Global, regional, and national disability-adjusted life-years (DALYs) for 359 diseases and injuries and healthy life expectancy (HALE) for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017

Understanding global trends in the health status of populations and changes in the leading causes of disease burden over time is crucial to tracking progress towards the Sustainable Development Goal to ensure healthy lives and promote wellbeing for all at all ages. 1 Robust assessment of these trends requires objective and comparable measures of population health that can help countries identify priorities and address challenges to achieving this goal. The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2017, the third annual update in the series, uses all available up-to-date epidemiological data and improved standardised methods to provide a comparative assessment of health loss across 359 diseases and injuries and 73 age and sex groups for 195 countries and territories. The availability of GBD 2017 data for years of life lost (YLLs) because of premature mortality and years lived with disability (YLDs) provides an opportunity to assess trends in population health over the past 28 years by analysing two complementary summary measures: healthy life expectancy (HALE), which quantifies the number of years expected to be lived in good health, and disability-adjusted life-years (DALYs), which quantifies the health loss due to specific diseases and injuries. HALE provides a snapshot of overall population health and DALYs are useful for quantifying and ranking disease burden due to specific causes. DALYs can be utilised to help decision makers and the public understand the leading causes of health burden and whether improvement occurs over time. The continuing trend of increasing life expectancy and decreasing mortality because of improvements in living conditions, income per capita, education, and medical practices is well known and understood. 2 – 5 Previous GBD papers have reported that increases in HALE have been slower than increases in life expectancy, resulting in more years of poor health, and suggesting an absolute expansion of morbidity. 6 – 9 However, details of how many of the additional years of life gained are spent in good health versus poor health across countries and sociodemographic groups have not been well characterised. As people live longer, such information becomes increasingly relevant for policy development, health systems planning, and resource allocation, the effects of which cannot be understated for population health. The estimates herein provide insight into the importance of access to services and appropriate health care, and the potential societal burden of caregiving and excess health-care expenditure for years lived in poor health. 10 In this study, we present GBD 2017 results for HALE and DALYs by age and sex from 1990 to 2017 for 195 countries and territories. GBD 2017 includes new morbidity and mortality data (epidemiological surveillance data, disease registry data, scientific literature sources, survey sources, verbal autopsy studies, vital registration systems, cancer registries, complete birth history sources, summary birth history sources, and sibling history surveys); refined methods; and new estimations at the subnational level for Ethiopia, Iran, Norway, and Russia, and stratified by ethnicity for New Zealand. Also, the disaggregation of larger cause categories (eg, diabetes) has allowed separate estimation for several additional diseases (eg, type 1 and type 2 diabetes). GBD 2017 provides a complete reanalysis of all available data by country from 1990 to 2017, and thus supersedes all previously published GBD estimations of HALE and DALYs.

The GBD study comprehensively and systematically quantifies the comparative magnitude of health loss due to diseases and injuries by age, sex, and location over time. We estimated all-cause and cause-specific mortality using the following key principles: identification of all data sources that are available, assessment of the quality of the data and correction for known bias, application of highly standardised analytical procedures, and assessment of model performance using cross-validation analysis. We used similar principles to identify, enhance comparability, and analyse data to estimate the incidence, prevalence, and YLDs of diseases and injuries. 7 Using the GBD 2017 results for YLLs and YLDs, we calculated DALYs for 359 diseases and injuries. 11 , 12 We used age-specific mortality and YLDs per person to calculate HALE, defined as the average number of years that a person at a given age can expect to live in good health, taking into account mortality and loss of functional health. 13 Additional details for computing HALE can be found in appendix 1 . We calculated years lived in poor health (ie, years lived with functional health loss) as life expectancy minus HALE. Estimations for GBD 2017 cover the period 1990 to 2017 for 195 countries and territories. We did analyses using Python versions 2.7.12 and 2.7.3, Stata version 13.1, and R version 3.2.2. For this study, we followed the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER), 14 which include recommendations on documentation of data sources, estimation methods, and statistical analysis ( appendix 1 ). Interactive online tools are available to explore GBD 2017 data sources in detail using our online sourcing tool, the Global Health Data Exchange . Data before and after adjustments and the fit of the model to the data for causes of death and non-fatal outcomes can be explored with the available data visualisation tool . In GBD 