The roles of the government andthe market in health

World spending on health totaled about $1,700 bil- tion, and other sectors important for health,as lion in 1990, or 8 percent of global income. Of this, well as through regulation of health systems, governments spent more than $1,000 billion, or health providers, and insurers. Governments fur- nearly 60 percent. Of the $170 billion spent on ther affect health by their impact on household health in the developing countries of Africa, Asia, income and educational levels (as discussed in and Latin America, governments spent half the Chapter 2), by financing public health services, total amount-2 percent of those regions' GNP. In and by providing care directly. What governments the established market economies, where total do varies enormously from country to country, but health spending was almost $1,500 billion,govern- every government plays an important role. ments spent just over $900 billionmore than 5 Three economic rationales justify and guidea percent of GNP (Table 3.1). The sheer size of these government role in health. They are discussed in expenditures on health makes it critical to under- greater detail in "The rationales for government stand the impact of government policies on peo- action," below. ple's health. But governments profoundly influ- The poor cannot always afford healthcare ence health in less direct ways, through their that would improve their productivity and well- policies toward education, water supply, sanita- being. Publicly financed investment in the health

Table 3.1 Global health expenditure, 1990

Public sector health Total health Health expenditure Percentage Per capita Ratio of Percentage expenditure expenditure as percentage of GNP health per capita of world (billions of as percentage of regional spent on expenditure spending Demographic region population dollars) of world total total health (dollars) (SSA = 1) Established market economies 15 1,483 87 60 9.2 1,860 78.9 Formerly socialist economies of Europe 7 49 3 71 3.6 142 6.0 Latin America 8 47 3 60 4.0 105 4.5 Middle Eastern crescent 10 39 2 58 4.1 77 3.3 Other Asia and islands 13 42 2 39 4.5 61 2.6 India 16 18 1 22 6.0 21 0.9 China 22 13 1 59 3.5 11 0.5 Sub-Saharan Africa 10 12 1 55 4.5 24 1.0 Demographically developing countries 78 170 10 50 4.7 41 1.7 World 100 1,702 100 60 8.0 329 13.7 Note: SSA, Sub-Saharan Africa. Source: Appendix table A.9.

52 of the poor can reduce poverty or alleviate its education bring. The second factor is the amount consequences. and effectiveness of expenditure in the health sys- Some actions that promote health are pure tem. The third factor is the range of diseases pre- public goods or create large positive externalities. sent, which is determined largely by climate and Private markets would not produce them at all or geography. Effective health policy takes account of would produce too little. different disease prevalences but is not simply de- Market failures in health care and health in- termined by them. surance mean that government intervention can Differences in health spending are an obvious raise welfare by improving how those markets starting point in the search for an explanation of function. differences in health. In 1990 total annual health Any potential benefits from greater public sector spending ranged from less than $10 per person in involvement in health must be weighed against several African and Asian countries to more than the risk that governments will in fact make matters $2,700 in the United States. There was also consid- worse. For example, to satisfy special interest erable variation within regions. In Africa, Tanzania groups, governments may adopt policies that re- spent only $4 per capita for health in 1990, while duce the general welfare. Even when they choose Zimbabwe spent $42 per person. In Asia, Ban- correct policies, they may fail to implement them gladesh spent $7 per person each year, as against properly. $377 in Korea. Since the share of GNP devoted to Governments have a responsibility to spend health tends to rise with income, rich countries well, to get "value for money," whenever they differ from poor ones even more in health expen- devote public resources to health. This means al- diture than in income. locating resources so as to obtain the most im- But health spending alone cannot explain all the provement in health per public dollar, taking into variation in health among countries. Nor can in- account the private market's response to public come and education, or even spending, income, sector spending. Because private health care mar- and schooling taken together. Figure 3.1 illustrates kets can also fail to achieve value for money, gov- the discrepancies. The vertical axis shows how far ernment policy has a role in providing information life expectancy in a country differs from the value and incentives to improve the allocation of re- predicted on the basis of that country's income sources by the private sector. In most of the world and average schooling. France, Haiti, Singapore, a great deal of additional health could be obtained and Syria have almost exactly the life expectancy from a relatively small number of cost-effective in- predicted. China, Costa Rica, Honduras, and Sri terventions that could be delivered at modest cost Lanka, in the top half of the figure, all achieve five and with little need for high-level facilities or years or more of life beyond what would be ex- medical specialties. pected. Egypt, Ghana, Malawi, Uganda, the United States, and Zambia, in the bottom half of Health expenditures and outcomes the figure, all have a life expectancy about five years lower than expected, given their levels of Chapter 1 showed how greatly health status dif- income and education. fers among populations. Life expectancy ranges The horizontal axis of Figure 3.1 shows how far from forty years or less in some countries of Sub- total health spending differs from the value pre- Saharan Africa to seventy-five or more in the es- dicted by income and education. Egypt, Morocco, tablished market economies. In Sub-Saharan Af- Paraguay, Singapore, and Syria, in the left half of rica half of all deaths occur under age 5; in the the figure, spend relatively little. France, Haiti, In- established market economies half occur after age dia, Mozambique, and the United States, in the 74. Child mortality rates exceed 200 per 1,000 in right half, spend more than expected. several African countries but are below 20 in the At any level of income and education, higher richest countries. The burden of disease is five health spending should yield better health, all else times higher, per capita, in the worst-off than in being equal. But there is no evidence of such a the healthiest regions. relation. Countries are scattered in all quadrants of Three factors help to explain these huge differ- the figure. The countries that appear in the upper- ences. The firstis human behavior. Chapter 2 left quadrant obtain better health for less money. showed that both health and the capacity to im- China, for instance, spends a full percentage point prove health are related to income and education less of its GNP on health than other countries at and to the changes in behavior that wealth and the same stage of development but obtains nearly

53 Health expenditure, income, and schooling only partly explain variation in life expectancy.

Figure 3.1 Life expectancies and health expenditures in selected countries: deviations from estimates based on GDP and schooling

Deviation from predicted life expectancy (years) 10 - 0 Better outcome, China Better outcome, lower expenditure higher expenditure Costa Rica Sri Lanka 0 0 o Honduras Greece 0 S . S Paraguay . 0 India 00 . . Morocco .. Haiti ..-% 0 0 S. 0 France Singapore0 Syria ... °Mozambique . 0 United States 0 Egypt Ghana 0 0 Worse outcome, Zambia Worse outcome, lower expenditure higher expenditure

-10 I I -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Deviation from predicted percentage of GDP spent on health

Source: World Bank data.

ten years of additional life expectancy. Singapore than predicted to achieve severalyears less of life spends about 4 percent less of its income on health expectancy than would be typical for its high in- than others at the same level of development but come and high educational level. achieves the same life expectancy. Other coun- Analyses using other measures of health status, tries, of which Costa Rica and India are examples, such as child mortality, yield similar results. This obtain relatively good health results but also spend raises obvious and important questions. Whatac- relatively more. (In the case of India health spend- counts for these large deviations? How much is ing is low and health status is poor, but even lower attributable to the characteristics of healthsys- spending and worse status would be expected for tems? How can public policy help to provide better a country with such low levels of income and health outcomes for a given national effort? schooling.) Egypt and Zambia, by contrast, get poor health for a lower-than-predicted level of The rationales for government action spending. Finally, it is possible both to spend more than predicted on health care and still achieve un- Public policy in health is successful if it leads to expectedly poor results. The United States is an increased welfare through better health outcomes, extreme case, spending 5 percent more of GNP greater equity, more consumer satisfaction, or lower

