Quantifying the Global Burden of Disease
We first give an overview of the GBD approach toward summarizing the health of populations and the disease and injury causes of loss of health through the use of a particular form of summary measure, the DALY, and discuss the value choices incorporated in the DALY.
The GBD Study
The simplest and most widely used method for producing population health statistics is to aggregate data on individuals to generate estimates of quantities, for example, the proportion of the population (or of a particular grouping by age or sex) suffering from a particular health problem, being in a particular health state, or dying from a specific cause in a defined time period. This approach rapidly becomes unwieldy when a number of problems are being monitored and the intent is to make comparisons over time, across population groups, or before and after specific health interventions, as in cost-effectiveness analyses. Policy makers then face an explosion in the number of statistics they must compare and difficulties in comparing indicators relating to different health states, mortality risks, or disease events. Such statistics on the health status of populations also suffer from several other limitations that reduce their practical value for policy makers:
-
Health statistics are partial and fragmented. In many countries, basic information on causes of death is not available for all important causes, and even where mortality data are available, they fail to capture the impact of nonfatal outcomes of disease and injury, such as mental disorders, musculoskeletal disorders, blindness, or deafness, on population health.
-
Analyses of incidence, prevalence, or mortality for single causes often result in overestimates, even when carried out by well-intentioned epidemiologists, if not constrained to fit within demographically plausible limits and to be internally consistent and consistent with information on other causes. These problems are compounded when estimates are carried out by groups in competition for scarce resources that are acting as advocates for affected populations or by groups carrying out program evaluation that are also responsible for program implementation ( Murray, Lopez, and Wibulpolprasert 2004 ).
-
Health statistics based on a compilation of separate measures of mortality and of incidence and prevalence rates for a large number of conditions do not allow analysts or policy makers to evaluate outcomes of policies or to compare the relative cost-effectiveness of different interventions.
The 1990 GBD study developed a comprehensive framework for integrating, validating, analyzing, and disseminating fragmented information on the health of populations so that it is truly useful for health policy and planning ( Murray and Lopez 1996b , 1996c , 1997a , 1997b ). Features of this framework included the incorporation of data on nonfatal health outcomes into summary measures of population health (described in the next subsection), the development of methods and approaches to estimate missing data and to assess the reliability of data, and the use of a common metric to summarize the disease burden both from diagnostic categories of the International Classification of Diseases (ICD) and the major risk factors that cause those disease and injury outcomes.
The basic philosophy guiding the burden of disease approach is that almost all sources of health data are likely to have information content provided that they are carefully screened for plausibility and completeness and that internally consistent estimates of the global descriptive epidemiology of major conditions are possible with appropriate tools, investigator commitment, and expert opinion. This philosophy remains central to the 2001 GBD study, which has expanded the framework of the 1990 GBD study to
-
quantify the burden of premature mortality and disability by age, sex, and region for 136 causes;
-
develop internally consistent estimates of incidence, prevalence, duration, and case fatality rates for more than 500 sequelae resulting from the foregoing causes;
-
analyze the contribution to this burden of major physiological, behavioral, and social risk factors by age, sex, and region.
Summary Measures of Population Health and the DALY
To address the problems described above and to provide an outcome measure for cost-effectiveness analyses and priority-setting exercises, a common metric is required for mortality and for loss of health that can be disaggregated into disease and injury causes and risk factors. Since the mid-20th century, analysts have generally agreed that time is the most appropriate metric: time in years lived or lost because of mortality and years lived in various health states. Investigators have developed a wide range of such time-based summary measures of population health, many of them generalizations of life expectancy, such as disability-free life expectancy or variants of the so-called quality-adjusted life year (QALY). For assessing the health of populations, summary measures of population health provide a simple and useful digest of the vast array of components of population health ( Murray, Salomon, and Mathers 2000 ; Wolfson 1999 ). Summary measures of population health do not replace the more detailed reporting of data for specific aspects of health and mortality or for specific causes of health problems; rather, they supplement these data by providing a metric that can be used to monitor trends and compare health across populations or for measuring health outcomes in cost-effectiveness analyses. The last two decades have seen a marked increase in interest in the development, calculation, and use of summary measures ( Field and Gold 1998 ; Murray, Salomon, and others 2002a ; Robine and others 2003 ).
