Difficult Health Policy Choices in a Severe AIDS Epidemic

AIDS will affect the health sector in two ways: by increasing demand and by reducing the supply of a given quality of care at a given price. As a result, some HIV-negative people who would have obtained treatment had there been no epidemic will be unable to do so, and total national expenditure on health care will rise, both in absolute terms and as a proportion of national product.2

Increased demand for care. Most people who develop AIDS are prime-age adults. Without AIDS, this 15-to-50 age group accounts for only 10 to 20 percent of all deaths in a developing country, but these deaths typically generate a disproportionate share of total health care demand (Over, Ellis, Huber, and Solon 1992; Sauerborn, Berman, and Nougtara 1996). Moreover, since several studies suggest that adults with AIDS use more health care prior to death than those who die of other causes, or even of other prolonged illnesses, the percentage increase in the demand for care by adults is likely to exceed the percentage increase in their mortality due to AIDS. As a result of these two factors, in a country where prime-age adults utilized one-quarter of all health care before AIDS, a given percentage increase in their demand for health care will increase total demand by at least one-quarter of that percentage. For example, a 40 percent increase in the mortality rate of prime-age adults will increase total demand by at least 10 percent, even though total mortality has increased by only 4 percent to 8 percent.3 If AIDS patients use expensive antiretroviral therapies, the increase in demand will be much greater.

How much the demand for care increases in the aggregate depends on the increase in the prime-age adult death rate, which in turn depends on the level of HIV prevalence and the median time from infection to death (table 4.3). A stable prevalence rate of 5 percent among prime-age adults eventually increases their annual mortality by about five deaths per 1,000 adults if the median time from infection to death is ten years, or by about ten deaths if the median time is only five years.4 A prevalence rate of 30 percent, such as is observed in Lusaka, Zambia, will increase the number of deaths per 1,000 adults by 30 to 60, depending on the median time to death. In Sub-Saharan Africa, where mortality rates in this age group were as high as five per 1,000 before the epidemic, even a 5 percent infection rate will double or triple the adult death rate. In a middle-income developing country with adult mortality of one per 1,000, the same endemic level of HIV infection will increase prime-age adult mortality five- or tenfold.

Table 4.3 Deaths per Thousand Adults Caused by a Constant Rate of HIV Infection

Given these parameters, how much will the epidemic increase the demand for care? In a country where adults consume one-quarter of health care prior to the AIDS epidemic, HIV prevalence is constant at 5 percent of adults, the median time to death is ten years, and the baseline mortality rate among prime-age adults is 5 per 1,000, the epidemic will cause a 26 percent increase in the demand for health care at every price.5 If the prevalence rate is higher, the median time to death shorter, or the baseline adult mortality rate smaller, the percentage increase in demand will be correspondingly greater.

A final important factor that may increase demand is insurance. This may take the form of private insurance, a government-run insurance program, or, more typically, health care financed through general taxation. Because a portion of health care costs is often covered by one or more of these types of insurance, the price paid by the patient is usually a fraction of the cost of providing the care. Since insurance enables patients to purchase more care than they would otherwise, it increases the demand for care arising from any given level of illness, thus magnifying the price shock of an AIDS epidemic. For example, if the proportion of cost of providing care paid by patients (i.e., the coinsurance rate) is 25 percent, patients will reduce their utilization in response to increased cost by only a quarter as much as they would if they had to pay the full increase.

Reduced supply of health care. In addition to increasing the demand for care, the AIDS epidemic will reduce the supply available at a given price, in three ways. The magnitude of these effects, discussed below, will generally be larger in the poorest countries with the largest epidemics.

