The world health report

Chapter 5

Technical considerations for cost-effectiveness analysis

The estimates, which provide the basis of the results reported in this chapter, were undertaken on a regional basis as part of the WHO CHOICE project. The six WHO regions were divided into mortality strata as described in earlier chapters, resulting in 14 epidemiological subregions. The total costs and total effects of each intervention were estimated separately for the 14 subregions. Eventually it is hoped that there will be sufficient data to make estimates at a country level, and even at the subnational level for large countries, but this is not currently possible.

Subregional analysis offers a valuable basis from which country analysts can work to calibrate the results to their settings. It is much more policy-relevant than a global analysis because the epidemiology, cost structures, and starting points (such as the availability of trained health staff and the history of health interventions) varies less within each subregion than across the world as a whole. The results are used here to identify interventions that are very cost-effective, cost-effective, and not cost-effective in each subregion.

Costs are reported in terms of international dollars rather than in US dollars, to account better for differences in cost structures between settings. Unit costs for most regions are higher using international dollars (I$) based on purchasing power parity exchange rates than they would be if official exchange rates had been used.5 Effectiveness is measured in terms of disability-adjusted life years (DALYs) gained by the intervention. A brief description of the methods is found in Box 5.1, while full details of the methods and the calculations can be found on the WHO web site.6

Box 5.1 Methods for cost-effectiveness analysis

The cost-effectiveness analysis on which this report is based considered what would have happened if a set of interventions had not been implemented and compared the result with what happens on their implementation. Through a four-state population model, the number of healthy life years lived over a period of a hundred years by a population in the absence of that set of interventions is estimated by inputting parameters of incidence, remission, cause-specific and background mortality, and health state valuations reflecting the natural history of the disease. The parameters reflecting the natural history of the disease were mostly estimated by back-adjusting current rates using coverage and known effectiveness of interventions. The same four-state population model can then be rerun, reflecting changes in the parameters due to interventions or combinations of interventions. For example, based on data from earlier chapters, vitamin A deficiency increases the risk of dying from diarrhoea. The impact of vitamin A supplementation is then mediated in the model by a decrease in case fatality rate for diarrhoea. Effectiveness data came from systematic reviews where available. The difference in the healthy life years gained by the population with and without the intervention is the impact of the intervention and is entered as the denominator of the cost-effectiveness ratio.

Costs covered in this analysis include expenses associated with running the intervention, such as administration, training and contact with the media. They also include costs incurred at the individual level such as counselling. Considerable effort was exerted to try to standardize the methodology used in collecting and classifying costs. The quantities of inputs required to run each intervention were estimated by experts in 17 regions of the world and validated against the literature. Some individual-level costs were obtained by multiplying unit costs of inputs by the expected utilization of those inputs by the people covered by the programme. Unit costs for outpatient visits and laboratory tests were obtained from a review of literature and supplemented by primary data from several countries. The total costs for implementing a programme for 10 years constitutes the denominator of the cost-effectiveness ratio.

Stochastic uncertainty analysis was carried out for key parameters in both the numerator and denominator.

Sources: (3,17,19)

It is not much value to provide decision-makers with information on the costs and effectiveness of interventions that are undertaken badly. Accordingly, the results reported here show what would be achieved if the interventions were undertaken in a relatively efficient manner. For example, we assume capacity utilization of 80% in most settings -- for example, staff and capital equipment are fully occupied for 80% of the normal working day -- except when estimating the effect of expanding coverage to very high levels. To reach 95% of the population it might be necessary to provide facilities in isolated areas where population numbers are insufficient to support such high rates of capacity utilization. The results, therefore, provide guidance on selected interventions that should be given high priority in the policy debate about resource allocation, but only if they are undertaken in an efficient manner.

Sets of interventions that interact in terms of effectiveness or costs are considered together, as stated earlier. For example, interventions to reduce risks associated with hypertension and high cholesterol interact. The analysis is based on estimates of the effects on population health of reducing blood pressure alone, reducing cholesterol levels alone, and doing both together.

In addition, many of the interventions are evaluated at different levels of coverage. For most, three levels were used (50%, 80% and 95%) and the impact on costs and effects of expanding coverage was incorporated.

The standard practice in this type of analysis is to discount both the health effects and the costs of the different programmes under consideration. There is no controversy about the appropriate discount rate to use for costs: the opportunity cost of capital. The discount rate for benefits is often thought to comprise two parts. One is a "pure" time preference for immediate over postponed consumption. The second relates to the fact that, as the prosperity of a society increases, the utility or benefit to it of a defined unit of consumption is less -- that is, there is declining marginal utility of a unit of consumption as income rises. Many cost-effectiveness studies have assumed that this applies to health benefits as well and have discounted future health at a rate between 3% and 5% per year. This practice has long been debated, and some people have argued that the discount rate for health benefits should be close to zero and certainly less than the discount rate for costs (20,22).

This question is important for the analysis in the following section as it can change the relative priority of interventions. Not all health care programmes achieve results at the same rate. Public health and health promotion programmes in particular may take many years to produce tangible results, and applying a discount rate to the benefits of such programmes will reduce their apparent attractiveness compared with programmes that produce rapid benefits of a similar magnitude.

Common practice remains to discount costs and benefits at the same rate, so we follow the same practice in our baseline calculations using a rate of 3%. To be consistent with the approach used in Chapter 4 for measuring the burden of disease, age weights are also included in the baseline calculations.

The recent report of the Commission on Macroeconomics and Health suggested that interventions costing less than three times GDP per capita for each DALY averted represent good value for money and that, if a country could not afford to undertake them all from its own resources, the international community should find ways of supporting them (23). This report's classification of interventions is based on this principle, and defines very cost-effective interventions as those which avert each additional DALY at a cost less than GDP per capita, and cost-effective interventions as those where each DALY averted costs between one and three times GDP per capita.

Finally, cost-effectiveness analyses can be found in the published literature for some of the interventions discussed in this chapter, which does not, however, simply report the published results. The methods used for estimating costs and effectiveness varies considerably across the published studies and their results cannot be compared. Moreover, most provide insufficient information on how they estimated costs to be sure that all possible costs were included and valued appropriately. This report, therefore, re-estimated costs and effects using a standard approach for all interventions, although each study that could be found was evaluated to determine if the parameters it used could be incorporated.