1. Investing in Health

Annex 1.A: The Burden of Disease in 2001

This annex provides estimates of the burden of different diseases and injuries in 2001. Alan Lopez, Colin Mathers, Christopher Murray, and their colleagues at WHO generated the estimates, aggregated them by World Bank regions, and provided final updates. A companion volume (Lopez and others 2006) will provide more comprehensive tables of results for a much finer disaggregation of conditions, a full exposition of methods and data sources, and sensitivity analyses (including assessments of the sensitivity of results to including stillbirth). All numbers in this annex are consistent with those in the companion volume.

This annex first provides a brief background on assessments of deaths by cause and disease burden and then an overview of the uses of such measures for health policy. It concludes with aggregated tables on deaths and on disability-adjusted life years (DALYs) by selected causes or groups of causes. The tables present estimates both with and without stillbirths, which constituted approximately 5.5 percent of deaths globally in 2001. Estimates are provided separately for high-income countries and for the low- and middle-income countries as a group.

Background

Many countries, including all high-income ones, maintain vital registration systems that provide data (usually annual) on the number of deaths by cause, age, sex, and sometimes race. Some countries additionally compute years of life lost (or YLL) by cause, which assigns a number of years of life lost attributable to each cause that depends on the age of death and some relevant measure of life expectancy. As of the early 1990s, no similar estimates existed for many developing countries or for regional groupings of them. Experts on individual conditions or the relevant disease program at WHO generated estimates for the diseases of interest to them. When added up across diseases, however, such estimates exceeded, often by a factor of 2 or more, any plausible estimate of the total number of deaths occurring in each age group. DCP1 and the World Development Report 1993 (Lopez 1993 and World Bank 1993) generated estimates of the number of deaths by cause that were consistent with demographically determined death totals for eight regional groupings of countries. WHO collaborated closely on this work. The number of deaths from a disease is one measure of the magnitude of its burden, and YLL constitutes for many purposes a better measure. Neither takes account of the disability or suffering associated with a nonfatal disease.

The 1993 World Development Report also developed a variant of the quality-adjusted life year (QALY) from the health economics literature to add a disability dimension to YLLs in order to generate a more comprehensive measure of burden. The result, called a disability-adjusted life year, measures burden from a specific cause as the sum of years of life lost from that cause and the equivalent years of life lost (in a sense that is made quite specific) from the disability caused by the condition. Original publications on disease burden included estimates that discounted future events at 0 percent or at 3 percent per year. They also included estimates that weighted the value of a year of life uniformly across all age groups and estimates that placed greater value on middle-aged groups. These are labeled DALYs (r, k) with the first number indicating the discount rate in percent per year and the second indicating whether uniform or nonuniform age weights were used. The most widely reported variant on the DALY is the DALY (3,1)—that is, one that uses a 3 percent discount rate and nonuniform age weighting. This chapter and the companion volume report DALYs (3,0)—that is, with discounting but uniform age weighting.

Estimates of DALYs by cause for 1990 first appeared as appendix B of the 1993 World Development Report and, in expanded form, in Murray, Lopez, and Jamison (1994). Christopher Murray, Alan Lopez, and colleagues later produced updated estimates for 1990 and a fuller account of the methods used (Murray and Lopez 1996a, 1996b). The relative burden of different conditions as measured by numbers of deaths correlates highly with DALYs, but important exceptions exist. The massive burden of major psychiatric conditions, for example, is captured by DALYs but missed in estimates of deaths by cause or YLLs. Table 1.A1 summarizes the current estimates of deaths by cause in 2001 from Mathers, Murray, and Lopez (2006). Table 1.A2 shows disease burden in DALYs. Additional columns in these tables show the effect of including stillbirths on the percentage distribution of burden across conditions while leaving unchanged the other numbers (see Jamison and others 2006).

