如何進(jìn)行醫(yī)學(xué)實(shí)驗(yàn)室風(fēng)險(xiǎn)評估

HowtoEvaluateRiskinthe

MedicalLaboratoryJamesH.Nichols,PhD,DABCC,FACBProfessorofPathology,MicrobiologyandImmunologyMedicalDirectorofClinicalChemistryandPoint-of-CareTestingVanderbiltUniversitySchoolofMedicineMedicalDirector,ClinicalChemistryNashville,Tennessee,USAjames.h.nichols@

1Objectives1.Discussthesourcesoferrorinthemedicallab2.ReviewQCasameansofcontrollinganalyticalerror3.Identifyriskmanagementasatotalqualityassuranceofthetestingprocess4.Describeregulatorychangesadoptingriskmanagementinthelaboratory

2MedicalErrorAnadverseeventornearmissthatispreventablewiththecurrentstateofmedicalknowledge.(theQualityInteragencyCoordinationTaskForce)

MedicalErrorsInstituteofMedicineoftheNationalAcademiesreport1999Medicalerrorskill44,000-98,000patientsinUShospitalseachyear.“Numberoneproblemfacinghealthcare”LucienLeape,HarvardProfessorofPublicHealthMedicalErrors2002CommonwealthFundreportestimatedthat22.8millionpeoplehaveexperiencedamedicalerror,personallyorthroughatleastonefamilymemberReinforcesthe1999IOMreport,“ToErrisHuman”Annualcostsestimatedat$17–29billionUSAgencyforHealthcareResearchandQuality(AHRQ)estimatemedicalerrorsarethe8thleadingcauseofdeathintheUS–higherthan:MotorVehicleAccidents(43,458)Cancer(42,297)AIDS(16,516)LaboratoryErrorsAreviewoftheliteraturefoundthemajorityoferrorsoccurinthepreandpostanalyticalphases.

