Optimizing resource allocation for HIV/AIDS prevention programmes: an analytical framework

Bautista-Arredondo, Sergioa; Gadsden, Paolaa; Harris, Jeffrey Eb; Bertozzi, Stefano Ma,c,d

doi: 10.1097/01.aids.0000327625.69974.08
Section I: Informing Scale-up and Resource Allocation

Introduction: Although investment in HIV/AIDS prevention has increased worldwide, it remains uncertain how the additional resources can be most efficiently allocated to maximize the number of infections averted, especially at the country, regional and local levels.

Methods: Data from developing countries in Africa, Asia and Latin America were reviewed on the allocation of HIV/AIDS prevention funds in relation to the prevalence of infection, as well as budgetary allocations for specific population groups at high risk of infection, such as sex workers, intravenous drug users and men who have sex with men. The variation in unit costs of voluntary counselling and testing in five countries was also examined.

Results: Evidence was found of three distinct sources of inefficiency in the allocation of HIV/AIDS prevention resources: inefficiency in the mix of interventions selected; inefficient targeting of key populations; and technical inefficiency in the production of HIV prevention services.

Conclusion: A general conceptual framework for evaluating the efficiency of HIV/AIDS prevention programmes at the country, regional and local levels is proposed. This framework stresses three equally important components of programme efficiency: cost-effectiveness (the choice of the mix of interventions); targeting (the choice of the mix of target populations); and technical efficiency (the delivery of prevention services at least cost).

Author Information

From the aNational Institute of Public Health, Cuernavaca, Mexico

bMassachusetts Institute of Technology, Cambridge, Massachusetts, USA

cCenter for Economic Research and Teaching (CIDE), Mexico City, Mexico

dUniversity of California, Berkeley, California, USA.

Correspondence to Stefano M. Bertozzi, Instituto Nacional de Salud Pública, Avenida Universidad 655, Cuernavaca, Morelos 62508, México. Tel: +52 777 329 3069; fax: +52 777 311 1156; e-mail: sbertozzi@correo.insp.mx

Article Outline
Back to Top | Article Outline


Although the past decade has witnessed significant investments worldwide in HIV/AIDS prevention [1–3], there remains a large gap between need and resource availability. UNAIDS estimated that US$27 billion was needed for HIV/AIDS prevention globally in 2005–2007, whereas only US$18 billion was likely be available [4]. Although more resources will undoubtedly be invested in HIV/AIDS prevention in the coming years [3], it remains uncertain how the additional resources can be most efficiently allocated to maximize the number of infections averted. At the global level, Stover et al. [5] estimated that an investment of approximately US$122 billion during 2005–2015 could prevent 28 million infections at a cost of US$3900 per infection averted, but with a concomitant saving of US$4770 in treatment costs per infection. At the regional, national and local levels, however, fundamental questions still need to be answered about the specific mix of interventions to be implemented, the specific populations to be targeted, and the specific strategies for monitoring of interventions to ensure performance.

Here we propose a conceptual framework for addressing these fundamental questions about resource allocation for HIV/AIDS prevention. The efficiency of a specific allocation of resources, we emphasize, can be evaluated along three distinct, critical dimensions: (i) the cost-effectiveness of its mix of interventions; (ii) its targeting efficiency, that is, the extent to which the interventions are directed to the appropriate target population; and (iii) its technical efficiency, that is, the extent to which the interventions are produced at least cost. Previous studies of resource allocation have focused primarily on cost-effectiveness. Targeting efficiency has been implicitly considered, in as much as many interventions are partly defined by their intended target populations. Technical efficiency, we suggest, has been largely ignored, with most analyses implicitly assuming that producers will be equally efficient. Our objective is to focus sharply and simultaneously on all three dimensions. For example, an intervention that is in principle highly cost-effective for intravenous drug users (IDU) may nonetheless be inefficient if it is implemented in a population with only a small proportion of IDU. Likewise, a prevention programme that is in principle highly cost-effective among prostitutes and is implemented in a population with a high proportion of prostitutes may still be inefficient if it is operating at only 10% of capacity.