54 total cost than would occur in the absence of public finance care: spending more can translate into action. Of course, the pursuit of one or more of more services for the poorest or the same services these objectives does not by itself justify govern- for more people, including the less poor. In prac- ment intervention. There must be a basis for be- tice, very poor countries must target if they are to lieving that the government can achieve a better offer the poor any meaningful health care. outcome than private markets can. There are three Public goods and externalities are forms of market broad reasons why that belief may be true: one failure that may justify government intervention. centers on poverty and the equitable distribution The key characteristic of public goodswhich may of health care and the other two involve market be products or servicesis that one individual can failures. use them or benefit from them without limiting Reduction or alleviation of poverty provides a others' consumption or benefit. As long as some- straightforward rationale for public intervention in body pays, everybody benefitswhich makes it health. Success in reducing poverty requires two difficult or impossible to find anybody altruistic equally important strategies: promoting the use of enough to pay. Many public health interventions, the most important asset of the poortheir labor such as wide-area control of disease vectors and and increasing their human capital through access radio-based health information campaigns, are to basic health care, education, and nutrition. As nearly pure public goods for which only the gov- Chapter 1 showed, investment in the health of the ernment can ensure provision. Another public poor raises their educability and productivity. It good, new scientific information, has contributed gives them both the assets they need to lift them- enormously to the rapid improvements in health selves from poverty and the immediate welfare during this century. Its continued creation will de- gain of relief from physical suffering. Further- pend at least in part on governments. The right more, in most societies providing health and edu- choice of interventions and the proper level of pro- cation for the poor commands a degree of political vision of any public good require careful analysis assent that is altogether lacking for transfers of in- of the health benefits in relation to the costs. Prices come or of assets such as land. Investing in the provide no indication of what benefits are worth health of the poor is an economically efficient because private markets do not supply public and politically acceptable strategy for reducing goods. Nonprofit nongovernmental organizations poverty and alleviating its consequences, as World (NGOs) may supply such goods but cannot fully Development Report 1990 emphasized. substitute for government action. If "the poor" are all those living on less than $1 Externalities, or spillovers of benefits or losses (in real purchasing power) per day, they can typ- from one individual to another, characterize cases ically neither afford much health care nor borrow in which a private market might function but to pay for it. Simply transferring small amounts of would produce too much or too little. For example, income to poor people would create relatively little curing an individual of tuberculosis also prevents additional demand for health care. But because the transmission of the disease. But an individual's poor are more sensitive to the price of medical care demand to be cured of tuberculosis (or of mild or and also suffer a greater burden of disease than the asymptomatic sexually transmitted disease)is nonpoor, access to free or low-cost care can pro- probably not affected by consideration of the risk duce large increases in their consumption of health to others. If the externality is not taken into ac- care. count, treatment will be priced too high in private To ensure that subsidized health services actu- markets, and too little treatment will be given. ally reach the poor, however, may require restric- Subsidies for treatment are therefore justified. An tions, particularly on the kind of care that is paid example of negative externalities is a person's use for by the public sector. Offering free care of all of antibiotics, which may, by increasing microbial kinds to everybody typically leads to rationing of resistance to a drug, reduce the drug's value to servicesgeographically or according to quality. others and increase their risks. Such universal programs may not reach the poor Failures in markets for health care and health insur- or improve their health. They may, however, com- ance provide a third rationale for government ac- mand more political support than targeting, and tion to improve efficiency and, in the case of fail- they more easily address the problems of insur- ures in the market for health insurance, to improve ance markets that are discussed below. Who equity. One source of market failure, "adverse se- should receive free care depends on the preva- lection," arises because individuals face different lence of poverty and on the country's capacity to risks. Customers who know themselves to be at

55 high risk are motivated to buy more insurance and There is some moral hazard in the markets for are more likely to use it. So it is in the insurer's house and vehicle insurance. The extreme form is interest to find out who the high-risk customers when somebody burns down a house to collect the are and either to exclude them or to compensate insurance or abandons a car and reports it as for their greater risk by charging them higher pre- stolen. But unlike consumption of too much health miums. (Higher prices for all customers would re- care, these actions are crimes, with penalties that duce demand by low-risk people and therefore may greatly exceed the value of the asset. In any push prices still higher.) Defensive efforts to ob- case, the insurer's potential liability is limited to tain valuable information about risks add to the the (easily determined) market value of the asset. cost of insured health care without improving All the limitations on moral hazard and adverse health outcomes. selection are weaker in health insurance. Itis Adverse selection presents a serious problem for harder to identify individual risks, and still harder risks existing at the time insurance is taken out, to attribute them to behavioral choices. There is no but an even more complex problem arises from the market value for the human body and no possi- fact that an initially low-risk person may become bility of abandoning one that is worn out and ac- high-risk later in life. In principle, there should be quiring a new one. The lack of a natural limit on insurance available specifically against this likeli- costs (since the asset being insured, the body, has hood of increased risk, or else insurance should no price with which costs can be compared) distin- cover a person's entire lifetime, with sharing of guishes health from other insurable risks. risks that may arise in the distant future, as well as The difficulties in insurance markets carry over of current ones. Neither solution is easy to imple- directly into markets for health care. If people have ment because of the extreme uncertainty; insur- "too much" health insurance, they will have an ance can cover known risks but not uncertainty incentive to use "too much" health care at too about risks. high prices. Unfortunately the difficulty of judging Another problem is the tendency of consumers health care risks and the impossibility of placinga to use more of a service when its marginal cost to value on a living body make it impossible to deter- them decreases. Insurance reduces or eliminates mine how much is "too much" in health care and the marginal cost of health care to consumers. So, health insurance. Nor is making a consumer pay providing insurance does not simply shift the way more for health care a sure way of reducing only a given amount of health care is paid for but in- "unnecessary" demand. creases the amount of care demanded. Failuresof information make matters even Because the financial cost of disease is reduced, worse. A patient who knew the likely outcome people may take less care of their health, leading and the cost to him or her of every possible treat- to more illness and more subsequent demand for ment might yet be able to choose rationally be- care. Or they may protect their health more by tween gains and costs. But patients do not have way of health care, paid for by insurance, and less such knowledge, and the medical professional through their own behavior. Passing costs on to generally knows far more than the customer. This others such as insurers because one does not bear asymmetry of information means that the provider the full consequences of one's actions is called not only provides services but also decides what "moral hazard." It arises because of uncertainty services should be provided. The result is a poten- and because insurers cannot fully monitor con- tial conflict of interest between what the provider sumers' behavior and make them responsible for stands to gain from selling more services and his their decisions. Moral hazard also results when or her duty to do what is best for the patient. The providers induce demand for services that neither patient is at even more of a disadvantage when they nor consumers will pay for. sick and unable to make decisions or when deci- Both adverse selection and moral hazard have sions must be made quickly because of threats to more pernicious effects in markets for health in- life. surance than in markets for insurance on houses The same potential for consumption of unneces- or cars. Risks to houses are higher in areas prone sary services can arise any time a supplier is better to earthquakes or hurricanes, but they are easy to informed than a customer. It is a notorious prob- determine, and insurers respond by charging lem in car repair and home improvement services. higher premiums in those areas. Similarly, car in- But in these sectors the insurer has more oppor- surance premiums are higher for young drivers tunity to supervise the service provider, and the and other identifiable groups at greater risk of ve- insurer may decide simply to replace the item hicular accidents. rather than to repair it. Health insurers have no 56 replacement option. They may try to review pro- Since poor people typically cannot buy such care fessionals' recommendations before agreeing to for themselves, there is a straightforward case for pay for services, but health professionals often dis- public finance. Public health measures and essen- agree about expected medical outcomes, and wait- tial clinical care together constitute a package of ing for a second opinion may cause pain, suffer- health care that might justifiably be financed by ing, and increased risk for the patient. general revenues, with perhaps some contribution These problems constitute the market failure pe- from user fees. This strategy is also compatible culiar to health: expenditure on medical care can with the argument that basic health care is a fun- be extremely high, yet not all justified care is pro- damental right. Although most of the population vided and much care of doubtful value is paid for. may be able to pay for such care, the government Some people are denied insurance, while others has a responsibility to ensure that the poor, too, may be overprotected. Those who do not pay the can exercise their rightat least to the extent that full costs of treatment may take poorer care of their society can afford. own health than they could. Many of the extra Third, the rationale that the government should costs are paid by society as a whole. intervene in health care markets because of signifi- The market for health care goods and services cant market failures applies particularly to the reg- can also fail through imperfect competition among ulation of health care and health insurance. The providers, which allows excess profits, inefficient government cannot finance all medical care for use of resources, poor quality, and too little pro- which insurance might be desirable without wors- duction. Sometimes governments themselves arti- ening the tendency toward higher costs and risk- ficially stifle competition. For example, govern- ing de facto rationing of health care, which par- ments may prohibit or interfere unduly with the ticularly hurts the poor. Beyond a well-defined operation of private health care providers, particu- package of essential services, therefore, the role of larly NGOs. Governments often also protect do- the government in clinical services should be lim- mestic producers of drugs and vaccines. In Ban- ited to improving the capacity of insurance and gladesh tetanus vaccine produced domestically at health care markets to provide discretionary care government insistence had such low potency that whether through private or through social insur- its use in 1989-92 risked thousands of lives before ance (earmarked taxes such as social security or it was replaced with imported vaccine. other mandated arrangements). Of course, the Economies of scale in productionwhich occur range of services included in the nationally de- when a single large producer is much more effi- fined essential package will vary substantially cient than many small onesalso lead to noncom- from country to country. To provide equitable ac- petitive situations. In many parts of the world hos- cess for the poor, to address problems of adverse pitals and specialists face little or no competition selection, and to contain costs, the governments of because of economies resulting from large-scale almost all OECD countries have made available a operation. Such situations may call for regulation comprehensive essential package with public (or of the private market. publicly mandated) financing. Poorer countries The three rationales for government interven- must, of necessity, define their essential packages tion in the health sectorprovision of public more narrowly. goods, reduction of poverty, and market failure Governments can further improve how markets correspond roughly to three different kinds of ser- function by providing information about the cost, vices. First, the services classified as public goods, quality, and outcome of health care. Simply by de- and some of those characterized by large exter- fining an essential clinical package, the public sec- nalities,constitute what is known as "public tor provides valuable guidance on what is and health." Public health includes those services pro- what is not cost-effective. This distinction may vided to the population at large or to the environ- then influence the design of private or social insur- ment, such as spraying to control malaria. It also ance packages and the behavior of individual pro- typically includes some services such as immun- viders or patients. Information on the relative cost- zations that are not public goods but that carry effectiveness of different discretionary procedures substantial externalities. is similarly valuable and might be used by insurers Second, the inclusion of health care as part of a and providers to reduce costs and attract clients. strategy for combating poverty justifies public fi- Neither theory nor experience points to a gen- nancing of "essential" clinical or individual ser- eral rule on the extent to which the public sector vices. These are highly cost-effective services that should provide health care directly, as distinct would greatly improve the health of the poor. from financing it. Governments might have to 57 Box 3.1Paying for tuberculosis control in China

