Discussion and Conclusions
The analysis presented in this chapter has confirmed some of the conclusions of the original GBD study about the importance of including nonfatal outcomes in a comprehensive assessment of global population health, and has also confirmed the growing importance of noncommunicable diseases in low- and middle-income countries. However, it has also documented dramatic changes in population health in some regions since 1990. The key findings include the following:
HIV/AIDS is now the fourth leading cause of the burden of disease globally and the leading cause in Sub-Saharan Africa.
In low- and middle-income countries, the epidemiological transition has resulted in a 20 percent reduction in the per capita disease burden due to Group I causes since 1990. Without the HIV/AIDS epidemic, this reduction would have been closer to 30 percent. Several of the "traditional" infectious diseases, such as TB and malaria, have not declined, in part because of weak public health services and the increased numbers of people with immune systems weakened by HIV/AIDS.
The per capita disease burden in Europe and Central Asia increased by nearly 40 percent during 1990-2001, meaning that this region now has worse health than all other regions except South Asia and Sub-Saharan Africa. The unexpected increase in the disease burden, and the concomitant reduction in life expectancy, in countries of this region appear to be related to such factors as alcohol abuse, suicide, and violence, which seem to be associated with societies facing dramatic social and economic changes. The rapidity of these declines has dramatically changed our perceptions of the time frames within which substantial changes in the burden of chronic disease can occur and of the potential for such adverse health trends to occur elsewhere.
Adults under the age of 70 in low- and middle-income countries face a greater risk of death from noncommunicable diseases than adults of the same age in high-income countries.
In Europe and Central Asia, Latin America and the Caribbean, and the Middle East and North Africa, more than 30 percent of the entire disease burden among male adults ages 15 to 44 is attributable to injuries, including road traffic accidents, violence, and self-inflicted injuries. In addition, injury deaths are noticeably higher for women in some parts of Asia and the Middle East and North Africa than in other regions, partly because of high levels of suicide and violence. Combined with higher rates of infant and child mortality for girls, this results in a narrower differential between male and female healthy life expectancy than in any other region.
Sense organ disorders, principally hearing and sight loss, contribute significantly to disability in all regions of the world.
Levels of nonfatal health loss are proportionately greater in low- and middle-income countries than in high-income countries, contrary to the perception that disability is associated with older populations. The gap between healthy life expectancy and total life expectancy is proportionately highest for the low-income countries.
The analysis presented in this chapter has aimed to produce a comprehensive and detailed assessment of the global burden of disease, based on all available relevant data. It has attempted to maximize the use of high-quality, population-based data, and for regions and causes for which data are sparse has used the available evidence and the best available methods to make inferences and to assess the uncertainty in resulting estimates (see chapter 5 ). The need for internal consistency between estimates of incidence, prevalence, case fatality rates, and mortality rates for a given disease and for consistency across diseases and injuries with known total levels of mortality are crucial strategies for making the best use of multiple sources of uncertain and potentially biased data.
The data inputs used for the GBD 2001 estimation of global and regional causes of death have been summarized in tables 3.1 , 3.2 , and 3.5 . In excess of 770 country-years of death registration data and more than 3,000 additional sources of information on levels of child and adult mortality and on specific causes of death were used to estimate global and regional patterns of mortality. Together with the more than 8,500 data sources used for the estimation of YLD, the GBD 2001 has incorporated information from more than 10,000 data sets relating to population health and mortality. This represents the largest synthesis of global information on population health carried out to date.
Despite the perceptions of some critics that the GBD study is inadequately empirically based for some regions, particularly Sub-Saharan Africa ( Cooper and others 1998 ), it is notable that fully one-third of the more than 10,000 data sources used relate to Sub-Saharan African populations, albeit with the serious limitations on the information available on mortality noted earlier. We believe that the GBD studies have demonstrated the importance of including assessments of all causes of the disease and injury burden, even in the face of limited or missing data, to ensure that a comprehensive overview is provided to gain a better understanding of the importance of specific diseases and risk factors in causing loss of health. Otherwise, limitations in the evidence base for certain causes or regions might lead to their omission, and hence to the conclusion that they cause no burden, thereby presenting health decision makers with a misleading picture.
Nevertheless, the fact that estimates are possible does not obviate the need to put a higher priority on addressing the serious lack of information on levels of adult mortality and causes of death in some regions, particularly Sub-Saharan Africa. The key need for countries is to establish a system that registers the most common causes of death for the entire population without serious biases (such as an emphasis on urban mortality), in which there is reasonable confidence, and which yields timely data. Complete VR with annual population updates is the ideal system to generate this information, but it is not essential. Recent experience in countries such as China, India, and Tanzania suggests that sample registration based on a representative set of surveillance sites, and with appropriate controls and reporting procedures, can yield extremely useful information about levels, patterns, and causes of mortality for large populations ( Setel and others 2005 ; Yang and others 2005 ). Low- and middle-income countries can benefit from the advantages of death registration without implementing a system of complete population coverage and medical certification ( Rao, Bradshaw, and Mathers 2004 ). To support such systems, priority needs to be given to developing a standardized reporting form for verbal autopsies and to implementing validation studies to assess the reliability and accuracy of verbal autopsy methods.
