3983-06 May1992 GeographicVariationintheVolume IntensityofMedicarePhysicianServices in1988: ADescriptiveAnalysis JohnHolahan StephenZuckerman SupportforthisresearchwasprovidedbytheHealthCareFinancingAdministrationtoThe UrbanInstitutethroughCooperativeAgreementNo. 17-C-99473/3-01. Thiscooperative agreementsupportedanumberofresearchhprojectsrelatedtotheMedicareVolume PerformanceStandardPolicy. Thetotalbudgetforalloftheseprojectswas$449,419and representsthesolesourceoffunding. Opinionsexpressedinthisdocumentarethoseofthe authorsanddonotnecessarilyrepresenttheviewsoftheHealthCareFinancingAdministration, TheUrbanInstitute,oritssponsors. f TABLEOFCONTENTS I. INTRODUCTION I II. VARIABLEDEFINITIONSANDMETHODS 3 IndicesDefined 3 LaspeyresIndex 3 PaascheIndex 4 Fisher'sIdealIndex 4 AreasDefined 5 Type-of-ServiceDefined 8 Methods 8 III. RESULTS 11 VariationsinAllowedCharges 11 VariationsinVolumeandIntensity 13 UtilizationinSelectedHighandLowVolumeandIntensityAreas 16 VariationsinAllowedCharges,Prices,andVolumeandIntensityby TypeofAreaandRegion 20 EvaluationandManagementServices 27 Procedures 29 ImagingandTests 30 IV. CONCLUSIONS 32 REFERENCES 35 APPENDDC ii ' r LISTOFTABLES Table 1 DesignationofMedicarePaymentLocalities 7 2 TypeofServiceClassificationSystem 9 3 AreaVariationinMedicareAllowedChargesPerEnroUee,byTypeof Service,1988 12 4 AreaVariationsinMedicareVolume/IntensityPerEnrollee,byTypeof Service, 1988 14 5 WhatTypesofServiceExhibittheMostVariations? VolumeandIntensityper BeneficiaryinAbsoluteDollars 17 6 WhatTypesofServiceExhibittheMostVariations? VolumeandIntensityper BeneficiaryRelativetotheNationalAverage 18 7 VariationAcrossAreasandRegionsinAllowedCharges,Prices,and Volume/Intensity—AllPhysicianServices, 1988 21 8 VariationAcrossAreasandRegionsinAllowedCharges,Prices,and Volume/Intensity-SummaryofResults,EvaluationandManagment Services,1988 23 9 VariationAcrossAreasandRegionsinAllowedCharges,Prices,and Volume/Intensity-Procedures, 1988 24 10 VariationAcrossAreasandRegionsinAllowedCharges,Prices,and Volume/Intensity—ImagingandTests, 1988 25 iii I INTRODUCTION I. Geographicvariationinhealthserviceutilizationhasbecomeamajorissueforresearchas weUasforpolicymakers. EvidenceonvariationsacrosssmallareasproducedbyWennbergand othershasresultedinamajorefforttoanalyzetheappropriatenessofawiderangeofhealthcare services(WennbergandGittelsohn, 1982;RoosandRoos, 1982;McPhersonetal., 1982;Moore, 1985). Onefocusofsuchresearchhasbeenonidentifyingtheamountofinappropriatecare (Chassinetal., 1987),witiithegoalofultimatelyimprovingthequalityofhealthservices availabletoindividualsintheUnitedStatesandelsewhere. Theissueofareavariationshasa largerpolicycontextaswell. Variationsinutilizationratesimpliesthatsomeindividualsmay usetoofewserviceswhileothersmayusetoomany. Lowlevelsofutilizationmayindicate problemsofaccesstonecessaryphysicianservices. Highlevelsofusemayimplyover- utilization;suchpatternsmaybetiedtoexcessiveMedicareexpenditures,amajorproblemina timeoffiscalstressingovernmentprograms. TheissueofwidevariationsinMedicareutilizationalsohasbeenrecognizedinthe PhysicianPaymentReform(PPR)initiative,legislatedinthe1989OmnibusBudget ReconciliationAct(OBRA89). ThatlegislationcalledfortheadoptionbyMedicareofarelative valuescale,withgeographicadjustmentstoMedicarefeesreflectingcostsofpractice,beneficiary protectionbylimitingcharges,andMedicarevolumeperformancestandards(MVPS),which placelimitsontherateofgrowthinMedicarefees. Thegeographicadjustmentstofeesandthe MVPSprovisionsofthePPRdirectiyandindirectiyaddressissuesofgeographicvariationsin accessandover-utilization. ThegeographicadjustmentstotheMedicarefeesthroughthe GeographicPracticeCostsIndex(GPCI),generallywillhavethedirecteffectofincreasingfees, jesc:\...