Tuberculosis kills or debilitates more adults than any Charging tuberculosis patients had perverse effects. other single infectious agent. Without appropriate When doctors and institutions expected to be reim- treatment, 60 percent of those with the full-blown dis- bursed by insurance, they provided excessive diagnos- ease will die. In China, itis estimated, more than tic tests and examinations during treatment and dis- 360,000 peoplemost of them poor peasantsdie of pensed higher-cost antibiotics that should have been tuberculosis every year. Tuberculosis is best prevented reserved for the most resistant cases. Daunted by the by curing infectious persons early in the course of dis- costs, many low-income victims failed to enter treat- ease, thus interrupting transmission to others. Well- ment or dropped out early. There were no incentives to run programs can cure 80 to 90 percent of patients; ensure that patients completed treatment or were poorly administered programs cure 30 percent or less, cured. Because health system records showed very leading to larger numbers of sustained cases of infec- high rates of cure for those who completed treatment, tion and related deaths and to new infections. the government remained largely unaware of the dete- China made substantial progress against tuber- riorating situation. The direct cost to the health system culosis during the 1960s and 1970s, using standard of a poorly functioning program was nil, but the indi- long-term (twelve to eighteen months) antibiotic ther- rect costs to the economyand the personal costs to apy that was essentially free of charge. Since the early patients and their familieswere enormous. 1980s, however, infection rates in about half the coun- An estimated I million to 1.5 million additional tu- try's provinces have stagnated or increased, despite berculosis cases remained infectious during the 1980s the adoption of an improved short-course (six to eight because treatment was no longer free. Tens of millions months) therapy. Much of the trend is attributable to of new infections were produced, and many of those changed health-financing policies and, in particular, to infected will fall victim to the disease later in life. the government's decision that health facilities should The development of drug-resistant strains was also ac- be encouraged to charge patients for virtually all ser- celerated. Given appropriate policies, many of the vices. Starting in 1981, health institutions had to earn more than 3 million persons who died of tuberculosis much of their operating costs from sales of drugs and in China during the decade could have been saved, services. Although base salaries were still funded from and the risk of infection for society could have been the public budget, workers' bonuses, housing, and re- halved. tirement benefits depended in part on institutional in- China, having recognized the problems caused by come from service provision. Managers' investment charging for tuberculosis therapy, has begun a major budgets were also linked to revenues from fees. A few national tuberculosis-control effort that provides sub- public health services such as immunizations remained sidies for treatment and appropriate incentives for pro- partially subsidized, but tuberculosis diagnosis and viders of care. Early results of this policy show dra- treatment were not, despite drug costs of $30 to $80 per matic increases in the number of cases cured. treatment. -I supply a package of essential health services di- In some circumstances market failure may im- rectly where private care would not be feasible pose only slight welfare losses, and the benefit of without high subsidiesfor example, in lightly correcting it may be outweighed by the costs of populated, very poor areas. (In many parts of the government action. In other cases the losses from developing world an alternative method of provid- failure to take account of positive externalities and ing such services is to subsidize an NGO.) In most supplier-induced demand can be enormous. Pol- circumstances, however, the primary objective of icy toward tuberculosis control in China provides public policy should be to promote competition an example: elimination of some free health care among providersincluding between the public and introduction of profit incentives in the provi- and private sectors (when there are public pro- sion of health services dramatically reduced treat- viders),as well as among private providers, ment rates, reversing progress against the disease whether nonprofitorfor-profit.CompetitionS and causing much needless suffering (Box 3.1). should increase consumer choice and satisfaction Failures of government intervention can arise, how- and drive down costs by increasing efficiency. ever, even when government action might be Government supply in a competitive setting may sound policy. improve quality or control costs, but noncompeti- Governments may misjudge how an interven- tive public provision of health services is likely to tion will work in practice. Governments have only be inefficient or of poor quality. partial control over how private actors respond,

58 and those responses can undermine the intended ments are most appropriate for affecting the objective. Since 1971 physicians' fees in all pro- behavior of insurers, providers, and patients. This vinces of Canada have been set by negotiation raises the question of how far the government with provincial governments, and fees are no should itself act as an insurer, through social insur- longer rising faster than the general price level. To ance, and how far it should regulate private in- protect their incomes, particularly during the infla- surers. Each of these decisions involves tradeoffs tionary period 1971-75, physicians carried out a among the objectives of health policy: better greater number of procedures. This reaction was health outcomes, lower costs, more equity, and strongest where real fees fell the most. The saving greater consumer satisfaction with the health sys- in government expenditure was therefore much tem as a whole and with individual care. less than had been anticipated. Governments may not have the capacity to Value for money in health administer or implement policies well. Indeed, they may suffer from corruption and from sheer No matter how health services are organized and incompetence. The examples of two donor- paid for, what they actually provide are health in- financed public hospitals, each with 500 to 600 terventions: specific activities meant to reduce dis- beds, in two Latin American countries illustrate ease risks, treat illness, or palliate the conse- the problem. One was simply too large to adminis- quences of disease and disability. Debates about ter and operate and therefore could not be used at whether health services should concentrate on more than 60 percent of capacity. The other was so "vulnerable groups" such as children, pregnant badly designed that it could not accommodate women, or the elderly, or about the relative roles more than one-third the planned number of of hospitals versus health centers, or about pre- patients. ventive versus curative activities, are at bottom de- Governments are vulnerable to special inter- bates concerning the proper mixture of interven- ests both within and outside the health system. By tions. In health, as in every other sector, customers financing the training of unneeded physicians, by want value for the money spent on such interven- paying for low-value discretionary services for bet- tionswhether they pay directly or indirectly, in ter-off patients, and by protecting domestic indus- their roles as taxpayers or as buyers of health tries, governments help create the interests that insurance. later impede good policy, especially when quick Knowing the cost-effectiveness of a health inter- responses are needed to meet changing circum- ventionthe net gain in health (compared with stances or new opportunities. Even when society doing nothing) divided by the costcan be ex- as a whole would gain, public action may fail be- tremely useful for both public and private deci- cause it does not overcome the resistance of those sions. Governments can generate such informa- who would lose as a result. tion, and they can use it in two ways. First, they Perhaps the most fundamental problem facing can use it in determining whether a particular pub- governments is simply how to make choices about lic sector intervention is cost-effective: this means health care. Too often, government policy has con- judging the improvement in health compared with centrated on providing as much health care as pos- what would have happened through private deci- sible to as many people as possible, with too little sions in the absence of public action. (Chapter 4 attention to other issues. If governments are to addresses these issues for public health measures finance a package of public health measures and and Chapter 5 for the public finance of essential clinical services, there must be a way to choose clinical services.) Second, they can supply infor- which services belong in the package and which mation about the outcomes and costs of different will be left out. (The next section describes a mea- health interventions to consumers, providers, and sure of cost-effectiveness for health interventions insurers, and this knowledge can increase the that helps with this choice.) If financing is public value per health dollar spent in the private sector, but provision is private, governments must decide including what is spent on discretionary services. how to subsidize private care. The question of in- Private providers have no more incentive than centives to providers raised by that issue also ap- public providers to measure health outcomes, but plies to paying for publicly provided careto the they do face greater incentives to know their costs. "internal market" in the public sector. And if gov- Cost information alone can promote allocative effi- ernments are to influence the market for discre- ciency, as the experience of a Brazilian nonprofit tionary services, they must decide what instru- maternal and child hospital demonstrates (Box

