Global, regional, and national disability-adjusted life years (DALYs) for 306 diseases and injuries and healthy life expectancy (HALE) for 188 countries, 1990-2013: quantifying the epidemiological transition.
OA: closed
Abstract
BackgroundThe Global Burden of Disease Study 2013 (GBD 2013) aims to bring together all available epidemiological data using a coherent measurement framework, standardised estimation methods, and transparent data sources to enable comparisons of health loss over time and across causes, age-sex groups, and countries. The GBD can be used to generate summary measures such as disability-adjusted life-years (DALYs) and healthy life expectancy (HALE) that make possible comparative assessments of broad epidemiological patterns across countries and time. These summary measures can also be used to quantify the component of variation in epidemiology that is related to sociodemographic development.MethodsWe used the published GBD 2013 data for age-specific mortality, years of life lost due to premature mortality (YLLs), and years lived with disability (YLDs) to calculate DALYs and HALE for 1990, 1995, 2000, 2005, 2010, and 2013 for 188 countries. We calculated HALE using the Sullivan method; 95% uncertainty intervals (UIs) represent uncertainty in age-specific death rates and YLDs per person for each country, age, sex, and year. We estimated DALYs for 306 causes for each country as the sum of YLLs and YLDs; 95% UIs represent uncertainty in YLL and YLD rates. We quantified patterns of the epidemiological transition with a composite indicator of sociodemographic status, which we constructed from income per person, average years of schooling after age 15 years, and the total fertility rate and mean age of the population. We applied hierarchical regression to DALY rates by cause across countries to decompose variance related to the sociodemographic status variable, country, and time.FindingsWorldwide, from 1990 to 2013, life expectancy at birth rose by 6·2 years (95% UI 5·6-6·6), from 65·3 years (65·0-65·6) in 1990 to 71·5 years (71·0-71·9) in 2013, HALE at birth rose by 5·4 years (4·9-5·8), from 56·9 years (54·5-59·1) to 62·3 years (59·7-64·8), total DALYs fell by 3·6% (0·3-7·4), and age-standardised DALY rates per 100 000 people fell by 26·7% (24·6-29·1). For communicable, maternal, neonatal, and nutritional disorders, global DALY numbers, crude rates, and age-standardised rates have all declined between 1990 and 2013, whereas for non-communicable diseases, global DALYs have been increasing, DALY rates have remained nearly constant, and age-standardised DALY rates declined during the same period. From 2005 to 2013, the number of DALYs increased for most specific non-communicable diseases, including cardiovascular diseases and neoplasms, in addition to dengue, food-borne trematodes, and leishmaniasis; DALYs decreased for nearly all other causes. By 2013, the five leading causes of DALYs were ischaemic heart disease, lower respiratory infections, cerebrovascular disease, low back and neck pain, and road injuries. Sociodemographic status explained more than 50% of the variance between countries and over time for diarrhoea, lower respiratory infections, and other common infectious diseases; maternal disorders; neonatal disorders; nutritional deficiencies; other communicable, maternal, neonatal, and nutritional diseases; musculoskeletal disorders; and other non-communicable diseases. However, sociodemographic status explained less than 10% of the variance in DALY rates for cardiovascular diseases; chronic respiratory diseases; cirrhosis; diabetes, urogenital, blood, and endocrine diseases; unintentional injuries; and self-harm and interpersonal violence. Predictably, increased sociodemographic status was associated with a shift in burden from YLLs to YLDs, driven by declines in YLLs and increases in YLDs from musculoskeletal disorders, neurological disorders, and mental and substance use disorders. In most country-specific estimates, the increase in life expectancy was greater than that in HALE. Leading causes of DALYs are highly variable across countries.InterpretationGlobal health is improving. Population growth and ageing have driven up numbers of DALYs, but crude rates have remained relatively constant, showing that progress in health does not mean fewer demands on health systems. The notion of an epidemiological transition--in which increasing sociodemographic status brings structured change in disease burden--is useful, but there is tremendous variation in burden of disease that is not associated with sociodemographic status. This further underscores the need for country-specific assessments of DALYs and HALE to appropriately inform health policy decisions and attendant actions.FundingBill & Melinda Gates Foundation.