Two classes of summary measures of population health have been developed: health expectancies, for example, disability-free life expectancy; active life expectancy, and healthy life expectancy; and health gaps, such as DALYs and healthy life years (
figure 3.1
). Health expectancies extend the concept of life expectancy to refer to expectations of various states of health or of the overall expectation of years of equivalent full health, not just of life per se. Health gaps are a complementary class of indicators that measure lost years of full health against some normative ideal. Measures of potential years of life lost due to premature mortality have been used for many years to measure the mortality burden of various causes of death. These all measure the gap in years between age at death and some arbitrary standard age before which death is considered premature (typically 65 or 75). The DALY, developed for the GBD study, is an example of a health gap indicator that extends the notion of mortality gaps to include time lived in states other than excellent health.
[Figure
3.1]
One of the fundamental goals in choosing a summary measure of population health for quantifying the global burden of disease was to be able to identify the relative magnitude of different health problems, including diseases, injuries, and risk factors. A health gap measure was chosen because it permits categorical attribution of the fatal and nonfatal burden of diseases and injuries to an exhaustive and mutually exclusive set of disease and injury causes ( Mathers, Ezzati, and others 2002 ; Murray, Salomon, and Mathers 2000 ). The lost years of health (or DALYs) are additive across such a set of disease or injury categories. By contrast, health expectancy measures do not naturally lend themselves to disaggregation by categorically defined causes. Instead, counterfactual methods such as disease elimination are required to quantify the contribution of disease causes to overall health expectancy measures, as well as for dealing with risk factors. Health gap measures also generally require counterfactual analysis to attribute the burden of disease to health determinants and risk factors, as discussed in chapter 4 .
DALYs for a specific cause are calculated as the sum of the years of life lost due to premature mortality (YLL) from that cause and the years of healthy life lost as a result of disability (YLD) for incident cases of the health condition as follows:
The YLL are essentially calculated as the number of cause-specific deaths multiplied by a loss function specifying the years lost as a function of the age at which death occurs. Ignoring for the moment other social preferences discussed later, the basic formula for YLL for a given cause c , age a , and sex s is as follows:
where N(c,a,s) is the number of deaths due to cause c for given age a and sex s and L(a,s) is the standard loss function in years for age a and sex s .
The 1990 GBD study did not use an arbitrary age cutoff such as 70 for the loss function used to calculate YLL, but instead specified the loss function in terms of the life expectancies at various ages in standard life tables, with life expectancy at birth fixed at 82.5 years for females and 80.0 years for males (Coale and Demeny West Model Levels 26 and 25, respectively, see Murray 1996 ), the highest observed life expectancies in the mid-1990s. The sex difference in the loss function was based on evidence of an intrinsic biological difference in life expectancy for males and females, but one that it is much less than the approximately five to seven years observed in developed countries ( Murray 1996 ). Chapter 5 presents a more detailed specification of the loss function used in the standard DALY calculation.
Because YLL measure the incident stream of lost years of life due to deaths, an incidence perspective is also taken for the calculation of YLD. To estimate YLD for a particular cause during a particular time period, the number of incident cases in that period is multiplied by the average duration of the disease and a weight factor that reflects the severity of the resulting health states on a scale from 0 (perfect health) to 1 (dead). Again without yet considering other social preferences, the basic formula for YLD is as follows:
where I(c,a,s) is the number of incident cases for cause c , age a, and sex s ; DW (c,a,s) is the disability weight for cause c , age a, and sex s ; and L(c,a,s) is the average duration in years of the case until remission or death.
The valuation of time lived in nonfatal health states formalizes and quantifies social preferences for different states of health as disability weights. Depending on how these weights are derived, they are variously referred to as disability weights, QALY weights, health state valuations, or health state preferences. Because the DALY is measuring loss of health (unlike the QALY, which measures equivalent healthy years lived), the disability weights for DALYs are inverted, running from 0 (ideal health) to 1 (state comparable to death). Health state valuations are discussed in more detail later.
DALYs are not unique to the GBD study. The World Bank used a variant of DALYs in its seminal study of health sector priorities ( Jamison and others 1993 ), which was derived from earlier work to develop time-based measures that reflected the public health impact of death or illness at different ages better than mortality or prevalence counts or rates ( Dempsey 1947 ; Ghana Health Assessment Project Team 1981 ). As noted, DALYs are an inverse form of the more general concept of QALYs, proposed by Zeckhauser and Shepard (1976) and widely used in economic evaluations. DCP2 ( Jamison and others 2006 ) and WHO's generalized cost-effectiveness analyses for more than 170 health interventions ( Tan-Torres Edejer and others 2003 ) use DALYs as the health outcome measure for their economic analyses.
