In the past 2 years, political commitment to respond to the HIV/AIDS pandemic has increased sub-stantially. The United Nations General Assembly Special Session on AIDS in 2001 and the recent creation of the Global Fund to Fight AIDS, Tuberculosis and Malaria are two indicators of this commitment at the global level. At the regional level, HIV/AIDS has been an issue on the agenda of the Inter-American Development Bank, the Asian Development Bank, and the Organization of African Unity, to name but a few institutions. At the national level, low-income and middle-income countries, home to over 95% of people living with HIV/AIDS, have made important progress in HIV/AIDS planning and program development [1,2].
In this policy environment, the importance of information on resources for HIV/AIDS prevention and care has increased. More and more, policy-makers are looking for data on the level and flow of current allocations to HIV/AIDS. They want to know where money for HIV/AIDS prevention and care is coming from, the services and commodities that are purchased with these funds, and the population coverage of implemented interventions. At the same time, to identify gaps between what is and what should be and to plan strategically, policy-makers are seeking information on the scale of resources required to prevent the further spread of HIV and to provide adequate care for those people living with HIV/AIDS.
Because the epidemic is so concentrated in low-income and middle-income countries, estimating HIV/AIDS resource allocations and requirements in these countries is key to responding effectively to HIV/AIDS world-wide. It is also in these countries that this task is most complicated. Program data, ideally produced by national health information systems, are required to monitor resource flows. In many of these countries, however, such systems are weak or nonfunctioning. Similarly, to derive estimates on resource needs, one requires a range of demographic, economic, and health data that are scarce or nonexistent in many developing countries.
Notwithstanding these obstacles, significant progress was made during the past year in both monitoring the level and flow of current allocations to HIV/AIDS and estimating HIV/AIDS resource requirements in developing countries. The present article reviews the latest data and examines their policy implications. It identifies the gaps and limitations of current research. The article also discusses future directions to strengthen the quality of data on resources for HIV/AIDS prevention and care.
Resources allocated to HIV/AIDS
Few countries, whether low, middle or high income, regularly monitor resource flows to the HIV prevention activities conducted by government and nongovernmental institutions within their territory. To date, no country has developed a system that regularly tracks expenditures on HIV/AIDS care.
The most well-established data collection activities documenting resource allocations to HIV/AIDS in developing countries are international initiatives. Each year, donors report their official development assistance (ODA) to HIV/AIDS and other sexually transmitted infections (STI) to the OECD Development Assistance Committee (DAC) . Similarly, the Netherlands Interdisciplinary Demographic Institute surveys donors (annually) and developing countries (biannually) for UNFPA and UNAIDS on their HIV/AIDS/STI expenditures as part of their Resource Flows Project. The most detailed information on resource allocations to HIV/AIDS in developing countries comes from indepth country studies conducted on an ad hoc basis. Most recently, the Regional AIDS Initiative for Latin America and the Caribbean (SIDALAC) and the Partnerships for Health Reform have investigated HIV/AIDS financing in several countries using the National Health Accounts (NHA) framework . The present section discusses the latest data from these sources.
Table 1 presents the HIV/AIDS/STI ODA data reported to the Resource Flows Project between 1998 and 2000, with data reported to the OECD DAC imputed when no data was reported to the Resource Flows Project. Donor countries disbursed a total of US$ 454 million in 2000. According to reports to the OECD DAC, this represents 87% of the US$ 521 million that were committed or allocated to HIV/AIDS/STI projects during 2000. Total disbursements for 2000 represent a significant increase from the US$ 294 million and US$ 279 million disbursed by donors in 1998 and 1999, respectively.
As in previous years, the United States was by far the largest donor of HIV/AIDS/STI ODA in 2000. However, when this ODA is broken down as a proportion of gross national income for each country, Luxembourg contributed the largest proportion of its gross national income (Fig. 1).