59 Box 3.2Cost information and management decisions in a Brazilian hospital The Instituto Materno-Infantil de Pernambuco (IMIP) is lations, it was evident that closing the intensive care a private, nonprofit hospital founded in 1960 to serve unit (except for newborns) and strengthening other the metropolitan area of Recife. In 1992 it received the services would save a greater number of children's first UNICEF award to a "child-friendly" hospital in lives. In particular, since the children who died in the Brazil in recognition of its work, particularly in the pro- hospital typically arrived very sick and often severely I motion of breastfeeding. IMIP depends for 95 percent malnourished, it appeared more cost-effective to try to of its revenue on contracts from the Instituto Nacional find high-risk children and treat them earlier. The strat- de Assistência Médica e Previdência Social (INAMPS) egy used was to expand the network of small commu- of Brazil's Ministry of Health. Annual spending runs nity health posts in the slum neighborhoods of Recife. about $6 million. IMIP opened the first such posts in 1983; by 1986 infant Starting in 1989, IMIP organized an accounting sys- mortality in those neighborhoods had fallen from 147 tem that divides services according to eleven cost cen- to 101 per 1,000 births. ters for final output. Administrative, laundry, food, ra- The experience of IMIP illustrates three lessons diology,laboratory,transport, and other nonfinal about cost-effective delivery of essential care. One is services were assigned to these final outputs in propor- that allocative efficiency can be improved without com- I tion to their measured or estimated use. plete information: medical professionals know much IMIP must match its average costs to average reve- about outcomes and often need only to know more nues determined by the price schedule of INAMPS, about costs. A second lesson is that autonomy facili- which is organized by treatment groups rather than by tates such changes: since private facilities generally individual services. Losses in any cost center must be have much more autonomy than public ones, this is an I offset by surpluses elsewhere. Gravely ill children are argument for more public finance of private provision referred to the hospital from all over northeast Brazil, or for decentralization of public systems. The third les- and there are three infant deaths per day among them. son is that even prices that are not based Ofl cost-effec- reduce mortality, IMIP created a pediatric intensive tiveness criteria can guide decisions about what care to care unit. The treatments provided, however, cost provide. It is more useful for government to set those much more than INAMPS would pay. And mortality prices correctly than to try to make all the allocative did not decline. Even without cost-effectiveness calcu- choices. Lb

3.2). By estimating costs for "cost centers" and value to human life, as would be necessary if costs relating them to outputs, the hospital discovered and gains were to be put in the same units. that its pediatric intensive care unit would drain Only in the past decade have costs and effective- resources from other departments, given the ness been systematically estimated for a wide prices the government paid for various services. range of health care interventionsalthough the The decision was made to limit the intensive care first such calculations had been made many years unit to newborns; community-level health posts earlier. Only a small share of the thousands of appeared more cost-effective for other cases. known medical procedures has been analyzed, but the approximately fifty studied would be able to Measuring the cost-effectiveness of health interventions deal with more than half the world's disease bur- den. Just implementing the twenty most cost- Given a common currency for measuring cost and effective interventions could eliminate more than a unit for measuring health effects, different inter- 40 percent of the total burden and fully three-quar- ventions can be compared by what it costs to ters of the health loss among children. achieve one additional year of healthy life. Out- The cost and effectiveness estimates used in this comes are measured in the same unit of disability- Report are based, as far as possible, on actual con- adjusted life years (DALYs) used to estimate the ditions in developing countries. Some fixed costs burden of disease. Nonhealth burdens, such as in- of operating a health system that cannot be attrib- come lost because of disease, are not included in uted to particular interventions are not consid- the measure. The ratio of cost and effect, or the ered, but the costs of intervention-specific capacity unit cost of a DALY, is called the cost-effectiveness are taken into account. Costs are assessed at mar- of the intervention; the lower that number, the ket prices. For inputs that cannot be traded inter- greater the value for money offered by the inter- nationally (such as semiskilled labor), costs will be vention. This approach avoids assigning a dollar lower in developing countries. For drugs, most 60 equipment, and high-level manpower, costs are either the public or the private sector costs lives. likely to be equal across countries, leaving aside An expenditure of $100,000 on chemotherapy for the effects of tariffs or other barriers. Indirect tuberculosis could directly save about 500 patients. costs, such as patients' costs of travel to treatment It would also prevent them from infecting others, or the income they forgo, can be substantial for for a total gain of about 35,000 DALYs. The same some interventions and perhaps particularly for expenditure on management of diabetes would women. Because these costs are difficult to deter- also benefit 500 patients but would save only 400 mine, they were largely ignored; more study is DALYs; each patient would gain less than one needed of how these barriers affect the utilization healthy year from a year of treatment, and there of health services. would be no benefit from reducing incidence. In- For some common health service packages such sisting on value for money is not only fully con- as immunizations, costs are computed on a joint sistent with compassion for the victims of disease, basis rather than separately for each intervention it is the only way to avert needless suffering. in the package. The estimates do not represent an The results of cost-effectiveness analysis confirm unattainable ideal; they assume that medically cor- the value of the primary health care interventions rect procedures are followed and that reasonable included in programs to reduce childhood malnu- care is taken as to quality, but they also allow for trition and mortality, chiefly from infectious dis- incomplete coverage or compliance. Whenever eases. Several hitherto neglected interventions are possible, actual experience is used to guide esti- also very cost-effective: chemotherapy against tu- mates of such things as how many patients will fail berculosis, integrated prenatal and delivery care, to complete a course of treatment. Future gains mass programs to deworm children, provision of from current interventions are discounted at 3 per- condoms along with information and education to cent per year, which has little effect on the ranking combat AIDS, and measures against smoking, of interventions the effects of which are felt such as education, consumer taxes on tobacco quickly, although it does reduce measured gains products (an effective deterrent for adolescents from interventions when the health effects are felt who are not yet addicted), and prohibition of only in the long run. smoking in public places. Many of the most cost- This Report found huge differences in both the effective health interventions are preventive in cost and the effectiveness of various health inter- character. But not all preventive measures are cost- ventions. Figure 3.2 presents both dollar costs and effective: spraying to control the mosquitoes that gains in DALYs for each of forty-seven different carry dengue is an example of relatively poor value interventions. Higher points represent interven- for money. At the same time, a small number of tions that are more effective in improving health; neglected but cost-effective clinical (mostly cura- points farther to the right represent lower-cost in- tive) interventions could eliminate a substantial terventions. Some interventions cost more than fractionof the burden ofdiseasein many $10,000 per person benefited, while others cost countries. less than $1. Some interventions add more than In general, most cost-effective interventions can ten years of healthy life; for others the gain is be performed outside hospitals. By treating a small equivalent to only a few hours or days of full number of severe cases of disease, however, hos- health. Both axes are scaled in logarithms so that pitals can sometimes improve health at a lower the diagonal lines show equal cost-effectiveness cost per DALY than lower-level facilitiespro- ratios in dollars per DALY. These ratios vary vided that clinics or health posts treat most cases widely, from as little as $1 to as much as $10,000. and refer to hospitals only those requiring more Higher lines represent more cost-effective inter- sophisticated care. ventions. Four specific interventions illustrate ex- treme combinations of cost and health gain: vita- Complications in the use of cost-effectiveness min A supplementation in areas where the risk of blindness from vitamin deficiency is high (very Both the cost and the effectiveness of an interven- low cost, high gain), chemotherapy for tuber- tion can be affected by the incidence and preva- culosis (high cost, very high gain), environmental lence of the disease and the probability of dying control of dengue (low cost, low gain), and treat- from it. Preventive interventions are less cost- ment of childhood leukemia (very high cost, mod- effective for relatively rare diseases because more erately high gain). people have to be reached to prevent one case. The Because interventions can differ so much in cost- fatality rate matters because preventing or control- effectiveness, making allocative decisions badly in ling a disease saves more lives if there is a high 61 The costs and effectiveness of health interventions vary greatly.