BoniniP,PlebaniM,CeriottiF,RubboliF.ClinChem2002;48:691-698.Manymistakesarecalled“l(fā)aberror”,butactuallyduetopoorcommunication,actionsbyothersinvolvedinthetestingprocess,orpoorlydesignedprocessesoutsidethelab’scontrol.Medicalerrorsoccurinprevention,diagnosisanddrugtreatmentoccur.Amongerrorsindiagnosis;50%werefailuretouseindicatedtests,32%werefailuretoactonresultsoftests,and55%involvedavoidabledelayindiagnosis.LeapeLL,BrennanTA,LairdN,etal.NEngJMed1991;324:377-84.7HistoricalQualityControlQualitycontrolhistoricallyusedtoreduceerrorsandpreventthereleaseofincorrectresultsfromthelaboratory!Bornfromthe1950’sindustrialmodelofqualityinanalyticalandfactoryprocessesQualitycontrolisastabilizedsurrogatesampleanalyzedlikeapatientsamplecontainingknownamountofmeasuredanalyte.IftheanalyticaltestsystemcanachievethedesiredresultusingaQCsample,thenthesystemisstableandqualitypatientresultsareassumedtobeproduced.8QualityControlAdvantagesQChastargetvalues,ifassayrecoverstarget,thensystemisassumedtobestable(instrument,reagent,operator,sample)ShiftinQC=Shiftinqualitypatientresults!WarnsoftestresultproblemsbeforereleasetophysicianQualityControlDisadvantagesIfQCnotanalyzedfrequentlyenoughorlabfailstotakeactionwhenQCfails,qualityofpatientresultsaffected!Withnewautoverificationorcontinuousreleaseofresultsbychemistryandhematologyanalyzers-PossibletoreleasepatientresultsbeforeaproblemisdetectedWhenaproblemisdetected,labmustgobackandreanalyzepatientssincelast“good”QCandpossiblysendoutcorrectedresults.Clinicalactioncouldbetakenonerroneousresultsintheinterimofcorrectinganunrecognizedissue!Despiteitslimitations,QCremainsanimportantmeansofdetectingandpreventinglaboratoryerrors!1011SystematicErrorsLiquidQCdoesagoodjobatdetectingsystematicerrors,like:ReagentdeteriorationorpreparationImproperstorageorshipmentconditionsIncorrectoperatortechnique(dilution,pipettesetting)Calibrationerrors–wrongsetpoint,factorsErrorswhichaffecteverytestinaconstantandpredictablemannerErrorsthatoccurfromonepointforwardorforalimitedperiodoftime12RandomErrorsLiquidQCdoesapoorjobatdetectingrandomerrorsunlesstheerrorspecificallyoccurswiththeQCsampleorincreasedresultvariabilityisnoted.Errorswhichaffectindividualsamplesinarandomandunpredictablefashion,like:ClotsBubblesInterferingsubstances13QCandLabErrorsQCQC09:0009:00LineLeak11:0001:15HemolyzedsampleQCdailyat09:00,twolevelsLineleakoccursat11:00,dispensespartialreagentcausing25%decreaseinvalues,notdetecteduntilnextdayQCat09:00.(SystematicError),mustreanalyzeallspecimenssincepreviousQCat09:00,thepreviousday.Hemolyzedspecimen(Randomerror)affectsonesample14HistoryofQualityControlQualitycontrolhasbeentraditionalmeansofmanagingerrorinthelaboratory!CLIA’67–requireddailyQCasintegralpartoftestingprocessCLIA’88-reaffirmed“twolevelsofQCpertestperday”(bloodgasesevery8hours)CAPadvocatesdailyQCthroughinspectionandaccreditationTheJointCommissionfocusesonpatientsafetyandpreanalytic,analyticandpostanalyticprocessesISOQCRecommendations15ISO15189–5.6.1Thelaboratoryshalldesigninternalcontrolsystemsthatverifytheattainmentoftheintendedqualityofresults.Specialattentionshouldbepaidtoeliminationofmistakesintheprocessofhandlingsamples,requests,examinations,reports,etc.Documentationshouldincludequalitycontrolproceduresbasedonmanufacturerinstructionsforuse.InternalQualityControl(internaltothelaboratory)isdefinedasasetofproceduresundertakenbylaboratorystaffforthecontinuousmonitoringofoperationandtheresultsofmeasurementsinordertodecidewhetherresultsarereliableenoughtobereleased.TheregularanalysisofQCmaterialscanserveasanessentialcomponentofalaboratory’sinternalcontrolsystem.16QualityinLaboratoryTestingDespiteanalyzingtheminimumrequired2levelsQC/day,wehaveallexperiencederroneouslaboratoryresults!QualityismorethanQC!Qualityis:TestingthatissafeandreliableAppropriateuseoftechnologytomeetmedicalneedsTestresultsthatcanbetrustedformedicalmanagementAlab’sreputationdependsonqualityofitsresults!Whatshouldbethelabgoalsforerrors?UnacceptableErrorsDeaths/106enplanements 0.2PPMDeaths/106generalanesthesiainductions 2-5Viraltransmissions/106transfusions 29Results>7SD/106automatedchemtests 100Lostbags/106airplanepassengers 5000Failures/106proficiencytests 14,000Difference>5%/106hematocrittests upto 190,000WitteD&VanNessS.FrequencyofUnacceptableResultsinPoint-of-CareTesting.ArchPatholLabMed123:761,1999SigmaMetricsSigmametricsareanindustrialmeasureofqualityinafactoryprocess

-estimatesrateofdefectsinproductsGoalofsigmametrics:ImprovecustomersatisfactionBalancenumberofrejectedproductswiththeincreasedcostrequiredtoimproveprocesses(tolowerdefectrates)SixsigmaconsensusforbalancingrejectionagainsttheincreasedcostsofimprovingerrorsforindustrialprocessesSixsigmawasadoptedandpopularizedintotheclinicallaboratoriesin1990’s18SigmaPerformanceNeedforQualityLaboratoryTestingLaboratorytestinginfluences>70%ofallmedicaldecisions.>7billionlabtestsconductedintheU.S.annually