What is revealed by these data and past surveys on HIV/AIDS/STI ODA [5–7] is that surveys provide reasonable information on these expenditures, albeit varying in comprehensiveness and accuracy. Past global surveys have also illustrated that questionnaires can provide relatively good data on HIV/AIDS resource allocations that flow into developing countries from the United Nations system and nongovernmental institutions in high-income nations. Although as HIV/AIDS is increasingly integrated into projects addressing broader health or development issues, HIV/AIDS expenditures are becoming more difficult to track even among these international institutions.
However, questionnaires sent to national HIV/AIDS coordinating institutions to collect data on domestic resource allocations to HIV/AIDS are much less efficient tolls for gathering quality data. In part, this is because regularly updated information systems do not exist and it is difficult for national HIV/AIDS coordinating bodies to gather expenditure data from the many ministries and organizations implementing HIV/AIDS interventions in a country. In addition, for large portions of HIV/AIDS expenditure, data must be estimated with special studies. To estimate domestic expenditure on HIV/AIDS care, for example, studies costing selected services must be undertaken. Likewise, to capture how much individuals themselves spend on HIV/AIDS services (out-of-pocket spending), which in many countries constitutes the majority of overall AIDS spending, requires household or clinic-based studies of people living with HIV/AIDS.
It is these information gaps that in-depth country studies using the NHA framework aim to fill. These studies attempt to account for all expenditures by looking not only at public sector financing, but also at spending within the private sector, including spending by individuals. They collect the data that are available and conduct special studies, such as limited household surveys, as necessary.
Most of the studies on HIV/AIDS resource allocations using the NHA framework have been carried out in Latin America and the Caribbean (LAC) . Referred to as National HIV/AIDS Accounts, they were first carried out in Brazil, Guatemala, Mexico and Uruguay in 1997/1998 [9–12], with substantial scaling-up of efforts occurring in the past year. SIDALAC recently completed studies in 12 countries (Argentina, Bolivia, Brazil, Chile, El Salvador, Guatemala, Mexico, Nicaragua, Panama, Paraguay, Peru, and Uruguay) and studies in three additional countries are underway (Costa Rica, the Dominican Republic, and Honduras).
Total HIV/AIDS spending in the 12 countries in Latin America studied (representing 75% of the population of the region)  in 2000 was estimated at US$ 1.04 billion. This represents an average of US$ 2.70 per capita for the 12 countries, with individual country per-capita expenditure ranging from US$ 0.30 and US$ 0.60 in Bolivia and El Salvador to US$ 5.60 and US$ 4.90 in Uruguay and Argentina (Table 2). In terms of average expenditure per person living with HIV/AIDS (PLWHA) in the 12 countries, this translates into a little over US$ 1000 regionally, with over US$ 3000 spent per PLWHA in Uruguay and only US$ 175 per PLWHA spent in Guatemala.
Overall, US$ 753 million (73%) of HIV/AIDS resources in the 12 countries were spent on care, with only US$ 283 million (27%) spent on prevention. This trend was true across countries with the exception of Bolivia and Nicaragua, countries with relatively lower prevalence rates, where only 34 and 36% of HIV/AIDS resources, respectively, were spent on care. Almost 72% of the resources spent on HIV/AIDS care in the 12 countries were spent on drugs, with the vast majority (almost 90% of drug expenditure) spent on antiretroviral drugs (ARV). Of course, social pressure for access to ARV in the region is high, and three of the 12 countries (Argentina, Brazil, and Uruguay) provide universal access to these drugs. However, this estimate does appear to be extremely high and may represent a bias since drug expenditures are easier to monitor than other components, especially when procurement of ARV is centralized through national HIV/AIDS programs, such as is the case in Brazil.