Full text
42,292 characters
· extracted from
pmc-nxml
· 4 sections
· click to expand
Intro
The Global Burden of Disease study 2013 (GBD 2013) seeks to bring together all available epidemiological data using a coherent measurement framework, standardised estimation methods, and transparent data sources to allow comparisons of health loss to be made over time and across causes, age–sex groups, and geographies. The GBD 2013 data for disease and injury incidence and prevalence, years lived with disability (YLDs), causes of death, and years of life lost because of premature mortality (YLLs) for 188 countries provide an opportunity to assess the effect of recent changes in population health by examining summary measures of health loss attributed to specific causes, expressed in DALYs, and summary measures of average population health, expressed as HALE. 1 , 2 These measures are crucial to track health progress, strengthen policy decisions, assess programme effects and results, and inform health service and research priorities. Such holistic measures of population health, encompassing both disability and mortality levels and patterns in populations, are also attracting interest as part of the discussion around the Sustainable Development Goals . 3 - 5
A hallmark of the GBD approach is an emphasis on making national data easier to compare by taking into account the extensive variation that exists in national medical certification and cause of death coding practices and widely varying case definitions and measurement methods used to track the incidence and prevalence of diseases and injuries. 1 , 2 The GBD not only provides detailed metrics for specific causes, but also generates summary measures, such as DALYs and HALE, which enable comparative assessments of broad epidemiological patterns across countries and different time periods. HALE is a useful summary of overall health for a country and DALYs allow assessment of both premature mortality and non-fatal outcomes by cause. These broad summary measures allow quantification of general trends, such as the epidemiological transition, while also making clear how countries and regions deviate from general patterns. 6 - 9 The unfolding of the HIV epidemic and the rise of adult mortality, especially among men in Eastern Europe and Central Asia, have called into question the notion of a universal pattern of epidemiological change that occurs with sociodemographic development. 2 , 10 - 13 However, the general notion of a shift from communicable to non-communicable causes of disease burden and injuries remains a powerful framework for global and regional health policy debates. 9 , 14 - 18 The GBD provides an opportunity to quantify these patterns and explore the extent to which epidemiological change is driven by sociodemographic change, reduction of health risks, improvement of health management, or other local factors.
GBD 2013 results for deaths, YLLs, incidence, prevalence, and YLDs by cause for 1990 to 2013 for 188 countries have already been published. 1 , 2 In this study we use these GBD 2013 results to calculate DALYs and HALE. These summary metrics are used to characterise broad patterns of lost healthy life and cross-country variations within these patterns. The GBD 2013 provides a complete re-analysis of each country’s data from 1990 to 2013 and thus supersedes all previously published GBD analyses of DALYs and HALE.
Methods
GBD 2013 uses a hierarchy of causes that organises 306 diseases and injuries into four levels of classification, the rationale for which has been described previously. 2 , 19 The first level distinguishes three broad categories: first, communicable, maternal, neonatal, and nutritional disorders; second, non-communicable diseases; and third, injuries. Level 2 has 21 mutually exclusive and collectively exhaustive categories, level 3 has 163 categories, and level 4 has 254 categories. The full cause list, including International Classification of Diseases tenth edition (ICD-10) codes, has been reported previously. 1 , 2 Mortality rates and causes of death for each country–age–sex–year group have been estimated in accordance with some general principles: identification of all available data sources, evaluation of the quality and correction for known bias in each data source, consistent statistical estimation including uncertainty analysis, and cross-validation analysis to assess model performance. Details of data sources and estimation methods used to deal with missing data and multiple measurements for the same country–age–sex–year group have been described previously. 2 Disease and injury incidence and prevalence and computation of YLDs have been estimated in line with similar principles of identification and assessment of the quality of all available sources for 2337 sequelae of the 301 diseases and injuries. 1 The discrepancy between the 306 diseases and injuries for which DALYs are calculated and the 301 diseases and injuries for which YLDs are calculated is attributable to five diseases that cause death but do not cause disability: sudden infant death syndrome, indirect maternal deaths, late maternal deaths, maternal deaths aggravated by HIV/AIDS, and aortic aneurysm. Various statistical estimation methods were used depending on the details of specific diseases, the most common approach being the application of a Bayesian metaregression model, DisMod-MR 2.0. 20 We used alternative methods when the basic susceptible, with disease, and dead states in DisMod-MR 2.0 were insufficient to capture the natural history of a sequela. We aggregated sequelae prevalence into YLDs first by estimating the distribution of comorbidities through microsimulation, and second by using disability weights derived from population-based surveys of the general public to assign disability weights to each sequela and combination of sequelae—details of both steps have been described previously. 1 , 21
We used the GBD 2013 results for YLLs 2 and YLDs 1 to calculate DALYs. To calculate HALE, we used YLDs per person and life tables. 1 , 2 We applied decomposition of variance using hierarchical regression to DALY rates by cause.
For each year for which YLDs have been estimated (1990, 1995, 2000, 2005, 2010, and 2013), we computed DALYs by adding YLLs and YLDs for each age–sex–country group. We assumed that uncertainty in YLLs is independent of uncertainty in YLDs. We did this by summing the first draw of the 1000 draws for YLLs and YLDs and then repeating for each subsequent draw. We calculated 95% uncertainty intervals (UIs) using the 25th and 975th ordered draw of the DALY uncertainty distribution.
