ContractNo.: 500-94-0062 MPRReferenceNo.: 8246-080 TheImpactof ProspectivePaymenton MedicareServiceUseand ReimbursementDuringthe FirstDemonstrationYear December16,1998 JenniferSchore Submittedto: Submittedbv: HealthCareFinancingAdministration MathematicaPolicyResearch.Inc. 7500SecurityBoulevard(C-3-21-06) P.O.Box2393 Baltimore.MD 21244-1850 Princeton,NJ 08543-2393 (609)799-3535 ProjectDirector: ValerieCheh ProjectOfficer: PrincipalInvestigator: AnnMeadow RandallS.Brown ACKNOWLEDGMENTS InpreparingthisreportIreceivedtheassistanceandsupportofanumberofmyMPRcolleagues forwhichIamgrateful:RandyBrownprovidedcriticalreviewofthereportandansweredcountless questionsaboutstatisticalmethodology;theearlierworkonthisprojectofValerieCheh.Barbara Phillips,andChrisTrenholmisembodiedinthisreport'sbackgroundchapters;RobCederbaum. ElizabethStuart,andAmyZambrowskiskillfullyandpatientlybuiltandanalyzedthedatafilesupon whichthereport'sanalysisisbased;MarjorieMitchellproducedthereportwiththeassistanceof CindyCastro,andStephanieCollins:andWalterBrowerprovidededitorialassistance. If.afterall thishelp,therearestillerrorsinthereport,theresponsibilityforthemismine. iii CONTENTS Chapter Page EXECUTIVESUMMARY xi I THEPER-EPISODEHOMEHEALTHDEMONSTRATION ANDEVALUATION 1 A. THEMEDICAREHOMEHEALTHBENEFIT 2 B. THEPER-EPISODEDEMONSTRATION 4 1. DemonstrationPaymentandIncentives 5 2. OtherDemonstrationProcedures 10 C. COMPONENTSOFANDAPPROACHTOTHEEVALUATION 13 1. AnalysisofAgencyDecisionsandOperations 13 2. AnalysisofProgramImpacts 15 D. HYPOTHESESCONCERNINGTHEEFFECTOFPER-EPISODE PAYMENTONTHEUSEANDCOSTOFMEDICARESERVICES ....17 E. GUIDETOTHISREPORT 20 II DATA.SAMPLE.ANDMETHODOLOGY 21 A. DATA 21 1. IdentifyingEpisodesandRiskPeriods 22 2. MedicareClaimsforOutcomesVariables 23 3. ControlVariables 26 B. ANALYSISSAMPLE 30 C. SUMMARYSTATISTICSFORCONTROLVARIABLES 32 D. STATISTICALMETHODOLOGY 41 1. StatisticalModelsforEstimatingOverallImpacts 42 2. HypothesisTestsfortheImpactEstimates 44 3. Weighting 45 4. DesignEffects 46 5. RobustnessChecks 47 v CONTENTS(continued) Chapter Page III IMPACTSONTHEUSEOFANDREIMBURSEMENTFOR MEDICARE-COVEREDSERVICES 51 A. INPATIENTANDEMERGENCYROOMSERVICES 51 B. SKILLEDNURSINGFACILITYANDHOSPICESERVICES 54 C. NONDEMONSTRATIONHOMEHEALTHSERVICESAND PARTBHOMEHEALTH 58 D. PARTBSERVICES 63 E. TOTALMEDICAREREIMBURSEMENT 66 F. CONCLUSION 68 REFERENCES 71 APPENDIXA: ESTIMATEDCOEFFICIENTSFORKEYOUTCOMES vi TABLES Table Page II.1 OUSUETCAONMDERVEAIRMIBAUBRLSEESMDEENSTCDRUIBRIINNGGMTEHDEI1C2A0R-ED-ACYOAVTE-RRIESDKSPEERRVIIOCDE..24 11.2 STANDARDCONTROLVARIABLESFORMULTIVARIATEANALYSIS. BYSOURCE 27 11.3 WEIGHTEDMEANSFORSTANDARDBENEFICIARY-SPECIFIC CONTROLVARIABLES,BYTREATMENTSTATUS 33 11.4 WEIGHTEDMEANSFORLAGGEDDEPENDENTVARIABLES. BYTREATMENTSTATUS 37 11.5 WEIGHTEDMEANSFORSTANDARDAGENCY-ANDAREA-SPECIFIC CONTROLVARIABLES.BYTREATMENTSTATUS 40 III.1 ESTIMATEDDIFFERENCESBETWEENPER-EPISODEPAYMENTAND COSTREIMBURSEMENTININPATIENTHOSPITALUSEAND REIMBURSEMENTDURINGTHE120-DAYAT-RISKPERIOD 53 111.2 ESTIMATEDDIFFERENCESBETWEENPER-EPISODEPAYMENTAND COSTREIMBURSEMENTINEMERGENCYROOMUSEAND REIMBURSEMENTDURINGTHE120-DAYAT-RISKPERIOD 55 111.3 ESTIMATEDDIFFERENCEBETWEENPER-EPISODEPAYMENTAND COSTREIMBURSEMENTINSKILLEDNURSINGFACILITYAND HOSPICEUSEANDREIMBURSEMENTDURINGTHE120-DAY AT-RISKPERIOD 56 111.4 ESTIMATEDDIFFERENCEBETWEENPER-EPISODEPAYMENTAND COSTREIMBURSEMENTINNONDEMONSTRATIONHOMEHEALTH USEANDREIMBURSEMENTDURINGTHE120-DAYAT-RISKPERIOD ..60 111.5 ESTIMATEDDIFFERENCEBETWEENPER-EPISODEPAYMENTAND COSTREIMBURSEMENTINPARTBSERVICEUSEAND REIMBURSEMENTDURINGTHE120-DAYAT-RISKPERIOD 64 111.6 ESTIMATEDDIFFERENCEBETWEENPER-EPISODEPAYMENTAND COSTREIMBURSEMENTINTOTALREIMBURSEMENTDURING