The first and largest effect is the increased cost of maintaining a given level of safety for medical procedures. Even without HIV, hospitals and clinics in poor countries may pose a risk to health. Needles and other instruments are not always sterilized, rooms are often overcrowded and poorly ventilated, and care providers may lack rubber gloves and sometimes even soap. Without modern blood banks, a transfusion might infect the recipient with hepatitis B. In such situations, infections of all types spread rapidly; some, including such common illnesses as pneumonia, may kill. Before HIV, however, infections picked up in a clinic or hospital were rarely fatal to persons not already in a seriously weakened state.6

Because the AIDS epidemic has greatly increased the risk to patients of existing medical procedures, simply maintaining the level of safety that existed before HIV requires additional hygiene and blood screening, both of which increase the cost of care. In middle- to high-income countries, where blood screening and sterilization of injecting equipment are already the norm, the impact of AIDS is confined to the incremental costs of adding an HIV test to existing tests and using rubber gloves and face masks in situations where they were previously not used. In poor countries, where blood screening and needle sterilization were lacking before the epidemic, the resources needed to maintain the quality of care in the face of the AIDS epidemic can be substantial. For example, the annual recurrent budget of the Ugandan Blood Transfusion Service, which was established in response to the epidemic and meets the demands of the entire Ugandan national health care system for clean blood, is estimated to be about $1.2 million, including capital and recurrent costs. This amounts to about 2 percent of national public health expenditures or about 1 percent of total national health expenditures (European Commission 1995a). Despite the potentially high costs of blood screening, HIV has greatly increased the justification for a government role in ensuring a safe blood supply. However, there is no convincing rationale for government to subsidize the entire cost of running such a service indefinitely (see box 4.1). Blood screening and improved collection procedures will protect blood donors and recipients. However, since average donors and recipients do not engage in unprotected sex with a large number of partners, a person infected while giving or receiving blood is not likely to pass the infection to many others. Thus, in developing countries where the cost of establishing a safe blood supply is high, blood screening will not be among the more cost-effective approaches to preventing an epidemic based on sexual transmission (see box 4.2).

To be sure, blood screening and better hygiene will help to prevent the spread of other infectious diseases besides AIDS. Such measures will also reduce the occupational risk of AIDS and other diseases that health care workers face, and therefore reduce the amount of additional compensation needed to offset their occupational risk—an issue we discuss below. A careful accounting of the net cost of protecting patients from HIV by screening blood would need to take into consideration these additional benefits, for which data are lacking. However, it seems likely that even if these benefits are taken into account, the remaining cost of screening blood and improving hygiene to protect patients from HIV/AIDS would substantially increase the unit cost of medical care.

The second factor reducing the supply of medical care at a given price is the increased attrition of health care workers who become infected with HIV. Like all adults, health care workers may become infected with HIV as a result of sexual contact or use of unsterile injecting equipment. They also face an additional risk of becoming infected in the course of their work; however, this risk is generally much smaller than the risk from sexual contact. Thus whether the AIDS mortality rate among health care workers is higher or lower than among the general population depends mostly on the effects of income, education, and social status on sexual behavior. Two studies of HIV prevalence among health care workers from Africa suggest that doctors and nurses are at least as likely to become infected as other people (Mann and others 1986, Buvé and others 1994). If this is true elsewhere, a country with stable 5 percent HIV prevalence can expect that each year between .5 and 1 percent of its health care providers will die from AIDS; a country with 30 percent prevalence would lose 3 to 7 percent of its health care workers to the epidemic. This attrition from AIDS deaths may substantially increase the cost of health care. For example, if labor costs are half of total health care costs, and training or recruiting a replacement worker requires a one-time expenditure equal to the worker’s annual salary, then a 7 percent increase in attrition will increase total costs in the health sector by 3.5 percent.

The third way in which AIDS reduces the supply of health care is through the additional risk it imposes on health care workers. Even though most HIV-infected health care workers acquire their infection through sexual contact, in a society with a large proportion of HIV-positive patients, health care work will be more dangerous than if there were no HIV. Some students who would have become doctors and nurses will therefore choose alternative occupations, unless they are compensated with higher pay for the increased risk. A recent survey of medical and nursing students in the United States found that AIDS had indeed reduced the attractiveness of specialties in which contact with HIV-positive patients was more likely (Bernstein, Rabkin, and Wolland 1990; Mazzullo and others 1990).This problem is likely to be most severe in hard-hit developing countries, where HIV prevalence is much higher and rubber gloves and other protective equipment are often in short supply. In Zambia, for example, some nurses have demanded special payments to compensate for increased occupational risk due to HIV (Buvé and others 1994).

The magnitude of increased costs of medical staff has not been estimated. As noted above, improved precautions in hospitals and clinics may reduce these costs. But because people respond to perceived risk rather than actual risk, such improvements may have little impact on the demand for increased compensation. Thus, it seems clear that health care workers’ perception of risk will increase the cost of care.

