WORKING PAPERS Population,Health,and Nutrition
Populauonand HumanResources Department
The WorldBank February1992 WPS852
How HealthInsurance Affectsthe Delivery of Health Care in DevelopingCountries JosephKutzin and HowardBarnum
The goal of delivering health services efficiently and equitably can be more effectively promoted by an insurance institution that actively organizes consumers' entry into the health system and removes the financial incentives that encourage providers to increase the volume and cost of services.
Policy Research Working lapers dissemninateLhcfindings of work in pnogressand encourage thecechange of ideas among Bank staff and all others interested in dcvelopment issues Thcsepapers, distrbuted by the Research Advisory Staff, carry thenames of the authors,reflect and conclusions are the authors own. They should interpretations, only their views, and should be used and cited accordingly.The rtndings, not be attributed to the World Bank, its Board of Directors, its management, or any of its member countries.
This paper - a product of the Population, Health, anid Nutrition Division, Population and Human Rcsources Department-is part of a largereffort in the department to develop efficient and equitable health sector pricing and insurance strategies. Copies of the paper arc available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Otilia Nadora, room S6-065, extension 31091 (20 pages'. February 1992. To alleviate financial crises, many developing countries ate considering health insurance as an option for increasing available resources in the health sector. But besides affecting revenues, insurance also affects how efficiently and equitably health services are delivered. To undersLandhow insurance affecLsthe delivery of health services, Kutzin and Barnum studied systems in Brazil, China, Korea, aiid Zaire. They looked at the following characteristics of insurance programs: the system for reimbursing providers, the services covered, the insurer's role, the extent to which beneficiaries help cover costs, and the proportion of the population covered by insurance. Kutzin and Bamum use the following indicators for efficiency and equity in the delivery of health services: cost escalation, resource allocation, the use of specific medical technologies, and equity of access to services. They conclude that insurers must take an active role in establishing institutional mechanisms (such as contractual obligations) that encourage health service providers to make efficient and equitable decisions about resource
allocation. Incentives to providers are important because they determine the supply of services and can also tremendously affect demand. As examples from Brazil, China, and Korea show, providers can increase the use of curative services so much that health care costs escalate rapidly, resource allocation in the sector is distorted, medical technologies are inappropriately used, and access to services is inequitable. To correct these distortions, a publ.c insurance institution should create incentives that encourage providers to behave in a manner consistent with social goals. This is achieved most easily with direct insurance, where the goals of the insurer and the provider are identical, but thirdparty payers can also take an active role. Problems promoting efficient and equitable delivery of health services are magnified when an insurer serves merely as a financial conduit for reimbursing providers. Goals of efficiency and equity can be more effectively promoted by an insurance institution that actively organizes consumers' entry into the health system and removes the financial incentives that encourage providers to increase the volume and cost of services.
ThePolicy Research Working PaperSeriesdisseminates the findingsofwork under way in the Bank. Anobjectiveofthc series is to get these findings out quickly, even if presentations are less than fully polishcd. The findings, interpretations, and conclusions in these papers do not necessarily represent official Rank policy. Produced by the Policy Research Dissemination Center
Table of Contents Introduction .............................
................ .1.... . I
Overviewof InstitutionalCharacteristics ..................... Country Cases.5.
Brazil Prior to 1983.. Brazil After 1983.. China ................. Republicof Korea.. Zaire's BwamandaHealth Zone.. Conclusionsand Recommendations..
6 9 12 15
.............. .............. .
Cost Escalation ............................................ ResourceAllocationand Referral ................ MedicalTechnologyIncentives ................ Accessand Equity ................ References .................
17 17 18 19 19 20
Figures and Tables Figure 1. Real Per Capita RecurrentHealth Expenditurein China, by Source .11 Table 1.
Character:sticsof Health InsuranceSystems.
Observedor ExpectedEffectsof Health InsurancePrograms.
Distributionof PaymentCategoriesin the BwamandaHealth Zone Populationand HospitalPatients, Dec. 1988- Oct. 1989 (maternitypatients excluded) . ...... ..
HOW HEALTH INSURANCE AFFECTS THE DELIVERY OF HEALTH CARE IN DEVELOPING COUNTRIES*
INTRODUCTION Financial crisis is a common state of affairs in the government health sector of many developing countries, and an increasing number are considering implementinguser charges and insuranceprograms to shift some of the financialburden for health servicesaway from direct budget allocations by a health ministry. Although they are often implementedas ways of mobilizing additionalresources, prices and insurancealso affect the allocationof health resources by changing the signals sent to producers and consumersof health services. Changes in incentivesengendered by these alternativefinancingprograms, therefore, have implicationsfor the efficiencyand equityof health servicesdelivery, in additionto their more obviousimpacton revenues. This paper examines institutionalaspectsof insuranceprograms in four developingcountries--Brazil,China, Korea, and Zaire--and assesses the impact of each on the efficiency and equity of the health sector. Much attention is given to insurance reimbursementof hospitals because these represent the largest componentof national health expenditures and are the focal point for much of the activity in the sector. The case study countrieswere selectedbecause their hospitals are financed largely, and in some cases entirely, by sources other than government budget allocations and because excellent descriptions of their health financing systems exist. Understanding how alternative financing programs have distorted the allocation of health resources and how these distortions might be mitigatedis importantfor these countries and others consideringchangesto their present system of health care financing. Different forms and combinationsof user fees and insurance are associated with different incentives to consumers and providers of health services. These incentivescan beneficiallyor perversely affect the performanceof both hospitalsand the health sector more generally. For each country case study, the characteristicsof the insuranceprogram(s) are described along a numberof dimensions,including: 1) reimbursementsystem; 2) servicescovered by insurance3) insurer role; 4) extent of consumer cost-sharing; and 5) extent of population coverage. These features are discussedin more detail in the next section of the paper, and Table 1 summarizesthem for each of the case study insuranceprograms. The incentivescreated by these specific characteristicsaffect the performance of the health sector. "Performance"can be interpretedmanyways, so we propose severalootentialefficiencyand equity effects which can be compared across programs. These include: 1) cost escalation; 2) the allocationof sectoral resources between secondary/tertiaryand primary care, and the effectiveness of referral; 3) the use of specific medicaltechnologies;and 4) equity of accessto health services. Where sufficient informationis available, we assess the impacts of insuranceprograms on each of thesedimensionsof performance. A discussionof lessonslearnedconcludeseachcountrycase study.
' The authors wish to thank Bill McGreevey for helpful comments on an earlier draft.
2 Table 1. Characteristicsof HealthInsuranceSystems Country/ Program Brazilbefore 1983
Reimburse-| ment System
Populatioi1 Coverage Near
Retrospective Curative; Gatekeeper case-based possibilityfor requirement preventive should with SUDS reduce
Sameas above;SUDS plans moveto universal
Retrospective Predominantly fee-for-service curative
Not for most of the insured
Retrospective Mostcurative fee-for-service services;some high-tech excluded
Substantial officialcostsharing;plus unofficial
Zaire's Bwamanda Health Plan
Directprepaid w/elementsof retrospective case-based
Hospitalcare if beneficiary has been referred
Active; 20% for most accessto services;none hospitalis for delivering effectively motherswho monitored obtained
60% of rural population voluntarily enrolled
The analysis is limited to the extent that insufficientdata exist to quantify the impact of a health insuranceprogram on any of these dimensionsof performance. However, where observed changesare consistentwith the incentivesin place, we infer causalityas to the direction (and in some cases, the magnitude)of changesin these indicators. Table 2 summarizesthe observed or expected effects of the case study insuranceprograms on the four indicatorsof health sector efficiency and equity.