With 60% of prevention expenditure on condoms and 14% on mass media campaigns, most of the prevention expenditure in the 12 countries was spent on interventions targeting the general population. This is notable in countries where HIV prevalence is still extremely low in the general population and the epidemic is concentrated among specific population groups . In Mexico, Nicaragua and most of the Andean region, sex between men is the most prominent route of HIV transmission. Similarly, in Argentina, Chile, and Uruguay, injecting drug use is the main route of transmission. Given the epidemiology of the epidemic in the region, one would therefore expect that significant proportions of HIV prevention expenditures would be allocated to interventions targeting these population groups that are key to the expansion of the epidemic. However, only limited funds in the region (7% of prevention expenditure) were reported as allocated to such interventions.
The major sources of HIV/AIDS funds also varied across countries. In Argentina, Brazil, Chile, El Salvador, Guatemala, Mexico, and Panama, the government health sector was the primary source of resources allocated to HIV/AIDS. In Paraguay, Peru, and Uruguay, private funds from enterprises, nongovernmental organizations and households were the primary sources of resources allocated to HIV/AIDS, while international sources provided most of the funding in Bolivia and Nicaragua.
Even though studies using the NHA framework provide the most detailed estimates on resource flows in countries, they remain estimates that vary in completeness and accuracy. Although the studies attempted to account for all expenditures, the quality of the estimates depended on the availabilityand quality of financial and accounting data within relevant government and non-governmental institutions, and on the capacity of personnel within these organizations to formulate estimates where data were not available. It is probable, as mentioned earlier, that certain categories of expenditures, including spending on information, education and communication interventions, were underestimated because they were more difficult to track than expenditure on commodities such as drugs and condoms. It is also probable that expenditure by nongovernmental organizations, especially those that were community based, were under-reported since donated goods and services were not quantified adequately. Likewise, it is probable that in decentralized governments, with weak financial tracking at lower administrative levels, funds were missed or, alternatively, double counting occurred with expenditures being reported at higher and lower administrative levels.
To estimate the expenditure on HIV/AIDS prevention and care for all countries in LAC, a regression was run to extrapolate the estimates for these 12 countries to the region. [A regression was run for these 12 countries with SIDALAC country totals as the dependent variable and the values from the care imputing exercise (described later) as the independnt variable.] The result is an estimate of US$ 1.4 billion for all countries in LAC.
The only country outside of LAC where the NHA framework has also been used to estimate expenditures on HIV/AIDS is Rwanda. A study conducted for 1998 concluded that a total of US$ 10 million or US$ 1.27 per capita were spent on HIV/AIDS during that year . This represents a total of US$ 25 per person living with HIV/AIDS, compared with the average of US$ 1000 per PLWHA in the 12 countries in Latin America already discussed.
While spending in Latin America is large compared with spending in Rwanda (and presumably other countries in sub-Saharan Africa), even this expenditure is very small compared with expenditure by high-income countries such as the United States. The Federal Government spent US$ 10.8 billion on HIV/AIDS in the year 2000 . If this amount is raised by the same proportion as that which prevails between public and total spending on health in the United States, then total HIV/AIDS spending can be estimated just below US$ 25 billion in 2000 [17,18]. This amount translates into nearly US$ 90 per capita, or just over US$ 30 000 per PLWHA. A check on the credibility of this seemingly high level of spending is provided by an analysis of spending on Medicaid-covered AIDS patients, which projected that expenditure would average almost US$ 36 000 per patient .
Resource needs for HIV/AIDS
Similar progress was made in the area of estimating resource needs for HIV/AIDS over the past year. Two major studies of resource requirements estimates were published. The first, undertaken for the Commission on Macroeconomics and Health (CMH), estimated resources needed to scale-up a package of core interventions to address HIV/AIDS and other priority illnesses in 83 low-income and middle-income countries (including all of sub-Saharan Africa) by the years 2007 and 2015 [20,21]. The second, carried out in preparation for the UN General Assembly Special Session (UNGASS), estimated the cost of HIV/AIDS prevention and care needs in 135 low-income and middle-income countries in 2005 .