There are almost no data on the performance of prevention campaigns around the world along each of the three dimensions. Examining recent data from developing economies in Africa, Asia and Latin America, however, we find some evidence to suggest high levels of inefficiency. Remedying each type of inefficiency, we suggest, may require very different types of investments. For example, improvements in cost-effectiveness may require investments in intervention trials at a national or global scale. Improvements in targeting efficiency may require investments in surveillance systems at a national or regional level. Improvements in technical efficiency may require investments in management training and performance monitoring at a regional or local level. Such investments, we further suggest, entail two types of costs– what we term ‘information costs’ and ‘cultural costs’– that have so far not been carefully characterized in the prevention literature.

Back to Top | Article Outline

Previous studies of resource allocation for HIV prevention

Our analysis takes advantage of previous literature addressing the allocation of resources for HIV prevention in developing countries [6–16]. Previous studies have, however, not clearly distinguished the respective roles of and complementarities between each of our three key components: the choice of target populations, the selection of cost-effective interventions within each chosen target population and the technical efficiency of delivery of the intervention in each chosen target population.

For example, the 1993 World Development Report [12] and its companion Disease Control Priorities in Developing Countries [7] recommended interventions for HIV/AIDS and other diseases based primarily on cost-effectiveness criteria. In their chapter on HIV/AIDS in the latter volume, Over and Piot [10] distinguished between cost-effectiveness as a criterion for choosing interventions and the targeting of those interventions among populations with different epidemiological profiles. The problem of reaching target populations was subsequently articulated in the ‘allocation by cost-effectiveness’ (ABC) model that was adopted in the 2000 report of the Institute of Medicine [11]. The 2004 World Bank Development Report [15], addressed the problem of inefficiency in public sector service delivery in low-income settings, and stressed the need to design mechanisms to ensure governmental accountability and monitor local service delivery. In 2005, the World Bank analysed the national HIV/AIDS strategic plans of the 21 African countries participating in the Multi-Country AIDS Programme by 2003 [16]. The report graded each national plan on the extent to which it relied on established interventions, aimed to achieve high coverage among groups at highest risk of infection, prioritized interventions along cost-effectiveness criteria, and considered the role of government regulation.

The 2006 edition of the Disease Control Priorities Project in Developing Countries [8] contained two pertinent chapters: ‘Cost-effectiveness analysis for priority setting’ [9] emphasized that lack of cost-effectiveness data specific to low and middle-income countries resulted in implicit reliance on ‘expert’ judgements, extrapolation from high-income settings and inefficient use of resources. Although the relative vulnerability of different populations was mentioned, the authors did not clearly distinguish between the choice of interventions and the selection of populations to target. The other chapter, ‘HIV/AIDS prevention and treatment’, emphasized the need to collect adequate data on the population distribution of HIV infection and risk behaviours, as well as to generate cost-effectiveness data for individual and community-level interventions [6]. The entire volume, however, was nearly silent on the technical efficiency of interventions.

In parallel with these efforts, UNAIDS has offered two basic recommendations for HIV/AIDS prevention: target resources towards the most vulnerable populations and select interventions appropriate to each country's epidemic profile as classified by UNAIDS (low, concentrated, generalized-low, or generalized-high) [17]. In essence, the more concentrated the epidemic, the more targeted to the relevant populations the interventions should be [18].

Back to Top | Article Outline

Three dimensions of efficiency

Cost-effectiveness: allocation among interventions

Figure 1 presents the intended distribution of funds for eight sub-Saharan African countries over the initial years of this decade [16]. Countries are arranged from low to high prevalence and range from concentrated to generalized high-level epidemics. The epidemics in all eight countries are characterized by heterosexual transmission.

Figure 1 reveals that the distribution of resources among interventions bears no obvious relationship to a country's epidemic profile. Even within countries with similar epidemic profiles, there is no consistent pattern of allocation of resources. In the subgroup with generalized low-level epidemics (Ghana, Burkina Faso and Cameroon), for example, Burkina budgeted 40% for condom promotion and distribution, whereas Cameroon distributed less than 10%. At the same time, Cameroon budgeted nearly 40% for information, education and communication (IEC), whereas Burkina Faso budgeted approximately 10%. In the subgroup of countries with generalized high-level epidemics (Uganda, Mozambique, Zambia), Zambia allocated 40% to voluntary counselling and testing (VCT) and 5% of its budget to IEC, whereas Mozambique allocated less than 5% to VCT and nearly 50% to IEC.