OVERVIEW OF INSTITUTIONAL CHARACTERISTICS The reimbursementsystem refers to the manner by which providers are paid out of an insurancefund. We classifythe reimbursementsystemsof insuranceschemesintotwo major types: third party retrospectivereimbursementand prepaidcapitationhealth care organizations. Third party retrospectiveplans involve remunerationof the providerby a separate insurancefund after services have been delivered, whereas prepaid capitation organizationseither serve as both insurer and provider (direct insurance)or pay a separateprovider (third party arrangement)a fixed amountout
Table 2. Observed or Expected Effects of Health Insurance Programs
Choic,. of Technology
Access and E, uity
Brazil before 1983
Skewed toward privately provided curative care
Provider induced use of profitable high-tech services and drugs
Very high payments f.r relatively few patients; supply skewed toward urban areas
Brazil after 1983
Probably still high, but few data
Case mix shift with AIH; SUDS should yield better mix of curative and preventive
Incentives for particular technologies remain, but to L lesser extent
No noticeable change with AIH; SUDS should provide universal access, but supply constraint likely to remain in rural areas
Skewed toward curative care in higher level facilities
Powerful financial incentives to provide new hightech services and drugs
Consumptionof services heavily skewed toward the insured
Consumers chose high-tech hospitals despite higher cost sharing; legal change requiring referral should help
Unregulated hightech services pushed as a means of competing for patients
Poor insured at disadvantage due to high copayments and unofficial payments to providers
Referral requirements effectively encouraging consumers to first use health centers
No available information
Zaire's Bwamanda Health Plan
Goal of increasing hospital resources being met
Strong evidence of moral hazard and adverse selection; possibly access problems for the uninsured
of which the provider will serve the health needs of the program's enrollees for a specified period. Retrospective reimbursement plans typically use prospectively determined prices in the form of fee schedules that indicate the units of service which are reimbursable under the program. The most common forms of retrospective reimbursement are fee-for-service payment systems, which involve charging for each individual service consumed by a patient, such as inpatient bed-days, drugs, and x-rays. An alternative to this for hospital inpatient care is case-based reimbursement. With this method, there is one global charge based on the category of patient admission (for example, lung cancer admissions, maternity admissions, etc.) irrespective of the type and quantity of specific services provided during the patient stay. The most well-known example of case-based reimbursement is the Medicare hospiial payment system in the United States. where admissions are
4 grouped into categories, called Diagnosis Related Groups (DRGs), based on their clinical characteristics. However, as is shown below, there are examplesof case-busedreimbursementin developingcountriesas well. Insuranceprogramsvary as to the coveredservicesthat are reimbursable. Differentincentives are likely to be generatedif the programcovers only curativeservicesrather than both curative and preventivecare. Another importantaspect of this feature is whetherspecific curaiive services are excluded from coverage, or if there are administrativeprocedures that must be followed (referral requirements,for example)for servicesto be reimbursedunder a program. This characteristiccan have implicationsfor resource allocation, technology choice, and the effectiveness of a referral system. An importantfeature of insuranceprograms is whetherthe insuring institutionplays an active or passive role with regard to the providers of care. Is the insuring institutionmerely a fir.ancial cornduit,or does it attemptto enforce cost disciplineon providers? A useful way of analyzingthis issue is to assess the extent to which the insurer and the provider have common interests. On a purely financial level, we might assume that the insurer wants to minimizepayments to providers, whereasproviders want to maximizepaymentsfrom the insurer's revenues. With direct insurance, the provider and insurer are the same institution,so this system lends itself to a very active insurer role. However, third party arrangementscan involvea whole spectrum of roles, from laissez-faire to active monitoringand review of provider activitiesby the insurer. Cost-sharingrefers to the amount that health insurancebeneficiaries(consumers)have to pay each time they use serv'rvs. Forms of cost-sharinginclude deductibles (a set amount which the beneficiarymust pay beforerec' iing insurancebenefits)and copayments(the amountof the charges for specific services which insurance does not pay, excludingthe deductible). Economistsoften recommendthat cost-sharingbe includedin insuranceprogramsto discourageoveruse of servicesby insured persons,who wouldotherwiseface a zero price at the point of service. Insurancecreatesan incentivefor behavioralchangeon the part of the insured, who are protected from severe financial loss and are thus likelyto demandmore servicesthan the uninsured. This additionaldemand is called moral hazard and is conceptuallysimilar to what is oftentermed "unnecessaryutilization"arising in free care systems. Cost-sharingis intendedto limit the effects of moral hazard. The extent of population coverage is the final characteristicof insurance programs to be considered in this paper. As with the other characteristics, this can exhibit a wide range, from universalcoverageto only a small percentageof the populationwith coverage. An importantfeature to consider is whether insurance is voluntary or is mandatedby law for specific groups of the population. Where coverage is voluntary, the viability of insurance can be threatened by adverse selection,the tendency for persons likely to consumea higher than average quantity of services to purchase insurance. Where coverage is mandatedfor specific groups of the population, equity problems may arise because the uninsured face much higher prices than the insured. This equity issue is also a concernin voluntarysystemswhere some peoplemay not be able to afford to purchase insuranceand may thus have limited financialaccessto services as a result.
COUNTRY CASES Brazil Prior to 1983 Mostcurative personalhealth care in Brazil was financedby the NationalInstitutefor Medical Care and Social Security (INAMPS), the health insurance componentof Brazil's payroll-financed social securitysystem. Preventivecare and basiccurativeservicesfor the poor were providedby the Ministry of Health, which was financed through governmentbudget allocations and administered separately from INAMPS. The social security system has grown rapidly, from coverage of 23 percent of the population in 1963 to over 90 percent in 1982 (McGreevey, 1982). Although insurance coverage has become nearly universal, there is great disparity in the supply of health providers betweenthe relativelywell off south and southeastern ivgonsof Brazil and the rest of the country (Briscoe, 1990). Until 1983, INAMPS reimbursedphysiciansand hospitalson a fee-forservice basis for each service providedto patients accordingto an official price list which included 2,600 items (the HospitalGuide (GIH) system). INAMPSpaid all costs; coveredpersons were not responsiblefor any cost sharing. Under this system, INAMPSserved largely as a third party payer for servicesileliveredby private providers. In 1981for example,85 percent of hospitalizationspaid for by INAMPSwere in private contracthospitals,while only about 9 percent (including2 percent in INAMPS' own facilities) were in public sector facilities (McGreevey, 1988). Thus, INAMPS under the GIH systemcouldbe characterizedas a retrospectivefee-for-servicereimbursementsystem covering curative services only. INAMPS served a3 a passive financial conduit; it did not try to imposecost consciousnesson providers, and the absenceof cost sharing meant that consumersalso had nothingto gain by limit-agtheir utilizationof services. The implicationsof these characteristics are describedbelow. Cost escalation. Largely as a res"lt of the incentivesof the curative care reimbursement system, health expendituresincreasedby more than 20 percent annuallyduring the 1970s,and public sector health expenditureas a share of GDP rose from I perzent in 1949to 3.7 percent in 1975and 5.6 percent in 1982(World Bank data). From the late 1960suntil the early 1980s, the number of private hospitalbeds increasedby over 40 percent, and the hospitalindustrygrew faster than the rest of the economy(Rodrigues, 1989). The fee-for-servicereimbursementsystem encouragedhospitals to increaseboth the number of patients and the amountof services providedper patient as a means of generating profits. The private provider network "responded to a set of incentivesthat paid [providers]whateverthey billed to providevirtuallyunlimitedservicesto patients who bore none of the costsdirectly" (Briscoe, 1990). The retrospectivefee-for-servicesystem coupledwith the passive role played by INAMPS with respect to monitoringprivate providers was probablyresponsiblefor the widespread fraud (e.g., billing for non-existentpatients and procedures by the private health provider network)discoveredin the mid-1980s. Resource allocation. The curative health subsystem, financed through payroll taxes, developed independentlyfrom the Ministry of Health's public health programs, and grew with increasedemploymentand economicexpansionin the 1960sand 1970s. The MOH budget's share of GDP declinedover this period. Thus, the explosivegrowthof total health care spending referred to above occurred entirely in the market for curativeservices. The result of these forces was that the share of government(MOH plus INAMPS)health expendituredevotedto curative (medicaland hospital) as opposed to preventiveservices rose from 36 percent in 1965to 85 percent in the early