These two studies built on prior work on estimating resource needs for HIV/AIDS and used similar methodologies [23,24]. The methodologies used have been detailed previously [20–22]. Both studies included a selection of interventions that were costed based on published and unpublished project assessments (Table 3). Estimates were then made intervention-by-intervention and country-by-country using demographic, economic and epidemiological data to adjust the estimates to different country contexts. The main difference in methodology between the two studies was the inclusion in the CMH study of the costs for infrastructure strengthening necessary for scaling-up. In addition, there were differences in assumptions, with the most important being differences in target population coverage rates for the different interventions.
The UNGASS study called for the annual spending of US$ 9.2 billion on HIV/AIDS prevention and care in low-income and middle-income countries by the year 2005, with up to US$ 6 billion coming from international sources. The CMH study concluded that, depending on coverage assumptions and price estimates, between US$ 13.6 billion and US$ 15.4 billion should be spent on HIV/AIDS prevention and care annually (including necessary infrastructure strengthening) in selected low-income and middle-income countries by the year 2007 in addition to what is already being spent, and that this should increase to between US$ 20.6 billion and US$ 24.9 billion by 2015.
The ranges of the results within and across studies underline the fact that these are estimates with limitations. They underscore the data gap in low-income and middle-income countries and the many assumptions required while building each model to derive parameter estimates for which no data exist. As discussed further later, they should therefore be interpreted with caution and be seen as works in process that can be refined as new information becomes available on cost data, current intervention coverage estimates, and country capacity to expand services.
Nonetheless, sensitivity analysis conducted usingthe UNGASS model confirms that study results provide a consistent estimate of the scale of resources needed. A probabilistic analysis that varied assumptions on intervention coverage, costs and country capacity to expand services produced a range of results that were comparable with the ranges reported in the CMH study.
So, in short, the UNGASS and CMH estimates provide policy-makers with consistent information on the scale of the resources needed. But it would be inappropriate to use them to guide resource allocations among interventions at the national level. Although these two studies did differentiate across countries whenever possible, data limitations did not allow them to pay significant attention to individual country characteristics and the way in which those may affect overall costs.
To improve the estimates so as to have them serve as tools for country strategic planning, both study teams recognized that additional country-level work would be necessary. This process has begun with individual country validations of the UNGASS estimates for the LAC region. The 10 countries to participate in a first phase of this effort were Brazil, Chile, the dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Jamaica, Mexico, and Trinidad and Tobago .
These countries increased the estimated resource requirement for HIV prevention by 15% and the estimated care requirements by 27%. The main differences in prevention estimates are accounted for by an increase in estimated resource needs for the social marketing of condoms and prevention of mother-to-child transmission, while the main differences in care estimates were due to important differences in the expected costs for highly active antiretroviral therapy. Total expected resource needs for highly active antiretroviral therapy increased by 45% compared with UNGASS estimates.
To estimate the HIV/AIDS prevention and care needs for all countries in LAC, the estimates for these 10 countries were extrapolated to the region using linear regressions. [Regressions were run for the prevention and care estimates separately. Because a strong correlation was found between the care data estimated by the countries and those estimated for UNGASS, we used Stata's impute command to impute the values. The command used sub-sets of regressions to impute values for the countries that have not yet revised their estimates. The prevention exercise was similar but, in addition, it used the results of the care imputation as an extra independent variable that was imputed with the linear regression command rather than with the impute command.] The revised UNGASS model called for US$ 1.1 billion, with US$ 550 million for prevention and US$ 550 million for care and support. [Global prevention numbers were refined following the Science publication of the model estimates lowering the estimate for Latin America and the Caribbean from US$ 590 million.] Country revisions increased this by US$ 160 million to almost US$ 1.3 billion, with US$ 480 million for prevention and US$ 780 million for care and support.
Since the discrepancy between the UNGASS estimates and those of the 10 specialists is reasonably small, these preliminary estimates provide further support for the overall consistency of the results.