We calculated HALE in accordance with the methods outlined by Salomon and colleagues. 8 In brief, we used Sullivan’s method 22 to incorporate information about average levels of health experienced at different ages into an abridged life table to produce estimates of life expectancy that are adjusted for reductions in functional health status relating to prevalent health conditions. Effectively, the cumulative years lived in an age group in the abridged life table (the life expectancy column) for each country–age–sex group is multiplied by the YLDs per person for that country–age–sex group. Calculation of HALE relies on three inputs from GBD 2013: life tables by sex, country, and year; estimates of the prevalence of 2337 sequelae by age, sex, country and year; and disability weights for 235 unique health states that collectively cover the range of functional health losses and symptoms associated with the 2337 sequelae. Wang and colleagues 2 have described data sources and methods to estimate mortality and life tables, and Vos and colleagues 1 have described these for the measurement of prevalence of sequelae and disability weights. 1 We combined information about prevalence and disability weights into measures of the overall rate of YLDs per person in each age–sex–country group. We make the strong assumption that uncertainty in YLDs per person is independent of uncertainty in age-specific death rates to calculate uncertainty distributions for HALE.
The epidemiological transition is an extension of the notion of the demographic transition. In demographic transition, a characteristic evolution occurs in populations over time towards reduced fertility rates, reduced mortality rates, and an older age distribution of the population. The widely used concept of the epidemiological transition adds the idea that, in addition to these changes, a characteristic change occurs in the contributing causes of death. The epidemiological transition has been broadened to encompass the more general health transition, including both morbidity and mortality. A single variable to characterise countries over time in terms of their demographic and epidemiological status would be crucial to describe the epidemiological transition. Some studies examine associations with income per person, whereas others use variables such as mean age of the population. 23 , 24 We aimed to construct a single composite variable to represent both demographic status and socioeconomic development to explore the patterns of the epidemiological transition. To construct this sociodemographic status variable, we assessed variables indicative of socioeconomic status and demographic change that were available for all 188 countries from 1990 to 2013. We did not include measures of income inequality, such as the Gini coefficient, because these were not available for all countries for each year. We used principal components analysis (PCA) of the log transformation of income per person (in constant international dollars), average years of schooling of the population after age 15 years, the log of the total fertility rate, and the log of the mean age of the population. The relationship between the PCA variables and DALY rates were highly non-linear, but became linear with respect to the log DALY rates after log transformation of three of the four sociodemographic status component variables. Before using PCA, we normalised each variable to have a mean of zero and a standard deviation of 1·0. Only the first component of the PCA had an eigenvalue greater than 1·0 and the weights were 0·471 for income per person, −0·517 for total fertility rate, 0·495 for education per person, and 0·516 for mean age. 25 As expected, the sign on the total fertility rate was negative, whereas the sign on the other three components was positive. We also tested all possible combinations of the four variables using the same PCA approach to confirm that the principal component of all four was the most predictive of variation in DALY rates by cause. We used the predicted value of the first component for each country–year in the subsequent ANOVA and predictions of the epidemiological change with sociodemographic status.
We used hierarchical regression to decompose variance in log DALY rates into components related to the sociodemographic status, intercountry variation, year, and fraction explained by the interactions of the other variables. This approach estimates a simple model with uncorrelated random effects for year, country, and sociodemographic status. 26 We divided sociodemographic status into vigintiles (20 equal interval bins) to allow for non-linear correlations between log DALY rates and sociodemographic status for a cause. We did tests with up to 50 bins for the sociodemographic status variable with no change in qualitative results. We divided the variance of each random effect by total variance to decompose variance into different factors. We did this variance decomposition analysis for GBD level 2 and level 3 causes. We use these regressions to predict the pattern of DALYs by cause (and through separate regressions for YLLs and YLDs) purely as a function of sociodemographic status, holding all other random effects (year and country) at zero. Because there could be lagged effects between sociodemographic status and DALY rate, we tested alternative models with sociodemographic status lagged from 1 to 10 years. Use of sociodemographic status from the same year as the DALY rates explained, on average, the highest proportion of the variance in DALY rates.
We selected the revised GBD 2013 world population standard for the age standardisation of rates. Details of the age standard, and its development, have been reported previously. 2
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had access to the data in the study and final responsibility to submit the paper.
Results
Global life expectancy at birth for both sexes combined increased from 65·3 years (95% UI 65·0–65·6) in 1990 to 71·5 years (71·0–71·9) in 2013, whereas during the same interval, HALE at birth for both sexes combined increased from 56·9 years (54·5–59·1) to 62·3 years (59·7–64·8). The survivorship curves shift up and to the right with increasing quintiles of country sociodemographic status ( figure 1 ). In the three groups of countries, defined as the lowest, middle three, and highest quintiles of country sociodemographic status in 2013, individuals are distributed across ranges of disability weights, with the majority of the population in most age groups living in health states with disability weights in the range 0·0–0·1. The fraction of individuals in the life table in full health (ie, living with a disability weight of zero) is 16·3% in the lowest sociodemographic status quintile of countries. Even in the most advantageous sociodemographic quintile of countries, the time lived in full health constitutes only a small fraction (17·5%) of the overall life course. At the other end of the spectrum, the expectation of years lived with disability weights greater than 0·5 is 3·59 years in the lowest quintile and 6·60 years in the highest quintile of countries.