In principle, many factors may have influenced the allocation of resources for HIV/AIDS prevention within these African countries, including fertility rates, the extent of urbanization, and the prevalence of sexually transmitted infections. Nonetheless, the data in Fig. 1 strongly suggest that the prevalence of HIV infection bears little relation to a country's allocation of resources among different interventions. If the African countries in Fig. 1 had been employing common selection criteria, one would expect those countries with similar epidemic profiles to have a similar mix of preventive responses.

Why do the resource allocation decisions depicted in Fig. 1 appear not to be evidence based? Among the possible reasons are: lack of convincing, sufficiently rigorous or sufficiently country-specific evidence; lack of translation or communication of evidence to policy makers; lack of skills for converting evidence into prioritized strategic plans; domination of political imperatives or cultural acceptability over effectiveness or cost-effectiveness evidence; and lack of monitoring data at the national or international level.

Back to Top | Article Outline
Targeting: allocation among subpopulations

Figure 2 presents the results of a 2003 coverage survey of HIV/AIDS services in 73 low and middle-income countries, representing 88% of individuals living with HIV/AIDS worldwide [18,19]. The figure shows the coverage of basic preventive services for each key population, by global region. Coverage was defined as the weighted average percentage of individuals in each subpopulation who utilized one or more of the relevant preventive services. Coverage for IDU thus represented the proportion of IDU who received risk-reduction information, education and counselling, needle and syringe exchange services, or drug substitution, weighted by the estimated number of IDU in each country within the region. Similarly, coverage for sex workers represented the weighted proportion in each region covered by outreach prevention programmes, whereas coverage for students in primary schools represented the weighted proportion in each region who received AIDS education.

The results in Fig. 2 generally show the highest coverage for prisoners and students in all regions. By contrast, much lower rates of coverage are observed for the populations driving the epidemic in each region. Coverage for IDU was less than 10% in the eastern Mediterranean and eastern Europe regions. Less than 20% of sex workers received basic prevention services in south east Asia. In the Americas, the coverage for men who have sex with men was approximately 30%. In the allocation of prevention resources across subpopulations, the reach of preventive efforts apparently does not correlate with the importance of the subpopulation in transmitting the virus.

The costs of reaching subpopulations may be critical determinants of coverage. Prisoners may have high coverage, for example, because they represent a captive population. Students may have high coverage because AIDS prevention education is less costly and students are also in some sense captive. We cannot exclude the possibility, however, that the mismatch between programme coverage and risk of infection is a consequence of inadequate epidemiological information and surveillance data. In the absence of transparent data on target populations, political, ideological, cultural and religious issues may dominate resource allocation decisions.

Back to Top | Article Outline
Technical efficiency: allocation among inputs

Technical efficiency refers to the ability of a facility to produce a given quantity of preventive services (condoms distributed, VCT sessions, etc.) at minimum cost. Technical efficiency implies not only the optimal mix of inputs (labour, office space, condoms, informational materials, advertising, etc.), but also their optimal use (with minimum wastage and corruption). Technical efficiency is usually not explicitly considered in cost-effectiveness analysis. Estimates of cost per unit service and cost-effectiveness are typically derived from experimental evidence or from highly monitored facilities. The assumption that such ideal conditions apply equally to real-world facilities will probably result in overly optimistic estimates of cost-effectiveness.

Figure 3 shows the average annual unit costs of VCT versus the annual number of clients served among 82 programmes in five developing countries, based upon data collected by the Prevent AIDS Network for Cost-Effectiveness Analysis (PANCEA) during 2003–2004 [20]. Both axes are on a logarithmic scale. The overall negative relationship between unit cost and volume of service is consistent with the presence of economies of scale, at least in the range of output depicted [20]. Even more significant is the substantial variability in unit cost among facilities at similar volumes both across and within countries, however, with cost per unit VCT ranging from less than US$10 to close to US$1000. It is implausible that the observed variation of two orders of magnitude is caused entirely by differences in wages or other input prices, or differences in the quality of the output– especially with a service as standardized as VCT. Nor is inadequate demand arising from high transportation costs in remote areas likely to be the critical source of variation. The data points corresponding to Mexico and Russia, for example, are derived entirely from urban VCT centres. More likely, Fig. 3 reflects large variations in technical efficiency.

