspayers’listsofnonreimburs-able‘‘neverevents’’continuetoincrease,failuretodosomaybecomeaverycostlydecision.

Conclusion

Inthisprospectivestudy,assessingtheeconomicim-pactsofhospital-acquiredAEsassociatedwithspinalsur-gery,anAErateof17.4%accountedfor16%ofthetotalcostofin-hospitalcareinourstudysampleof1,815pa-tients.IncreasingclinicalseverityofAEshadaprogres-sivelygreaterimpactonthecostandLOSonaper-casebasis;however,themorefrequentlowerseverityevents(ie,GradeIandII)alsohadasubstantialaggregateimpactonbothcost(43%oftotalincrementalcosts)andLOS(49%oftotalincrementalbeddays).Theseresultssuggestthatastrongbusinesscaseexistsforpatientsafetystrate-giesthatfocusnotonlyonthemostsevereAEsbutalsoonthereductionoflowerseverityeventsthatmaybemoreamenabletopreventionefforts.Approximately,50%ofAEsoccurringinthiscohortarepotentiallymodifiable;consequently,atargetedevidence-basedapproachcouldresultinsignificantcostsavings.References

[1]KohnLT,CorriganJM,DonaldsonMS.Toerrishuman:building

asaferhealthsystem.Washington,DC:InstituteofMedicine,1999.[2]ZhanC,MillerMR.Excesslengthofstay,charges,andmortality

attributabletomedicalinjuriesduringhospitalization.JAMA2003;290:1868–74.

[3]ZhanC,FriedmanB,MossoA,PronovostP.Medicarepaymentfor

selectedadverseevents:buildingthebusinesscaseforinvestinginpatientsafety.HealthAff2006;25:1386–93.

[4]Wilson-MacDonaldJ,FairbankJ,FrostH,etal,SpineStabilization

TrialGroup.TheMRCspinestabilizationtrial:surgicalmethods,outcomes,costs,andcomplicationsofsurgicalstabilization.Spine2008;33:2334–40.

[5]WeissHR,GoodallD.Rateofcomplicationsinscoliosissurgery-asystematicreviewofthePubMedliterature.Scoliosis2008;3:9.[6]NasserR,YadlaS,MaltenfortMG,etal.Complicationsinspinesur-gery.JNeurosurgSpine2010;13:144–57.

[7]LeeMJ,KonodiMA,CizikAM,etal.Riskfactorsformedicalcom-plicationafterspinesurgery:amultivariateanalysisof1,591patients.SpineJ2012;12:197–206.

[8]SchoenfeldAJ,OchoaLM,BaderJO,BelmontPJJr.Riskfactorsfor

immediatepostoperativecomplicationsandmortalityfollowingspinesurgery:astudyof3475patientsfromtheNationalSurgicalQualityImprovementProgram.JBoneJointSurgAm2011;93:1577–82.[9]StreetJT,LenehanBJ,DipaolaCP,etal.Morbidityandmortalityof

majoradultspinalsurgery.Aprospectivecohortanalysisof942con-secutivepatients.SpineJ2012;12:22–34.

E.K.Hellstenetal./TheSpineJournal13(2013)44–53

53

[10]WhitmoreRG,StephenJ,SteinSC,etal.Patientcomorbiditiesand

complicationsfollowingspinalsurgery:asocietal-basedcostanaly-sis.Spine2012;37:1065–71.

[11]RampersaudYR,NearyMA,WhiteK.SpineAdverseEventsSever-itySystem:contentvalidationandinterobserverreliabilityassess-ment.Spine2010;35:790–5.

[12]MirzaSK,DeyoRA,HeagertyPJ,etal.Towardsstandardizedmea-surementofadverseeventsinspinesurgery:conceptualmodelandpilotevaluation.BMCMusculoskeletDisord2006;7:53.

[13]WardleG,WodchisWP,LaporteA,etal.Thesensitivityofadverse

eventcostestimatestodiagnosticcodingerror.HealthServRes2012;47:984–1007.

[14]NewPW,JacksonT.Thecostsandadverseeventsassociatedwithhos-pitalizationofpatientswithspinalcordinjuryinVictoria,Australia.Spine2010;35:796–802.

[15]WinslowC,BodeRK,FeltonD,etal.Impactofrespiratorycompli-cationsonlengthofstayandhospitalcostsinacutecervicalspinein-jury.Chest2002;121:18–.