Figure 2 shows the need to understand global epidemiological change in terms of numbers, rates, and age-standardised rates. The number of DALYs caused by communicable, maternal, neonatal, and nutritional disorders has declined steadily from 1·19 billion (95% UI 1·15 billion to 1·24 billion) in 1990 to 769·3 million (725·5 million to 814·9 million) in 2013, whereas DALYs for non-communicable diseases (NCDs) have increased steadily, rising from 1·08 billion (0·97 billion to 1·20 billion) to 1·43 billion (1·26 billion to 1·61 billion) during the same period ( figure 2A ). The year of crossover, during which global DALYs for NCDs exceeded those for global communicable, maternal, neonatal, and nutritional causes, was 1994. DALYs due to injuries have remained relatively constant, decreasing slightly from 269·6 million (251·6 million to 286·7 million) to 247·6 million (231·3 million to 265·1 million). Figure 2B shows crude DALY rates per 100 000 people for these three broad cause groups, thereby removing the effect of global population growth during the period. The DALY rate for NCDs has remained fairly constant, while substantial declines have occurred in DALY rates for communicable, maternal, neonatal, and nutritional disorders (52·2%, 50·1–54·4) and injuries (32·0%, 27·0–35·9). The analysis of age-standardised DALY rates, shows that, after controlling for changes in population size and composition, NCD disease burden worldwide, has continued to decline, falling by 14·5% (11·6–17·3) between 1990 and 2013 ( figure 3C ). During the same period, worldwide age-standardised DALY rates fell by 42·4% (40·0–45·0) for communicable, maternal, neonatal, and nutritional disorders and 30·9% (26·1–34·7) for injuries.
We decomposed the changes in the number of DALYs into trends for two periods: 1990–2005 and 2005–2013 ( figure 3 ). 10 Between 1990 and 2005, the number of global DALYs changed only slightly, from 2·54 billion (95% UI 2·40 billion to 2·70 billion) to 2·51 billion (2·33 billion to 2·72 billion). Looking at disease-level details within this relative stagnation reveals important trends for specific diseases ( figure 3A ); the earlier period was characterised by decreases in the number of DALYs from diarrhoea, lower respiratory infections, measles, neonatal causes, tuberculosis, and tetanus, with smaller contributions from declines in congenital causes and some injuries. Conversely, large increases in disease burden were recorded for HIV/AIDS and malaria, with smaller increases for road injuries and a diverse range of NCDs, including ischaemic heart disease, diabetes, low back and neck pain, stroke, and depression, in addition to several types of cancer. From 2005 to 2013, total DALYs worldwide decreased from 2·51 billion (2·33 billion to 2·72 billion) to 2·45 billion (2·23 billion to 2·68 billion; figure 3B ). Decreases were recorded for diarrhoea, malaria, HIV/AIDS, lower respiratory infections, measles, tuberculosis, and neonatal causes, and nearly all injuries. Increases were noted for a wide range of NCDs, especially low back and neck pain, ischaemic heart disease, diabetes, chronic obstructive pulmonary disease (COPD), depression, stroke, and sense organ disorders. Although they were not large contributors to the number of DALYs, notable increases were seen for dengue, food-borne trematodes, and leishmaniasis. Separate analyses of changes in age-standardised DALY rates for the period 2005–13 (data not shown) suggest that most of the increases shown in figure 3 are caused by ageing of the population and population growth.
We assessed changes in the age-standardised DALY rates of the leading GBD level 3 causes for 1990–2005 and 2005–2013 ( figure 4 ; level 4 of the GBD cause hierarchy is reported in the appendix p 2). Between 1990 and 2005 huge reductions occurred in measles, meningitis, iron-deficiency anaemia, congenital anomalies, tuberculosis, drowning, protein-energy malnutrition, and some neonatal disorders, whereas disease burden from HIV/AIDS and malaria substantially increased ( figure 4 ). From 2005 to 2013, age-standardised DALY rates for ischaemic heart disease, lower respiratory infections, and cerebrovascular disease have declined, although not sufficiently for these conditions to be replaced as the leading causes of disease burden worldwide. The ranks of low back and neck pain, road injuries and COPD have all increased since 2005. Age-standardised rates decreased significantly for 16 of the 25 leading causes of DALYs in 2013; for the remaining nine causes (Alzheimer’s disease, chronic kidney disease, congenital anomalies, depressive disorders, diabetes, low back and neck pain, migraine, neonatal sepsis, and skin diseases), age-standardised rates did not significantly change.