\3983\06wp.ut5/22/9210:44am » relativetopre-PPRlevels,inruralareasandreducingtheminurbanareas. OnePPRgoalisto . increaseaccessintheformer. TheMedicarevolumeperformancestandardpolicy(MVPS)willtiethefutureupdatesin physicianfeestotherateofgrowthinthevolumeandintensityofphysicianservices. Asthe MVPSpolicyhasbeenimplemented,itwilltreatallgeographicareasthesame,regardlessof ^whetherareashavehighorlowlevelsofutilization. TheMVPSisrequired,however,totake intoconsiderationevidenceofproblemsofaccessandover-utilization,butitisnotspecifiedhow thisistobedone. PossiblerefinementsoftheMVPStoincludesubnationalvolumeperformance standardsmightdirectlyaddressproblemsofaccessandover-utilization. AsubnationalMVPS mightdistinguishbetweenareaswithlowlevelsofutilizationbyrecommendinghigherstandards thanforareaswithhistoricallyhighlevelsofutilization. Thispaperpresentsacomprehensivedescriptionoftheextentofcross-sectional geographicvariationinMedicarephysicianserviceuseratesacrossallservicesandacrossthe entirecountry. Weuse 1988datafromthePartBMedicareAnnualData(BMAD)systemto deriveameasureofvolumeandintensitythatcapturesdifferencesinuseratesacrossallservices, andthatcanbedisaggregatedbytype-of-service(seeBerensonandHolahan, 1990). These indicesrepresentamajorimprovementoverearlierstudiesthatfocusedonalimitednumberof procedures(e.g.,Holahan,Berenson,andKachavos, 1990)and/orasmallsetofgeographicareas (e.g.,Stano, 1986). Ingeneral,dataormethodsthatwouldallowforacomprehensiveassessment ofthisissuehasnotbeenavailabletoearlierresearchers. jesc:V..\3983\06wp.m5/22/9210:44am 2 > » II. VARIABLEDEFINITIONSANDMETHODS ??» Thispaperusesthe 1988BMAD5percentsampleofMedicarebeneficiariesto decomposeareavariationsinexpendituresperenroUeeintothecomponentduetopriceandthat duetovolumeandintensity. Becausevolumeand intensitydifferencescannotbeobserveddirectly,theyaremeasuredasthevariationin expendituresthatremainsafterpricedifferenceshavebeennettedout. Thisisanimprovement overChassinetal.(1987)thatreliedonactualservicecounts-alesscomprehensivemeasure thanvolumeandintensity. Inordertomeasurevolumeandintensity,wedeveloppriceindices thatcanbeusedtodeflateMedicarephysicianexpenditures(allowedcharges)thattakeinto considerationthedifferencesinservicemixacrossareas. Areasaredefinedasaggregationsof the240pre-PPRMedicarepricinglocalitiesthatreflectfundamentaldifferencesincommunity types(discussedbelow). IndicesDefined LaspevresIndex. Themoststraightforwardapproachtothepriceindexwouldbaseiton thepricesofanationally-representativebasketofMedicareservicesineachareaofthecountry(a Laspeyresform). However,notallservicesthatareimportantinaggregatespendingdataare providedineachareaofthecountry. Therefore,itisnotpossibletopricetherepresentative marketbasketineacharea. Second,sincetherelativeimportanceofeachservicevariesfrom areatoarea,differencesinexpendituresdeflatedbythepriceindexreflectservicemixaswellas volumeandintensitydifferences. Servicemixmayvarybecauseofgeographicvariationin relativeprices,whichisextensive(Escarce, 1990). Preliminaryanalysisofthedatasuggeststhat jesc:V..