2017, as in previous GBDs, causes of mortality and morbidity are structured using a four-level classification hierarchy to produce results that are mutually exclusive and collectively exhaustive. GBD 2017 estimates 359 causes of DALYs, 77 of which are a source of disability but not a cause of death (eg, attention-deficit hyperactivity disorder, headache disorders, low back pain, and neck pain), and five of which are causes of death but not sources of morbidity (sudden infant death syndrome, aortic aneurysm, late maternal deaths, indirect maternal deaths, and maternal deaths aggravated by HIV/AIDS). In the GBD hierarchy, the number of mutually exclusive and collectively exhaustive fatal and non-fatal causes in each level for which GBD estimates is three at Level 1, 22 at Level 2, 169 at Level 3, and 293 at Level 4. The full GBD cause hierarchy, including corresponding International Classification of Diseases (ICD)-9 and ICD-10 codes and detailed cause-specific methods, is in GBD 2017 publications on cause-specific mortality 11 and non-fatal health outcomes 12 in the corresponding appendices. GBD 2017 includes 195 countries and territories that are grouped into 21 regions on the basis of epidemiological similarities and geographical proximity. 15 For the purposes of statistical analyses, we further grouped regions into seven super-regions (central Europe, eastern Europe, and central Asia; high income; Latin America and Caribbean; north Africa and Middle East; south Asia; southeast Asia, east Asia and Oceania; and sub-Saharan Africa). Each year, GBD includes subnational analyses for a few new countries and continues to provide subnational estimates for countries that were added in previous cycles. Subnational estimation in GBD 2017 includes five new countries (Ethiopia, Iran, New Zealand, Norway, and Russia) and countries previously estimated at subnational levels (GBD 2013: China, Mexico, and the UK [regional level]; GBD 2015: Brazil, India, Japan, Kenya, South Africa, Sweden, and the USA; and GBD 2016: Indonesia and the UK [local government authority level]). All analyses are at the first level of administrative organisation within each country except for New Zealand (by Māori ethnicity), Sweden (by Stockholm and non-Stockholm), and the UK (by local government authorites). All subnational estimates for these countries were incorporated into model development and evaluation as part of GBD 2017. To meet data use requirements, we present all subnational estimates excluding those pending publication (Brazil, India, Japan, Kenya, Mexico, Sweden, the UK, and the USA); these results are presented in appendix tables and figures ( appendix 2 ). Subnational estimates for countries with populations larger than 200 million people (as measured according to our most recent year of published estimates) that have not yet been published elsewhere are presented wherever estimates are illustrated with maps but are not included in data tables. We estimated age-specific mortality using data from vital registration systems, sample registration systems, household surveys, censuses, and demographic surveillance sites. 13 We estimated cause-specific mortality and YLLs using the GBD cause of death database, composed of vital registration and verbal autopsy data, survey and census data for injuries and maternal mortality, surveillance data for maternal and child mortality, cancer registries, and police records for interpersonal violence and road injuries. 11 The quality and comparability of the cause of death data were evaluated and improved through several steps, including adjustment of data from vital registration systems for incompleteness, conversion of causes found in the original data to the GBD 2017 cause list, and redistribution of deaths assigned to ICD codes that cannot be underlying causes of death. Detailed methods for each step are available in the appendix of the GBD 2017 causes of death paper. 11 We estimated cause-specific mortality using standardised modelling processes, most commonly the Cause of Death Ensemble model (CODEm), which uses a covariate selection algorithm to generate several plausible combinations of covariates that are then run through four model classes—namely, mixed effects linear models and spatiotemporal Gaussian process regression models for cause fractions and death rates. For a given cause, we categorised covariates into three groups on the basis of the following criteria: evidence of proximal or causal association (Level 1), strong evidence for an association but without adequate evidence of a causal link (Level 2), and covariates that are distal in the causal pathway and therefore might be mediated by other factors in Levels 1 or 2 (Level 3). 16 The programme then selects an ensemble of models that performs best on out-of-sample predictive validity tests for each cause of death. Ensemble models have been shown to produce smaller errors in estimated cause-specific mortality and more accurate trends than single-component models. 16 Additional detail, including model specifications and data availability for each cause-specific model, can be found in the appendices of the GBD 2017 causes of death 11 and mortality 13 publications. We calculated YLLs from the sum of each death multiplied by the standard life expectancy at each age. The standard life expectancy was taken from the lowest observed risk of death for each 5-year age group in all populations greater than 5 million people. For consistency across all fatal and non-fatal estimates in GBD 2017, we calculated our own population and fertility estimates. 