Improved verbal autopsy methods will also contribute to improving the accuracy of estimates of the causes of child deaths under five, the majority of which occur in countries without useable death registration data. As discussed in chapters 5 and 6 , new data and syntheses for major causes of child death may result in future revisions to the estimates of child deaths for certain causes.
There is also a lack of good population-based epidemiological data for developing regions, particularly for noncommunicable diseases. For example, even though IHD and stroke are among the leading causes of the burden of disease in most regions, few recent and reliable sources of information on the prevalence and severity distribution of chronic cardiovascular conditions and long-term disability following stroke are available outside the high-income countries. Similarly, even in high-income countries, few population-based studies of the prevalence of chronic lung disease or musculoskeletal conditions have been carried out. Cross-national surveys, such as WHO's World Health Survey conducted during 2002 and 2003 in 73 countries, will fill some information gaps for some chronic diseases and mental disorders ( Ustun, Chatterji, Mechbal, and others 2003 ). However, there remain significant issues that will need to be addressed relating to the comparability of prevalence data derived from self-reported survey data on symptoms of mental disorders, angina, and other chronic diseases.
Lack of information has resulted in limitations in the disease models used to estimate the burden of disease for some causes. Future iterations of burden of disease analysis will need to review disease models and sequelae chosen for estimating YLD to ensure that the best available estimates of the disease burden for each cause continue to remain based on current knowledge and data.
A particular difficulty is how to measure and characterize the average health states associated with sequelae. This is partly an issue of valuation of health states for the construction of disability weights, and partly an issue of lack of information on the population-level distribution of outcomes and the severity of health states. To a large extent, the disability weights used here derive from the original GBD study ( Murray 1996 ), where typically an average disability weight was estimated for disease sequelae averaged across the distribution of outcomes, in some cases separately for treated and untreated cases. Stouthard and others (1997) have gone further in assessing disability weights for a range of severity levels of outcomes for a particular sequela, thereby allowing the overall final disability weight for a sequela to take account of regional variations in the severity distribution of outcomes.
The 2001 WHO Multicountry Survey Study collected health state valuation data on more than 500,000 health states from respondents in 71 countries, which Salomon and Murray (2004) used to construct a health state valuation function. The World Health Survey also included a health state valuation module, and analysis of the resulting data is under way ( Salomon, Murray, and others 2003 ). In the next iteration of burden of disease analysis, it should be feasible to use health state valuations based on such survey data, together with descriptions of outcomes associated with disease sequelae, to produce updated disability weights that take into account not only the available information on health state distributions for disease sequelae, but also the health state preferences of people from all regions. A particular issue is the measurement of disability weights for low severity but highly prevalent conditions, such as anemia and hearing loss, where the current disability weights are small but quite uncertain and are multiplied by large prevalences (see chapter 5 ).
Burden of disease analysis provides a comprehensive, comparative overview of the state of population health and the factors affecting the health of populations. The 2001 GBD study is an expanded effort compared with the original 1990 study, with the incorporation of much new data and a greater understanding of the limitations of routinely available data sets. Nevertheless, substantial uncertainty remains about the comparative burden of diseases and injuries in many parts of the world that has significantly greater consequences for policy than the inclusion or otherwise of social choices such as age weighting in the basic burden of disease metric. We can conclude with some certainty that major causes of death and disability, such as tobacco and HIV/AIDS, are global pandemics and are likely to become more widespread unless control programs are more widely implemented. However, we remain substantially uncertain about the true levels of the disease burden from chronic lung disease, heart disease, stroke, mental disorders, various forms of injury, and a number of other key health concerns. International health agencies such as WHO and public health and epidemiological researchers need to make a concerted effort to improve data collection, and hence knowledge, about the true extent of the disease burden worldwide. Even efforts that substantially reduce uncertainty will be a major advance toward this goal.
With rising pressure on resources for health in all countries, priority setting in the health sector will increasingly depend on comprehensive, comparative information about the impact of diseases, injuries, and risk factors on population health. The burden of disease framework, with 15 years of development and application in numerous countries across the globe, offers the best, indeed the only, approach to comprehensively assess the impact of conditions and exposures that health systems need to deal with if population health is to improve rapidly. Yet to be even more useful for setting and monitoring global health priorities, a more concerted effort is needed to obtain and critically assess data sets on the health of populations in all countries. This must be a key focus of future efforts to assess the burden of disease. With WHO now giving greater emphasis to working with countries on capacity building and on specific organizational intervention priorities, new global partners such as the Ellison Institute for Global Health ( Horton 2005 ) are urgently required to provide stewardship and guarantee that the evidence base for health policy and priority setting will develop at a pace commensurate with need.