The total impact of these three effects—increased cost of preventing infection in medical facilities, attrition of health care workers due to HIV, and additional pay that health care workers demand to compensate them for increased risk—will depend most importantly on HIV prevalence and whether modern blood banks and hygiene were already in place. In a country that has 5 percent HIV prevalence among prime-age adults and lacked blood banks and blood screening before the epidemic, a conservative guess is that the cost of providing care of a given quantity and quality will rise by about 10 percent.

Scarce care, higher expenditures. Taken together, increased demand and reduced supply have two related impacts: first, health care becomes scarcer and thus more expensive; second, national health care expenditure rises. The size of the increases in health care prices and national health care expenditure depends partly on the price-responsiveness, or "elasticity," of the demand for and supply of care. For most goods, higher prices reduce demand, as consumers switch to substitutes or forgo an intended purchase altogether. This same principle holds true for health care, but the price-responsiveness or elasticity of demand for adult health care is usually small, since there are no close substitutes, and people who are sick and who have the ability to pay will often pay whatever is needed to get well. For the purposes of our simulation, we assume that a price increase of 8 percent would decrease utilization by only about 8 percent, for an elasticity of 0.8.7

Higher prices also generally increase supply. Here, too, however, the nature of the health sector affects the supply response. In the very short run, perhaps a month, the supply of care is unlikely to change much. Over the long run, the supply of physicians and inputs to health care can expand as much as necessary. Over the medium run, five years or so, we would expect the supply of care to respond somewhat to increased demand and the resulting higher price. One response observed in Canada, Egypt, India, Indonesia, and the Philippines is that physicians who work in the public sector rearrange their schedules to offer more health care privately, after their obligations to the government have been met. The elasticity of this response has been estimated at about 0.5, meaning that every 10 percent increase in the price of care elicits a 5 percent increase in supply (Chawla 1993, 1997; Bolduc, Fortin, and Fournier 1996).

We have argued in the previous two subsections that a constant 5 percent seroprevalence rate would eventually increase the demand for care by about one-quarter and the cost of care of a given quality by 10 percent. Drawing on the assumptions in this subsection about the elasticities of the demand and supply responses, and assuming that patients pay half the cost of health care, box 4.3 shows that total national health expenditure, and also the government’s share of expenditure, would both increase by about 43 percent. The increase would be less in a country like India, where only about one-fifth of the cost of care is paid by the government, and substantially more in countries like those of Latin America and Eastern Europe, where three-quarters or more of the cost are subsidized.

Does the available empirical evidence support these conclusions? Although there are significant data problems, the short answer is yes.

Measuring the scarcity of medical care through changes in the price of care of given quality is problematical because of the difficulties in measuring quality. This is especially true in developing countries, where a general lack of data is compounded in the health sector by government subsidies and nonprice forms of rationing. In such cases, the effective price of care may rise even though nominal prices remain constant (see box 4.4). Furthermore, because of the lag between infection and death, the time between the attainment of a given HIV prevalence rate and the full impact of that rate on the demand and supply of health care can be ten to 20 years. For these reasons, we cannot accurately assess changes in scarcity of health care in developing countries by observing changes in nominal price. Nonetheless, we can get some sense of the extent to which HIV/AIDS increases the effective price of health care by considering whether the epidemic makes it more difficult to obtain care. Studies of hospital admissions data strongly suggest that this is the case.

Table 4.4 shows the percentage of beds occupied by HIV-positive patients in six referral hospitals in developing countries with large epidemics. The hospitals are the top health care institutions in each country, providing the best care available outside of a few expensive private clinics. Because these hospitals are at the apex of their health care pyramids, we would expect that AIDS patients account for a significant proportion of their patients. Even so, the percentage of beds occupied by HIV-positive patients is striking, ranging from 39 percent in Nairobi, Kenya, to 70 percent in Bujumbura, Burundi.