17 We then used the GBD world population age standard to calculate age-standardised rates for cause-specific deaths and YLLs. The GBD world population age standard and the standard life expectancies are available in the appendix of the GBD 2017 mortality publication. 13 Changes we have implemented since GBD 2016 for cause-specific mortality include the addition of important sources of new mortality data (detailed at the beginning of this section) and the expansion of the GBD location hierarchy, refinements in the calculation of Socio-demographic Index (SDI), and disaggregation of specific causes into subgroupings to provide additional detail. We estimated the following specific causes separately for the first time: invasive non-typhoidal salmonella disease; liver cancer due to non-alcoholic steatohepatitis (NASH); cirrhosis due to non-alcoholic steatohepatitis; myelodysplastic, myeloproliferative, and other haemopoietic neoplasms; benign and in-situ intestinal neoplasms; benign and in-situ cervical and uterine neoplasms; other benign and in-situ neoplasms; subarachnoid haemorrhage; non-rheumatic valvular heart disease; non-rheumatic calcific aortic valve disease; non-rheumatic degenerative mitral valve disease; other non-rheumatic valve diseases; gastro-oesophageal reflux disease; type 1 diabetes; type 2 diabetes; chronic kidney disease due to type 1 diabetes; chronic kidney disease due to type 2 diabetes; poisoning by carbon monoxide; and poisoning by other means. Specific data sources are available in the appendices of the GBD 2017 non-fatal diseases and injuries 12 and causes of death 11 publications. Additional information on data sources used can be found in our online source tool. For estimation of non-fatal health loss, we most commonly used the Bayesian meta-regression tool DisMod-MR 2.1, which synthesises variable data sources to produce internally consistent estimates of incidence, prevalence, remission, and excess mortality. 18 If DisMod-MR 2.1 did not capture the complexity of the disease, or if incidence and prevalence needed to be calculated from other data, we used custom models; detailed methods for each cause are in the appendices of the GBD 2017 non-fatal diseases and injuries publication. 12 We estimated each non-fatal sequela separately and assessed the occurrence of comorbidity for each age group, sex, location, and year separately using a micro-simulation framework. 12 Disability estimated for co-morbid conditions was distributed to each contributing cause during the comorbidity estimation process. Although the distribution of sequelae and the severity and cumulative disability per case of a condition might be different by age, sex, location, and year, previous studies have found that disability weights do not substantially vary between locations, income per capita, or levels of educational attainment. 19 , 20 Additional details, including model specifications, data availability, data adjustments to enhance comparability for each cause-specific model, and the development of disability weights by cause and their use in the estimation of non-fatal health loss, are available in the appendices of the GBD 2017 non-fatal diseases and injuries publication. 12 To calculate HALE, we used the following inputs from GBD 2017: age-specific mortality rates; estimates of the prevalence of sequelae by age, sex, location, and year; and disability weights for all unique health states. We used the method originally developed by Sullivan 21 to estimate HALE ( appendix 1 ). We calculated DALYs as the sum of YLLs 11 and YLDs 12 for each location, year, age group, and cause, by sex. We calculated 95% uncertainty intervals (UIs) on the basis of 1000 draws from the posterior distribution of each step in the estimation process using the 2·5th and 97·5th percentiles of the ordered 1000 values. We attributed the uncertainty associated with estimation of mortality and YLLs to multiple sources, including sample size variability in data sources, adjustment and standardisation methods applied to data, and model specifications. We attributed the uncertainty associated with estimation of YLDs to sampling error of data inputs, adjustment and standardisation methods applied to data, the uncertainty in coefficients from model fit, and the uncertainty of severity distributions and disability weights. The SDI is the geometric mean of three rescaled components: total fertility rate under age 25 years (ie, the number of births expected per woman aged 10–24 years), lag-distributed income per capita, and average educational attainment in populations aged 15 years or older. The methods we used to calculate the SDI are in appendix 1 . SDI scores were scaled from 0 (lowest income, fewest years of schooling, and highest fertility) to 1 (highest income, most years of schooling, and lowest fertility). We estimated the association between SDI and cause-specific mortality using a generalised additive model with a Loess smoother on SDI; we then used this association to calculate expected YLLs. Expected YLDs were calculated on the basis of the relationship between SDI and YLD rates. We then calculated expected DALYs as the sum of expected YLLs and YLDs, and expected HALE using expected YLDs and expected life tables. All results are available both in appendix 2 and through our online visualisation tool. 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 full access to all the data in the study and had final responsibility for the decision to submit for publication.