6 1980s. Hospitals alope accountedfor nearly 70 percent of public health expenditures,up from less than 40 percent in the 1960s(Briscoe, 1990). M .ical technologyincentives. There were no discerniblefinancialor institutionalincentives for private providers financed by INAMPS to limit the provision of certain treatment technologies which were the most profitable. Indeed, McGreevey (1988) suggeststhat the 1982 INAMPS fee schedulepayment rate for a physicianconsultationwas quite low relative to the value of a doctor's time, whereas capitalizedmedicaltechnologyembeddedin pharmaceuticalsand medical equipment was priced well above cost. Doctors were thus financiallyencouraged by the system to overuse services, drugs, tests, facilities, and operations in order to enhance their incomes. One service encouragedby the reimbursementsystem was caesareandeliveries, which in 1981accountedfor 31 percent of all births, giving Brazil the highestrate of any countryin the world. Another exampleis diagnostic services, particularly x-ray, use of which grew at a very rapid annual rate from 1970to 1981. Reportedly, many of these tests could have been avoidedwithoutdetracting from treatment. In 1979,the Rio de Janeiro State UniversityHospitalreducedthe numberof x-raysby 40 percent and found no loss in diagnostic efficiency. These and other types of complementaryexaminations(the number of which grew at 22 percent per year between 1979 and 1981)could have been eiiminated with no effect on the treatment ultimatelyadministeredto patients (McGreevey, 1988). Access and equity. The incentivesof the reimbursementsystem encouragedrapid expansion in the volume of services provided to individual patients. One result was the allocation of a substantialshare of total public health expenditurestoward costly high technologyservices, such as renaldialysis,heart bypassoperations,and computedtomography(CT) scans, which benefitrelatively few people. The magnitudeof this effect was such that in 1981, total expenditureon 12,000 high cost patients was greater than the amount spent to providebasic health services and disease control for 41 million people in the poor north and northeast regions of the country (McGreevey, 1988). Thus, the incentivesfor usingexpensivehigh-technologyservicesclearly led to an inequitablepattern of expenditures. Lessons learned. The experienceof Brazil's social securityfinancedhealth insurancesystem prior to 1983sheds light on a numberof importantissues. Fee-for-servicereimbursementof private providers with no beneficiarycost-sharingby a public agency exercisingno control over utilization led to provider-inducedincreases in the volume of medical services beyond what was medically necessary. The use of specific technologiesand services with the greatest profit marginsgrew at particularly rapid rates. This resulted in rapid growth of expenditureson curative services and distortions away from a cost effective mix of preventive and curative care. This distortion was especiallypronouncedbecausethese serviceswere financedseparately(INAMPSfinanced curative, the MOH financed preventive); therefore, cross-subsidizationof preventive care by the rapidly growing curative services was not possible. Brazil After 1983 The Brazilian government implementeda number of reforms to its hospital reimbursement system during the 1980sin an attempt to control the growth in hospital costs and utilizationwhich had arisen from the fee-for-serviceinsurancesystem. A pilot programof financingand administering health serviceswas implementedin the relativelyprosperousstate of Parana. This system, calledthe Curitiba Plan for the state capital, had two distinguishingfeatures:
7 1) Initial patient contact with health providers (triage) occurred at an INAMPS or other public (state or municipal) facility, where walk-in patients would be examined by a physicianauditor (employed by INAMPS). The physician-auditor either treated the patient or referred him./her for hIigherlevel set vices as needed. Walk-in patients were not authorized to go directly to private facilities. 2) The INAMPS physician-auditor assigned the patient to a service group, and case-based reimbursement rates were tied to assignment, regardless of the treatment setting (e.g., public clinic, private hospital, etc.). Facilities and providers are reimbursed a fixed amount per case, and only for the procedures specified by the physician-auditor. If a hospital treated a patient at less cost than the reimbursement amount, it profited. If treatment costs exceeded the case payment rate, the hospita oore the financial loss. These features, which were implemented more broadly throughout INAMPS in 1983 as part of the Hosp;.al Certification System (AIH), were aimed at many of the problems inherent under the GIH system. Prior *o implementation of the Curitiba Plan, nmanywalk-in patients were treated by private hospitals as emergency cases, whick received substantially higher fee-for-service reimbursements from INAMPS. The public triage requirement created a structured referral system to address this. In addition, by separating the decision on the course of treatment from the facility which would actually provide (and be reimbursed for) the care, the financial incentive to overprescribe and overuse complementary examinations and services in private facilities was diminished.' Under the initial Ctiritiba project, hospitalizations were reduced to 5 percent of initial consultations, compared to a Brazilian average of 6.5 percent. By mid-1982 (the Curitiba pilot began in 1981), INAMPS declared that the plan had reduced hospital admissions by 30 percent (McGreevey, 1988). Unfortunately, however, we are not aware of any subsequent follow-up studies that would demonstrate ti- long term effects of these changes. INAMPS reforms altered some, but not all, of the key characteristics of the health insurance system. The reimbursement system remained retrospective, but on a case basis rather than fee-forservice. There were still no cost-sharing provisions for INAMPS beneficiaries. By requiring entrance to services via public primary care providers (i.e., its own employees), the role of INAMPS in enforcing cost discipline on private providers became more active.2 Under the AIH system, the financing and administration of curative and preventive care remained separate, but in 1987 the Unified and Decentralized Health System (SUDS) was announced with the aim of correcting this situation. The objective of SUDS was to unify the functions and resources of INAMPS and the MOH
I This would be true to the extent that the INAMPSphysician-auditordid not also work in pnvate practice or for a privatehospital, or was immuneto influencefrom privateproviders. One problemof the systemprior to 1983 notedby McGreevcy(1988) was a conflict of interestarising because most physicianswho were employedin public clinicswere also holdingpositionswith at least one privatehealth care provider. Indeed,a 1981survey of physiciansin GreaterSao Ptulo foundthat they average3.5 jobs each. This conflictof interestcontinuesto be an issue fc. he post-1983INAMPS physician-auditors,as the reformed system places tremendousresource allocatior. -owers in t.. hands of these public employees. Private contractorsclearly ha-e much to gain by trying to influencethe I hysician-auditors'decisions. 2 Other activities undertakenby INAMPS reflectiveof its more active role vis-a-vis the private provider network includedimplementationof more systematicand thoroughauditsof privatecontractors,establishmentof publicreference centers for high-technologyservices, and modificationof proceduresby which high-costitems were procured by private contractors(Briscoe, 1990).
8 at the state and municipallevel. In this way, curativeand prevartive serviceswould be financedby the same iistitution. While the incentivesinherent in this reform fo, redressing resource allocation problemsare e!ear, Brazil has beenslow to implementthe plan (Briscoe, 1990), and thus its effects have not yet been observed. Cost escalation. There are few empiricaldata with uhich the impactof INAMPSreforms on health care cost escalationsince 1983can be ar:alyzed. A study of Brazilianhospitalizationduring the 1980s(Rodrigues, 1989)concludedthat the changeto cas.-based reimbursementdid not appear to be effective in controllinghealth care costs becduseit created an incenitivefor private hospitalsto increase certain types of admissions. In addition, although a system of prospectively set reimbursementrates encouragesprovidersto economizeCathe levelof resourcesallocatedto the care of an individualpatient, such a system also provides incentivesfor hospitalsto maximizethe volume of profitableadmissions. Howeve. to date there have beenno expendituredata availableto sompare the effects on total hospital costs of the fee-for-servicesystem with the case-basedsystem.' Resource allocation. The case-basedreimbursementsystem created inceiitivesfor private hospitalsto adjusttheir inpatientcasernix. Witha limitednumberof case paymentgroups, there will be patients of varying severity within each category. INAMPS reportedlyset payment rates based on average reimbursementwithin a category (Rodrigues, 1989). One way that hospital, could minimizetheir resource inputs per case would be to treat the less severe patients within a category and try to shifLsicker patients to public hospitals. Rodrigues concludedthat private hospitalsdid, in fact, shift more costlypatients to public and universityhospitals, leaving themselveswith a less severe case mix for which profit marginsper patient were greater. Supportingevidenceis provided by Briscoe (1990), who examined inpatient mortality data and concluded that patients in public hospitalshad significantlymore severe conditions,on average, thai. those in private hospitals. In terms of resource allocation between cuiative aid preventive caru, the trend towards a decreasing preventive share apparentlyreversed itself, as 78 percent of health expenditures were devoted to curative care in 1986, down from 85 percent in 1982 (World Bank data). This change occurred before the SUDS was announced,and although this approach would seem to allow for a reallocationof health resources toward a more cost effective mix of preventive and curative care services, there are as yet no data upon which an evaluationof this reform can be made. Medical technology incentives. There is little informationavailable on the utilizationof specific medicaltechnologiessince 1983. The case-basedpaymentsystem should have diminished the overuse of complementarydiagnostic tests and other services which was prevalent under the earlier system. In addition, some controlsover the use of high-technologyserviceswere instituted. Nevertheless,Briscoe(1990)reports that INAMPSpaid over three percent of its total budgetin 1988 on just four outpatienttreatment procedures(hemodialysis,hemodynamicstudies, endoscopy, and computerizedtomography). Thus, further controls or limitations of INAMPS financing of hightechnologyservicesmay be needed.