If the hospitals were operating well below capacity before the epidemic, they might have accommodated the HIV-positive patients without reducing care for HIV-negative clients. Although no data on occupancy prior to the epidemic are available for these specific hospitals, bed occupancy rates in such hospitals typically were well above 50 percent even before AIDS.8

The best evidence that AIDS is making it more difficult for people not infected with the virus to get medical treatment comes from an in-depth study of Kenyatta National Hospital (KNH), the premier teaching hospital in Nairobi, Kenya. The KNH study compared all patients admitted during a sample 22 days in 1988 and 1989 with all patients admitted during a sample 15 days in 1992 (Floyd and Gilks 1996). Panel A of figure 4.2 shows that while the average number of patients admitted per day increased from 23 to 25, the number of HIV-positive patients more than doubled, while the number of HIV-negative admissions shrank by 18 percent. Since the number of HIV-negative people in the hospital’s "catchment area" could not have shrunk by this much, this evidence suggests that the AIDS epidemic did in fact result in some HIV-negative patients being dissuaded or barred from admission to the hospital.

Table 4.4 Evidence of Possible Crowding Out of HIV-Negative by HIV-Positive Patients, Six Countries, circa 1995

b. Since Floyd and Gilks found the average length of stay to be identical across HIV-positive and -negative patients, the ratio of HIV-positive to total admissions is a useful estimate of the proportion of beds occupied by HIV-positive patients. Thus this entry is calculated from figure 4.2 as 9.6/24.9.

Sources: First four hospitals, van Praag 1996; Kenyatta Hospital, Floyd and Gilks 1996; Rubaga Hospital, Tembo and others 1994.

Since the HIV-infected make up an increasingly large fraction of the sick people in a severely affected country, it is appropriate that they occupy an increasing share of hospital beds and consume an increasing share of health care resources. The pressure of this increased demand for care will naturally be felt by all citizens, whether or not they are HIV-infected. However, the extent of the shift in health care resources away from the HIV-negative can be exaggerated, as indeed it may have been in Kenyatta National Hospital, if the government provides special subsidies for people with HIV.9 We discuss this issue, and the broader issue of how the level of government health care subsidies affects the demand for care and health care expenditure, in the next section.

Scarcer and more expensive care and increased total health expenditure present society with difficult choices. Because a large share of the increased expenditure is typically financed through tax revenues, governments and their constituencies will confront tradeoffs along at least three dimensions:

• treating AIDS versus preventing HIV infection
• treating AIDS versus treating other illnesses
• spending for health versus spending for other objectives.

The need to confront these difficult choices can be reduced somewhat if a government is willing and able to increase tax revenues. But few countries will be able to avoid the choices entirely, especially developing countries facing a severe epidemic. Unable to pay for everything, most governments will subsidize some goods and services more than others, thereby disproportionately benefiting certain groups of citizens.

As the number of AIDS cases increases, governments are likely to face mounting pressure for two responses that on first consideration seem rational and humane. One is to pay a larger share of health care costs; the other is to provide special subsidies for the treatment of HIV/AIDS. Unfortunately, these responses can have unintended consequences. For reasons discussed below, governments that wish to minimize the impact of HIV on the health sector should try to avoid both courses of action. However, this does not mean that governments should do nothing to help alleviate the suffering caused by HIV/AIDS. The section concludes with a list of compassionate and affordable measures that governments can and should undertake to mitigate the health sector impact of an HIV/AIDS epidemic.

No increase in the overall subsidy to health care. One obvious and politically appealing response to the HIV/AIDS epidemic is to increase the government share of health care costs and thus the overall subsidy for health care. Such a course of action may be especially attractive early in the epidemic, when few people are sick with AIDS. There is an argument for it on economic grounds as well: it would fill the gap created by the failure of the private market to offer health care insurance in poor countries. However, increasing the subsidy to curative care increases the demand for a limited supply. As a result, both effective price and total expenditure will rise by a greater proportion than the increased subsidy alone, or the increased demand arising from the epidemic alone, or even the sum of the two, would suggest. As more and more people become sick with AIDS, this effect becomes evident in escalating health care expenditures; in a severe epidemic, the burden on the government budget is likely to become unsustainable.

To understand how changes in the level of government subsidies affect the impact of the epidemic on the health care sector, we first look at the extent to which governments already subsidize care. Then, taking India as an example, we project the impact of an expanding epidemic at the current subsidy level and an increased subsidy level. As we shall see, increasing the overall subsidy to care can greatly exacerbate the impact of the epidemic on the health sector.