Countries and health development agencies alike have widely adopted the burden of disease approach as the standard for health accounting, as well as for guiding the determination of health research priorities ( Baskent University 2005 ; Bradshaw and others 2003 ; Bundhamcharoen and others 2002 ; Lozano and others 1995 ; Mahapatra 2002 ; Mathers and de Francisco 2004 ; Mathers, Vos, and Stevenson 1999 ; McKenna and others 2005 ; Vos and others 1995 ; WHO 1996 ).
Making Social Value Choices Explicit
In developing the DALY indicator, Murray (1996) identified three additional value choices that he argued should be made explicit in the formulation of the summary measure:
-
How long "should" people in good health expect to live? This choice determines the loss function L(a,s) for age a and sex s . Should the loss function be determined at the national level or globally? The DALY uses a global loss function that is the same for all people of a given age and sex, irrespective of other characteristics such as race, socioeconomic status, or occupation.
-
Is a year of healthy life gained now worth more to society than a year of healthy life gained in 20 years' time? In other words, should time discounting be applied to the stream of incident lost healthy years represented by the DALY?
-
Are lost years of healthy life valued more at some ages than others? Is a year of life at young adult ages valued more than in old age or infancy? In other words, should unequal age weights be applied to years of healthy life lost at different ages?
Much of the comment on and criticism of the GBD study focused on the explicit social value choices incorporated in the DALY ( Anand and Hanson 1997 , 1998 ; Hyder, Rotllanat, and Morrow 1998 ; Williams 1997 ; Williams 1999 ), particularly the social choices pertaining to age weights and severity scores for disabilities, rather than on the uncertainty of the basic descriptive epidemiology. The latter, particularly in the least developed regions, is likely to be far more consequential for setting health priorities (see chapter 5 ). See Murray and Acharya (1997) and Murray and Lopez (2000) for responses to the criticisms of the value choices made for the 1990 GBD study.
Murray (1996) argues on equity grounds for use of the same life expectancy "ideal" standard for specifying years of life lost for a death in all population subgroups, whether or not their current life expectancy was lower than that of other groups. In addition, he argues that the same disability weight should be used for people of the same age in the same health state.
The DALY measures the future stream of healthy years of life lost due to each incident case of disease or injury and for each death. It is thus an incidence-based rather than a prevalence-based measure. The GBD study applied a 3 percent time discount rate to years of life lost in the future to estimate the net present value of years of life lost. With this discount rate, a year of healthy life gained in 10 years' time is worth 24 percent less than one gained now. Discounting future benefits is standard practice in economic analysis and the following specific arguments can be made for applying discounting to the DALY when measuring population health ( Murray and Acharya 1997 ):
-
to be consistent with the measurement of health outcomes in cost-effectiveness analyses;
-
to prevent giving excessive weight to deaths at younger ages;
-
to address the disease eradication and research paradox, that is, assuming that investment in research or disease eradication has a nonzero chance of succeeding, then without discounting, all current expenditure should be shifted to such investment because the future stream of benefits is infinite.
Chapter 5 examines the sensitivity of the burden of disease results to the choice of discount rate.
In addition to 3 percent time discounting, the 1990 GBD study ( Murray 1996 ) and the GBD results reported in recent world health reports ( WHO 2000 , 2002d , 2004b ) used nonuniform age weights that give less weight to years lived at younger and older ages in calculating DALYs. The inclusion of nonuniform age weights was based on human capital arguments and on a number of studies that suggest the existence of a broad social preference to value a year lived by a young adult more highly than a year lived by a young child or an older person ( Murray 1996 ). At its extreme, age preference manifests as a lack of policy interest in any deaths at ages where the death is not considered premature.
The particular age weights used in the GBD study result in greater weight being given to all deaths below age 39 compared with deaths at older ages. Age weights have perhaps been the most contentious social value incorporated into the DALY ( Anand and Hanson 1997 ; Murray and Acharya 1997 ), and some national burden of disease studies have chosen not to use them ( Mathers, Vos, and Stevenson 1999 ). The editors of DCP2 decided that uniform age weights should be used. Chapter 5 examines the sensitivity of the burden of disease results to different age weighting choices.