Table 1 shows DALYs for each cause in 2005 and 2013 and changes in numbers and age-standardised rates of the DALYs (for the same information for 1990 to 2013 see appendix p 4).
We decomposed the variance of DALY rates for GBD level 2 causes into contributions from sociodemographic status, year, country, and unexplained sources (residual; table 2 ). Sociodemographic status explained more than 50% of the variance for diarrhoea, lower respiratory infections and other common infectious diseases; maternal disorders; neonatal disorders; nutritional deficiencies; other communicable diseases; musculoskeletal disorders; and other NCDs. Furthermore, sociodemographic status explains between a fifth and a half of the variance for HIV/AIDS and tuberculosis; neurological disorders; mental and substance use disorders; transport injuries; and forces of nature, war, and legal intervention. Sociodemographic status explained little of the variance in the DALY rates for neglected tropical diseases and malaria, for which time-invariant country differences account for 84·98% of the variance. Notably, less than 10% of the variance in the burden of several level 2 causes could be related to sociodemographic status, including cardiovascular diseases; chronic respiratory diseases; cirrhosis; diabetes, urogenital, blood, and endocrine diseases; unintentional injuries; and self-harm and interpersonal violence. Year explained less than 7% of variance for all causes. By contrast, time invariant intercountry variation was an important component of the variance in DALY rates for all causes, ranging from a low of 11·29% for neonatal disorders to 91·23% for self-harm and interpersonal violence. Intercountry variation explains more than two-thirds of the total variance in DALY rates for HIV/AIDS and tuberculosis; neglected tropical diseases and malaria; neoplasms; cardiovascular diseases; chronic respiratory diseases; cirrhosis; digestive diseases; mental and substance use disorders; diabetes, urogenital, blood, and endocrine disorders; unintentional injuries; and self-harm and interpersonal violence. Notably, together, sociodemographic status and country account for more than 90% of the variance for 17 of 21 GBD level 2 causes; indeed, the lowest fraction of variance accounted for by these three factors is 80·2% for forces of nature, war, and legal intervention.
Figure 5 shows the predicted global composition of YLLs and YLDs for level 2 causes at different levels of sociodemographic status, controlling for year and country. YLLs from diarrhoea, lower respiratory infections, and other common infections, and neonatal disorders fall substantially as sociodemographic status increases. Other YLLs that fall noticeably with rising sociodemographic status include YLLs from maternal causes, nutritional deficiencies, other non-communicable causes (including congenital causes), and unintentional injuries. YLLs from cardiovascular diseases at first increase slightly with increasing sociodemographic status, but then decrease at the highest levels of country sociodemographic status. Some important causes of global YLLs are not strongly related to sociodemographic status because they are largely country-specific, such as neglected tropical diseases and malaria, neoplasms, and intentional injuries. By contrast, overall YLDs decline slightly at first, but then increase substantially, showing the opposite trend to YLLs. The large increases in YLDs are related to musculoskeletal disorders; mental and substance use disorders; diabetes, urogenital, blood, and endocrine diseases; and neurological disorders. As sociodemographic status rises, the steady decreases in YLLs and increases in YLDs cause the proportion of total DALYs attributable to YLDs to steadily rise from 9·9% at the lowest level of sociodemographic status to 49·1% in the highest vigintile. Above the tenth vigintile of sociodemographic status, the rise in YLDs and fall in YLLs nearly compensate for each other so that DALY rates have remained largely constant.
In 1990, life expectancy ranged from 46·9 years (95% UI 45·1–48·2) in the Central African Republic to 80·7 years (78·7–82·5) in Andorra, while HALE ranged from 40·4 years (38·2–42·5) in the Central African Republic to 70·2 years (67·7–72·5) in Japan. By 2013, life expectancy ranged from 48·3 years (46·5–50·1) in Lesotho to 83·9 years (82·3–85·5) in Andorra, and HALE ranged from 42·0 years (39·8–44·2) in Lesotho to 73·4 years (70·5–76·0) in Japan. Disaggregating by sex, in 1990, there were no countries in which men had attained a HALE of 70 years or more, and only in Japan and Andorra did women attain this. By 2013, HALE for men exceeded 70 years in only two countries (Japan and Singapore), whereas HALE for women exceeded 70 years in 40 countries ( table 3 ). Of the 21 GBD regions, nine contained at least one country in which female HALE was at least 70 years in 2013. For most countries, changes in HALE were positive for both men and women between 1990 and 2013, and the differences were significant. However, HALE in 2013 was not significantly higher than it was in 1990 in 43 countries for men and in 32 countries for women. As life expectancy increases, the gap between life expectancy and HALE widens, increasing to more than 10 years at a life expectancy of 77 years for women and 78 years for men. The life expectancy minus HALE divided by life expectancy (the percentage of life expectancy lost because of poor health) ranged between 11·5% and 15·0%.