Figure 3.2 Benefits and costs of forty-seven health interventions

Increase in DALYs (log scale) 100 . Chemotherapy for tuberculosis 10 Vitamin A . supplementation . 0.1 Treatment of $1/DALY leukemia S

0.01 $10/DALY Greater effectiveness Environmental 0.001 Lower control of dengue $100/DALY cost $10,000/DALY $1,000/DALY 0.0001 10,000 1,000 100 10 1 0.10 Cost per intervention or per intervention-year (dollars, log scale)

Target: Children under age 15 Adults age 15 or older a Note: DALY, disablility-adjusted life year. Interventions are specific activities intended to reduce disease risks, treat illness, or palliate the consequences of disease and disability; an intervention-year is an intervention repeated throughout the year rather than provided only once. a. Jncludes some interventions that benefit all age groups. Source: Jamison and others forthcoming; Worlc Bank data.

probability of dying. Immunization in an environ- information is needed to judge which interven- ment in which children are undernourished and tions should have priority. National or regional as- many die from preventable diseases is more cost- sessments are also important for estimating the effective than if children are otherwise healthy and expenditure levels required and the probable im- face little risk of dying. (Box 3.3, on measles and pact on the national burden of disease. tuberculosis, illustrates these issues.) Fortunately, If providing an intervention did not impose any differences in cost-effectiveness between one in- fixed costs in infrastructure and program adminis- tervention and another are often much larger than tration, then a low cost per DALY saved would either the variation from one locale to another or suffice to justify the intervention. In practice, there the uncertainty in the estimates. Where this is not may be substantial fixed costs to share over a num- the case, as exemplified by malaria, detailed local ber of interventions, and administrative capacity

62 Box 3.3Cost-effectiveness of interventions against measles and tuberculosis The costs and effects of measles vaccination were esti- average cost of the lower-coverage strategy, but the mated for a model urban area based on data from higher-coverage approach continues to be extremely Lagos and Kinshasa. Data from Matlab, Bangladesh, cost-effective. Similar calculations were made for che- were used to model measles in rural areas. In each motherapy for tuberculosis, for both a standard course area, 36,400 cases of measles were assumed to occur in of treatment (twelve to eighteen months) and a short the absence of vaccination, with 1,452 urban deaths course (six to eight months), each with and without and 806 rural deaths. All the health damage from mea- hospitalization. Using data from Malawi, Mozam- sles comes from deaths, each of which costs thirty bique, and Tanzania, the average incremental cost DALYs. The simulations considered three different (marginal cost plus the average cost attributable to the strategies:no vaccination,immunizationatnine fixed costs of the tuberculosis control program but ex- months (the earliest age at which the standard vaccine cluding other fixed costs of the health system) was esti- is effective) with 60 percent coverage, and nine-month mated at about $80 to $110 per cure for ambulatory vaccination plus efforts to increase coverage to 80 per- treatment, and $160 to $300 when hospitalization was cent, which raises supervision costs by 10 percent. For required. Cost per death directly averted was in the the last strategy, incremental as well as average costs range of $75 to $275, but cost per total death averted, were calculated to highlight the effect of raising cover- taking account of the interruption of transmission, can age. Costs were related to each of three effects: cases be as low as $20 and never exceeds $100. These very averted, deaths averted, and DALYs gained. Box table low costs translate into costs per DALY saved of about 3.3 shows the results. $1 to $3, making chemotherapy for tuberculosis one of Measles strikes later in childhood in rural areas, so the most cost-effective of all interventions. These costs cases are easier to prevent. But because earlier cases (in do not vary with the annual rate of infection. The cost- urban areas) cause more deaths, the cost per death effectiveness of BCG vaccine, by contrast, is extremely averted or per DALY is higher. In both urban and rural sensitive to infection rates; the vaccine is cost-effective areas, the marginal cost of raising coverage exceeds the only when the risk of infection is high.

Box table 3.3 Health costs and gains from measles immunization

Urban, by percentage vaccinated Rural, hr percentage vaccinated Item 60 80 60 80 Cases prevented (thousands) 10.7 14.2 16.2 22.6 Total cost per case prevented (dollars) 17 18 11 12 Incremental cost (dollars) - 22 - 15 Deaths averted (thousands) 0.4 0.6 0.4 0.5 Total cost per death averted (dollars) 432 462 525 561 Incremental cost (dollars) 552 670 DALYs gained (thousands) 12.3 16.4 10.2 13.5 Total Cost per DALY (dollars) 15 16 18 19 Incremental cost (dollars) - 19 - 23 Source: Foster, McFarland, and John forthcoming.

may be limited. Spending on interventions that Since only relative ranking is possible, the at- are very cost-effective but resolve very small dis- tractiveness of an intervention also varies accord- ease burdens could waste resources by making it ing to what other health problems and treatments difficult to deal with diseases that impose much are locally prevalent. Oral rehydration therapy is larger burdens. Priority should go to those health an example: in environments in which child mor- problems that cause a large disease burden and for tality is low, it is much less cost-effective than im- which cost-effective interventions are available munization because it may have to be given re- (Box 3.4). If a particular health problem causes the peatedly during a child's first few years, butas loss of many healthy life years but there are no mortality rises it becomes more cost-effective. In meansor only very costly meansfor dealing general, the cost-effectiveness ratio varies not only with it, then it should be a priority not for health with local conditions but also with the degree to care but for research on development of cost-effec- which an intervention penetrates or covers a tive interventions. population. Cost-effectiveness is also influenced

63 r Box 3.4Priority health problems: high disease burdens and cost-effective interventions It is easy to determine which health problems among ease burden in women ages 45-59 and for smaller children under age 5 deserve priority. As Appendix shares in other age and sex groups. Among communi- table B.6 shows, nine diseases each account for more cable diseases tuberculosis, AIDS, and respiratory in- than 1 percent of the total disease burden in this age fections deserve priority, but they cause less than 10 group. These diseases range from acute respiratory in- percent of all ill health after age 45 and only 20.1 per- fections (more than 17 percent in both boys and girls) cent in men ages 15-44. to iodine deficiency (1.2 to 1.3 percent). Of these prob- Large disease burdens and cost-effective interven- lems, which cause fully 80 percent of young children's tions coincide for only one group of adults, women ill health, eight can be addressed by interventions cost- ages 15-44. Six of the ten main sources of ill health can ing less than $100 per DALY saved. The only exception be prevented or treated for less than $100 per DALY. is congenital problems, which are responsible for more These range from maternal health problems (18.0 per- than 6 percent of the disease burden but for which no cent of the burden) to respiratory infections and ane- cost-effective interventions are known. mia (2.5 percent of the burden each) and account in The situation is much more complicated for adults total for 44 percent of ill health among women of repro- (Appendix table B.7). For example, cerebrovascular ductive age. Two other problemsdepression and self- disease is the leading cause of healthy life years lost in inflicted injuryeach cause at least 3 percent of the both sexes after age 60 and in women ages 45-59, but disease burden, but dealing with them is much more interventions to deal with it cost $1,000 or more per problematic. DALY saved. Ischemic heart disease is the second or These calculations illustrate the chief problem a third leading cause of ill health in both sexes after age health care system faces as the population ages: the 45, but the cost per DALY of dealing with it is $250 to marginal cost of a year of healthy life gained rises $1,000. Among the ten principal noncommunicable sharply, leading to difficult choices between increased causes of ill health in this age group, interventions spending and lower health gains. However, many costing less than $100 per DALY saved exist only for health problems of the elderly that cannot be fully re- I cataracts, anemia, and cancers of the respiratory sys- solved may be palliated at low cost. And much can be tem (through reduction of smoking) and the cervix. done at earlier ages to improve the health of future These problems account for only 7.9 percent of the dis- generations of old people.

