How Can Policy Makers Use Cost-Effectiveness Analysis?
To provide good policy guidance, cost-effectiveness must be complemented with essential information about the larger context, in particular, the prevailing burden of diseases, the existing coverage of health interventions, and the overall capacity of the health system.
An essential contextual factor in using information on the cost-effectiveness of any intervention is the burden caused by a disease. Some interventions may be highly cost-effective but affect only a small number of people or provide a small improvement in health (figure 3.1). For example, leishmaniasis treatment is relatively cost-effective, but is only applicable to a relatively small number of cases. By contrast, antimalarials and insecticide-treated bednets are cost-effective measures that, in certain countries, would avert a large burden of disease. If possible, countries would finance all measures that would improve health, but as every country faces a tight budget or constrained capacity to deliver services, the avertable burden of disease is an essential piece of information that policy makers require when choosing between otherwise similarly cost-effective interventions.
[Figure
3.1]
". . . the avertable burden of disease is an essential piece of information that policy makers require when choosing between otherwise similarly cost-effective interventions."
Health interventions that are preventive will generally be more cost-effective in places where the burden of the targeted disease or risk factor is high and, consequently, where the intervention will avert more cases. Yet current prevalence is not always a good indicator of whether an intervention will be cost-effective, particularly in places where effective public health programs are responsible for the low rate of prevalence. For example, the prevalence of diphtheria, tetanus, pertussis, and measles is generally low in countries with effective vaccination programs, yet the cost-effectiveness of the vaccination program, which is needed to avoid a resurgence of these illnesses, is still quite high.
Prevalence also has a large effect on the cost-effectiveness of screening for illnesses and, indirectly, on the cost-effectiveness of a package to address a certain ailment. For example, screening and treatment for helicobacter, a bacterial risk factor for stomach cancer, is not cost-effective in the United States, but is cost-effective in Colombia, because the prevalence of stomach cancer is higher in Colombia and many of the costs of treatment are lower (DCP2, chapter 29).
The cost-effectiveness of screening for cancers and many other illnesses depends on the costs of identifying cases, on how many people do not follow up with treatment, and on the direct costs of treatment. Of course, if no treatment is available, screening is pointless. Testing for anemia among people with AIDS is cost-effective among those treated with zidovudine not only because screening is relatively cheap (less than US$0.02 per anemia test) but also because anemia occurs in 10 percent of these patients. When costs are higher or the likelihood of encountering conditions is small, screening may not be cost-effective.
" . . . universal blood screening for HIV is costly, yet it is also cost-effective, even in countries with a low prevalence of HIV/AIDS . . ."
Cost-effectiveness is also sensitive to the probabilities of transmission. For example, universal blood screening for HIV is costly, yet it is also cost-effective, even in countries with a low prevalence of HIV/AIDS, because receiving contaminated blood has such a high probability of leading to infection—almost 100 percent.
An appropriate time horizon is also imperative in assessing the weight of a disease burden and the value of an intervention for several reasons. One is that the gains from the intervention may accrue only in the long term, so the intervention appears to be effective with a long horizon but not a short one. The discount rate matters greatly to this comparison because it makes the distant future less valuable. Another reason is that the intervention may have to be repeated for several years to assure the potential health gains. This is the case for ORT, which may need to be given many times over several years to prevent diarrheal disease deaths among young children, and for penicillin prophylaxis, to prevent deaths from infection in children with sickle cell disease (DCP2, chapters 19 and 34). Finally, an intervention may have substantial start-up costs that must be amortized over some period. DCP2 uses 10 years as the standard in such cases.
The coverage of existing interventions is another crucial contextual factor in making use of cost-effectiveness analysis. When policy makers decide how to allocate resources, they can compare interventions that are relatively more or less cost-effective in light of the current supply of services. For example, some interventions may be extremely cost-effective but have low coverage. These are neglected opportunities that policy makers should look at more closely. Barring other contravening factors, these are likely to be interventions that would have a large effect on health for relatively little cost.
DCP2 mostly reports cost-effectiveness ratios as if they were independent of the level and scale of interventions, yet the incremental cost-effectiveness of most interventions will also vary with the level of service coverage. The cost of reaching the first 1 percent of a population may be quite high when the fixed costs of purchasing equipment, training staff, and setting up management systems are taken into consideration and may yield relatively few health gains. As coverage increases, however, the average cost may fall and health improvements may increase, resulting in a substantial improvement in the cost-effectiveness of reaching an additional group, for example, extending from 50 percent coverage to 51 percent coverage. Once coverage is high, reaching the remaining, and often marginalized, segments of the population may again be quite costly without a correspondingly large health gain, and consequently cost-effectiveness will worsen. Consider the experience of eradicating smallpox. At a certain point in the campaign, large parts of the world were free of smallpox and eradication became contingent on identifying the last few redoubts of the virus and responding massively and quickly to quarantine those infected and vaccinate everyone else in those areas. Today the polio campaign faces a similar challenge: reaching and vaccinating a few children in rural parts of India and Sudan is much costlier than treating many more in urban areas, but elimination of the disease can justify those high costs. A similar process is at play with the provision of basic health care in that it is generally less costly per person in areas with dense rather than sparse populations.
In addition to disease prevalence and existing coverage, policy makers need to take other local factors into consideration. DCP2 provides estimates based on regional averages of unit prices,1 but local prices and the availability of inputs may vary substantially from regional averages. Therefore a first consideration is whether a particular country's prices are near to or diverge sharply from the regional average. A second consideration is whether prices of key inputs have changed since the original analysis. One of the most dramatic changes since the earlier edition of Disease Control Priorities in Developing Countries (Jamison and others 1993) has been the fall in prices of antiretroviral drugs. Consequently, antiretroviral therapy is substantially more cost-effective today than it was a decade ago. Further reductions in the costs of diagnostic testing and alternative forms of delivery may increase the cost-effectiveness of antiretroviral therapy even further in the near future.
Finally, the cost-effectiveness of most health interventions also depends on how well the health system functions (DCP2, chapter 3). Most DCP2 chapter authors calculate cost-effectiveness ratios based on the assumption that a functioning health system is available to deliver the intervention; however, this is an assumption whose validity varies greatly across countries. If a country has a particularly weak health system, then interventions that rely heavily on medical professionals, complex treatments, or sophisticated information systems will not be as cost-effective in practice as they would be in countries with stronger health systems.
The experience of introducing IMCI (DCP2, chapter 63) demonstrates the extent to which health system functioning can influence the cost-effectiveness of health interventions. Experiences in several districts in Brazil and Tanzania show that the IMCI package of interventions not only improves children's health outcomes but can actually be cost saving by reducing improper care and excessive use of medications. However, in most low- and middle-income countries the IMCI package has encountered difficulties in implementation and failed to realize its promise of cost-effectiveness because of high rotation and attrition of trained staff, inadequate supplies, and insufficient funds.