Globally, life expectancy at birth for both sexes combined increased by 7·4 years (95% UI 7·1–7·8), rising from 65·6 years (65·3–65·8) in 1990 to 73·0 years (72·7–73·3) in 2017 ( appendix 2 ). The increase in life expectancy at birth was 7·3 years (6·9–7·7) for males (from 63·2 years [62·9–63·4] in 1990 to 70·5 years [70·1–70·8] in 2017) and 7·5 years (7·2–7·9) for females (from 68·0 years [67·8–68·3] in 1990 to 75·6 years [75·3–75·9] in 2017; table 1 ). During the same period, global HALE at birth for both sexes combined increased by 6·3 years (5·9–6·7), from 57·0 years (54·6–59·1) in 1990 to 63·3 years (60·5–65·7) in 2017 ( appendix 2 ). The increase in HALE at birth was 6·2 years (5·8–6·7) for males (from 55·6 years [53·5–57·5] in 1990 to 61·8 years [59·4–64·0] in 2017) and 6·4 years (6·0–6·8) for females (from 58·4 years [55·7–60·8] in 1990 to 64·8 years [61·7–67·4] in 2017; table 1 ). The number of years lived in poor health from birth increased globally by 1·1 years (95% UI 0·9–1·4), in the low SDI countries by 1·5 years (1·2–1·9), and in the high SDI countries by 1·3 years (1·0–1·7; appendix 2 ). The increases from 1990 to 2017 in life expectancy and HALE varied across SDI quintiles and countries, with the greatest increases in life expectancy and HALE seen in the low SDI quintile and the smallest increases seen in the high SDI quintile. The increase in life expectancy varied from 5·1 years (5·0–5·3) in high SDI countries to 12·0 years (11·3–12·8) in low SDI countries. HALE increased for both sexes in low SDI countries by 10·5 years (95% UI 9·8–11·2; by 10·1 years [9·3–10·9] for males and by 10·9 years [10·1–11·7] for females), and in high SDI countries by 3·8 years (3·4–4·1; by 4·4 years [4·0–4·8] for males and by 3·2 years [2·9–3·5] for females; table 1 , figure 1 ; appendix 2 ). Increases in HALE in the middle SDI quintiles for both sexes follow a gradient, with increases of 6·8 years (6·3–7·4) for low-middle SDI, 5·7 years (5·3–6·2) for middle SDI, and 5·5 years (5·1–5·8) for high-middle SDI ( figure 1 ; appendix 2 ). Increases in HALE at birth between 1990 and 2017 were smaller than increases in life expectancy over the same period in all SDI quintiles, resulting in an increase in years lived in poor health. Of the additional years of life expected at birth, 26·3% (20·1–33·1) were spent in poor health in high SDI countries compared with 11·7% (8·8–15·1) in low-middle SDI countries ( figure 2 ). The largest increase in years lived in poor health was for females in low SDI quintile countries, by 1·6 years (1·2–2·1), and the smallest increase was for males in middle SDI quintile countries, by 0·8 years (0·6–1·0; figure 1 ). Although males and females in all SDI quintiles had increases in HALE at birth, the increase was disparate between the sexes. In 1990, high SDI quintile countries had the largest gap in HALE between males and females, at 4·1 years (3·4–4·8; figure 3 ); this gap decreased to 2·9 years (2·3–3·5) by 2017. In 2017, the high-middle SDI quintile countries had the largest gap in HALE between males and females (3·8 years [3·1–4·5]); this gap had increased from 1990 to peak in 1995 at 4·8 years (4·5–5·2) and then decreased gradually thereafter. Low SDI quintile countries had the smallest gap in HALE between males and females in 2017 ( figure 3 ). The difference in HALE at birth between the sexes during the period 1990–2017 increased in countries in the low SDI, low-middle SDI, and middle SDI quintiles. South Asia was the only region where males had higher HALE at birth than females did in 1990, and this trend lasted until 2000, after which point females had higher HALE at birth ( appendix 2 ). In all other regions, females had higher HALE at birth than males did in all study years. Eastern Europe had the biggest difference in HALE at birth between females and males throughout the period 1990–2017, with a difference of 7·7 years (6·8–8·4) in 2017. North Africa and the Middle East had the smallest difference in HALE at birth between females and males in 1990, and south Asia had the lowest difference in 2017 ( appendix 2 ). At the country level, countries that had the largest increases in HALE at birth for both sexes combined between 1990 and 2017 were Eritrea (23·7 years [95% UI 21·9–25·6], Ethiopia (18·5 years [17·2–19·9], and Uganda (16·3 years [13·5–19·0]). The smallest increases were seen in Saint Vincent and the Grenadines (1·0 years [0·4–1·7]), South Africa (1·1 years [0·3–1·8]), and Uzbekistan (1·4 years [0·4–2·4]). Decreases in HALE at birth for both sexes combined occurred in Lesotho (5·0 years [3·1–6·8]), Swaziland (eSwatini; 2·3 years [0·4–4·3]), and Guam (1·2 years [0·5–1·9]). In 2017, the highest HALE at birth was estimated to be in Singapore for both females (75·8 years [72·4–78·7]) and males (72·6 years [69·8–75·0]; table 1 ). The lowest HALE at birth in 2017 for both males and females was estimated to be in the Central African Republic, at 47·0 years (43·7–50·2) for females and 42·8 years (40·1–45·6) for males. The largest differences in HALE at birth in 2017 between males and females were in Ukraine [8·7 years (7·5–9·8), Lithuania (7·8 years [6·6–8·9]), and Russia (7·4 years [6·6–8·1]). Globally, HALE at age 65 years for females was 11·9 years (95% UI 10·8–12·9) in 1990 and 13·7 years (12·4–14·9) in 2017, an increase of 1·8 years (1·6–2·0; table 2 ). For males, HALE at age 65 years increased by 1·7 years (1·5–1·9), from 10·3 years (9·4–11·1) to 12·0 years (10·9–12·9). At the SDI level, both males and females in the high SDI quintile had the largest increases in HALE at age 65, increasing from 13·9 years (12·7–15·1) in 1990 to 16·2 years (14·6–17·6) in 2017 for females, and increasing from 11·3 years (10·3–12·1) in 1990 to 13·8 years (12·6–15·0) in 2017 for males. The countries with the largest increases in HALE