To plan and implement effective responses to the epidemic, policy-makers require data on resources for HIV/AIDS prevention and care. Without a representative picture of public and private spending on the epidemic, governments are unable to track and assess the impact of their response to HIV/AIDS. Without some estimate of the scope of potential resource needs to address the epidemic adequately, they are unable to plan strategically and mobilize resources for the future. Those data that are available highlight the need for additional resources and for greater efficiency in the utilization of those resources available. But to date, few policy-makers in low-income and middle-income countries (those countries that need this information most urgently) have these data for their countries.
Data are available on the annual official development assistance allocations to HIV/AIDS by high-income countries with some limitations. These data suggest that there was a significant increase in the flow of HIV/AIDS funding from high-income countries to developing countries in 2000, although US$ 450 represents only a fraction of the estimated US$ 5–6 billion in international resources required annually to respond to the epidemic in these countries.
Similarly, in-depth country studies (most of which have been carried out in LAC) highlight the vast inequalities in spending on people living with HIV/AIDS that exist globally. These studies suggest that an average of US$ 1000 per person living with HIV/AIDS was spent in Latin American countries in 2000. This is significantly more than estimates for sub-Saharan Africa (where 70% of people with HIV/AIDS live) of less than US$ 50 and far less than estimates for the United States at over US$ 30 000.
Comparing the results of the UNGASS resource-needs model for LAC (almost US$ 1.3 billion for the year 2005) with those of the National HIV/AIDS Accounts studies expanded to the region (US$ 1.4 billion for 2000) further illustrates some of the important policy implications that are raised by these type of data. The higher estimates of current expenditure are due in part to the fact that the two estimates do not measure the resources associated with identical packages of interventions. The two exercises also make different assumptions on costs of inputs, and estimates of current intervention coverage. In addition, they are also based on different sub-sets of countries and there are therefore different levels of uncertainty in expanding these estimates to the region. However, the greatest difference probably lies in the assumptions made in the UNGASS model that the procurement of commodities such as ARV, condoms, and HIV test kits is efficient, and that there is technical and managerial efficiency in the implementation of HIV interventions. The fact that estimates of current expenditure are higher than those for future resource needs in part reflects the efficiency gains that could result from improved purchasing arrangements that enabled countries in the region to pay less for condoms and ARV.
To strengthen the quality of HIV/AIDS financing data in developing countries, necessary next steps include the institutionalization of studies to track National HIV/AIDS Accounts. SIDALAC has demonstrated that the process is feasible. The studies conducted in the region linked with the National Health Accounts promoted by the World Health Organization and the OECD have cost between US$ 25 000 and US$ 55 000 per year per country depending on country size and existing capacity. This demonstrates that limits in financial resources need not be a barrier to the execution of a fairly complete system for monitoring HIV/AIDS resource flows.
At the same time, estimates of country-level resource needs should be compiled for all low-income and middle-income countries. Workshops are planned for the rest of the LAC region. Similar initiatives should be undertaken in sub-Saharan Africa, Asia, and Eastern Europe.
Finally, one important component regarding HIV/AIDS prevention and care resources in developing countries is still missing. Ideally, estimates of resource needs should go a step further, approximating not only resource needs, but also the benefits provided by the resources spent. Focusing on the benefits expected from new investment would provide data critical to decisions on the distribution of resources within and across countries and sectors.
The authors are grateful to Eric Lief and Lisa Regis of UNAIDS for the use of the HIV/AIDS/STI data reported to the Resource Flows Project. They acknowledge the contributions of the SIDALAC country consultants from Argentina, Bolivia, Brazil, Chile, El Salvador, Guatemala, Mexico, Nicaragua, Panama, Paraguay, Peru, and Uruguay who collected the data on the National HIV/AIDS Accounts. They also thank Nicolas Noriega of SIDALAC for his assistance with the compilation of the National HIV/AIDS Accounts data.