The World Bank's LAIPH divisionis proposingto analyze newly accessibledata on costs and utilizationin private hospitalsin Brazil coveringthe period 1981-1990. Hopefully,this will yield conclusionsas to the effects of the change in reimbursementsystemson cost escalation.
Lessonslearned. Brazil's attemptsto addressthe deficienciesof its pre-1983health financing system appearto generatebetter incentivestor a more efficienthealth care delivery system, but there are insufficientdata availableto clearly evaluatethe impactof various reforms (thoughthis may be remediedby the proposedstudy mentionedin the previous footnote). The CuritibaPian and later the AIHicreated an institutionalframeworkfor movingconsumersthrough the health care systeni in an organized manner and for limitingprovider-induceddemandby requiringthat the gatekeepersto the system be employeesof the insurtr, and that these employeesdeterminethe total course of treatmrent to be provided. Pixed reimiburse..i,nton a case basis rather than fee-for-serviceshould reducc the number of unnecessary procedures that are perfornied. Finally, unifying the resou,ces and administrationof the institutionsresponsiblefor deliveringboth curative and preventive services should allow for cross-subsidizationand the possibilityof a more cost effectivehealth care delivery system. 4 China
In 1981, China's health facilities were instructed to cover their non-staff recurrent costs through user charges, paid either out-of-pocketor through insurance. Although facilities are all publiclyowned, this change inducedthem to operate in a mannersimilar to private institutions. By 1988, 82 percent of recurrent health expenditures were estimated to have been made through a combinationof patient fee payments and insurancereimbursement. In that year, about 20 percent of the populationwascoveredunder governmentand employer-sponsored(thoughheavily subsidized tl,rough the tax system) health insurance that reimbursed health providers 100 percent of billed charges. Another IOto 15 percentof the populationis estimatedto have more limitedcoverage(i.e., with cost-sharingprovisionF). Reimbursementunder these programs is on a retrospectivefee-forservice basis, and insurers are third-partypayers separate from providers and playing no role other than financing services. Insurance primarily pays for curative services, but disease prevention activitiesfor which prices are chargedare coveredas well. The pricingof health serviceshas create,iincentivesfor providersto favor the acquisitionand use of particulartypes of services and technologies. Fees for technologiesand serviceswhich were in use in the late 1950s (e g., outpatientconsultations,inpatientcare, surgical operations)were set below costs (withoutan accurate idea of what cests were) at that time, and even with attempts to increase prices for this category of services in 1985, fees for most of these "older" medical techno'ogies, procedures,and services remain considerablybelow cost. Older price guidelines do not exist for new technologies, and a number of these (e.g., CT scans, laborato±'yservices, ultrasound,coronarycare, ar' dialysis)appearto be priced abovecosts. Therefore, hospitalsbenefit financially when these newer services are consumed. Another category of prices is for drugs. Hospitalsare explicitly allowedto mark up the price of drugs by 10-18percent over the wholesale price for whichthey acquirethem. The mark-upfoe traditionalmedicinedrugs averages20 percent. While the mark-up is intendedto cover storage and distributioncosts, there are clear incentivesto overprescribe. Cost escalation. Total healthexpendituresgrew at an annualaveragerate of 17percent in real terms frorn 1980 to 1988, increasingfrom 2.( percent to 3.2 percent of GDP during these years. Mostof this growth has resultedfrom increasesin patient fee payments(average annualgrowth rate
A forthcomingWorldBank study (Bumgarner,ed.) is the source of informationprovidedon China in this section.
of 23 percent) and insurancereimbursements(15 percent annual growth). Although there are no availabletime series utilizationdata, the need for most health institutionsto earn revenuessufficient to coveroperationalcostsand some capitalcosts has almostcertainlyresultedin provider-encouraged consumptionof those health services on which a profit can be earned. The pricing of health (and particularlyhospital)services, coupledwith the knowledgethat for most of the insured population, no monetarycost willbe borne for theseservices,has encouragedhealth care providersto offer more services, and services with a greater margin betw -en price and marginalcost. Res- Arceallocation and referral. Another consequenceof the incentivesof the financial decentralizationreforms in China is that governmenthealth expenditures, and to a much greater extent total health expenditures,have been increasinglydevoted to higher level hospitals, while a decreasingshare has been devotedto lower level hospitals. Dataon governmenthealth expenditures show that the higher level hospitals' share rose from 38 percent in 1980to 44 percent in 1987, while basic hospitals fell from 23 to 17 percent over the same period. Although anti-epidemicand maternal/childhealth activitiesremainedaround 17 percent of publicexpenditures,the growth in the percentageof total health expendituresfinancedby fees and insurance (from 72 percent in 1980 to 81 percent in 1987), and the preponderanceof fee/insurancepaymentsdevoted to hospital services meant that the share of total health expendituresdevotedto tasic and preventiveservicesundoubtedly fell over the period. The insuranceprogramsdo not exerciseany influenceto steertheir beneficiariesto appropriate levels of the referral network. The lack of effectivereferral systemshas resulted in fee paying (and especiallyinsured)patientsseekingcare from higherlevelfacilitieswhen lesscomplexfacilitieswould have been sufficient. To the extent that future investmentdecisionsare based on (distorted)historical utilizationtrends, too great a share of the public investmentbudgetwill be devotedto higher level hospiials. Currentdistortions,therefore, not only imposeshort run costs in terms of cost ineffective utilizationbut also long run inefficiencyvia investmentdecisionsbased on current patternsof facility demand. Medical technology incentives. Hospitalsnegotiatefees for new technologicalserviceswith local health and pric- bureaus, typicallyon the basis of the likely "average cost" of the service and assumptionsregardingthe volume of services likely to be provided. Accordin-gto the World Bank study, when hospitalsare engagedin the price settingnegotiations,they have an incentiveto project moderatelevelsof utilization,but after the price is established,they have a powerful incentiveto use the service intensivelyto generate financial profits. Case studies of the service costs and prices charged for specifictechnologiesshow there to be strong non-medicalincentivesto provide several services, includingelectroencephalography,coronarycare, CT scans, ultrasound,and renal dialysis. Pricingpolicy for drugs has led to a defacto subsidyfor the domesticpharmaceuticalindustry by the health system, and particularly the insurance systems, whicb merely reimburse providers without any efforts to monitor the cost-effectivenessof utilization patterns. The effects of the incentivesto overprescribeare reflectedin an average of 2.3 drugs prescribed per patient visit and by the nearly 50 percent of total health expendituresgoing for drugs. This is not just a financial problem; polypharmacyis clearlycontraryto sound medicalpracticeand can have deleterioushealth effects. Accessand equity. The combinationof high user fees and insurancecoveragefor only part of the populationhas resulted in an inequitat -pattern of health care utilizationwhereinthe provision
11 1980 Chinese FRv3
1980 Chinese FIv1B
20 -020 0
01980 1981 1982 1983 1984 1985 1986 1987 1988 _
Source Bunrgarner,ed. (forthccrning)