by the presence of other interventions that might An important limitation on the use of cost-effec- affect costs (through sharing of joint costs) or out- tiveness analysis of resource allocation in health is comes. Sometimes combining two interventions, that a number of interventions with important one preventive and one curative, is the appropri- health consequences also affect income or welfare ate way to deal with a particular disease, as is the in other ways. Chemotherapy for tuberculosis has case for tuberculosis and malaria. Some treatment no value beyond the DALY gain associated with for malaria is necessary because preventive mea- curing tuberculosis, but investing in girls' school- sures do not protect everyone; even if treatment is ing has both important consequences for health more costly, both interventions should be applied. (as documented in Chapter 2) and for income and In exceptional circumstances it may be worth status later in life. Similarly, family planning, in paying high marginal costs to extend coverage of addition to its health benefits, permits family an intervention to the entire population because choice about the number and spacing of children; the disease can be eradicated permanently (as has improved water supply and sanitation create ame- occurred with smallpox and may now be possible nity and time-saving benefits; increased food con- with polio). The gains in such cases include not sumption allows higher levels of physical activity; only the DALYs saved at the margin from the last and improved road safety reduces property dam- people immunized but all the healthy years that age and saves lives. For some of these interven- would otherwise be lost to the disease in the tions (for example, family planning and girls' future. A similar argument holds if a low-cost schooling) the cost per DALY is sufficiently low to intervention that has to be applied continually make them attractive on health grounds alone; isreplaced by one with largeinitialcosts other benefits only strengthen the case. For other but permanent effects, such as sanitation to reduce interventions the cost per DALY gained may be the need for treating many fecally borne diseases. too high to justify investment on health grounds

64 alone, but consumer willingness to pay for non- has the most easily measured output. Coverage in health benefits means that costs to the health sys- many regions remains incomplete. Immunization tem can be low. Many water and sanitation invest- against measles and against diphtheria, pertussis, ments are in this category. and tetanus has reached 90 percent or more in Using cost-effectiveness to select health inter- Chile, China, Cuba, Korea, and Saudi Arabia, but ventions for public financing does not necessarily it is still below 50 percent in some Asian and many mean spending the most resources where the bur- Sub-Saharan African countries. Many of the other den of disease is greatest. Instead, it means con- components of an adequate public health package centrating on the interventions that offer the great- scarcely exist. est possible gain in health per public dollar spent. As far as clinical services are concerned, the The relevant comparison is usually not with a situ- principal government failing in most countries is ation in which nothing is done but with the situa- the attempt to provide everything to everybody, tion created by privately financed health interven- with no distinction between more and less essen- tions. The most justified public measures will tial care and more or less needy patients. For some therefore combine a strong rationale for public ac- health services provided by the public sector, the tion with a cost-effective health intervention. Be- system of provision is so grossly inefficient that it cause individuals differ in how they value the is unlikely to be cost-effective no matter what in- present in relation to the future and in how they terventions the system tries to provide. Such inef- judge the seriousness of different health condi- ficiencies have been criticized so clearly and for so tions, a uniform ceiling on what the government long that it is evident they will only be overcome pays to gain one DALY may leave some people by radical changes in the organization of health with more publicly financed care than they would caresuch as a shift in the government's role from choose and some with less. But of all possible uni- providing care to financing care and stimulating form criteria by which to judge what interventions competition among providers. These changes will to pay for, cost-effectiveness appears to yield the in turn require a clear distinction between essential most efficient distribution of health resources. and discretionary spending and a new determina- tion by governments to achieve value for money in Data limitations health services. There is no other equitable way to control government spending. Cost-effectiveness analysis requires data on expen- Most governments also perform poorly in regu- ditures for particular interventions and on out- lating markets for private services, including in- comesinformation that health facilities and sys- surance. As recent research in Brazil has shown, tems,particularly those in the public sector, the quality of medical care could be substantially typically do not collect. Such information could improved at low cost if government discharged promote substantial gains in efficiency, but it will this role better. The rapid growth and almost total take considerable time and effort for most public lack of regulation of private insurance in such systems in developing countries to learn how to countries as Brazil and Korea present another chal- gather and use it. Budgets often disaggregate only lenge for which governments are ill-prepared. by inputs, not by programs, and costs per consul- tation or per bed-day mix many different interven- Allocation of spending to cost-effective services tions. Outcomes are seldom quantified. For inter- ventions the cost-effectiveness of which vary There are no calculations of how many years of greatly with local conditions, there is no substitute healthy life are currently saved by health systems. for information on both costs and results. Nonetheless, it is clear that many of them perform much worse than they might. Many governments Health policy and the performance of health spend too much on sophisticated hospital services systems of low cost-effectiveness and too little on essential public health and clinical services. The share of All governments subscribe to the view that the public expenditures for health absorbed by tertiary state must ensure the provision of certain basic and secondary care hospitals, for example, is as public health services. But few achieve this goal high as 70 to 75 percent in Jordan and Venezuela. even for immunization, which is probably the Tertiary care hospitals alone may consume 30 to 50 health intervention that has received the greatest percent of the health budget. Only a quarter of government attention and donor support and that government spending, and often less, is devoted

65 Table 3.2 Actual and proposed allocation of public expenditure on health in developing countries, 1990 (dollars per capita)

Estimated Spending under the proposed package actual Low- Middle- All spending, all income income developing developing Package component count ries count ri cs' countries countries Contents Public health 4 7 5 1 EPI Plus; school health programs; tobacco and alcohol control; health, nutrition, and family planning information; vector control; STD prevention; monitoring and surveillance Essential clinical services 8 15 10 4-6 Tuberculosis treatment; (minimum package) management of the sick child; prenatal and delivery care; family planning; STD treatment; treatment of infection and minor trauma; assessment, advice, and pain alleviation Total, public health and 12 22 15 5-7 minimum essential clinical services Discretionary clinical services' 6 40 6 13-15 All other health services, including low-cost- effectiveness treatment of cancer, cardiovascular disease, other chronic conditions, major trauma, and neurological and psychiatric disorders