Figure 6 shows the ten most common causes of DALYs for each country in 2013. The leading causes of DALYs vary substantially across regions, representing both different levels of sociodemographic status and distinct regional patterns. In high-income regions, low back and neck pain, ischaemic heart disease, cerebrovascular disease, and tracheal, bronchus, and lung cancer are often among the four most common causes, although major depression, COPD, and diabetes come into the top four slots in some countries. In central and eastern Europe, cardiovascular diseases rank consistently in the most common causes of DALYs. Self-harm and depression frequently rank higher in eastern Europe than in central Europe or elsewhere. In central Asia, representative of the mixed levels of sociodemographic status present in the region, leading causes include neonatal encephalopathy and congenital causes. In central Latin America, violence, ischaemic heart disease, diabetes, low back and neck pain, and road injuries comprise the top five causes. Other examples of distinct regional patterns include the high ranking of COPD in east Asia, the dominance of malaria in west Africa, and the dominance of HIV/AIDS in eastern and southern sub-Saharan Africa. Within some regions, such as north Africa and the Middle East, the leading causes varied substantially.
Discussion
Global health is improving: life expectancy at birth rose by 6·2 years between 1990 and 2013, while HALE at birth increased by 5·4 years during the same interval; worldwide, age-standardised DALY rates fell by 27%. Global progress has accelerated since 2005 because of major reductions in HIV/AIDS and malaria, in addition to continued progress against other major communicable, maternal, neonatal, and nutritional disorders. Although the total volume of DALYs is down by only 3·6% over the 23 year period, this is largely explained by population growth and ageing driving up numbers of DALYs. Declines in age-standardised DALY rates are counterbalanced by population growth and ageing, so that, despite improvements in age–sex-specific health status, demands on health systems remain high. An example of these demands is the fact that the number of individuals in the world living in states of health characterised by a disability weight greater than 0·1 has increased by 43% from 1990 to 2013.
In 1971, Omran 27 outlined the concept of the epidemiological transition to describe the changing pattern of causes of death that results from sociodemographic development. The notion of the epidemiological transition has been expanded to recognise the phase in transition that leads to double burden of disease 9 , 28 , 29 and the countertransitions of the HIV/AIDS epidemic and the rise of mortality in the former Soviet Union. 2 , 10 , 11 , 13 , 30 - 32 Many studies have recognised regional and national variation in the epidemiological transition. 9 , 14 , 33 - 35 Taking advantage of the database of the GBD 2013 country-level results from 1990 to 2013, we have quantified the extent to which sociodemographic status accounts for changes in epidemiological patterns, as opposed to other factors changing over time or static variation between countries. Isolation of this component of the variation of DALY rates allows examination of the shifts in disease and injury patterns that would be expected purely as a function of changing sociodemographic status. As sociodemographic status rises, YLLs from diarrhoea, lower respiratory infections, neonatal causes, maternal mortality, and other infectious causes decline substantially, while at the same time, there has been a slower rise in YLDs from musculoskeletal disorders, mental and substance use disorders, neurological disorders, and diabetes, urogenital, blood, and endocrine diseases. DALY rates for neoplasms and cardiovascular disease are minimally related to sociodemographic status; instead local factors have a profound effect. However, with rising sociodemographic status, the proportion of DALYs due to these causes increases because of decreases in other causes of YLLs. Although, DALY rates for cardiovascular disease seem not to be related to sociodemographic status, large declines have been recorded for these causes in high-income countries in age-standardised rates over the past decades. The wide variation between some high sociodemographic status countries, such as Japan and Finland, shows how other factors, such as diet, physical activity, and other risks, vary substantially within the same level of sociodemographic status and also affect cardiovascular disease outcomes. Furthermore, our analysis of the epidemiological transition is based on crude population rates, so reductions in age-specific cardiovascular rates associated with rising sociodemographic status are countered by shifts towards an older population. Our analysis of the epidemiological transition shows decreases in DALY rates for cardiovascular disease at the very highest levels of sociodemographic status. Notably, the predictable rise in YLD rates for some causes (such as musculoskeletal disorders, diabetes mellitus, and mental and substance use disorders) driven by population ageing is not well recognised in the literature about the epidemiological transition.