Figure 1. Real Per Capita Recurrent Health Expenditurein China, by Source of serviceshas been skewed away from the majorityof the population. Insurancereimbursements were the source of 44 percent of health finance in 1988, yet only about 20 percent of the population was covered under official insuranceprograms in that year. Figure 1 shows that, on a per capita basis, consumptionof health care by insured persons has been muchgreater (and is rising) than by uninsured persons, suggestingsome moral hazard on the part of the insured and possibly limited access for low income uninsuredpersons. Uninsuredpeople are receivingless health care, and the rapid growth in fee payments during the 1980shas exacerbatedthe equity problem because these persons are using a greater share of their incomesfor health servicesthan are the insured. Lessons learned. As in Brazil under the GIH system, China experiencedrapid cost escalation resulting from third party fee-for-servicereimbursementof 100 percent of health service charges. In a system where only some of the populationhas insurancecoverage, there is clear evidenceof differential utilizationof services, suggestingmoral hazard on the part of the insured and possibly limited access for the uninsured. China's experience also highlights the importance of price incentivesfor the use of particularmedicaltechnologies. The signals sent to providershave resulted in the expansionof profitablehigh-technologyservices and drugs. In addition, because investment decisionsare based on existing signals, the long run consequencesof today's distorted price signals are likely te be an even more inefficientmix of facilitiesand equipment. In short, althoughChina's decentralizedpricing system and insuranceprogramshave been effectivein generatingrevenuesfor
12 health facilities,thev have done so at the cost of equityin the use of servicesand inefficiencyin their organizationand production. Republicof Korea' The first compulsorysocial securityhealth insurancescheme coveringless than 15 percent of the populationwas introducedin 1977,and the beneficiarypopulationhas been rapidly increaseduntil universalcoveragewas achievedin 1989. Insuranceis providedthrougha large numberof nonprofit, noncompetingsocietiesorganized at the firm, firm-group,or countylevel. 'Thereare programs for the unemployed and self-employed,plus public assistance programs which essentially pay the insurancepremium for indigent persons (about 10 percent of the population). There are two large administrativeagencies which handle claims processingand payment, data collection, and program monitoring for all of the insurance societies. They work closely with the Ministry of Health and occasionallyengage in monitoringof claims and utilizationreview (Yang, 1991). Services are largely delivered in private facilities, and the insurance agencies and the government(through the MOH) are somewhatmore active than are similar agencies in Brazil and China vis-a-visproviders. In mid-1989,a law was passed requiringpatients to have a referral slip from clinics beforetheyare allowedto use hospitalcare, changingthe prior situationwhereinpatients couldchoose the providerlevel at which they entered the health care system. The MOH also defines the servicesthat are covered by insuranceand ultimatelydeterminesthe prices for covered services. However,a numberof high-technologyexpensiveservices, such as CT scanners and Extra-corporeal Shock Wave Lithotripsy(ESWL)are not covered, and providers can charge unregulatedprices for these (Yang, 1991). A nationalscheduleof fees is negotiatedand then set on a cost-plusbasis, and providers are retrospectivelyreimbursedon a fee-for-servicebasis accordingto this schedule. Un!ikethe programs in Brazil and most health insurancein China, there are significantcost sharing requirementsin the Korean program, except for certain categoriesof the medicallyindigent. The combinationof third party benefit paymentsand cost sharing serves to finance health care, which is largely providedby private facilities. A combinationof deductiblesand copaymentsyield effective coinsurancerates of 62, 65, and 41 percent for outpatientservices in general hospitals(highest level), hospitals, and clinics, respectively. The inpatientcoinsurancerate is 20 percent at all facilities. However, extra paymentsto senior physiciansand other undocumentedexpensesmean that actualcost sharing rates are even higher (Yang, 1991). Cost escalation. Attemptsto sort out the effects of the expansionof insurancecoverage on total Korean health expenditures are confoundedby the effects of the rapid growth in per capita income (ahout eight percent annually) that Korea has experiencedduring the last two decades. Nevertheless,it is Iikelythat the expansionof health insuranceplayed an importantrole in the growth of total health spending from 2.8 percent of GNP in 1976to 4.2 percent in 1985 and an estimated 7.0 percent in 1989 (Yang, 1991). For example, a study (Kim et al., 1986)of one rural province
s A World Bank Technical Paper (De Geyndt, 1991) has recently been published providing insights into cost escalation and Korea's national health insurance program. The paper assesses alternative strategies for mitigating the rise in health care expenditures. We were not able to incorporate this paper's findings into our analysis because the technical paper was not available at the time this section was completed.
13 found that insurancecoverage effecteda 37 percent increase in inpatientadmissionrates and a 66 percent increase in inpatientdays per 100persons. In addition, the number of visits to outpatient departmentsat clinics and hospitals . -eased by 71 percent after the introductionof insurance coverage. Yang (1991) found that total expendituregrowth resulted from a combinationof increasesin cost per case (a nearly threefold increase for inpatient services and a near doubling of outpatient services from 1980to 1988, comparedto a consumerprice index increaseof only 60 percent over the same period) and in the volume of servicesused (outpatientvisits per capita nearly doubledand hospital admissionsper capita rose by about40 percent). He suggeststhat one importantcause of the increases in both the average cost and volume of services has been the retrospectivefee-forservice reimbursementsystemusing cost-pluspricing. Under this system, cost-increasingbehaviors are always rewarded, and this is exacerbatedover the long run as higher costs induce higher fees. Despitethe presence of significantcopaymentsin the system, hospital utilizationdata indicatethat there has been a growing tendency to increase the volumeof services provided; the frequencyof repeat visits is rising and lengthsof hospitalstay are increasing. This suggeststhat incentiveson the demandside (i.e., high copayments)have not been very effectiveat limitingdemand in the face of rapidly rising incomes. Resourceallocationand referral. There has been no clear distinctionamong the roles of clinics, hospitals, and general hospitalsbecause these have all been competingwith each other for patients (Yang, 1991). The insuredshow a markedtendencyto use large general hospitalseven for primary care; these hospitalstend to have higher proportions of specialists and advanced medical eqtuipmentand are accordinglymore costly. Using hospitalsis also the most expensiveoption for a consumer: they are allowed to charge fees which are about 15 percent more expensivethan at clinics, and the perceintagecopaymentfor outpatientcare is 50 percent greater. Despitethese higher charges, however, the utilizationof general hospitalshas risen at almost the same rate as that of clinics in recent years (Anderson, 1989). The pattern of overuse of the more sophisticatedgeneral hospitalswhich is becoming establishedin Korea may give signals that there is a need for more of thesefacilities, which couldgeneratestill more undesirablecost escalation. The referral requirement introducedin mid-1989appearsto be an appropriatedevice for steering patients toward lower level providers for entry into the health care system, but it is too early to tell how effectivethis has been in practice. Medicaltechnologyincentives. Becauseprices are set by the government,Korean hospitals cannot competefor patientson the basis of price and must do so on some other basis. By and large, this other basis has been the availability of senior medical staff and of sophisticatedmedical equipmentbecause of the strong societal perceptionthat anything less is not good medical care. Therefore, it is not surprisingthat hospitalsare acquiringlarge quantitiesof modernequipment. One exampleis CT scanners, which are now in nearlyevery hospitalwith morethan 200 beds. The costs of these devices generally are greater than the revenuescollected from their use, but hospitalsuse them as loss leaders to attract patients. There are 38 hospitals with approved heart transplant facilities, and Korea also has 26 ESWLs. That these figures are high is evidencedby a comparison with Canada, which has 32 heart transplantcenters and only four ESWLs (Yang, 1991). Access and equity. Results from studies undertakenprior to the universalizationof health insurancesupportthe conclusionthat insurancecoverageleadsto moralhazard, and coverageof only part of the populationresults in inequitableaccessto care. Anderson(1989) reports that two years