for females at age 65 years from 1990 to 2017 were the Maldives (by 7·0 years [6·3–7·8]), Singapore (by 4·9 years [4·4–5·5]), and Ethiopia (by 4·6 years [4·0–5·3]). For males, the largest increases were in Iraq (by 4·7 years [3·8–5·6]), the Maldives (by 4·5 years [4·0–5·1]), and South Korea (by 4·5 years [4·0–5·0]). In 2017, the largest difference between HALE at age 65 years between females and males was in Bermuda (17·7 years [16·2–19·2] for females, and 13·4 years [12·4–14·5] for males). Although life expectancy at birth increased for both males and females globally during the period 1990–2017, the proportion of additional years spent in poor health was 15·3% (95% UI 11·8–19·2) of the additional 7·5 years (7·2–7·9) for females and 14·9% (11·3–19·0) of the additional 7·3 years (6·9–7·7) for males, with substantial variations across SDI quintiles, regions, and countries ( figure 2 ). The proportion of additional years spent in poor health varied from 11·1% (8·3–14·3) of the additional 8·2 years (7·5–8·8) among females in low-middle SDI countries, to 27·3% (20·8–34·4) of the additional 4·4 years (4·2–4·6) among females in high SDI countries. From 1990 to 2017, life expectancy at birth increased the most for both males and females in eastern sub-Saharan Africa compared with other regions ( table 1 ), but of the additional years gained, 11·1% (8·3–14·3) of 13·8 years (12·8–14·8) for males and 13·1% (10·0–16·7) of 14·6 years (13·7–15·5) for females were spent in poor health. At the country level, 179 countries gained additional years of life from 1990 to 2017, and 16 did not ( appendix 2 ). The most additional years of life were gained by males in Eritrea (28·6 years [25·7–31·4]), Ethiopia (21·1 years [19·5–22·8]), and Equatorial Guinea (18·6 years [15·1–21·8]). Of these years gained, the proportion spent in poor health was 11·3% (8·0–14·9) for Eritrea, 12·1% (9·2–15·6) for Ethiopia, and 13·7% (9·4–18·7) for Equatorial Guinea. Females gained the most additional years of life in Eritrea (22·4 years [19·7–25·4]), Ethiopia (21·5 years [20·0–23·2]), and Rwanda (19·5 years [17·1–22·1]). Of these years gained, the proportion spent in poor health was 10·4% (7·1–14·6) for Eritrea, 13·6% (10·2–17·4) for Ethiopia, and 16·1% (11·8–21·1) for Rwanda. Males spent over a third of the years of life gained in poor health in Belize (45·9% [5·5–155·6]), Ukraine (39·3% [14·1–128·7]), the Federated States of Micronesia (34·2% [7·0–55·9]), and the USA (33·6% [25·1–42·2]). Similarly, females spent over a third of their years of life gained in poor health in the USA (47·9% [35·2–61·6]), Northern Mariana Islands (40·7% [25·9–70·7]), and the UK (34·3% [25·8–43·6]; appendix 2 ). Globally, between 1990 and 2017, life expectancy for females at age 65 years increased by 2·5 years (95% UI 2·4–2·6) and they spent 27·6% (21·2–35·0) of those years in poor health ( appendix 2 ). Similarly, males at age 65 years gained 2·4 years (2·2–2·5) of life and spent 26·6% (20·1–34·0) of those years in poor health. Between 1990 and 2017, at the SDI quintile level, females in the high SDI quintile spent the largest proportion of years gained in poor health (29·9% [22·9–37·5] of 3·2 years [3·1–3·4] gained), and females in the low-middle SDI quintile spent the smallest proportion of years gained in poor health (23·8% [17·1–31·6] of 2·0 years [1·7–2·3] gained; appendix 2 ). Similarly, males in the high SDI quintile spent the largest proportion of additional years in poor health in the same period (30·0% [22·9–37·8] of 3·7 years [3·6–3·8] gained), and males in the low-middle SDI quintile spent the smallest proportion of additional years in poor health (23·7% [17·2–30·9] of 1·5 years [1·2–1·8] gained). At the country level, life expectancy between 1990 and 2017 for females at age 65 years increased the most in the Maldives (9·3 years [8·6–9·9]), Singapore (6·4 years [5·9–6·8]), and Iraq (6·2 years [5·0–7·4]), with the proportion of years spent in poor health being 24·4% (18·2–31·3) for those in the Maldives, 23·2% (17·4–29·7) for those in Singapore, and 28·5% (21·3–36·3) for those in Iraq ( appendix 2 ). For males at age 65 years, life expectancy increased the most for Iraq (6·8 years [5·6–7·9]), South Korea (6·1 years [5·6–6·6]), and the Maldives (5·9 years [5·3–6·4]), with the proportion of those years spent in poor health being 30·3% (22·9–38·2) for Iraq, 26·1% (19·6–33·1) for South Korea, and 22·5% (16·6–29·4) for the Maldives ( appendix 2 ). In 2017, females had longer life expectancy than males in 180 of 195 countries ( appendix 2 ). Of these countries, the largest differences (>10 years) between female and male life expectancy at birth were in Ukraine (76·5 years [95% UI 75·8–77·2] for females and 64·7 years [63·9–65·4] for males), Lithuania (80·2 years [79·4–81·0] for females and 69·6 years [68·7–70·5] for males), Russia (77·2 years [77·1–77·4] for females and 66·8 years [66·6–66·9] for males), and Swaziland (eSwatini; 65·1 years [62·1–68·4] for females and 54·9 years [52·6–57·6] for males). For the other 15 countries, life expectancy was not significantly different between the sexes. Globally, in 2017, females lived an additional 5·1 years (95% UI 4·9–5·3) compared with males, and of those extra years 57·5% (46·2–67·4) were spent in good health and 42·5% (32·6–53·8) in poor health ( figure 4 ). The proportion of additional years spent in poor health varied a lot across SDI quintiles, regions, and countries. Compared with all other SDI quintiles, females in the high-middle SDI quintile lived the most extra years of life compared with males, living longer than males by 6·1 years; however, 37·8% (29·1–47·6) of those