Most governments subsidize a large share of health care expenditures. The balance includes payments by private insurers and all "out-of-pocket" payments at private or government-subsidized facilities, whether traditional or modern. The average overall subsidy to health care varies widely but generally rises with GDP. As can be seen in figure 4.3, the poorest countries, with average per capita income of about $600, typically subsidize less than half of the cost of health care, while upper- income countries subsidize about three-quarters of the cost.

In India in 1990 the government subsidized about 21 percent of total health care expenditures, a small share even compared with other low income countries. The bottom line of figure 4.4 projects government health expenditure if India were to have no AIDS epidemic and continued to spend 6 percent of a constantly growing GDP on health care, of which the government continued to finance 21 percent. In this baseline scenario, India’s government health expenditures grow from $3.2 billion in 1991 to $8 billion in 2010. The second line from the bottom shows the increase in government health spending if India’s current steep rise in HIV prevalence continues until 2000, then levels off at a stable 5 percent. This is about the growth in prevalence seen in countries such as Zambia and Botswana, where focused prevention was not implemented early in the epidemic. The result in India would be to increase the government’s expenditure on health in the year 2010 by about one-third, from $8 billion to $10.5 billion.

What if India in 1990 had increased health care subsidies to about 50 percent, the level seen in many Latin American countries? The top pair of projections in figure 4.4 shows the impact of the higher subsidy on expenditure. Even without an AIDS epidemic, expenditure more than triples to $11 billion in 1991 due to the more than doubling of the government’s share of existing expenditure combined with the demand stimulus caused by the greater subsidy. Subsequent growth of health expenditure proportional to GDP brings health expenditure to $27 billion in 2010 (third line from the bottom). Now again suppose a serious AIDS epidemic that reaches a stable 5 percent HIV prevalence rate in 2000. The fourth line from the bottom of figure 4.4 gives the projected result: health expenditures in 2010 would reach $39 billion. Thus, not only has the increased subsidy tripled health care spending, as might have been expected, but it has also increased the vulnerability of the budget to the AIDS epidemic, adding $12 billion (43 percent of $27 billion) rather than just $2.5 billion (31 percent of $8 billion) to government health care expenditures.

The large expenditure shocks that will result from the AIDS epidemic will create new pressures on health budgets, especially in countries that enter the AIDS epidemic with higher subsidy rates. For example, although Mexico’s infection rate was estimated to be only 0.4 percent in 1994 and it subsidized only 49 percent of the cost of AIDS treatment, compared with 76 percent for other sicknesses, AIDS was already consuming 1.2 percent of its health budget. In contrast, Tanzania has kept the subsidy rate for AIDS treatment down to 28 percent in line with the subsidy it provides to other illness categories. As a result, despite a prevalence rate of 5 percent, more than ten times higher than Mexico’s, the AIDS share of total government health care expenditure is only 3.5 percent, just three times larger in Tanzania than in Mexico.10

Although a discussion of the design of health financing systems is beyond the scope of this book, the evidence suggests that countries in the nascent or concentrated stages of the epidemic, like India, should carefully consider not only the immediate budgetary consequences of any expanded commitment to fund curative care, but also the multiplication of these consequences that would occur if the AIDS epidemic spreads. A prudent course would be to consider any expansion of government-financed health care subsidies or insurance only in conjunction with vigorous prevention programs that enable people most likely to contract and spread HIV to protect themselves and others.

Equal subsidy rates regardless of HIV status. A second common health sector response to the HIV/AIDS epidemic is to offer a different subsidy rate depending on whether or not the person receiving care is infected with HIV. Especially in the countries in the nascent stage of the epidemic, HIV-infected people all too frequently experience discrimination, including restricted access to or higher effective prices for health care. As the epidemic advances, however, governments are often pressed to provide special subsidies for the treatment of HIV/AIDS. This section points out the government’s role in limiting discrimination against the HIV-infected in health care settings and then considers the consequences of preferential subsidies for HIV treatment.