14 after a 1981 demonstrationproject introducedinsurancecoverage in three rural provinces,inpatient service use increased by nearly 75 percent and outpatient use by 80 percent. Comparison of utilizationand chargesamongstinsuredand uninsuredinpatientsat one universityhospital foundthat the average volumeof servicesand lengthof stay for five diagnostic categorieswere greater for the insur..d, whereasout-of-pocketcharges were greater for the uninsured (Yu, 1983). Universal coverage under Korea's national health insurance program should theoretically provide for universalaccess and better equity implicationsthan had previouslyexisted. However, as in Brazil, smaller per capita supply of physicians and facilities in rural areas has led to geographically-basedinequity. Another equityconcern is the very high coinsurancerates which are charged in this system. With the exceptionof some categories of the medically indigent, these charges are levied irrespectiveof patient incomeand thus are more of a burden for poorer families. Data on per capita health service utilizationby income class for insured persons in 1987showsthat utilizationby the poorer insured was considerablyless than that of higher income insured persons (Yang, 1991). In addition,the capacityto make informal "under the table" paymentsand to pay for non-coveredhigh-technologyservices has led to a two-class system of health care. Moreover, accordingto Yang, manyproviders are unwillingto serve public assistanceprogram beneficiaries, even thoughthey are legally entitledto care. Anothersource of inequityin this system of universal coveragederives from the large numberof insurancesocieties. Becauseeach has a relativelysmall numberof beneficiariesand serves a small geographicarea, some serve largely healthy, high-income persons, and others have to contendwith a less healthy risk pool. With no mechanismin place for premium to be transferred across insurancesocieties, the higher-riskpools will tend to have higher paymentsand fewer resources than lower-riskpools. Lessons learned. Universalcoverageshould provideuniversalaccess, but the heavyofficial cost-sharing obligations which are not income-related(except for the poorest groups) and the unofficialpayments demanded by private providers weigh most heavily on lower-incomeinsured persons. Despite these significantbeneficiarycost sharing requirementsin Korea's fee-for-service reimbursementsystem,however, the introductionand expansionof insurancehas been associatedwith rapid growth in health and hospitalcosts. Demandside approachesto encourageappropriateuse of referral facilities have not been very effective. This suggests that the key to encouraginga more efficient service delivery system is on the supply side, by providingstrong incentivesto providers to limit the volume of services which they provide and to use more cost effective treatment technologiesand settings. Clearly, such incentivesare not present in Korea's health insurance system, with the exceptionof the recent requirementfor patientreferral. AlthoughKorea's insurance program uses its strengthas the principlepurchaserof servicesto negotiate a nationalhospital fee schedule, an unintendedconsequenceof this has beenthat private hospitalscompetefor patientson the basis of perceivedquality. Becauseconsumers' understandingof hospital and medicalpractice is usuallylimited, there may often be great differencesbetweenperceivedqualityand actualquality. As a result, this competitionhas taken the form of hospitalsaccumulatingtechnologicallyadvanced equipmentto attract patients. This has createda distorted pattern of investmentwhich will lead to an increasinglyinefficientmix of services in the future.
Zaire's BwamandaHealthZone6 The focus of this case study is only on one small region of Zaire, not a national system of hospital financing as was the case in the other examples. In many ways, it is not comparablewith the other programsdiscussedin this paper becauseit is focused aroundone district reference hospital rather than an array of providers that includessophisticatedtertiary hospitals. Nevertheless,several importantlessonscan be drawn from the health plan in this region. In 1975, Zaire developeda plan to organizeits health servicesaround a large numberof health people with a reference hospital and satellite health zones, each serving approximately 100,UOO given a considerable degree of autonomy, includingresponsibility The health zones were centers. to develop cost recovery programs to meet their operating and maintenancecosts (Bitran et al., 1987). The managementof one zone, Bwamanda,created an insurance program as a means of generatingrevenuefor its referencehospitalaiid organizingthe delivery of servicesin the zone. The insurance plan is managed by the health zone and is thus a direct insuranceprogram rather than a third party arrangement, and it is administered as a prepaid capitation health care organization. Enrollmentin the plan is voluntary, but in an attempt to limit adverse selection, all membersof a family are required to join if one memberjoins. For residentsof the zone who are not enrolled in the plan, case-based fees are charged, with 16 case payment rates related to the diagnostic characteristicsof the patient. Insured persons who have been referred from a health center pay a copaymentequal to 20 percent of the case price charged to uninsured residents of the zone. In addition, the hospital charges non-residentstwice the resident rate and salaried employees persons (whoseemployersare requiredby Zairian law to pay the full cost of their medicalcare) 250 percent of this rate. In a sense, there are two insuranceplans in the zone; one for the voluntary enrollees (60 percent of the zone's populationin 1989)and employer-paidcoverage(4.6 percent). Unlikethe other insuranceprogramsdiscussedabove,the Bwamandainsuranceplan only covershospitalservices (plus chronic care treatment in health centers). However, hospitalizationbenefitsare only available to memberswho can documentthat they were referred from a satellitehealth center. The BwamandaHealthPlan and the requirementfor employersto pay the health care costs of salaried employeesgeneratedistinct financialincentivesto providersand consumersfor each of three populationgroups: the insured, the uninsured,and the employer-paidgroup. The insuredgroup has paid its premium and thus wishesto maximizeits utilization,though this is limitedby its 20 percent cost sharing and the requirementto pay for care at health centers. The hospital and the insurance plan (the same organization)wish to minimizethe use of services by this group, as it does not want claims paymentsto exceedpremium income(the prepaid capitationincentive). The uninsured try to avoiduse of the hospital becausethey face high prices if they are admitted. The hospital wants this group to use the hospital, but hopes to minimizethe level of inputs providedto any person admitted so that treatment costs do not exceed the case payment rate (the retrospectivecase-basedincentive). The employer-paidgroup does not face any charges, so they will maximizetheir use of the hospital. The hospitalalso has incentivesto admit employees(but to still minimizeresourceinputsduring their stay), as employerspay at a muchhigher rate than that paid by other patients. Cost escalation. The impact of the BwamandaHealth Plan on cost escalation cannot be discussedin the same contextas the impactsof the nationalinsuranceprograms previouslydescribed.
Information on the Bwamanda health zone in Zaire is derived from Shepard et al., 1990.