[16]ClassenDC,PestotnikSL,EvansRS,etal.Adversedrugeventsin

hospitalizedpatients:excesslengthofstay,extracostsandattribut-ablemortality.JAMA1997;277:301–6.

[17]ParadisAR,StewartVT,BayleyKB,etal.Excesscostandlengthof

stayassociatedwithvoluntarypatientsafetyeventreportsinhospi-tals.AmJMedQual2009;24:53–60.

[18]DeyoRA,MirzaSK,MartinBI,etal.Trends,majormedicalcompli-cations,andchargesassociatedwithsurgeryforlumbarspinalsteno-sisinolderadults.JAMA2010;303:1259–65.

[19]RampersaudYR,MoroERP,NearyMA,etal.Intraoperativeadverse

eventsandrelatedpostoperativecomplicationsinspinesurgery:im-plicationsforenhancingpatientsafetyfoundedonevidence-basedprotocols.Spine2006;31:1503–10.

[20]RosenbaumP,RubinDB.Thecentralroleofthepropensityscorein

observationalstudiesforcausaleffects.Biometrika1983;70:41–55.[21]EncionsaJ,HellingerFJ.Theimpactofmedicalerrorsonninety-day

costsandoutcomes:anexaminationforsurgicalpatients.HealthServRes2008;43:2067–85.

[22]FuKM,SmithJS,PollyDW,etal,ScoliosisResearchSocietyMorbid-ityandMortality.CorrelationofhigherpreoperativeAmericanSociety

ofAnesthesiologygradeandincreasedmorbidityandmortalityratesinpatientsundergoingspinesurgery.JNeurosurgSpine2011;14:470–4.[23]

SchroederRA,MarroquinCE,ButeBP,etal.Predictiveindicesofmorbidityandmortalityafterliverresection.AnnSurg2006;243:373–9.

[24]BasuA,ManningWG,MullahyJ.Comparingalternativemodels:logvs.Coxproportionalhazard?HealthEcon2004;13:749–65.

[25]

BrennanTA,LeapeLL,LairdNM,etal.Incidenceofadverseeventsandnegligenceinhospitalizedpatients.ResultsoftheHarvardMed-icalPracticeStudyI.NEnglJMed1991;324:370–7.

[26]

BakerGR,NortonPG,FlintoftV,etal.TheCanadianAdverseEventsStudy:theincidenceofadverseeventsamonghospitalpatientsinCanada.CanMedAssocJ2004;170:1678–86.

[27]JacksonT.Onedollarinseven:scopingtheeconomicsofpatientsafety.Ottawa,ON,Canada:CanadianPatientSafetyInstitute,2009.[28]MilsteinA.Endingextrapaymentfor‘‘NeverEvents’’—strongerin-centivesforpatients’safety.NEnglJMed2009;360:2388–90.

[29]

KimS,MortazS,CoytePC,RampersaudYR.Cost-utilityoflumbarde-compressionwithorwithoutfusionforpatientswithsymptomaticde-generativelumbarspondylolisthesis.SpineJ2012;12:44–.

[30]KalanithiPS,ArrigoR,BoakyeM.Morbidobesityincreasescostandcomplicationratesinspinalarthrodesis.Spine2012;37:982–8.

[31]

ParkerSL,AdogwaO,WithamTF,etal.Post-operativeinfectionaf-terminimallyinvasiveversusopentransforaminallumbarinterbodyfusion(TLIF):literaturereviewandcostanalysis.MinimInvasiveNeurosurg2011;:33–7.

[32]

AmbrossiGL,McGirtMJ,SciubbaDM,etal.Recurrentlumbardischerniationaftersingle-levellumbardiscectomy:incidenceandhealthcarecostanalysis.Neurosurgery2009;65:574–8.

[33]

GoodeAP,CookC,GillJB,etal.Theriskofrisk-adjustmentmea-suresforperioperativespineinfectionafterspinalsurgery.Spine2011;36:752–8.

[34]

BatesDW,SpellN,CullenDJ,etal.Thecostsofadversedrugeventsinhospitalizedpatients.AdverseDrugEventsPreventionStudyGroup.JAMA1997;277:307–11.

[35]

RivardPE,LutherSL,ChristiansenCL,etal.Usingpatientsafetyindicatorstoestimatetheimpactofpotentialadverseeventsonoutcomes.MedCareResRev2008;65:67–87.