Disease burden can be assessed by risk factor as well as by disease or condition. An initial assessment of risk factor burden appeared in the 1993 World Development Report and later in Murray and Lopez (1997) and WHO (1996). WHO published a much fuller set of estimates in its 2003 World Health Report (WHO 2003b). Ezzati and others (2006) provided a substantial update adjusted to the same methodological assumptions as for deaths and DALYs, including use of DALYs (3,0).

Uses of Disease Burden Measures

DALYs are useful for informing health policy in at least six ways. Estimates of deaths by cause or YLL serve these same purposes, but for some uses less well.

1. Assessing performance. A country-specific (or regional) assessment of the burden of disease provides an outcome indicator that can be used over time to judge progress or across countries or regions to judge relative performance. The most natural comparison is to the development of national income and product accounts (NIPAs) by Simon Kuznets and others in the 1930s, which culminated in 1939 with a complete NIPA for the United Kingdom prepared by James Meade and Richard Stone at the request of the U.K. Treasury. NIPAs have, in the subsequent decades, transformed the empirical underpinnings of economic policy analysis. One leading scholar has put it this way:

   The national income and product accounts for the United States (NIPAs), and kindred accounts in other nations, have been among the major contributions to economic knowledge over the past half century. . . . Several generations of economists and practitioners have now been able to tie theoretical constructs of income, output, investment, consumption, and savings to the actual numbers of these remarkable accounts with all their fine detail and soundly meshed interrelations. (Eisner 1989, 1)

   Disease burden measures have the potential of serving a similar purpose for health policy.

2. Generating a forum for informed debate of values and priorities. The assessment of disease burden in a country in practice involves participation of a broad range of national disease specialists, epidemiologists, and, often, policy makers. Debating the appropriate values, say, for disability weights or for years of life lost at different ages helps clarify values and objectives for national health policy. Discussing the interrelations among diseases and their risk factors in the light of local conditions sharpens consideration of priorities.

3. Identifying national control priorities. Many countries now identify a relatively short list of interventions, the full implementation of which becomes an explicit priority for national political and administrative attention. Examples include interventions to control TB, poliomyelitis, HIV infection, smoking, and specific micronutrient deficiencies. Because political attention and high-level administrative capacity are in relatively fixed and short supply, the benefits from using those resources will be maximized if they are directed to interventions that are both cost-effective and aimed at problems associated with a high burden. Thus, national assessments of disease burden are one input in establishing a potential short list of control priorities. In the summary of the cost-effectiveness analyses reported in this volume, chapter 2 pays particular attention to identifying cost-effective interventions capable of averting a large disease burden.

4. Allocating training time for clinical and public health practitioners. Medical schools offer a fixed number of instructional hours; training programs for other levels and types of practitioners are likewise limited. A major instrument for implementing policy priorities is allocating this fixed-time resource well. Again that means allocation of time to training in cost-effective interventions in which disease burden is high.

5. Allocating research and development resources. Whenever a fixed effort will have a benefit proportional not only to the size of the effort but also to the size of the problem being addressed, estimates of disease burden become essential for formulating policy. Developing a vaccine for a broad range of viral pneumonias, for example, would have perhaps hundreds of times the effect of a vaccine against Hanta virus infection. Thus, information on disease or risk factor burden is one vital input (of several) to inform research and development resource allocation, as discussed in chapters 4 and 5.

6. Allocating resources across health interventions. Here disease burden assessment often plays a minor role; the task is to shift resources to interventions, which, at the margin, will generate the greatest reduction in DALY loss. When there are major fixed costs in mounting an intervention, as is the case with political and managerial attention for national control priorities, burden estimates are required to improve resource allocation. Likewise, major fixed costs may be associated with making the use of an intervention universal (or expanding it to cover a major percentage of the population), and if so, the cost-effectiveness of the expansion will depend in part on the size of the burden.

Results

Tables 1.A1 and 1.A2 convey summaries of deaths by cause and burden of disease in 2001, respectively.