16 Data on the real level of the annual premium show that it has increased at an annual rate of 9.5 percent since 1986. This is not indicativeof any problem; rather, the insurance program was established in 1986 with the goal of increasing resources for the reference hospital of the zone. Evidencefrom the first few years under the plan show that this goal has been reached. 100percent of hospital charges for plan beneficiarieswere coveredby premium income in 1987 and 1988, with additionalresources left over to cover administrativecosts. Cost re(c3veryin the hospital increased from about 48 percent of operatingcosts in 1985to 79 percent in 1988. Resource allocation and referral. Even though the insurance plan does not cover charges in health centers (exceptfor care of chronic conditions),it includesstrong incentivesto discourage self-referral to the hospital. Specifically,the plan will pay for hospitalcare only if a beneficiaryhas a referral slip from a health center. Insured patientswho self-referto the hospital must pay the full (non-covered)fee for a private physicianvisit. The incentivefor first using less complex facilities is thus built directly into its reimbursementrules. Another aspect of the plan which encourages appropriateuse of primarycare facilitiesis whatinsured womenare entitledto a free hospitaldelivery only if they have receivedprenatal care. The health plan takes a very active role in managingthe use of facilities in the zone, a task simplifiedbecausethe providerand insurer are the same entity, thus makingreferral rules relativelyeasy to enforce. Access and equity. The incentives in place for the three Table3. Distribution of PaymentCategoriesin the groups of the population are borne BwamandaHealthZonePopulationand HospitalPatients, out by hospital utilizationdata. The Dec. 1988- Oct. 1989(matemitypatientsexcluded) results of a sample of hospital _ ____ H i patients by paymentcategory during Payment Population Hospital Ratioof 1988-1989are presentedin Table 3. Category t%) Patients Patientsto The table clearly showsthat salaried l (%) Population employees were heavily Employed 4.6 17.3 3.76 overrepresented (relative to their I populationshare) in the hospital,and Insured 60.2 76.7 1.27 the insured plan members were also Not insured 35.2 6.1 0.17 overrepresented. The non-insured were very much underrepresented. Source: Shepardetal., 1990 Based on the hospital register of admissionsby paymentsource, 1989 hospital admission rates were also found to be lower amongthe uninsuredthan those insured throughthe plan (by nearly 700 percent), and much lower (nearly twenty-fold)than those whose fees were paid by their employers. This suggestsa combinationof limited access for the very poor (as there are apparently no exemptions from payment), moral hazard arising from the financial protection provided by insurance and employer payment, and the hospital's incentiveto admit salaried employeesto receive the highest payment rate. The enrollment in the plan of individualswho believed they were likely to use the hospital (adverseselection)was also probablya factor in the overrepresentationof plan membersin the hospital. Lessons learned. The purpose of establishinga health insuranceprogram in Bwamandawas to increasethe level of resourcesavailablefor health services. Increaseshave been achieved,yet the allocation of resources has been tightly managed and utilizationof non-hospitalservices strongly
17 encouragedby the program. The direct insurancemodel whereinthe insurer and provider are the sameentity offers advantagesover third party payment in terms of organizingthe health care system efficiently. For example,no specific referral incentivesto providers are needed; therefore, referral rules are more easiiy enforced. The principal drawbackof the voluntary insuranceapproach used in Bwamanda is differential access to care between the uninsured and the insured (including employer-paid). The greater use of services by the insured reflects moral hazard arising from insurancecoverage, but may also mean that accessfor the uninsured is limited by their inabilityto pay. The latter is of greater concern, but the magnitudeof tlhisproblemis unknown. Another issue related to the BwamandaPlan is that it has a monopolyon the provision of hospital services in the region. This may limit the replicabilityof this modelin other regions unlessthey have the resources and the will to assure maintenanceof the quality of serviceprovision. Nevertheless,the successof the BwamandaHealth Plan in organizing and sustainingthe services provided in the health zone demonstratethat insurancecan be a viable option for financinghealth services in a rural area with a populationprimarily composedof self-employedfarmers.
CONCLUSIONS AND RECOMMENDATIONS The principalrecommendationfor the design of health insuranceprograms derived from these case studies is that insurers should take a very active role in establishinginstitutionalmechanisms (such as contractualobligations)that encourageproviders of health services to make efficient and equitable resource allocation decisions. Incentives to providers are most importantbecause they determinethe supplyof servicesand also have tremendousleverageover demand, due to consumers' dependenceon providers for assistancein definingtheir personalhealth care needs. By controlling supply and strongly influencingdemandfor services,providers can increasethe use of curativecare servicesin a mannerthat, as the examplesof Brazil, China,and Koreashow, leadsto rapid escalation in health care costs, distortion of resource allocation in the sector, inappropriate use of medical technologies,and inequitableaccessto the services available. To correct these distortions, a public insuranceinstitutionshouldactivelyinterveneby creatingincentivesto encourageprovidersto behave in a manner consistentwith socialgoals. This is most easily achievedwith direct insurancebecause the goals of the insurer and providerare identical, but it is possiblefor third party payers to take an active role as well. The followingparagraphssuggestadditionalconclusionsand recommendations derived from the case studies. Cost Escalation
In each of these countries, alternativefinancing mechanismswere introduced as a means of increasingthe level of resources availablein the health sector and/orreducinggovernmentsubsidies. Thus, it is not surprising that each has been associated with rapid growth in health and hospital expenditures. Cost escalation is problematicwhere the incentivesof a specific financing program cause expendituresto spiral out of control. Lessonslearned from these case studies suggest that a fee-for-service pricing system coupled with third party reimbursementof private providers (or providers who behave as if they were private, i.e., exhibit profit- or revenue-maximizingbehavior, as in China) for billed services leads to large and rapid increases in health care costs via provider incentivesto increasethe volumeand sometimesthe average cost of services. Theory indicatesthat
18 beneficiary cost-sharing should limit this tendency, but the presence of very high copayment requirementsin Korea apparentlydid little to mitigatecost escalation.(An incomeelastic demandfor health care services in Korea explains some but not all of their health cost increases since the late 1970s.) This suggests that providers have a very important impact on the quantity of services ultimatelydemandedby health consumers(giventhe ignoranceof most consumersof their medical treatmentneeds and options),and that to achievegreater cost containment,incentivesmust discourage providers from expandingthe volume and cost of services beyond what is cost effective. To the extent possible, insuring institutionsacting in the public interest should create incentivesso that the fipancialinterestsof providers correspondwith their ova.i. One possible model for this is the aspect of the Brazilianiusurancesystem after 1983that gives public primary care providers responsibility for defining the course of a patient's treatment. This componentof the program separates the decision on the serv to be provided to a patient from the provider who will ultimately be reimbursedfor pros u;' the defined services. Retrospectivecase-basedreimbursementis presumedto result in less cost escalationthan a feefor-servicesystem, but evidencefrom Brazilto date is inadequateto supportthe presumption. Casebased reimbursementdoes not eliminatethe incentiveto increasecosts; the incentiveis to maximize profitable admissions and minimize the quantity of services provided per admission. This arrangementis likely to somewhatlimit cost increasesbut may have negative implicationsfor the quality of care and require either competitionor regulationto maintain standards. Furthermore, where a mix of private and public providers is financedon this basis, as in Brazil, private hospitals have an incentiveto alter their case mix toward less severe, more profitablepatients, sending sicker patients to the public hospitals. This can result in an overly strainedpublic system. Resource Allocation and Referral
Becausethird partyfee-for-servicereimbursementby a passiveinsurer generatesrapidly rising curativehealth care costs, it tends to lead to a declinein the relativeshare of health resourcesdevoted to preventive services. This does not necessarily imply a decline in resource availability for prevention,but for most countries,the marginalbenefitof additionalresources in preventiveservices would be greater than in curative services. This impact can be mitigated if the provision of preventiveservicescanbe subsidizedout of the revenuesgeneratedby the expansionof curativecare. Such cross-subsidizationis most feasible when there is a unitary authorityresponisiblefor all levels of health care, from primary prevention to hospital services. Two models for this are the direct insuranceapproachused in Bwamanda,where the healthzone is responsiblefor all levels of care but generatesrevenuesthrough the provision of curativecare, and the SUDSin Brazil, which unifiesthe resourcesand administrationof curativeand preventiveservices. Both approachesshould be closely monitoredto determinehow effectivelythey allocateresources. Financingsystemscan generate incentivesfor the use of specific types of facilities via price incentivesto consumersor by fiat. Evidencefrom Korea suggeststhat price incentivesmay not work very well if perceivedquality differencesbetweenlevelsof facilitiesis great. Making reimbursement contingentupon a patient's use of specified points of entry into the provider network appearsto be more effective. In Bwamanda, this functions as a price incentive, with a very large penalty tor insured persons who self-refer to the hospital. The post-1983Brazilianplan to require entry to the health system via a public primary care provider also creates a structured pattern of referral and generates provider incentivessimilar to that of the direct insurance model (since the provider is employedby the insurer).