AIDS treatment subsidies vary greatly from country to country. Figure 4.5 presents data on the percentage of AIDS-related and total 1994 health care expenditure funded by the government. In three of the five countries, the subsidy rates for AIDS treatment are significantly different from that for total health care expenditure. For example, although Mexico subsidized a generous 49 percent of the cost of AIDS treatment, this was much less than the 76 percent share of total health care expenditure. Brazil and Thailand subsidized AIDS care at a higher rate than all types of care, while Tanzania and Côte d’Ivoire subsidized AIDS treatment and total health care expenditure at roughly the same rate.

A bias against those with HIV/AIDS can take many forms, ranging from a singling out of AIDS-specific drug therapies for exclusion from public funding, to outright refusal of service. There are many anecdotes about discrimination against the HIV-infected in health care settings. In some hospitals, the HIV-infected were placed in special AIDS wards, which were subsequently shunned by fearful health care workers. In others, the HIV-infected were required to pay extra costs for rubber gloves or a private room. In still other cases, the HIV-infected have been denied treatment for common illnesses, perhaps because doctors and nurses mistakenly believed that nothing could be done to help a person with HIV/AIDS. Such discrimination is unfair, unprofessional, and unethical. Moreover, it displays ignorance of the many ways, discussed above, in which inexpensive treatments for symptoms and opportunistic illnesses can prolong and improve the lives of people with HIV/AIDS. Government has an important role to play in training medical personnel in order to eradicate all vestiges of discrimination against HIV-infected patients.

Yet it is equally unfair, and also inefficient, for government to subsidize a higher proportion of the costs of care for patients with HIV than for other patients. Aside from the issue of poverty, to be addressed in the next section of this chapter, there are three ways to justify government subsidies for curative health care: (1) as an incentive for those with an infectious disease to seek a cure and avoid infecting others, (2) as health care insurance with universal coverage and mandatory participation through general taxes, or (3) as government support for a "merit good" or "basic need." No treatment has yet been shown to reduce the infectivity of sexual contact with an HIV-infected person (see box 4.5). AZT treatment of HIV-infected pregnant women has been shown to reduce transmission at birth, but is still too costly an approach to preventing secondary infections in the poorest countries (see box 4.6). With the prominent exception of TB, the treatment of which should be subsidized in all countries, most opportunistic illnesses that afflict the HIV-infected are infectious only to other equally sick HIV-infected people. Thus the argument for treating the diseases on the grounds that they are infectious is weak. If government subsidy is considered as an insurance payment, efficiency criteria argue for a higher coinsurance (i.e., lower subsidy) rate for any condition in which the patient is likely to be highly price-responsive.11 The first section of this chapter established that the drugs and medical services to treat AIDS can amount to a great deal of money, although some of the most expensive of these treatments purchase the patient little additional life span and decrease, rather than improve, the quality of life. Thus on efficiency grounds, where the objective is to limit the responsiveness of expenditure to insurance, AIDS patients should face somewhat lower subsidies, not higher ones. The final possibility, that AIDS treatment is a basic need, is difficult to justify in poor countries where the opportunity cost of treating one adult for AIDS may be measles vaccines for 100 to 200 children or, as shown in figure 1.8, ten student-years of primary school. Thus none of these economic arguments justifies higher subsidy rates for AIDS.

What policy recommendations can be drawn from these two observations? The prudent, efficient, and equitable course is to place the financing of health care for HIV/AIDS on the same footing as other diseases. The treatment of particularly infectious illnesses striking HIV-infected people, including TB and STDs, should be subsidized relatively generously because of the secondary infections treatment will prevent. Other health care problems of the HIV-infected should be subsidized at the same rate that applies to other adult health problems that are equally infectious. Assuming that Brazil subsidizes about one-third of other health care costs (as figure 4.5 indicates it does in São Paulo) and that the infectious proportion of illness episodes is similar among the HIV-infected and uninfected populations, this policy would lead Brazil to reduce its subsidy to antiretroviral therapy from 100 percent to one-third. Similarly Thailand would reduce its subsidy of antiretroviral therapy from 100 percent to about 20 percent. Mexico, on the other hand, would increase its subsidy to AIDS patients to approximately the same rate it offers other patients.