THE120-DAYAT-RISKPERIOD 67 vii FIGURES Figure page III.1 HOWPER-EPISODEPAYMENTMIGHTINCREASEOVERALL MEDICARESPENDING 52 ix EXECUTIVESUMMARY Aspartofitsongoingefforttostudymethodsofprovidingmorecost-effectivecare,theHealth CareFinancingAdministration(HCFA)hasrecentlyimplementedthePer-EpisodeHomeHealth ProspectivePaymentDemonstration. Underthedemonstration,participatinghomehealthagencies receiveafixed,lump-sumpaymentforthefirst 120daysofeachepisodeofcareprovidedto Medicare beneficiaries and a predetermined rate for each visit thereafter. This method of compensationdifferssubstantiallyfromthecurrentmethodofMedicarereimbursementforhome healthservices,underwhichagenciesarereimbursedforactualcostsincurred,uptoaspecificlimit. Byallowingagenciestoretainmostofanysurpluspaymentsovercost,prospectivepaymentgives agenciesafinancialincentivetoprovidehomehealthcareinamorecost-efficientmannerthan undertraditionalcost-basedreimbursement. Ninety-oneagenciesinfivestatesenteredthethree-yeardemonstrationatthestartoftheir1996 fiscalyears. Priortothestartofthedemonstration,theparticipatingagencieswererandomly assignedtoeitherthetreatmentgroup(whichispaidunderthedemonstration'sprospectivepayment method)oracontrolgroup(whichcontinuestobepaidunderMedicare'snormalmethodofcost- basedreimbursement). Thepaymentstreatmentgroupagenciesreceiveforthefirst120daysofa patientepisodearebasedoneachagency'sowncostsinthefiscalyearimmediatelyprecedingits entryintothedemonstration,adjustedattheendofeachyearforchangesinitscasemix. Whileeach agencyis"atrisk"duringthefirst120daysafteradmissionforallhomehealthvisitstheagency provides.HCFAreimbursestreatmentagenciesforupto99percentoffiscal-yearlossesuptothe Section223paymentlimits.' ProfitsinexcessofspecifiedlimitsmustbesharedwithHCFA. RESEARCHQUESTIONSANDMETHODOLOGY Inthisreportweexaminedatafrom(roughly)the firstyearofthedemonstrationtotest hypothesesaboutthepossibleeffectsofprospectivepaymenton theuseofMedicare-covered servicesby agencypatientsandonreimbursement forthose services. Giventhe limiteddata currentlyavailable,thispreliminaryreportfocusesonlyonservicestakingplaceduringthe"risk" period(thatis.thefirst120days)ofahomehealthepisode. Asmoredatabecomeavailable,final evaluationreportswillexaminedemonstrationeffectsduringthemonthsfollowingtheriskperiod andduringtheseconddemonstrationyear. Forthisreportwetestedhypothesesconcerningthe impactsofthedemonstrationontheuseofandreimbursementforMedicareservicesbytvpe: inpatienthospital,skillednursingfacility(SNF).hospice,nondemonstrationhomehealth,outpatient hospital,physicianandotherpractitioner,durablemedicalequipment,andotherPartBservices. (Theimpactsofthedemonstrationontheuseofhomehealthservicesdeliveredbydemonstration agenciesandonthecostofdeliveringthoseservicesarethesubjectsofotherpreliminaryevaluation reports.) 'TheSection223paymentlimitsarecost-per-visitpaymentlimitsthatapplytoallagenciesin theMedicareprogram. xi Theanalysispresentedinthisreportisbasedonapproximately51.000homehealthepisodes takingplacein87ofthedemonstrationagencies(4ofthe91agencieswereexcludedbecausethey droppedoutofthedemonstration).Alladmissionsoccurringbetweenanagency'sdemonstration startdateandAugust 1996areincluded. Medicareclaimsfilesprovideddataontheoutcome variablesdescribingtheuseofandreimbursementforservicesduringtherisk,period. Datacollected atadmissionforcase-mixadjustmentandfromearlierMedicareclaimsprovidedmeasuresof preadmissioncharacteristicsofpatientsadmittedtodemonstrationagencies. Dataonagency characteristicswereobtainedfromtheagencycostreportsandthedemonstrationimplementation contractor. Ordinary least squares models and logit models were usedto estimate program