Back to Top | Article Outline

Measuring overall programme efficiency

Combining the three dimensions of efficiency

Consider a cost-effectiveness trial of a specific preventive intervention in a defined target population. Under ideal conditions, let us say that the minimum unit cost of the intervention is US$50 per individual in the target population. Moreover, under these ideal conditions, one case of HIV infection is prevented for every 100 individuals in the treated target population, that is 0.01 infections per individual in the target population. Then the theoretical cost-effectiveness ratio is US$50/0.01 = US$5000 per case of HIV averted or equivalently 0.0002 infections averted per US$ invested. More generally, the foregoing illustrative cost-effectiveness calculation can be written in the form of a verbal equation as:

The idealized conditions embodied in this equation entail three assumptions: the prevention intervention is the best possible choice for the specific target population; the target population is covered at the optimal level, given the available resources; and the intervention is implemented using the optimal cost-minimizing combination of inputs. In real-world applications, however, these assumptions generally do not hold. Let us posit, for example, that practitioners must treat 10 individuals to reach one member of the target population, and that the actual unit costs per treatment are twice the unit cost observed under ideal conditions. Then the actual cost-effectiveness ratio is equal to US$5000 × 10 × 2 = US$100 000 per case of HIV averted. This computation captures the three different dimensions of inefficiency described in the examples of the previous section: cost-effectiveness; targeting inefficiency; and technical inefficiency. We refer to inefficiency because the larger is any one of the three components, the larger will be the cost per HIV infection averted in actual practice.

Expressing the overall efficiency of the intervention in terms of infections averted per dollar invested, we obtain the following verbal equation:

The first term on the right-hand side (‘Inverse of theoretical cost-effectiveness ratio’) corresponds to the efficiency of the intervention under ideal conditions. In our illustrative numerical example, this quantity would equal one HIV infection averted per US$5000 invested. The last two terms on the right-hand side, which correspond respectively to the targeting efficiency and the technical efficiency of the intervention, can be expressed in percentage terms. In our example, practitioners had to treat 10 individuals to reach one member of the target population, and therefore the targeting efficiency was 10%. Moreover, the actual unit costs per treatment were twice the unit cost observed under ideal conditions, and therefore the technical efficiency was 50%. The overall efficiency was therefore equal to: 0.0002 infections averted per US$ invested, multiplied by a targeting efficiency of 10%, multiplied by a technical efficiency of 50%, or equivalently, 0.00001 infections averted per US$ invested. This result illustrates the potentially high costs of targeting and technical inefficiencies. Compared with the theoretical result of 0.0002 infections averted per US$ invested, the intervention in this example prevents 20-fold fewer infections (0.0002/0.00001) than the theoretical result.

We present a more detailed graphical description of our basic three-component theoretical model in Appendix 1.

Back to Top | Article Outline
Investing in improved efficiency: informational and cultural costs

Improvements in any one of the three dimensions of efficiency require the investment of resources. The costs of these investments are not necessarily captured in the efficiency formula we have proposed. Such costs are sometimes hidden and not explicitly considered in the policy calculus, yet they may be quite substantial and undoubtedly real. For example, the adoption of an intervention that is in principle highly cost-effective may entail overcoming cultural, religious or ideological barriers. There are many possible ways to overcome such barriers, from education of opinion leaders, to marketing and information campaigns, to the implementation of successful pilot projects that show existing beliefs to be unfounded. We label the costs involved in surmounting such barriers as ‘cultural costs’.