SilversteinMD.Anapproachtomedicalerrorsandpatientsafetyinlaboratoryservices.AwhitepaperpreparedfortheQualityInstituteMeeting,MakingtheLaboratoryaPartnerinPatientSafety.DivisionofLaboratorySystems,CentersforDiseaseControlandPrevention,April2003./mlp/qiconference/

BeyondSixSigmaWith>7billiontestsconductedintheUSannually,sixsigma(3.4defects/milliontests)meansreleasing23,800incorrectlabtestsannuallyintheUS!

Anysingleresulterrorcouldmeanamisdiagnosis,failuretotreatanillnessorincorrecttreatmentselection!Viruscontaminationofbloodproduct,IncorrectcancerdiagnosisMissedheartattack!Whatifthatwrongtestresultbelongedtoyouorafamilymember?Foramedicallaboratory,sixsigmaisanunacceptablegoal!Weneedtogo“beyondsixsigma”forlaboratoryperformance,withagoalofZEROerrorsthatreachthepatient!2122RiskDefinitionRisk–thechanceofsufferingorencounteringharmorloss(Webster’sDictionaryandThesaurus,1993Landoll,Ashland,Ohio)RiskisthepotentialforanerrortooccurRiskDefinitionISO/IECGuide51–RiskcanbeestimatedthroughacombinationoftheprobabilityofoccurrenceofharmandtheseverityofthatharmRISK=(Frequency)x(SeverityHarm)x(Detection)↑numberoferrors=greaterrisk↑harm=greaterriskInabilitytodetectanerror=greaterriskRiskissimplyaratingscaletoranktheorderforaddressingpotentialerrorsinlab(withlimitedavailableresources)23MedicalErrorsThePersonEasiertoblameapersonthananinstitutionforerrors.Inaviation,90%ofqualitylapsesarejudgedtobeblameless.TheSystemActivefailuresduetopersonalinteractionwithsystemLatentconditions,weaknessesinsystemduetodesignflawsorheirarchicaldecisionsNeedtoengineersystemsthatpreventdangerouserrorsandareabletotolerateerrorsandcontaintheireffectsReasonJ.BMJ2000;320:768-770.IdentifyingRiskintheLaboratoryThereisno“perfect”device!AnydevicecanandwillfailundertherightconditionsEverydeviceisdifferent!(universal2levelsQC/daydoesn’tfiteverysituation,mayneedmorefrequentQC–especiallyiffrequentstaffturnoverorunstablereagent)Anydiscussionofriskmuststartwithwhatcangowrongwithatest(errors)Labtestsarenotfool-proof,forasufficientlytalentedfool!26CLSIProject:EP23LaboratoryQualityControlBasedonRiskManagement.JamesH.Nichols,Ph.D.,ChairholderEP23describesgoodlaboratorypracticefordevelopingaqualitycontrolplanbasedonmanufacturer’sinformation,applicableregulatoryandaccreditationrequirements,andtheindividualhealthcareandlaboratorysetting27EP23LaboratoryQCBasedon

RiskManagementMedicalRequirementsforTestResultsTestSystemInformation:ProvidedbythemanufacturerObtainedbytheLaboratoryInformationaboutHealthCareandTest-SiteSettingInputInformationProcessRiskAssessmentOutputLaboratoryDirector’sQCPlanPostImplementationMonitoringContinuousImprovementRegulatoryandAccreditationRequirementsCLSIEP23TableWhereistheRiskin