Total 6 62 21 21 Note: Current spending on essential clinical services is estimated to be 20-30 percent of total public expenditure on health on the basis of estimates in World Bank health sector reports. The numbers reported should be regarded as approximations. Estimated for an income level of $350 per capita. Estimated for an income level of $2,500 per capita. Estimated residually. The negative number for low-income countries reflects total spending below the cost of the package. Source: World Bank calculations; World Bank sector reports on Ghana (1989), India (1992), Indonesia (1991), Jamaica (1989), Jordan (1989), Mexico (1989), Nigeria (1991), Pakistan (1992), Turkey (1990), and Venezuela (1992). to cost-effective public health measures and essen- propriate package of cost-effective care would ad- tial clinical care, delivered mainly in health centers dress. Even the best-designed care package could and communities. In many countries the share of not prevent all the health damage from these dis- public spending devoted to these basic services eases because of the low cost-effectiveness of some has been falling in recent years. In Brazil 64 per- interventions and the increasing marginal costs of cent of public spending on health in 1965 was for even the best ones. Still, because of the size of the preventive and public health activities, but by the burden and the low cost per DALY of the interven- mid-1980s the share had dropped to 15 percent, tions, it is reasonable to conclude that public ex- and hospitals absorbed fully 70 percent of expendi- penditure on health should initially be concen- ture. The resulting weakness of the primary care trated on those conditions. What this implies for network leads patients to seek care in hospitals; Up the distribution of spending by type of input or to 80 percent of the cases crowding hospital emer- level of facility is less clear, but it probably means gency rooms could be treated as effectively, but that facilities above the district hospital level more cheaply, at the primary level. should account for only a small share of the total, In the world as a whole, almost half the existing primarily for dealing with referrals. disease burden is from communicable diseases, Table 3.2 illustrates the degree of misallocation nutritional disorders, and maternal and perinatal of health spending by comparing estimated actual causes. It is primarily these problems that an ap- expenditure with what would be spent for a pack- 66 age of health services designed to address most cluding private spending, is about $14, about the effectively the burden of disease in the developing same as the proposed package. This means that world. This package consists of public health ser- either substantial private resources will have to be vices that would cost just over $4 per capita in poor used or additional government resources will be countries (with average income per capita of $350) needed; even if all public expenditure on discre- and a minimum package of essential clinical ser- tionary services were eliminated, current govern- vices that would cost about $8 per capita more. In ment spending on health would not meet the costs middle-income countries (with average income per of the package. capita of $2,500), the same package of public Since the minimum package would cost only health measures and essential clinical services about $60 billion for all developing countries to- would cost about 80 percent more, or $22 per per- gether, the task is to reallocate resources in mid- son. This difference partly reflects different epi- dle-income countries and to find additional re- demiological conditions, but input costs, particu- sourcesofabout $10billioninlow-income larly salaries, would also be higher in middle- countries. The $8 to $10 per capita needed in extra income countries. Many countries will define the spending on public health measures and essential essential clinical package much more broadly than clinical services is less than the $13 to $15 per cap- the minimum discussed here. Even in relatively ita now spent, on average, on discretionary or poor countries, targeting public finance of essen- nonessential clinical services. In fact, spending on tial services to the poor would allow creation of a these less cost-effective services is now roughly broader and more generous package. Such a pack- double the amount that countries spend on the age can be built up by adding interventions in or- recommended package of public health measures der of decreasing cost-effectiveness until the addi- and essential clinical care. The right combination tional health gain is judged not to be worth the of reallocation and additional expenditure would cost, given the country's resources. To ensure po- allow governments to achieve a large improve- litical support and to deal with problems of market ment in overall health. failure and equity, countries may choose to finance Table 3.3 indicates how large this gain could be. the essential package universally from public, or Properly allocated, an expenditure of only $12 per publicly mandated, sources. person in low-income countries (excluding China) Governments in developing countries spend an would be enough to reduce the disease burden by estimated $21 per capita on health, for a total of almost one-third. This is 226 million DALYs, about $84 billion. It is estimated that only a little equivalent to 7.0 million infant deaths per year. In more than $1 per person, or a total of $5 billion, middle-income countries the proposed package goes for cost-effective public health measures. To could deal with only 15 percent of the disease bur- buy the package described here, countries would den, despite the higher expenditure per person. need almost to quintuple what they spend on pub- The total reduction in ill health in middle-income lic health. About $4 to $6 per capita, or $17 billion countries would be about 45 million DALYs, the to $25 billion total, goes for clinical services deliv- equivalent of 1.4 million infant deaths per year. ered through lower-level facilities or classified as The smaller gain in these countries reflects the fact primary health care. These services commonly in- that they have already eliminated much of the bur- clude many of those in the essential clinical care den from easily controlled communicable diseases. package, but they are usually not delivered to the A large part of the remaining burden is caused by entire population. And this expenditure includes chronic disease and disability. some less cost-effective services that should be re- It is assumed that the disease burden would de- garded as discretionary. Paying for a minimum cline by the same share in China as in a middle- package of essential clinical care would require ex- income country because China has already sub- penditure of an additional $16 billion to $24 billion, stantially reduced the burden from the diseases doubling the current expenditure level. If total addressed by the package. About 30 million spending did not change, this would imply a re- DALYs could be gained, the equivalent of 930,000 duction by about half in what is now spent on infant deaths prevented. The cost per capita would discretionary services. be the same as in low-income countries. For some countries, paying for the proposed Full coverage with the minimum package would package of services poses a severe challenge. In cost an estimated $22 billion in low-income coun- fact, in the poorest countries total current public tries, $14 billion in China, and $26 billion in spending of $6 per person is about $6 short of the middle-income countries. The total cost would be cost of the package. Total per capita spending, in- about $62 billion, or $15 per person in the develop- 67 Table 3.3 Total cost and potential health gains ofa package of public health and essential clinical services, 1990

Cost per ca pit a Total cost (billions Reduction in disease burden Country group and package component (dollars) of dollars) Percent Millions of DALYs Low-income countrjes 12 22 32 226 Public health 4 8 8 57 Essential clinical services 8 14 24 170 Chinab 12 14 15 30 Public health 4 5 4 8 Essential clinical services 8 9 11 22 Middle-income countrjes 22 26 15 45 Public health 7 8 4 12 Essential clinical services 15 18 11 33 All developing countries 15 62 25 301 Public health 5 21 6 77 Essential clinical services 10 41 19 225 Estimated from data for Bangladesh, Egypt, India, Indonesia, Pakistan, and Sub-SaharanAfrica. China is shown separately because its cost per capita is assumedto be that of a low-income country but its percentage reduction in disease burden is assumed to be that of a middle-income country. Estimated from data for Latin America and the Caribbean, Other Asia and islandsexcept for Indonesia and Bangladesh, and the Middle Eastern crescent except for Egypt and Pakistan. Source: World Bank calculations.

ing world as a whole. This figure includes what covered by a combination of greater public spend- countries are already spending on the services in ing, increased donor contributions, andmore pri- the package, estimated at $20 billion to $30 billion. vate expenditure by those able to pay. Shifting The incremental cost would therefore be only $30 some part of the cost to higher-income con- billion to $40 billion a year. sumersfor example, through private or social in- The gain in health would be about 300 million surancewould allow for an expansion of the DALYs, which is equivalent to 9.3 million infant packageorareducedburdenofpublic deaths.Universal application of the package expenditure. would therefore yield about the same health gain There are several reasons why developingcoun- as eliminating nearly all the infant deaths in the tries fail to allocate sufficient resources to cost- world today. These gains could be achieved foran effective health interventions. Health providers average cost of about $50 per DALY for the public often lack incentives to provide cost-effectiveser- health measures and about $100 per DALY for the vices. Doctors' pay, promotion, and professional minimum package of essential clinical services. recognition are enhanced by specialization and by The cost per DALY of the interventions in the the use of expensive new medical technologynot package ranges from less than $5 to more than by serving as public health doctorsor district $200; average costs also reflect those public health medical officers in poor ruralareas. Badly de- measures that do not improve health directly but signed government salary schedules and pricesys- that are essential to the functioning of a health tems may exacerbate this trend. In China hospitals system. currently have a strong incentive to usenew diag- In both low-income and middle-incomecoun- nostic and therapeutic technologies, for which full tries the marginal cost per DALY would be less costs can be charged, instead of older and lessex- than the average cost because all the fixed costs of pensive technologies for which government-set infrastructure are included in these estimates and prices are far below actual costs. there would be spare capacity for producing small Consumer demand for cost-effective services is additional amounts of most services. Because of often weak. This may reflect lack of information. joint costs, it is difficult to separate the costper In rural Africa, for example, goiter and impaired DALY for every intervention in the package. mental abilities from iodine deficiency have in In middle-income countries the package could many places become accepted as the normal state be entirely financed by reallocating current public of affairs. Low demand may also reflect deficien- spending. In poor countries there would bea cies in supply. Most cost-effective interventions shortfall of about $10 billion a year, which could be can be delivered at primary care sites, but in poor