In mid-1997, none of these countries appeared to be following this recommendation precisely. Brazil and Mexico were continuing their former policies, with a tilt of subsidies toward AIDS treatment in Brazil and away from it in Mexico. Having spent $108 million on antiretroviral medication in 1996, Brazil was projecting an expenditure four times that large for 1997 (Chequar 1997). Thailand had recently embarked on an experiment that held out the possibility of an equal percentage subsidy for the treatment of AIDS and other diseases, on average, if not for the individual patient. In 1996, the Thai Ministry of Public Health found that, given the rising patient load, its policy of a 100 percent subsidy for antiretrovirals and drugs for the opportunistic illnesses would soon consume considerably more than the entire budget allocated to the National AIDS Program (Prescott and others 1996). As a result, the government revised its policy to provide free antiretroviral therapy only to HIV-positive pregnant women, where it might prevent mother-to-child transmission, and to participants in nationally approved clinical trials, where patients receive the support they need to maximize compliance (Kunanusont 1997). This policy makes sense for antiretroviral therapy, on the assumption that low levels of compliance outside clinical trials would have little therapeutic effect on patients and might cause negative externalities in the form of drug-resistant strains of HIV. Furthermore, participants in clinical trials produce a positive externality in the form of the knowledge that can be used to benefit many other patients, and therefore should receive a higher subsidy than other patients. Thailand’s decision to subsidize AZT for prevention of mother-to-child transmission can be justified as a "merit good," which might be affordable in a middle-income country (see box 4.6).

Affordable, humane responses to the epidemic. We have argued that governments should avoid two types of health care responses to the epidemic: increasing the overall subsidy to all types of treatment, and providing disproportionately large subsidies to treatment of HIV/AIDS. There are nonetheless several ways in which governments can intervene to mitigate the health impact of HIV/AIDS on infected individuals and their families and on the overall health sector. Each of these interventions is justified on public economic grounds, either because it has large positive externalities, or because it improves the efficiency or the equity of the health care market in other ways.

• Provide information about the efficacy of treatments. Because people with HIV/AIDS are often desperate for treatment and cannot easily research what works, they are especially vulnerable to quackery. Governments can serve the interests of everyone by promptly investigating unproven treatments and providing credible information about their validity. So long as this is done through existing media channels—for example by issuing press releases and arranging media interviews with credible experts—it can be done quite inexpensively.

• Subsidize the treatment of infectious opportunistic illnesses and STDs. Subsidized treatment is especially appropriate for tuberculosis, one of the most common opportunistic illnesses to infect AIDS patients, since curing a single case can avert many secondary infections. Treatment of gonorrhea, syphilis, and the other classic STDs should be subsidized, not only because they are highly contagious, but also because they exacerbate HIV transmission, as discussed in chapter 3. Because few people are susceptible to them, treating toxoplasmosis, cryptococcosis, or one of the other infectious opportunistic illnesses that develop only in people with severely disabled immune systems prevents few secondary cases and thus should be subsidized at a lower rate, closer to the subsidy rate for chronic, noninfectious disease. Whether a subsidy to antiretroviral treatment of HIV itself is justifiable as a way to prevent secondary HIV infections will depend on the efficacy of the treatment and on its cost relative to the cost of other HIV prevention measures. In mid-1997, such treatments were far too expensive and uncertain to warrant subsidies on these grounds (see box 4.5).

• Subsidize the start-up costs for blood safety and AIDS care. The AIDS epidemic has increased the willingness of individuals to pay for certain types of services, such as screening of blood for transfusions and care for the terminally ill. Where these services are lacking, government help with the start-up costs is justified, just as governments subsidize other large indivisible investments, such as an electric utility or a water system, so long as the users then pay for the services they receive. Thus, governments in poor countries should establish blood banks but should not indefinitely provide free blood. Similarly, government should help establish AIDS treatment facilities, especially community-based home care programs, but should not permanently subsidize the care they provide.

• Provide special assistance to the poor. Most countries already make special provision for medical care to the poor. As the AIDS epidemic increases the demand for care, governments may wish to focus such assistance even more on those who can least afford it. Sliding fee scales and other measures to make care available to the poor should apply to people with HIV/AIDS just as they do to people with other illnesses. This principle of providing assistance to those who need it most, regardless of their HIV/AIDS status, is discussed more fully in the next section on ways to mitigate the impact of HIV on poverty.