effects, controllingforpreexistingdifferencesbetweentreatmentandcontrolagenciesinpatientandagencv characteristics. Thisapproachprovedcrucialtoobtainingvalidimpactestimatessince,despitethe randomizationofparticipatingagencies,therewereseveralsignificantdifferencesbetweentreatment andcontrolagenciesasidefromthemethodofpayment. Observationswereweightedsothateach agencywasrepresentedequallyintheanalysis. Standarderrorsofimpactestimateswerecalculated usingspecialsoftwaredesignedtoaccountfortheeffectsofsampleclusteringandweighting,soas toavoidoverstatingtheprecisionoftheestimates. Analysesoftherobustnessofourregression estimatesshowedthattheywerenotsensitivetotheweightingschemeorstatisticalmethodsused. FINDINGS Althoughpreliminaryanalysissuggeststhatper-episodepaymentmarkedlyreducedthenumber ofvisitsdemonstrationhomehealthagenciesprovidedduringthefirst120daysafteradmission (Chehetal.1997).itdidnotaffecttheuseoforreimbursementforotherMedicare-coveredservices. ThisabsenceofaneffectontheuseofotherMedicare-coveredservicessuggeststhatareductionin homehealthuseatthelevelobservedunderthedemonstrationdoesnotadverselyaffectcarequality or shift costs to services in other settings. The following illustrates the lack ofeffect the demonstrationhadontheuseofdifferenttypesofMedicareservicesduringtheriskperiod. Control Treatment/Control P-value NumbersperPatientDuringthe120-DayRiskPeriod GroupMean Difference InpatientAdmissions 0.48 0.00 0.99 EmergencyRoomEncounters 0.54 -0.01 0.63 DaysinSkilledNursingFacility 2.4 -0.0 0.99 VisitsfromNondemonstrationHomeHealthAgencies 3.3 0.4 0.40 OutpatientHospitalVisits 2.5 -0.1 0.50 xii InpatientandEmergencyRoomServices Akeyconcernabouttheimplementationofper-episodepaymentwasthatthequalityofhome healthcaremightsufferasaresultoffinancialincentivestoreducethenumberofvisitsprovided orthecostofprovidingvisits.Themostseriousadverseeffectsonqualitywouldbereflectedin increasesinhospitalandemergencyroomuse,butnosuchincreaseswereobserved.Patientsof treatmentandcontrolgroupagencieshadnearlyidenticallevelsofinpatientserviceandemergency roomuseduringthe120daysfollowinghomehealthadmission.Roughlyathirdofallpatientsof treatmentandcontrolgroupagencieswereadmittedtothehospitalduringtheriskperiod,andabout athirdwenttoahospitalemergencyroomforcare. SkilledNursingFacilityandHospiceServices Evenifper-episodepaymentdidnotadverselyaffectcarequality,itmighthavecausedcareto beshiftedfromhomehealthtosomeothersetting,likeanursinghomeorhospice,particularlyifa homehealthagencyviewedapatientasparticularlycostlytoserve. However,wefoundnoevidence thatcareinSNFsorhospiceswassubstitutingfortheobservedreductioninhomehealthvisits. Roughly10percentofpatientsfromtreatmentandcontrolgroupagencieswereadmittedtoaSNF duringtheriskperiod,andunder3percententeredahospice. NondemonstrationHomeHealthServices Anumberofeventscouldleadapatientofatreatmenthomehealthagencytoreceiveservices fromanotheragencyduringthe120-dayperiodcoveredbyper-episodepayment. Someofthese eventsarenotrelatedtothedemonstrationandthusshouldbeequallylikelytooccurtopatientsof controlagencies. Forexample,itsometimeshappensthathospital-basedhomehealthagencieswill provideservicestopatientsrequiringhomehealthcarefollowinghospitalization,evenifthepatient hadbeeninthecareofanotheragencybeforethehospitalization. Ontheotherhand,theearlier dischargeofpatientsbytreatmentagencies,aclearincentiveunderper-episodepayment,isanevent thatmayleadtreatmentagencypatientstousemorenondemonstrationhomehealthservicesthan controlagencypatients. Infact.Chehetal.