theProcess?28WhatCouldPossiblyGoWrong?IdentifyingRiskintheLaboratoryFollowthesamplethroughthetestingprocess!Ateachstep,ask“Whatcouldpossiblygowrong?”Foreachhazardidentified,defineamitigationstep,awaytominimizetheriskoferroroccurring!Summarizethehazardsandactionsinaqualitycontrolplan!29StepsoftheTestingProcessPatientNeedforTestOrderPatientPreparationSpecimenCollectionLabReceiptAnalysisResultVerificationAcknowledgeResultOrderTreatmentTreatmentPreanalyticalPostanalyticalTransporttoLabAnalyticalProcessingFishboneDiagramfromCLSIEP23…RefertoAppendixAinCLSIEP18formorecomprehensivelistoferrorsourcesSourcesofLaboratoryErrorTestSystem:Reagentcontamination,deterioration,lotvariationReactionfluctuationsInadequatesamplingImproperorlossofcalibrationElectronicormechanicalfailurePowersupplyEnvironment:TemperatureandairflowHumidityLightintensityAltitudeOperator:Improperspecimenprep,handlingIncorrecttestinterpretationFailuretofollowtestsysteminstructionsCLIAInterpretiveGuidelinesforLaboratories.AppendixC,SubpartK.§493.1256Standard:ControlProcedures.InterpretiveGuidelines§493.1256(a)–(c)./Regulations-and-Guidance/Legislation/CLIA/Interpretive_Guidelines_for_Laboratories.html

33RiskManagementHolisticorglobalapproachtoqualityassuranceincorporatingpreanalytic,analyticandpostanalyticphasesoftestingEveryinstrumentordeviceisdifferentHazardanalysisandmitigationHazardanalysisdefinesthesourcesofpotentialerrorforaninstrumentordevice

andameansofdetectingorpreventingtheerror(mitigation)34ManufacturerChecks–AlternativeQCSomedeviceshaveinternalcheckswhichareperformedautomaticallywitheveryspecimen:Developmentofaline(Pregnancytest,Occultblood)Sensorsignal(bloodgasanalyzer,clots)Flowresistanceandliquidsensors(clotsorbubblesinchemistryandhematologyanalyzerpipettes)Otherchecksengineeredintoinstrument:Barcodingofreagentexpirationdates(preventsuse)LockoutfeaturesthatrequiresuccessfulQCDisposableanalyzercuvettes/pipettetips(carry-over)Temperatureindicatorinshippingcarton35QCandRiskManagementAnalyzerReagentFluidics,pipetting,electronics,motorsSomeinternalchecksdetectmechanicalandelectronicfailuresChemistryofreactionIntricatelylinkedwiththeanalyzerSusceptibletoenvironmenteffects36QCandRiskManagementTwolevelsofQCeachday,detectserrorsfrommanysourcesAnalyzer(pipets,mechanical,electrical)Reagent(storage,drift)Operator(calibration,dilution)Environment(reagentandsample)Automatedanalyzershavemanypartsandsourcesoferror,notallofwhicharedetectedbyinternalchecks.QCdoesagoodjobofdetectingsystematicerrorsthatpersistintime,butmayfailtodetectrandomerrorsoccurringwithsinglesamples.Randomerrors,likehemolysis–requiresserumindicesonchemanalyzer,visualsampleinspectionand/orphlebotomisttrainingforsamplecollectionissues37PointofCareQCandRiskManagementDeviceTestCartridgeReadstestcartridgesignalInternalelectronicchecksmaintainreaderwithinnarrowspecificationsChemistryofreactionContainsfluidics,calibrant,internalQC38QCandRiskManagementDeviceOnlyreadssignalElectronicscheckenvironment(temp)anddeviceintegrityLocksouttestingwhenoutsidespecificationsTestCartridgeActsastheanalyzerRiskfromenvironmentandoperatoreffectsQConcartridgeevaluateschemistrywitheachtestAdditionalcontrolprocessesforsampleandforoperatortechnique39QCandRiskManagementQCisexpensive,consumesreagentandlaborwithoutproducingresultsthatcanbebilled.ThepresenceofengineeredcontrolprocessesonnewerPOCTdevicesraisethequestionofwhenandhowthesecontrolscansubstitutefortwolevelsofliquidQC?Howtobestcontrolunit-usePOCTdevices?LabDirectormustbalancerequirementsforQC(2levels/day)againstcostandavailabilityofalternativecontrolswithinthetestsystemRiskmanagementandwalkingthroughthetestingprocesstoidentifyhazardsandsourcesoferrorcanhelplabsidentifytherightbalanceofliquidQCwithengineered-builtincontrolprocessesonthedevicesIndividualizedQualityControlPlanIndividualizedQualityControlPlanRiskAssessmentQualityControlPlanQualityAssessment40CLIANewIQCP