Similarly, reaching the target population in an efficient manner may require improvements in communication as well as surveillance studies to locate populations at high risk of infection. We label the expenditures involved as ‘information costs’. Such information costs can be global, local, or even facility specific. Efficacy trials are often carried out at a global level, entailing a prospective design that may require years to see results. Information on the effectiveness of an intervention assessed in one country can be employed to guide policy decisions and further intervention trials in other countries. Such information is a global public good that will typically require international cooperation to ensure adequate funding. On the other hand, surveys of risk behaviours to improve targeting efficiency are typically carried out at the local level, requiring months of planning and execution. Epidemic and behavioural profiles differ among settings. The location of brothels in Mumbai is of little use in Dar es Salaam. Finally, the information costs for improving technical efficiency are facility specific. If a programme manager knows that the Red Cross clinic on Mandela Street costs $23 per client served, it may be of little use in predicting why the cost in the Medecins Sans Frontieres clinic on Machel Avenue is 50% higher. Assessment of the technical efficiency of individual facilities, however, can sometimes be carried out in days.

The measurement of cultural and information costs is undoubtedly difficult, but it is essential that we understand how these hidden but real costs impede the attainment of efficient preventive strategies. In Fig. 4, we offer some speculative yet plausible possibilities. The figure shows hypothetical total cost curves for improvements in only one dimension, namely, targeting efficiency. As we move to the right on the horizontal axis, the target population represents a higher proportion of the total population reached by the intervention. In the figure, we posit that information costs display a pattern of rising marginal cost. For example, ascertaining the locations of the most popular gay bars and brothels in a city may entail little or no cost. However, further improvements in targeting men who have sex with men and sex workers will entail costly epidemiological or behavioural surveys. By contrast, we posit that cultural costs display a pattern of diminishing marginal cost. In a country with strong cultural and religious barriers, there may be significant political resistance to targeting interventions to men who have sex with men or to IDU. Changing the situation may entail years of educational campaigns and pilot studies, that is, significant investments before any result can be observed. As it becomes socially acceptable to intervene in such populations, however, the political and cultural barriers decline. In the figure, we have depicted cultural costs as relatively large compared with information costs at low levels of targeting efficiency. There is so little mapping of risk populations and behaviours in most countries, despite the relatively low information cost, that cultural costs are likely to be binding.

The same distinction between rising and falling marginal costs can apply to the other two dimensions: cost-effectiveness and technical efficiency. In the cost-effectiveness dimension, informational marginal costs may similarly rise because more sophisticated studies are needed to evaluate the effectiveness of additional interventions in new settings, whereas cultural marginal costs fall for reasons similar to those described in the targeting dimension in Fig. 4. In the technical efficiency dimension, by contrast, informational marginal costs may decrease because of the high initial setup costs and lower maintenance costs of performance monitoring systems. With respect to cultural costs, increased transparency and reduced corruption may also entail high initial setup costs, so that marginal costs may also decrease.

Back to Top | Article Outline


To assess the relative efficiency of alternative interventions to prevent HIV/AIDS, we have proposed a conceptual three-dimensional model that distinguishes between cost-effectiveness, targeting efficiency and technical efficiency. We have also stressed the importance of hidden but significant informational and cultural costs in moving from inefficient to efficient points within our proposed three-dimensional efficiency calculus. Our hope is that such a framework will prove useful in the allocation of existing resources for HIV prevention in the short run and the design of new prevention strategies in the long run.

Since the 1993 World Development Report, there have been repeated calls for additional information on the relative costs and benefits of the range of available prevention interventions in different settings. This point was strongly underscored in the 2006 Disease Control Priorities Project [10]. Acquiring such information is costly and time-consuming. It will require complex, long-term studies of cost-effectiveness. Sufficient information already exists in many countries, however, to improve the reach of available interventions to target populations. As suggested by our three-dimensional framework, improvements in targeting efficiency may have payoffs even greater than the potential gains from greater knowledge of cost-effectiveness. To the extent that there are political, religious and cultural impediments to improved target, the challenge is not to acquire more information about key populations, but to learn how to surmount what we have called ‘cultural costs’. What is more, donors and national governments alike have largely ignored the technical efficiency of their implementation partners. Recent evidence, including the data in Fig. 3, suggests enormous variation in unit cost per service delivered. Our three-dimensional framework suggests that improvements in management training and monitoring at the local level could in principle have even greater payoffs.

The main limitation of this study is the absence of adequate information to make our model fully operational at the quantitative level. At the same time, our purpose is to highlight where the gaps in information lie. Given that decisions have to be made in the absence of complete information, we propose that it is better to make decisions within the framework outlined here.