68 countries such as Burkina Faso and Mali, more health care center or doctor's office. In Indonesia than half of the population lives more than 10 kilo- in 1991, for example, rural households in the top meters from the nearest primary care center. income decile were three times more likely to live At a more fundamental level, the distribution of in a village with a health center than those in the political power explains much of the misallocation bottom decile. of government resources for health. The urban Partly because of difficulties in access, the poor population is better organized than rural groups in developing countries generally consume fewer and more vocal in demanding health facilities and health services. Household surveys from Sub- services. Similarly, middle-class workers in wage Saharan Africa and Latin America demonstrate employment, who frequently belong to powerful clearly that among people who report themselves trade unions, are more effective than self-em- to be sick, those in urban areas seek and obtain ployed farmers and workers in the informal sector medical care more often than those in rural areas, in lobbying for government-subsidized health and the wealthy contact a care provider more often benefits. Health professionals are also often better than do the poor. The differences can be large: in organized than the population they serve, and in Côte d'Ivoire in the mid-1980s, for example, an promoting their own interests they may make the urban household was nearly twice as likely to seek health system less efficient. Despite these prob- care as a rural household (60 versus 36 percent), lems, many countries have succeeded in dramati- and within the rural population a family in the top cally improving the health of their people. This income quintile was almost twice as likely to seek success can be accelerated, as newly available in- care as a family in the bottom quintile (44 versus 23 formation makes it clear how costly misallocation percent). is and how much health can be gained for rela- A study of Peru showed similar inequalities tively modest levels of spending. amonggeographicregionsandeducational groups. There was little variation in self-declared Equity in health status, utilization, and finance illness, but the likelihood of obtaining medical care when sick was nearly three times higher in some Data on health status, physical access to health parts of the country than in others. Regional dif- services, consumption of health care, the distribu- ferences in immunization rates were highest for tion of the financial burden of health care spend- uneducated mothers, whose children were only ing, and public expenditures for health all tell the one-third as likely to be fully immunized as the same story of severe inequities in developing children of women with secondary schooling. The countries. In Bangladesh, for example, the infant 36 percent of all the self-declared sick who lived in mortality rate for the urban poor (13.4 percent) is the capital city accounted for 53 percent of all Min- nearly twice the urban average (6.8 percent) and istry of Health ambulatory consultations, 41 per- about 50 percent higher than the average rate for cent of hospital admissions, and 47 percent of all the entire country (10 percent). In China, despite public expenditure attributable to care for individ- remarkable overall progress in health (infant mor- ual patients. At the other extreme was Piura, a tality fell from 20 percent in 1950 to 4.6 percent in poor mountainous region with 10 percent of the 1982), there is considerable geographic variation, sick but only 4 percent of public spending and of which is strongly related to income. Poor regions consultations. such as Yunnan, Xinjiang, and Tibet have infant Inequity in public spending for health both ac- mortality rates of more than 7 percent, compared counts for and reflects marked inequalities in ac- with less than 2 percent in more affluent Beijing, cess to and utilization of care. In Indonesia, for Guangdong, Shanghai, and Tianjin. To take an- example, despite significant investments in lower- other example, in Kenya the probability of a level health facilities in the 1980s, only 12 percent child's dying before age 2 varied among ethnic of public spending for health in 1990 went for ser- groups from 7.4 to 19.7 percent, and in Cameroon vices consumed by the bottom 20 percent of theseprobabilities ranged from 11.6to20.5 households, while the top 20 percent obtained 29 percent. percent of the government subsidy. This bias in The poor also have considerably worse access to favor of the wealthy was mainly a result of the health care. A number of surveys show that low- distribution of government spending for hospital income households, especially in rural areas, have inpatient and ambulatory care, services that were to travel considerably farther or longer to reach the used more frequently by the rich. Much more un- first level of referral services, usually a primary equal situations can be found in many countries

69 that concentrate government spendingeven more Household surveys systematically show that on high-level facilities. people choose whether to seekcare and which The few countries in which public spendingon provider to consult on the basis of many factors health is biased toward the poor show thatgovern- hours of service, travel time or cost, waiting time, ment policies can help reduce inequities inaccess availability of doctors or of drugs, and howpa- and health status. In Malaysia thegovernment has tients are personally treated. The time requiredto followed a pro-poor policy since the 1970s, with get care can be valued according to localwages the lowest-income groups receivinga larger share and treated as a cost of service together with of public subsidies for health than the middle class money payments. On this basis, free public medi- and the wealthy. Similarly, in Costa Ricagovern- cal care often is more costly than unsubsidized pri- ment spending for health has continued to favor vate care for which patients do not have to travel the poor, despite economic shocks anda major so far or wait so long. It is not surprising that, adjustment program in the 1980s that entailedcuts under these circumstances, evenpoor people ex- in public expenditure. In 1988 about 30 percent of press their dissatisfaction with public services by government spending for health went to the poor- paying for a great deal of private outpatientcare. est 20 percent of households and just over 10per- In both El Salvador and the Dominican Republic cent of spending to the richest 20 percent. The residents of the poorest quintile of the capital city poverty-oriented pattern of public spending for obtain more than half their ambulatorycare from health in Costa Rica can be explained largely by private physicians. Although the price of private the high degree of coverage by the social security care to the poorest quintile is, on average, half that health system (the entire population is covered in for patients in the richest quintile, it is still fifteen principle, even though only 63 percent of the times higher than Ministry of Health fees. Differ- working population contributes) and the relatively ences in waiting timeone hour for private pa- equal access to and quality of care enjoyed by all tients as against two and a half hours at Ministry Costa Ricans. It also helps that the wealthyget of Health facilitiesaccount for much of this effect. most of their outpatient care from the private Sensitivity to price and travel time is also found sector. in rural Peru and Côte d'Ivoire. But private hospi- Consumer satisfaction with health care tal care is still much too expensive for thepoor; even those who use private doctors go to public How satisfied people are with theirown health hospitals. The excessive use of hospitalcare in re- and their health care can be only partly explained lation to ambulatory services oftenseen in public by objective criteria; subjective expectationsmat- health systems partly reflects dissatisfaction with ter. People can also be pleased with theirown the high cost in time and the perceivedpoor qual- health care and dissatisfied with theircountry's ity of ambulatory care. In the absence of incentives health system as a whole. A comparison often to improve lower-level facilities and service, this OECD countries with different healthsystems overuse reinforces the tendency to concentrate re- found that in eight countries public satisfaction sources on hospitals, urban areas, and less cost- was related to the level of spending. Canada, with effective interventions. the second-highest expenditure, had the highest The importance of public satisfaction witha satisfaction rating, and people were generally bet- health care system raises two issues for the pack- ter satisfied with the costlier health systems of age of publicly financed services proposed here. France, Germany, and the Netherlands than with First, it suggests that qualitycan be maintained the lower-spending systems in Australia, Italy,Ja- only if coverage is broad enough. Services de- pan, and the United Kingdom. Both very highex- signed only for the poor will almost inevitablybe penditure and great dissatisfactionwere found in low in quality and will not receive the politicalsup- the United States. The study also showed that port necessary for adequate provision. This isa having a unified national health system didnot difficult political issue because itmay be hard to guarantee a high level of satisfaction. In most maintain equity and control costs ifcoverage is countries 30 to 50 percent of those polledsup- universal. The proper balance betweenmore care ported "fundamental changes" in the healthsys- for fewer people and thesame amount of care for tem. In Italy and the United States many people more people depends on ensuring that thepoor thought that such changes would not sufficeand have access to the same quality ofcare as everyone that the health system should be "completely else and on limiting public financeto cost-effective rebuilt." services for which there is a sound rationale. Sec-

70 ond, reform of public provision alone, important more cost-effective care, but they also cannot as it may be, may have much less effect on health provide equitably what care they do offer be- outcomes, costs, and satisfaction than reforms that cause facilities will inevitably be geographically also try to stimulate competition and improve peo- concentrated. ple's access to a variety of providers. One of the principal responsibilities of govern- ment is to match the available instruments of poi- Matching means and ends icythe levers the public sector actually controls to the objectives. Much of governments' failure to The objectives of a health system are to improve achieve better health outcomes derives not from outcomes, control costs, increase equity, and sat- the wrong choice of objectives but from the wrong isfy users. Policy instruments, however, do not choice of instrumentsin particular, from too correspond to individual objectives. What govern- much reliance on direct provision of care and cen- ments actually do is build facilities, buy equipment tral control of health facilities and too little use of and supplies, hire and train people, set fees or the financial, informational, and regulatory instru- other service conditions, regulate providers and ments at the disposal of the government. These insurers,disseminateinformation,determine instruments are particularly important for improv- overall policy, and maintain surveillance of disease ing performance in the private market. When gov- conditions or other variables. Misallocation and in- ernments pay for health care in addition to regulat- equity are caused by mistakes in deciding what ing it, they have a further responsibility to provide facilities to build, where to locate them, how to value for money by ensuring that public resources staff them, and what services to provide. If gov- go first to cost-effective public health and essential ernments spend too much on tertiary care, for ex- clinical services so as to buy the largest health gain ample, not only can they not adequately finance possible.

71