Our decomposition of variance analysis shows that, for many NCDs, the main determinants of variation in DALY rates are country-specific effects, not the epidemiological transition. Global health can be understood in terms of a general theme in which change in epidemiological patterns is related to sociodemographic status, upon which country-specific patterns are overlaid. Little of the variation in DALY rates was attributable to the year, which contrasted with previous findings showing that the association between life expectancy and income and education has shifted over time. 24 , 36 - 38 Our analysis only covers a 23 year period, which might be too short to fully capture the effects of changing sociodemographic status. Some of the country effects, such as those noted for neglected tropical diseases and malaria, might have been related to sociodemographic status in a longer run analysis. The substantial effect of country variation on the epidemiological transition pattern reinforces the importance of estimating the burden of disease for each country individually. The GBD results can be used productively in the future to characterise the deviation of individual countries from the general epidemiological transition. Our analysis of the association between crude DALY rates and country sociodemographic status does not provide insights into within-country disease associations across individual levels of socioeconomic status. For example, findings from multiple studies have shown that individuals of lower socioeconomic status have increased rates of cardiovascular and circulatory diseases. 39 - 47
HALE varies widely between countries and over time. As a single summary measure of population health, HALE is fairly simple to explain and provides an indicator that is affected by any changes in mortality rates or prevalence of disease or injury. HALE has been proposed as an indicator for the Sustainable Development Goals. As calculated through the GBD, HALE is an attractive measure that is sensitive to intervention and comparable over time and between populations. Although HALE needs input about the prevalence of multiple sequelae, the annual revisions of the GBD provide a widely available source for regular updates. By contrast, some other variants of health expectancies might be less appropriate for intertemporal or cross-country comparison. With measures that define disability according to arbitrary thresholds of disability weight, such as disability-free life expectancy (DFLE), 48 even slight changes in the disability weight threshold will profoundly affect conclusions about levels and changes in healthy life expectancy, as can be seen in figure 1 . Moreover, these measures are non-standardised and hence not comparable: in some implementations of DFLE, the choice of the severity threshold to define disability is left to individual respondents in the surveys. For example, the European Union 49 has adopted a measure of healthy life expectancy based on survey responses to a single item on activity limitations. Such a measure is susceptible to variation in the meaning attached to categorical descriptions of limitation levels, both between individuals and over time, as seen in related survey items on health status. 50 - 52 An example of the sensitivity of DFLE measured in Europe is the reported decline in DFLE in Italy after 2004, which was caused by a change in question phrasing; we noted an increase in HALE for Italy in this study. 53 Although HALE and DFLE both use disability weights, the continuous scale of disability weights used in HALE makes it less sensitive to measurement error than are the dichotomous (zero or one) weights used in DFLE.
Our findings support those of Salomon and colleagues, 8 which showed that HALE is increasing more slowly than life expectancy: ie, as life expectancy increases, the expectation of years lived with multiple sequelae increases as well. The difference between life expectancy and HALE has increased, whereas the ratio of this difference to life expectancy has remained fairly constant. Whether or not this change should be termed an expansion of morbidity is not the issue. Rather, we saw that while health loss because of YLDs from cardiovascular and circulatory diseases and neoplasms might be decreasing, the real drivers of the difference between HALE and life expectancy are musculoskeletal disorders, mental and substance use disorders, neurological disorders, and diabetes, along with vision loss and hearing loss. Prevalence for all of these conditions is strongly age-related, and age-standardised rates for them are not declining. Even though the age of onset for mental health and substance use disorders is not strongly age-related, the prevalence of these conditions tends to rise with age. Prevalence of musculoskeletal disorders, neurological disorders, diabetes, and hearing and vision loss rise even more profoundly with age. As individuals increasingly survive to 80 years and older, the amount of time spent with a combination of these disorders increases, even though age-standardised rates have not increased over time. According to our analysis, the available therapies have not led to significant declines in age-standardised YLD rates. Very few, if any, of these disorders receive the attention they deserve in public policy discourse about health and health research priorities.
Global health progress has been driven by impressive progress in reducing age-standardised rates for a wide range of causes of death. Age-standardised YLD rates, however, are not declining. Many potential reasons exist to explain the general success for mortality and absence of success for disease prevalence. Research and development investments by funders such as the US National Institutes of Health (NIH) and the pharmaceutical industry have focused on cardiovascular diseases, neoplasms, and endocrine disorders. 54 - 59 As we report, in the early phases of the epidemiological transition, the major driver of change in disease and injury patterns is progress in the reduction of YLLs. We believe that the historical focus of health research funding on causes of YLLs was probably appropriate. However, health progress now means that more research investment is needed for the disorders that debilitate, rather than kill. With each passing year, the shift towards YLDs as the leading causes of disease burden will be more evident. Action is needed now to develop preventive, curative, and ameliorative strategies for these conditions rather than waiting until this shift is even more obvious.