extra years were lived in poor health. At the regional level, the proportion of extra years lived in poor health among females in 2017 ranged from 21·9% (16·2–28·4) in central Europe to 92·2% (87·2–96·2) in south Asia ( figure 4 ). Extra years lived by females varied from 11·9 years (10·9–12·9) in Ukraine to 1·4 years (0·6–2·3) in Algeria ( appendix 2 ). Of the extra years lived, the proportion spent in poor health varied largely at the country level, with less than 20% of the additional years spent in poor health for females in Bosnia and Herzegovina, Burundi, and Slovakia, whereas females lived all extra years in poor health in Bahrain (1·6 years [0·2–3·1]; appendix 2 ). Globally, at age 65 years, women lived an additional 2·7 years (95% UI 2·6–2·9) compared with men, but 63·9% (55·2–72·2) of those years were spent in good health and 36·1% (27·8–44·8) in poor health ( appendix 2 ). Compared with other SDI quintiles, the difference in life expectancy was greatest between females and males in the high SDI quintile, with females living 3·3 years (3·1–3·5) longer than males, but 28·4% (21·3–36·1) of those years were spent in poor health. In the low SDI quintile, females are estimated to live the largest proportion of extra years in poor health compared with other SDI quintiles (58·0% [43·8–74·6] of 1·1 years [0·7–1·4]). For females aged 65 years in Iran and India, of the extra years lived compared with males, more than 60% of extra years were spent in poor health, 70·1% (53·5–88·2) for those in Iran and 63·4% (49·0–79·6) for those in India ( appendix 2 ). The global level of all-age all-cause DALYs in 2017 was 2·50 billion (95% UI 2·29–2·74). The estimated DALY counts and age-standardised rate of DALYs, and changes in these metrics between 2007 and 2017, are in table 3 . The largest contribution to global DALYs was from non-communicable diseases, which, combined, accounted for 62·0% (60·3–63·8) of total DALYs, whereas communicable, maternal, neonatal, and nutritional (CMNN) causes accounted for 27·9% (26·4–29·4), and injuries 10·1% (9·7–10·5) of total DALYs. For non-communicable diseases, global DALYs increased by 40·1% (95% UI 36·8–43·0), although age-standardised DALY rates decreased by 18·1% (16·0–20·2). Between 2007 and 2017, 12 causes at Level 2 of the GBD cause hierarchy (diabetes and kidney diseases, sense organ diseases, neurological disorders, neoplasms, musculoskeletal disorders, substance use disorders, cardiovascular diseases, chronic respiratory diseases, skin and subcutaneous diseases, mental disorders, digestive diseases, and self-harm and interpersonal violence) had significant increases in DALY counts. The greatest number of DALYs among Level 3 non-communicable disease causes in 2017 were estimated for ischaemic heart disease (170 million [167–174]), stroke (132 million [126–137]), and chronic obstructive pulmonary disease (81·6 million [76·0–86·8]) which, combined, accounted for 15·4% (14·2–16·4) of all-cause DALYs ( table 3 ). By contrast with the overall decrease in DALYs during 2007–17 for many causes, total all-age DALYs from non-communicable diseases increased from 1·11 billion (1·00–1·23) in 1990 to 1·34 billion (1·19–1·51) in 2007, and continued to increase to 1·55 billion (1·38–1·75) in 2017 ( appendix 2 ): a change of 20·7% (18·4–22·8) from 1990 to 2007, and 16·0% (15·1–16·9) from 2007 to 2017. At Level 4 between 1990 and 2007, the largest significant increases in total DALYs were estimated for HIV/AIDS and multidrug-resistant tuberculosis co-infection without extensive drug resistance (5329·0% [3143·5–8492·9]), multidrug-resistant tuberculosis without extensive drug resistance (700·6% [335·0–1,431·6]), and HIV/AIDS resulting in other diseases (454·3% [389·5–529·2]; appendix 2 . From 2007 to 2017, the largest significant increases in DALYs were for liver cancer due to NASH (37·4% [32·8–42·9]), chronic kidney disease due to type 2 diabetes (34·3% [30·9–37·2]), and diabetes type 2 (34·0% [30·3–38·1]). Relative to changes in the population, the age-standardised DALY rate for non-communicable diseases decreased by 13·2% (11·7–14·7), from 24 017·4 DALYs (21 861·4–26 509·6) per 100 000 in 1990 to 20 852·2 DALYs (18 710·2–23 388·5) per 100 000 in 2007, and decreased by a further 5·6% (4·8–6·6) to 19 676·5 DALYs (17 509·8–22 177·9) per 100 000 in 2017. From 1990 to 2017, the greatest significant decreases in age-standardised DALY rates were for visceral leishmaniasis (97·8% [97·0–99·2]), maternal haemorrhage (78·7% [73·4–83·0]), and ascariasis (70·8% [66·0–75·3]). Significant increases in age-standardised DALY rates from 1990 to 2017 were observed for 17 Level 4 non-communicable diseases, with other benign and in-situ neoplasms (35·1% [7·8–81·7]), type 2 diabetes (28·2% [24·1–32·3]), and opioid use disorders (24·8% [20·1–31·0]) having the most pronounced changes. Among CMNNs, at Level 3 of the GBD cause hierarchy the greatest contributors to global DALYs in 2017 were neonatal disorders (186 million [95% UI 175–197] DALYs), lower respiratory infections (106 million [100–112] DALYs), and diarrhoeal diseases (81·0 million [70·1–97·2] DALYs). Globally, total DALYs from all CMNNs decreased from 1·19 billion (1·14–1·23) in 1990 to 0·946 billion (0·910–0·987) in 2007, to 0·696 billion (0·660–0·740) in 2017, a decrease of 41·3% (38·8–43·5) from 1990 to 2017. In parallel, the age-standardised DALY rate for all CMNN causes decreased by 49·8% (47·9–51·6) from 1990 to 2017. For 2007–17, the largest decreases were for visceral leishmaniasis (by 63·9% [40·0–84·5]), HIV/AIDS and drug-susceptible tuberculosis co-infection (by 54·9% [51·0–57·9]), and maternal