Another limitation is that our theoretical discussion of cost-effectiveness refers only to a single prevention intervention, whereas in reality, prevention campaigns always consist of a mix of multiple interventions. We presented our theoretical model for a single intervention primarily for clarity of exposition. We think that the basic principles of our model can be readily generalized to the analysis of multiple interventions.

What is more, our model treats targeting efficiency as a simple proportionality factor. That is, our cost-effectiveness equation includes the simple multiplicative term ‘proportion of treated individuals in the target population’, with a value that can range from 0 to 100%. We know from the vaccination literature, however, that 100% coverage is not necessary in many cases. Reaching the last 5% of the target population can be extremely expensive yet confer relatively low benefits in terms of infections averted.

Finally, our presentation does not squarely confront the problem of minimum scale. We have implicitly assumed that a region or a country seeks to maximize the number of HIV infections prevented within a fixed prevention budget of sufficient size. We do not discuss the possibility that a country's prevention budget may fall below the level at which investments in greater cost-effectiveness are infeasible. Determining the minimal necessary size of a specific country's prevention budget is not a simple problem and requires further discussion.

Back to Top | Article Outline


The authors acknowledge helpful comments received at the following meetings: International AIDS Economics Network: Pre-Conference Meeting, Toronto, 2006; Seminario Internacional de Prioridades en Salud, The World Bank, 2007; Impact Evaluation Workshops, The World Bank, 2006–2007; XII Congreso de Investigación en Salud Pública, Cuernavaca, Mexico, 2007. The authors also wish to thank reviewers from Harvard University for their helpful comments on a previous version of the manuscript, as well as the insights of the journal referees and editorial board.

Conflicts of interest: None.

Back to Top | Article Outline


1. President's Emergency Plan for AIDS Relief (PEPFAR). Global health: spending requirements presents challenges for allocating prevention funding under the President's Emergency Plan for AIDS Relief. United States Government Accountability Office. Second annual report to congress. Washington, DC: PEPFAR; 2005.
2. The Global Fund. Annual report 2004. Geneva: The Global Fund to Fight AIDS, Tuberculosis and Malaria; 2005.
3. World Bank. The World Bank's global HIV/AIDS program of action. Washington, DC: World Bank; 2005.
4. UNAIDS. Resource needs for an expanded response to AIDS in low- and middle-income countries. Geneva: UNAIDS, 2005.
5. Stover J, Bertozzi S, Gutierrez JP, Walker N, Stanecki KA, Greener R, et al. The global impact of scaling up HIV/AIDS prevention programs in low- and middle-income countries. Science 2006; 311:1474–1476.
6. Bertozzi SM, Padian N, Wegbreit J, DeMaria LM, Feldman B, Gayle H, et al. HIV/AIDS prevention and treatment. In: Jamison DT, Breman JG, Measham AR, Alleyne G, Classon M, Evans DB, editors. Disease control priorities in developing countries. 2nd ed. New York: Oxford University Press; 2006. pp. 331–369.
7. Jamison DT, Mosley WH, Measham AR, Bobadilla JC, editors. Disease control priorities in developing countries, 1st ed. New York: Oxford University Press; 1993.

Fig. 4. Investment in improved efficiency: information and cultural cost curves for improved targeting efficiency.