\3983\06wp.Rt5/22/9210:44am 3 servicemixvariesmuchmorecross-sectionallythanitdoesovertime,makingitessentialto addressservicemixaswellasvolumeinthispaper. Standardeconomictheorysuggeststhatanareawillreducetheconsumptionofits relativelyhigh-pricedservicesandexpandthelow-pricedservices. Thiswouldleadtolesscross- sectionalvariationintheweightsforhigh-pricedthanforlowpricedservices,ifcomputed separatelyforeacharea. Thissubstitutionbiasissuewouldsuggestthat,formostgoodsand services,across-sectionalLaspeyresindexwouldoverstatetruepricedifferences. Ifprices differentialswereoverstated,thenvolumeandintensityvariationwouldbeunderstated. However,inthecaseofphysicianservices,itappearsthatarea-levelweightsforhigh-priced servicesvarymorethanthoselow-costservices. Possibly,thisisduetothefactthatnew technologiestendtohavehigherpricesandareintroducedatdifferentialratesacrossareas. This meansthataLaspeyresindexthatreliedonnationalweightswouldtendtounderstateprice differentialsandoverstatevariationinvolumeandintensity. PaascheIndex. UsingaPaascheindexinwhichtheweightsweredifferentforeacharea wouldofcourseresolvetheproblemthatsomeservicesthatareimportantoverallarenot producedineveryarea. APaascheformoftheindexwouldprovideinformationonthepricesof thearea'smarketbasketofservicesifitwerepurchasedatthenationalaverageallowedcharges. However,therewouldbenolinktothenationalaverageimportanceofeachserviceasthereisin aLaspeyresindex.Giventhegreaterextentofgeographicvariationintherelativeimportanceof high-pricedservices,thePaaschewouldleadtounderstatementofpricevariationand overstatementofvolumeandintensitydifferences,thereverseoftheLaspeyresproblem. Fisher'sIdealIndex. ThisreasoningleadsustoprefertheFisher'sIdealindex,as discussedinZuckermanandHolahan(1991). TheindexisformedastheproductofaLaspeyres jesc:\...\3983\06wp.at5/22«210:44am 4 indexthatusesnationalweightsfortheservicesproducedinagivenareaandaPaascheindex . thatusesarea-specificweights. Animportantpointtokeepinmindinthinkingaboutthisindex isthattheservicesincludedinthecomputationforaparticularareamaydifferfromthose includedinanotherarea. Theresultingpriceindexwillmeasurethedifferenceinpricesbetween theareaandthenationforthoseservicespurchasedinthearea. (Oneshouldnottrytocompare indicesbetweenareas,ascomparisonsareonlyvalidbetweenanareaandthenationalaverage.) Thecross-sectionalFisher'sIdealwillbecomputedforeachtypeofserviceandforallservices. TheFisher'sIdealpriceindicesforeachareaforallservicesarepresentedintheAppendix. ThepriceindicesarethenappliedtodataonallowedchargesfromtheBMADbeneficiary filetocreatemeasuresofvolumeandintensityforeachgeographicarea. Individualsonthe beneficiaryfilemayuseservicesintheirownlocalityaswellasanyotherlocalityinthecountry. Todeflateallowedchargesonthebeneficiaryfileforindividualslivinginagivenareabyprice indices,forservicesproducedinthatsamearea,willleadtobiasedmeasuresofvolumeand intensity. Itisnecessarytodeflateallowedchargesforservicesusedineachareabythe appropriatepriceindexforthatarea. Thisinvolvesaggregatingallowedchargesforbeneficiaries foreachareainwhichservicesarereceived. Wethendeflatebytheappropriatepriceindexand finally,sumacrossallareastocreatemeasuresofrealexpendituresbasedontheareainwhich individualsreside. Inthiswaywehavethemeasureofvolumeandintensityofservicesusedby beneficiariesineacharea. AreasDefined The121geographicareasthatserveasthebasisofthisstudyareaggregationsofthe240 pre-PPRMedicarepricinglocalitiesandcanrepresentbroadcategoriesofareas. Specifically,we identifysetsoflocalitiesthatrepresentverylargecities,largecitiesadjacenttoverylargecities, jesc:V..