19 Medical TechnologyIncentives Incentivesthat influencethe use of a medicaltechnologyarise from how thesespecificservices are priced, as is evident from the experienceof Brazil and China. Providers have an incentiveto maximizeutilizatiunof those servicesfor which the marginof reimbursementover costs of provision is greatest, and consumers both rely on the medicaljudgmentof providers and often tend to equate new high-technologyproceduresand a greater quantity of pharmaceuticalswith better health care. Whenthe price or reimbursementrate differ from the marginalcost of the service, the incentivemay distort the provision of services away from what is socially optimal and medically appropriateby introducingpersonal or institutionalfinancial considerationsinto the choice of technologydecision. The long term efficiency consequencesof this distortion can be severe, as investmentin future capacityis often basedon current utilization. The problemscan be mitigatedto some extentby using case-based rather than retrospective reimbursement,but overprovision incentives may remain, dependingon the nature of the case paymentcategories. Insurersshouldstriveto remove either price distortionsthrough improvedunderstandingof the marginalcostsof provisionor be consciousof the incentivesthey wouldlike to createregardingthe acquisitionand use of specificmedicaltechnologies, includingpharmaceuticals. Access and Equity Insurancecoverageis a meansof accessto medicalserviceswhich might otherwisebe difficult for manypersons if prices for services, especiallythose provided in hospitals,were high enoughto recover costs. When coverage is universal, as is practicallythe case in Brazil and has recently becometrue for Korea, differencesin accessarisingfrom insurancestatusshould not occur, although the poor appearto suffer a degree of limitedaccess due to a relative lack of providers in poor rural areas, and, in Korea, a heavier burden of formal and informalcost-sharingobligations. However, when only part of the populationhas such coverage, as in Chinaand Zaire's Bwamandahealth zone, disparities are likely, as the cost of using health servicesfacing the uninsured as compared to the insured population is radically different. Partial insurance typically exacerbates equity problems becausethe insured also tend to have higher incomesthan the uninsured. The disparity in utilization of servicesdoes not necessarilyimply that accessfor the uninsuredis inadequate,but access for the most vulnerable groups should be examined in detail and measures should be taken, such as exemptions from payment or public assistance for the purchase of insurance (as in Korea), to guaranteea minimumstandardof access to care for all citizens.
REFERENCES Anderson,G.F. (1989). "UniversalHealthCare Coveragein Korea." HealthAffairs 9(Summer):2434. Bitran, R., M. Mpese, , S. Bavugabgose,M. Kasonga, N. Nsuka, T. Vian, and K. Wambenge (1987). Zaire Health Zones FinancingStudy. Resourcesfor Child Health (REACH)Project, USAID, Arlington, Virginia. Briscoe, J. (1990). Brozil: 7he New Challengeof Adult Health. A World Bank Country Study. Washington,D.C. Bumgarner, J.R., ed. (forthcoming). China: Long-Term Issues and Options in the Health Transition. A Wor!d Bank Country Study. WashingtonD.C. De Geyndt, W. (1991). ManagingHealthExpendituresUnderNationalHealthInsurance: 7he Case of Korea. World Bank TechnicalPaper No. 156. Washington,D.C. Kim, I.S., S.H. Yu, H.J. Kim, Y.M. Chae, K.Y. Rhee, and M.S. Sohn (1986). "Impact of Regional Health Insurance on the Utilization of Medical Care by the Rural Population of Korea." YonseiMedical Journal 27(2):138-146. McGreevey,W.P. (1982). "BrazilianHealth Care Financing and Health Policy: an International Perspective." Population, Health and Nutrition Technical Note No. GEN 6. Washington, D.C.:The World Bank. McGreevey,W.P. (1988). "The High Costs of Health Care in Brazil." PAHO Bulletin 22(2).145166. Rodrigues, J. (1989). "Hospital Utilizationand ReimbursementMethod in Brazil." International Journal of Health Planningand Management4(1):3-15. Shepard, D.S., T. Vian, and E.F. Kleinau(1990). "HealthInsurancein Zaire." Policy, Research, and External Affairs WorkingPaper No. 489. Washington,D.C.:The World Bank. Yang, B.M. (1991). "HealthInsurancein Korea: Opportunitiesand Challenges." HealthPolicyand Planning 6(2):119-129. Yu, S.H. (1983). "Studyon Hospital Care ServicesBetweenInsured and Non-insuredPatients for SelectedDiagnoses in Korea." YonseiMedicalJournal 24(1):6-32.
Policy Research Working Paper Series
Conitact for paper
WPS828 Howthe MacroeconomicEnvironmentArvil VanAdams RobertGoldfarb AffectsHuman Resource TerenceKelly Development
V. Charles 33651
WPS829 Regulationof SecuritiesMarkets: SomeRecentTrendsand Their Implicationsfor EmergingMarkets
TerryM. Chuppe MichaelAtkin
F. Harbottle 39616
WPS830 Fixed Parityof the ExchangeRate and EconomicPerformancein the CFAZone: A ComparativeStudy
V. Barthelmes 39175
WPS831 Real Overvaluation,Termsof Trade IbrahimEibadawi Shocks,and the Cost to Agriculture in Sub-SaharanAfrica
V. Barlhelmes 39175
RichardB. Norgaard WPS832 Sustainabilityand the Economics of AssuringAssetsfor FutureGenerations
J. Shin Yang 81418
WPS833 Stabilizationand GrowthRecovery in Mexico:Lessonsand Dilemmas
L. Franchini 38835
WPS834 Scenariosfor Growthin the 1990s
J. Queen 33740
PatricioArtau Stijn Claessens
S. King-Watson 31047
WPS836 Sourcesof IncomeInequalityin Rural RichardH. Adams,Jr. Pakistan:A DecompositionAnalysis HaroldIlderman
C. Spooner 30464
WPS837 ManpowerPlanningin a Market Arvil VanAdams Economywith Labor MarketSignals John Middleton Adria;.Ziderman
S. Khan 33651
WPS838 MeasuringTrade Policy Intervention: BrianJ. Aitken A Cross-CountryIndex of Relative Price Dispersion
R. Martin 39065
WPS839 RegionalIntegrationUnderVERs: DavidG. Tarr WhenTradeDiversionis Unambiguously Beneficial
D. Ballantyne 37947
WPS840 PublicSector Debt,Fiscal Deficits, and EconomicAdjustment:A ComparativeStudyof Six EMENA Countries
L. Ly 37352
0. Nadora S6-065
AlfredoE. Thorne Azita Dastgheib
WPS841 How Accessto ContraceptionAffects SusanCochrane Fertilityand ContraceptiveUse in David K. Guilkey Tunisia
Policy Research Working Paper Series
Ccntact for paper
WPS842 CapitalFlowsto SouthAsiaand lshrat Husain ASEANCountries:Trends, KwangW. Jun Determinants,and Policy Implications
S. King-Watson 31047
WPS843 How FinancialMarketsAffect Long- EjazGhani Run Growth:A Cross-CountryStudy
A. Nokhostin 34150
WPS844 Heterogeneity,Distribution,and Cooperationin CommonProperly ResourceManagement
WPS845 InflationStabilizationin Turkey: Luc Everaert An Applicationof the RMSM-XModel
B. Mondestin 36071
WPS846 IncorporatingCost and CostEffectivenessAnalysis into the Developmentof Safe Motherhood Programs
LarryForgy DianaM. Measham Anne G. Tinker
0. Nadora 31091
WPS847 Copingwiththe Legaciesof SubsidizedMorigageCredit in Hungary
Silvia B. Sagari Loic Chiquier
M. Guirbo 35015
WPS848 HowEC 1992and Reformsof the MerlindaD. Ingco CommonAgriculturalPolicyWould Donald0. Mitchell Affect DevelopingCountries'GrainTrade
P. Kokila 33716
WPS849 FinancialStructuresand Economic Development
W. Pitayatonakarn 37666
WPS850 FiscalAdjustmentand the Real Kazi M. Matin ExchangeRate:The Caseof Bangladesh
D. Ballantyne 38004
WPS851 Sourcesof WorldBankEstimates of CurrentMortalityRates
0. Nadora 31091
WPS852 How Health InsuranceAffects the Joseph Kutzin Deliveryof HealthCare in Developing HowardBarnum Countries
0. Nadora 31091
WPS853 Policy Uncertainty,Information Asymmetries,and Financial Intermediation
W. Pitayatonakarn 37664
EdujardBos My T. Vu PiNtienceW. Stephens