CLIA,CAPandJointCommissionRequirementsTwolevelsofliquidQCrequiredeachdayoftesting

ORLaboratorydevelopsanIndividualizedQualityControlPlan(IQCP):Balancebuilt-incontrolprocesseswithliquidQCReducefrequencyofliquidQCtominimumrecommendedbymanufacturerMaximizeclinicaloutcome,availablestaffresourcesandcosteffectivenessinthelabConsidersthelaboratoryprocessfortestingandriskoferrorsateachstepofthetestingprocess!41WhatHaveWeLearnedFromOurIQCP?QCwasexcessive.Past,i-STAT,QCthedevice(x27i-stats)-separatetestsystemsLotshipments(testoneachdevice)MonthlyQCcartridges(oneachdevice)6monthcorrelations/linearities(eachdevice)BUT…Chemistryisinthecartridge–shouldvalidateeachlotofcartridgesusingsubsetofi-stats(whicharejustreaders)Rotatei-statsHighmidandlowQClevelssameas3vialsinlinearityset–sodon’tneedtopurchaseseparatelinearity42Before:(QCthedevice)Shipments= 3QCx5daysx6x/yr= 90testsQCmonthly= 3QCx27i-stats= 81testsLinearities= 27i-statsx3levelsx3repsx2x/yr= 486tests6mocorrelations=10patientsx27i-stats= 270tests

TOTAL= 927testsAfter:(QCthecartridges,centralizedshipments)Shipments= 3QCx6x/yr= 30testsQCmonthly= 3QCx5lotcartridges(max)x12=

180testsLinearities= alreadydonemonthlyasQC= 0tests6mocorrelations=5pts/shipmentx2old/newx6x/yr= 60tests TOTAL=270testsSavingsof657cartridgesperyear43WhatHaveWeLearnedFromOurIQCPs?FalselyDecreased

GlucoseMeterResultsComplaintfromanintensivecareunitofsporadicfalselydecreasedglucosemeterresultsImmediaterepeattestonsamemeter,gavesignificantlyhigher“clinicallysensible”valuesInspectionofunitfoundnursestakingproceduralshortcutstosavetimeBottlesofteststripsdumpedoncounterinspareutilityroomSomestripsnotmakingitintotrash,fallingbackoncounterandbeing“REUSED”44RiskofErrorfrom

OpenReagentsGlucoseteststripsexposedtoairforaslittleas2hourshavebeenshowntocause-26%bias.1Stripsleftoncountersposeriskofreuse,leadingtofalselylowresults.Somemeterscatchreuseand“error”preventingaresult.Othermetersdonot!245KefferP,KampaIS.Diabetes1998;47;abs0170.SilvermanBC,HumbertsonSK,StemJE,NicholsJH.Operationalerrorscauseinaccurateglucoseresults.DiabetesCare2000;23:429-30.ReagentErrors:ExpiredReagentsCentersforDiseaseControl“Checkandrecordexpirationdatesofreagents/kits,anddiscardanyreagentsorteststhathaveexpired.”1U.S.FoodandDrugAdministration“Checktheexpirationdateonthet