haemorrhage (52·4% [44·3–59·5]; table 3 ). Significant decreases of more than 50% in age-standardised DALY rates between 2007–17 were estimated for five Level 4 CMNN causes, specifically, visceral leishmaniasis (66·2% [44·3–86·0]), HIV/AIDS and drug-susceptible tuberculosis co-infection (59·8% [56·3–62·5]), HIV/AIDS and multidrug-resistant tuberculosis without extensive drug resistance co-infection (56·6% [40·1–69·1]), maternal haemorrhage (56·4% [49·1–62·8]), and HIV/AIDS resulting in other diseases (53·7% [51·1–56·1]). Against this general decreasing trend, seven CMNN causes at Level 4 were estimated to have Significant increases in age-standardised DALY rate between 1990 and 2017, specifically, HIV/AIDS and multidrug-resistant tuberculosis without extensive drug resistance co-infection (1805·0% [1016·3–3049·2]), HIV/AIDS resulting in other diseases (103·1% [76·3–134·9]), and cutaneous and mucocutaneous leishmaniasis (40·5% [7·0–106·0]). In 2017, the top five leading causes of DALYs were communicable diseases (lower respiratory infections, malaria, diarrhoeal diseases, HIV/AIDS, and tuberculosis) and neonatal disorders ( table 3 ). For low-middle SDI countries, both communicable and non-communicable diseases ranked highly as leading causes of DALYs, and the top five causes were neonatal disorders, lower respiratory infections, ischaemic heart disease, diarrhoeal diseases, and stroke ( figure 5 ). In high SDI countries, non-communicable diseases (ischaemic heart disease, low back pain, stroke, lung cancer, and chronic obstructive pulmonary disease) were the leading causes of DALYs ( figure 5 ). Younger age groups contributed more to total DALY counts in low SDI countries than did older age groups, and mostly from communicable diseases. In low SDI countries in 2017, 136 million (95% UI 130–143) DALYs were estimated for neonatal age groups to be from communicable diseases, with a further 18·9 million (17·2–21·0) from non-communicable diseases, and 2·95 million (2·60–3·26) from injuries. For low and low-middle SDI countries, the burden is more equal across age groups than in other SDI quintiles. The burden from DALYs in high-middle and high SDI countries occurs mostly from age 15 years and consists largely of non-communicable diseases ( figure 5 ). Globally in 2017, the leading causes of DALYs were predominantly CMNN and non-communicable causes for both men and women ( figure 6 ). Of the top 30 causes of DALYs for men, four were injuries: road injuries (49·8 million [95% UI 47·3–52·1] DALYs), self-harm (22·9 million [20·9–24·0] DALYs), falls (21·0 million [17·6–24·9] DALYs), and interpersonal violence (19·8 million [17·8–21·0] DALYs; table 3 ). By contrast, two of the top 30 causes of DALYs for women were injuries: road injuries (18·0 million [16·6–19·4] DALYs) and falls (15·0 million [12·3–18·0] DALYs). Our results showed disparities in disease burden between males and females. Females are more likely than males to have a higher burden from disabling conditions such as most musculoskeletal disorders except for gout, iron-deficiency anaemia, and major depressive disorder. Iron-deficiency aneaemia is common especially at reproductive ages for females, and for boys and girls (aged 5–14 years) equally ( appendix 2 ). Males are more likely than females to be affected by fatal conditions including different types of cancer, injuries, and ischaemic heart disease. Globally, the all-age DALY count of HIV/AIDS for females increased sharply from 1990 to 2007, rising from 36th to third leading cause of DALYs with an increase of 610·7% (548·5–680·6), then decreased from 2007 to 2017 by 53·9% (51·6–53·9), dropping to 12th leading cause of DALYs ( figure 6 ). Similarly, the global age-standardised DALY rate for females increased from 1990 to 2007 by 483·0% (431·4–540·5), and then decreased from 2007 to 2017 by 58·8% (56·8–60·6; appendix 2 ). Males had similar patterns in all-age DALY counts and age-standardised DALY rates for HIV/AIDS during the same period. The all-age DALY count of HIV/AIDS for males increased by 297·3 (259·8–338·4) from 1990 to 2007, rising from the 25th leading cause of DALYs to seventh ( figure 6 ; appendix 2 ). By 2017, all-age DALYs from HIV/AIDS for males dropped to 13th leading cause of DALYs, a decrease of 44·8% (42·8–46·6) between 2007 and 2017. Similarly, the age-standardised DALY rate of HIV/AIDS for males increased sharply from 1990 to 2007, with an increase of 212·9% (182·8–246·4), and then decreasing by 51·1% (49·0–52·6) in 2017. The global all-age DALY count of malaria for females increased from 1990 to 2007 by 28·6% (11·7–51·2), rising from the tenth leading cause of DALYs to the seventh leading cause of DALYs. Malaria age-standardised DALY rates for females also increased, with an increase of 23·2% (7·4–44·2) between 1990 and 2007, and then decreased from 2007 to 2017 by 40·2% (30·7–49·2). DALYs from malaria for males also increased from 1990 to 2007, increasing by 30·5% (12·9–55·7) from the 11th to the tenth leading cause of DALYs. Similarly, the age-standardised DALY rate for malaria in males increased between 1990 and 2007 by 23·6% (7·0–46·0), and then decreased by 37·7% (27·6–46·8) in 2017. The leading 30 causes of DALYs for men and women varied between SDI quintiles ( appendix 2 ). In the low SDI quintile countries, males and females had similar CMNN causes as the leading causes of DALYs in 2017, with the greatest total DALYs for both sexes being from neonatal disorders, lower respiratory infections, diarrhoeal diseases, and malaria for both sexes, and thereafter for females the fifth leading cause was congenital defects and for males was ischaemic