\3983\06wp.ut5/22/9210:44iim 5 otherlargecities,smallcities,ruralareasandsmallcities,ruralareas,andstates. Thisisthe sameareaclassificationschemeusedinHolahanandZuckerman(1991). Table 1indicatesthe citiesorthegeographicareasthatarewithineachtypeofarea. Verylargecitiesincludetheten largestMSAsintermsofpopulation. Sinceourdesignationofcitiesisconstrainedbythe definitionofMedicarelocalities,insomeinstanceswewereforcedtogroupcitieswith SurroundingareastoapproximateanMSA. Forexample,inSanFranciscowehaveincludedSan MateoandMarincounties. Washington,D.C.includesadjacentcountiesinMarylandand NorthernVirginia. LargePennsylvaniacitiesincludePittsburgh,PhiladelphiaandScranton. Largecitiesadjacenttoverylargecitiesisagroupofcitieswithmorethan 100,000 MedicareenrolleescontiguoustooneofthetenlargestMSAs. Forexample,Oakland/Berkeleyis adjacenttoSanFranciscoandSanBemadino/RiversideisadjacenttoLosAngeles. Non-adjacent largecitiesaresimilarlysizedcities,intermsofMedicareenrollment,thatarenotcontiguousto averylargeMSA. TheseincludeBirmingham,Alabama;NewOrleans,Louisiana;Baltimore, Maryland;Dallas,Texas,andothers. SmallcitiesincludeMSAswithlessthan 100,000Medicareenrollees. Theseinclude Mobile,Alabama;StocktonandBakersfield,California;IowaCity,Iowa;Tucson,Arizonaand Tulsa,Oklahoma. SmallcitiesandruralareasincludesbothsmallMSAsandruralareaswhere itwasimpossiblebecauseoftheMedicarelocalitystructuretoseparatethesmallcityfromthe ruralarea. Ruralareascouldbedefinedin 11statesandareentirelyrural,inthattheycontainno countiesthatarepartofanMSA. Mostofthe22statesthatcouldnotbesubdividedaresmall states,butsomearenot,e.g..NorthCarolina,NewJersey,Tennessee,andColorado. jesc;V..\3983\06wp.ttt5/22/9210:44am 6 1 Table 1 Designation of Medicare Payment Localities VERY LARGE CITIES NONADJACENT LARGE CITIES SMALL CITIES SMALL CITIES/RURAL STATEWIDE San Francisco, Cal metro area Birmingham, Ala Small cities, Ala Small cities/rural Iowa Arkansas WMCUiharasbimhaciinang,ogM.taFoslnsa1,11DC metro area NNSSeoaawrncttraheOmrelCanenltadaonr,saN/o,MrCotaLlhnateCreenyt,ralCalFla MMSSeomtrbaocilcellkdet,,ocni,tACilalaeCsal, La SSSSmmmmaaaallllllll cccciiiittttiiiieeeessss////rrrruuuurrrraaaallll NMKKeayabsns MCDNooeenwlltaoawHrnaaaardmeopshire NDLeYaCrtgreomeittcri,otiMeiascr,heaPenn KBIuanfndfsiaaalsnoa,CimteNyYtrometarroeasarea MBFoarneksetnreosr,feiyeC,ladlC,alCal NSSmomaarlt1lheaccsiitttiieeTssex//rruurraall MMiincnh VNSeoorurttmhhontDDaakkoottaa Houston, Tex Rochester, NY Small cities. 111 Southeast Tex New Jersey Los Angeles, Cal SDaanllaAsn,tonTieox, Tex SImowaallCictiyt,iesI,owaInd WSemsaltlerncitTeixes/rural Wash RShooudteh CIasrloalnidna Seattle, Wash Omaha, Neb Small cities/rural Wise Utah Milwaukee, Wise Lexington/Louisville, Ky Small cities/rural Ore Alaska ADJACENT LARGE CITIES APthloaenntiax,, AGariz NS.malClentcriatliesc,itiPeesn,n NY SSmmaallll cciittiieess//rruurraall CVaonn HNaewvaadiai Oakland/Berkeley, Cal Portland, Ore Fort Worth, Tex Small cities/rural Mo New Mexico SFoarntBeLarundaedirdnaol/eR,iveFlraside, Cal SHaanrtfDoiredg,o,CoCnaln TSumcasloln,citAireisz, Ga SSmmaallll cciittiieess//rruurraall WOhVi TIednanhoessee VAEPenaonsautthguherkiSapmt,.s/SiaeCLna/otlNuaiosr,Atnha1,1N1YCCalsuburbs TANSSCktteii.wdnr.ceoiwHPLnnaaoa/ntvuuYaeeilotnrs/u,iMn,,,ignsVMCaontOoeohnawinopno,liOsh,ioMinn OTSRCSutihkalacnalstmhraaafmlh,ooeornsmBddtaOa,,okrnbC,aViCratoaynW,,nv OCkal RSUmaRlAlL cities/rural MD WMMNiyaoirosntmsehiinsgCsairppoliina Cleveland, Ohio Columbus, Ohio Rura1 Ala Dayton/Springfield, Ohio Rural La Baltimore, MD Northern Rural Cal Rural Fla Rural 111 Rural Ind Rural NY Rural Penn Rural Ariz Rural Ga Rural Ok 1. Areas are 121 aggregations of 240 Medicare Carrier Pricing localities. Source: Urban Institute, Health Policy Center jh.3900:3983-06 tl