Controlling for sociodemographic status, substantial variation exists for DALY rates between countries. In our decomposition of variance, the importance of intercountry variation fluctuated by cause; for example, country level variation explains little variation for neonatal causes, but most variation for self-harm and interpersonal violence. These findings raise the question of whether the division of countries into 21 regions in GBD based on geographic contiguity and the levels and patterns of adult and child mortality rates is the best way to explain the variance in DALY rates. With country-specific results now available, more sophisticated clustering methods could be used, with various constraints, to propose a more empirically calculated set of regions. Regions have two dimensions: analytical and presentational. The presentational dimension is easily addressed because results for any set of countries can easily be generated from the country-specific results. In fact, the GBD Compare data visualisation tool provides several alternative presentational groupings, such as WHO political regions or World Bank regions. However, regions have an analytical effect on the results if the super-region and region hierarchy have been used in the Bayesian modelling. In the GBD cause of death analysis, spatiotemporal Gaussian process regression models use the GBD hierarchy to borrow strength. In DisMod-MR 2.0, the regional hierarchy also affects estimation of the prior for each country analysis. More simply, where data are sparse or not available, the GBD regional groupings can have an important analytical effect. More research is needed on two fronts. First, to explore the extent to which alternative regional groupings would explain more of the variance in the DALY results (or any other GBD indicator). Second, whether analytical tools, such as cause of death ensemble modelling (CODEm) 60 or DisMod-MR 2.0, 1 could be modified to easily allow for different regional hierarchies for different causes.
This study has all the limitations previously reported for the GBD 2013 analysis of YLLs and YLDs. 1 , 2 Additionally, a key limitation is the assumption that uncertainty is independent for YLLs and YLDs. In fact, for diseases modelled with DisMod-MR 2.0, we estimated a correlation between the uncertainty in condition-specific mortality and the uncertainty in prevalence. However, DisMod-MR 2.0 estimates excess mortality related to a cause, not the mortality that would be assigned to a cause according to the ICD principles of underlying cause. In future iterations of the GBD, it might be useful to attempt to use the correlation structure produced from DisMod-MR 2.0 as a proxy for the correlation between the underlying cause and prevalence. By assuming independence, we might be underestimating overall uncertainty in DALYs. However, more careful examination of the uncertainty in YLDs reveals that most uncertainty stems from uncertainty in disability weights and not from uncertainty in prevalence. We have no reason to assume that the uncertainty associated with valuations of health states in population surveys would be correlated with either disease or death outcomes. Inclusion of the smaller component of the correlation of prevalence with mortality in the estimation of DALY uncertainty would probably not substantially alter the large uncertainty already recorded for DALYs. Another key limitation, is that our assessment of the burden of disease related to sociodemographic status was affected by the choice of variables that we have included in our sociodemographic status indicator (income per person, average years of schooling after age 15 years, total fertility rate, and mean age of the population). Alternative measures of sociodemographic status could be developed that might explain more of the variance seen for some GBD level 2 causes of DALYs. However, we experimented with alternative formulations, including the addition of urbanisation and all combinations of the four variables in our index, and established that the approach we used had the most explanatory power. Other variables that would be interesting to include in a composite country-level sociodemographic status measure, such as the Gini coefficient, were not available for all countries in all years. Notably, where both measures are available, the correlation between the sociodemographic status indicator and the UN Human Development Index during the 23 year time period was 0·95.
In the post-Millennium Development Goal (MDG) era, there is much interest in identification of appropriate goals for population health and how these goals should be monitored. 4 Increasingly, measurement frameworks to assess levels of health in populations have moved away from measures of survival to encompass more holistic measures of disability as well as mortality. Our findings suggest that this broader focus is probably going to be increasingly relevant to guide countries’ public policy interventions. Large, impressive, and sustained gains are being made against the majority of leading causes of death in most countries, but as our findings show, these gains are not being accompanied by commensurate declines in age-standardised rates of disability, especially from major musculoskeletal disorders, mental and substance use disorders, neurological disorders, and diabetes. Moreover, the failure of health information systems to reliably describe trends in these disorders not only severely hampers policy responses, but contributes greatly to their further neglect and insufficient awareness of the significant part they now play in overall population health. Despite the important country variation in measures of population, now is the time for the global health community, donors and countries alike, to maximise the opportunities for health that have resulted from the data revolution, and ensure that priority is given to the development of scientifically valid, feasible, and informative data systems to measure progress in reducing disability. Improved data and monitoring can help decision makers to reduce DALYs from a wide range of causes by pursuing the most important opportunities for prevention, treatment, and rehabilitation.
Text is read by the "Ask this paper" AI Q&A widget below. Extraction quality varies by source — PMC NXML preserves structure cleanly, OA-HTML may include some navigation residue, and OA-PDF can have broken hyphenation. The publisher copy (via DOI) is the canonical version.
My notes (saved in your browser only)
Ask this paper
Answers must be backed by verbatim quotes from this paper's full text. Hallucinated quotes are dropped automatically; if no verbatim passage answers the question, we say so. How this works
Citation neighborhood (sparse)
Too few in-corpus citations on either side for a chart; here are the lists.
Cited by (1)
Cited by (1)
Source provenance
- europepmc
- last seen: 2026-08-04T06:16:37.499272+00:00
- unpaywall
- last seen: 2026-06-02T02:00:03.124865+00:00