8. Jamison DT, Breman JG, Measham AR, Alleyne G, Classon M, Evans DB, editors. Disease control priorities in developing countries, 2nd ed. New York: Oxford University Press; 2006.
9. Musgrove P, Fox-Rushby J. Cost-effectiveness analysis for priority setting. In: Jamison DT, Breman JG, Measham AR, Alleyne G, Classon M, Evans DB, editors. Disease control priorities in developing countries. 2nd ed New York: Oxford University Press; 2006. pp. 271–285.
10. Over M, Piot P. HIV infection and sexually transmitted diseases. In: Jamison DT, Mosley WH, Measham AR, Bobadilla JC, editors. Disease control priorities in developing countries. 1st ed. New York: Oxford University Press; 1993. pp. 455–527.
11. Ruiz MS, Gable AR, Kaplan EH, Stoto MA, Fineberg HV, Trussell J, editors. No Time to lose: getting more from HIV prevention. Washington DC: Committee on HIV Prevention Strategies in the United States, Institute of Medicine, National Academy Press; 2000.
12. World Bank. World development report 1993: investing in health. Washington DC: World Bank; 1993.
13. World Bank. Optimizing the allocation of resources for HIV prevention: the allocation by cost-effectiveness (ABC) model. Guidelines. Washington DC: World Bank; 2002.
14. World Bank. Optimizing the allocation of resources among HIV prevention interventions in Honduras. Washington DC: World Bank; 2002.
15. World Bank. World development report 2004: making services work for poor people. Washington DC: World Bank; 2004.
16. World Bank. Background paper for the OED evaluation of the World Bank's assistance for HIV/AIDS control. Review of national HIV/AIDS strategies for countries participating in the World Bank's Africa Multi-Country AIDS Program (MAP). Washington DC: World Bank; 2005.
17. UNAIDS, WHO. Guidelines for second generation HIV surveillance. Geneva: World Health Organization; 2000.
18. UNAIDS. Intensifying HIV prevention: a UNAIDS policy position paper. Geneva: UNAIDS; 2005.
19. USAID, UNAIDS, WHO, UNICEF, the POLICY Project. Coverage of selected services for HIV/AIDS prevention, care and support in low and middle income countries in 2003. Washington DC: USAID; 2004.
20. Marseille E, Dandona L, Marshall N, Gaist P, Bautista-Arredondo S, Rollins B, et al. HIV prevention costs and program scale: data from the PANCEA project in five low and middle-income countries. BMC Health Serv Res 2007; 7:108. Available at: http://www.biomedcentral.com/1472-6963/7/108. Accessed: May 2008.
Back to Top | Article Outline
Appendix 1

Here, we convert all of the components of our efficiency calculus into percentage terms. First, let us assume that there is a benchmark cost-effectiveness against which all potential interventions are to be gauged. For purposes of exposition, we assume a benchmark of US$1000 per HIV infection averted, which is considerably lower than the currently attainable value of US$3900 estimated by Stover et al. [5]. In our illustrative numerical example, the intervention had a cost-effectiveness of US$5000 per HIV infection averted. Therefore, in percentage terms, its intervention efficiency was US$1000/US$5000 = 20%. The overall efficiency of the intervention was therefore 20% × 10% × 50% = 1%. Abbreviating ‘cost-effectiveness ratio’ to CER, we get the following verbal equation:

Overall efficiency of prevention intervention = Benchmark CER as a proportion of intervention's theoretical CER × Proportion of treated individuals in the target population × Minimum unit cost as a proportion of actual unit cost

Each term in the foregoing equation is expressed in percentage terms. The overall efficiency is equal to the product of the intervention efficiency, the targeting efficiency, and the technical efficiency.

Figure 5 presents a graphical representation of this conceptual framework. Each axis measures one of the three components on a scale from 0 to 100%. The origin 0 represents zero efficiency in each dimension. The point A corresponds to 100% overall efficiency. The point B corresponds to an intervention whose cost-effectiveness ratio equals the benchmark CER, whose treated population consists entirely of individuals from the target group. The technical efficiency is nearly zero, that is, the unit cost of service far exceeds the efficient minimum unit cost. Our illustrative numerical example would be located at the point C. Whether or not the three dimensions are explicitly considered in the decision-making process, any allocation of resources for HIV prevention necessarily corresponds to some point in the three-dimensional space.

The data presented in Fig. 1, Fig. 2 and Fig. 3 suggest that many countries are not operating near the overall efficient point A in our three-dimensional model. If the data presented were representative of the global state of resource allocation for HIV prevention, then point D might not be far from reality in 2003. At that point, 50% of funds were allocated to cost-effective interventions, whereas 20% of key target populations were covered by preventive services and 60% of technical feasible outputs were being produced with the funds expended. This is of course completely speculative, but illustrates the need to consider all three aspects of efficiency when allocating and using prevention resources.


AIDS; budget; cost-effectiveness; efficiency; HIV; targeting

© 2008 Lippincott Williams & Wilkins, Inc.