To denote different choices for discounting and age weights, we use the notation DALYs( r,K ), where r is the discount rate in percent (not a fraction as in the GBD 1990) and K is the age-weighting modulation factor. The age weights used in the GBD are denoted by K = 1 and the nonuse of age weights (that is, uniform age weighting) is denoted by K = 0. Thus DALYs(3,0) denotes the DALY with 3 percent discounting and uniform age weights as used in DCP2 and DALYs(3,1) denotes the 3 percent discounting and varying age weights as used in the GBD study. Using DALYs(3,0), an infant death represents the loss of 30.3 DALYs(3,0) for males and 30.5 DALYs(3,0) for females, compared with 26.0 and 26.6 DALYs(3,0) at age 30 for males and females, respectively. A death at age 60 represents 16.0 DALYs(3,0) for males and 17.5 for females.
Comparing Time Lived in Different States of Health
To use time as a common currency for nonfatal health states and for years of life lost due to mortality, we must define, measure, and numerically value time lived in nonfatal health states. The valuation of time lived in nonfatal health states formalizes and quantifies social preferences for different states of health as health state weights. While death is not difficult to define, nonfatal health states are. They involve multiple domains of health that relate to different functions, capacities, or aspects of living. During the last three decades, there has been general acceptance of an approach to describing individuals' health states in terms of multiple domains of health and to developing self-reporting instruments that seek information on a core set of these domains, typically no more than five to eight, that capture most of the important variations in health states across individuals ( McDowell and Newell 1996 ; Sadana 2002 ).
One common approach is to describe health as a profile of levels on a series of domains. The Medical Outcomes Study (MOS) Short Form 36 is an example of such an instrument, with eight domains covering self-perceived health, vitality, bodily pain, mental health, physical functioning, social functioning, physical role limitations, and social role limitations ( Ware and Sherbourne 1992 ). MOS Short Form 36 domains are scored on continuous scales from 0 to 100, resulting in a large number of potential health states. Health state profiles intended for use with health state valuations tend to use a more limited number of levels in each domain.
Murray and colleagues argue that health state valuations should be conceptualized and operationalized as judgments about the overall level of health associated with a multidimensional description of an individual's health state, not about overall levels of well-being, quality of life, or utility ( Murray, Salomon, and others. 2002b ; Salomon, Mathers, and others 2003 ). In this conceptualization, health state valuations formalize the intuitive notion that health levels lie on a continuum and that we may characterize one individual as being more or less healthy than another individual at a particular moment in time. Health state valuations quantify departures from perfect health, that is, the reductions in health associated with particular health states. Note that these weights do not measure the quality of life of people with disabilities and do not measure the value of people to society.
By assigning a single number to an individual's health state with reference to ideal health, health state valuations permit aggregating individual health levels over time and comparing health across individuals, and thereby provide the critical link that allows individuals' nonfatal health experience to be combined with information about mortality in summary measures of population health. Researchers have developed a number of choice-based methods to measure preferences for health states ( Salomon and Murray 2004 ).
The 1990 GBD used two forms of the person trade-off method and asked participants in weighting exercises to make a composite judgment about the severity distribution of the condition and the preference for time spent in each severity level ( Murray 1996 ). This was largely necessitated by the lack of population information on the severity distribution of most conditions at the global and regional levels. The disability weights used in the GBD 2001 are still based in large part on the GBD 1990 disability weights ( Murray 1996 ). Disability weights may vary by age, sex, and region, reflecting variations in the severity distributions of health states and the proportions of cases treated. A common global valuation function is assumed for the underlying health state valuations for specific health states. Despite the assertion by some commentators that valuations for certain health states are likely to be extremely heterogeneous across individuals and populations, empirical evidence suggests otherwise. Valuation studies carried out with deliberative small groups from a wide range of countries have found surprising consistency in valuations across cultures ( Salomon and Murray 2002b ). More recently, valuation studies carried out as part of the WHO multicountry survey study have also found reasonable consistency in health state valuations for most health states ( Salomon, Murray, and others 2003 ).
Following the GBD terminology, the term disability is used here broadly to refer to departures from optimal health in any of the important domains of health, including mobility, self-care, participation in usual activities, pain and discomfort, anxiety and depression, and cognition and social participation. We thus refer to disability weights and healthy years lost due to disability as shorthand terms for health state preferences and years of healthy life lost because of time lived in states other than the reference state of optimal health, respectively. Note that with this usage, disability, that is, states other than ideal health, may be short term or long term: a day with a common cold is a day with disability.