(1997)foundthattreatmentagenciesdischargedpatients anaverageof10dayssoonerthandidcontrolagencies. However,earlierdischarge(andsubsequent useofotherhomehealthservices)mayreflecteitherbetterorpoorercarefromtreatmentagencies. Earlydischargewouldreflectbettercareifitresultedfrommore-efficientpatientteachingandbetter coordinationwithcommunityservices. Nonetheless,patientsmaybecomeaccustomedtoreceiving homehealthcareandinahighlycompetitivehomehealthmarket,maybeabletofindotherhome healthagenciestoprovideservices—eventhosenolongerstrictlynecessary. Ontheotherhand,if treatmentagenciesdischargedpatientsinappropriatelyearly,theuseofotherhomehealthcarewould reflectpoorercare. Alltheseevents—thoseinducedbyper-episodepaymentreflectingbetteror worsecare,aswellasthoseexternaltothedemonstration—arereflectedinourmeasuresofhome healthprovidedtodemonstrationpatientsbynondemonstrationagencies. Treatment/control differences in the receipt of and reimbursement for services from nondemonstrationhomehealthagenciesweresmallandgenerallynotstatisticallysignificant. Eight xiii percentofcontrolagencypatientsreceivedhomehealthservicesfromanagencyotherthantheir demonstrationagencyduringtheriskperiod,ascomparedwithninepercentoftreatmentagency patients. Thisdifferencewasstatisticallysignificantonlyatthe10percentlevelanddisappeared whenwedroppedonelargetreatmentgroupagencywhosepatientshadarateofnondemonstration agencyserviceuseaboutfourtimestheaverage. Furthermore,treatmentandcontrolagencypatients whodidreceiveservicesfromnondemonstrationagenciesreceivedroughlysimilarnumbersofvisits fromsuchagencies(about40). PartBServicesandOverallReimbursement Per-episodepaymentalsohadnoeffectontheuseofPartBservices. Nearlytwo-thirdsof patientsfromtreatmentandcontrolagencieshadanoutpatienthospitalvisitduringtheriskperiod, oftenforlabtestsandx-rays. Nearlyall(92percent)sawtheirphysicians(orotherpractitioners); almosthalf(46percent)purchaseddurablemedicalequipment:andmost(80percent)usedotherPart Bservices. Becauseper-episodepaymenthadnoeffectontheuseofMedicare-coveredservices,itcomes asnosurprisethatithadnoeffectonPartA,PartB,ortotalMedicarereimbursement. Onaverage, patientsoftreatmentandcontrolagencieshadpartAreimbursements(exclusiveofdemonstration homehealth)ofroughly$4,600andPartBreimbursementof$2,000duringthe120daysfollowing homehealthadmission,foratotalofjustunder$6,600(orabout$1,650permonth). Wedidnotestimatedifferencesindemonstrationhomehealthservicesinthisanalysisandthus havenotincludeddemonstrationhomehealthreimbursementinourtotals. Wedidnotestimate differencesintheuseofdemonstrationhomehealthservices,becausetheywerewasthesubjectof anotherprojectreport(Chehetal.1997). Estimatesoftheeffectofper-episodepaymentonthecost ofhomehealthcare(thatis.thecosttothehomehealthagencyofprovidingcare)willbethesubject ofafuturereport. Regression-adjustedestimatesoftheeffectofper-episodepaymentonhome healthreimbursement—thecostofcaretotheMedicareprogram-arenotinformative,because reimbursementtotreatmentgroupagencieswasset,bydesign,a:predeterminedlevelsbasedon agencyreimbursementpatternsduringtheyearbeforetheagencyjoinedthedemonstration.By contrast,controlgrouppaymentwasbasedonthenumberofvisitsprovidedtocurrentpatients. Thus,becausethepaymentmechanismsforpatientsoftreatmentandcontrolagencieswerenot comparable, estimating regression-adjusted impacts on home health reimbursement was not appropriate. We note, however, that Medicare reimbursement for home health services provided by demonstrationagencies duringtheriskperiodaveraged $3,067 forpatientsofcontrol group agencies,comparedwith$3,090forpatientsoftreatmentgroupagencies.(Theunadjusteddifference betweenthesetwoamounts,weightedforagencysize,wasnotstatisticallysignificant.) Therefore, totalMedicarespendingduringthe120-dayriskperiodwasjustover$9,600.orabout$2,400per month. xiv