heart disease. In the low SDI quintile, males had more injury sources of DALYs than females, whereas CMNN causes of DALYs were more common among the leading DALY causes for females than males. An epidemiological shift in sources of DALYs was evident in the leading causes for both men and women at middle to high SDI levels, which were dominated by non-communicable disease causes. In high SDI quintiles, two CMNN causes of DALYs were in the leading 30 causes for both men and women, lower respiratory infections and neonatal disorders, and the age-standardised DALYs rate for these causes decreased between 1990 and 2017, by 36·2% (95% UI 33·9–38·6) for lower respiratory infections and 30·1% (25·1–36·7) for neonatal disorders for men, and by 37·8% (35·8–39·9) for lower respiratory infections and 27·5% (22·3–34·1) for neonatal disorders for women ( appendix 2 ). Globally in 2017, males had 1·34 billion (95% UI 1·24–1·46) DALYs and females had 1·16 billion (1·04–1·29) DALYs. For males, 29·4% (23·8–34·9) of DALYs were from YLDs and 70·6% (65·1–76·2) were from YLLs. For females, 39·4% (33·0–45·6) of DALYs were from YLDs and 60·6% (54·4–67·0) were from YLLs. A greater proportion of DALYs were from YLDs for females than for males, particularly for those older than 65 years ( figure 7 ). Globally, among females YLDs were the larger fraction of DALYs for many age groups, whereas for males this was only the case for those aged 10–14 years. Across SDI quintiles, the composition of DALYs varied. Females aged 1–69 years in the high SDI quintile had more DALYs from YLDs than from YLLs, whereas in the low SDI quintile all age groups except those aged 10–29 years had more DALYs from YLLs than from YLDs. In all quintiles, women older than 65 years had more DALYs from YLDs than men older than 65 years did. Maps of observed minus expected HALE can highlight which countries have worse or better HALE than expected based on SDI quintile, as shown in figure 8 . The three countries with the greatest difference between expected HALE at birth and observed HALE at birth for males, with observed HALE much lower than expected on the basis of their SDI value, were Lesotho, with a difference of 13·4 years, Swaziland (eSwatini), with a difference of 11·7 years, and Central African Republic, with a differences of 7·9 years. For females, the countries with the greatest differences were Lesotho, with a difference of 10·0 years, Congo (Brazzaville), with a difference of 9·3 years, and Swaziland (eSwatini), with a difference of 8·1 years. The three countries that had the greatest difference between observed and expected HALE at birth for males, where HALE was much better than expected, were Niger, Nicaragua, and the Maldives, with differences of 11·9 years for Niger, 9·5 years for Nicaragua, and 9·2 years for the Maldives. For females, these countries were Niger, Nicaragua, and Ethiopia, with differences between observed and expected HALE at birth of 11·5 years for Niger, 8·3 years for Nicaragua, and 8·2 years for the Maldives. We saw a similar trend in countries having differences in observed and expected HALE at age 65 years for both males and females ( appendix 2 ). The greatest differences in observed and expected HALE at age 65 years for males, where HALE was worse than expected, were seen in Lesotho (3·1 years), Swaziland (eSwatini; 2·8 years), and Papua New Guinea (2·3 years), and for females were seen in Congo (Brazzaville; 2·7 years), Papua New Guinea (2·6 years), and Marshall Islands (2·2 years). HALE at age 65 years was better than expected on the basis of SDI level for males in Peru (by 5·0 years), Colombia (by 5·0 years), and Nicaragua (by 4·9 years), and for females in Colombia (by 4·5 years), Peru (by 4·3 years), and Nicaragua (by 3·7 years). Higher ratios of observed to expected age-standardised DALY rates indicate more DALYs experienced than expected on the basis of SDI ( appendix 2 ). Males had the highest ratio of observed to expected age-standardised DALY rates in Lesotho (1·94), Swaziland (eSwatini; 1·86), and Central African Republic (1·60). Females had the highest ratio of observed to expected age-standardised DALY rates in Congo (Brazzaville; 1·67), Lesotho (1·62), and Equatorial Guinea (1·57). Alternatively, males in the Maldives, Niger, and Nicaragua experience DALYs at a much lower rate than expected, with observed to expected ratios of 0·54 in the Maldives, 0·56 in Niger, and 0·57 in Nicaragua. Females had low ratios of observed to expected age-standardised DALY rates in Niger, Nicaragua, and Ethiopia, with ratios of 0·57 in Niger, 0·58 in Nicaragua, and 0·62 in Ethiopia.

Globally, HALE at birth increased by 6·3 years (95% UI 5·9–6·7) and the number of years lived in poor health from birth increased by 1·1 years (0·9–1·4) during the period 1990–2017, with substantial variation across socio-demographic quintiles and countries. People in low SDI countries gained an additional 10·5 years (9·8–11·2) of life in good health and 1·5 years (1·2–1·9) in poor health from birth, whereas people in high SDI countries gained an additional 3·8 years (3·4–4·1) in good health and 1·3 years (1·0–1·7) in poor health from birth. Women are expected to live longer than men (both at birth and at age 65 years) in most countries, but the number of extra years lived and the proportions of extra years spent in poor health vary greatly across countries. In 2017, the global leading causes of DALYs were neonatal disorders, ischaemic heart disease, stroke, lower respiratory infections, and chronic obstructive pulmonary disease. Communicable diseases and neonatal disorders were the leading causes of DALYs in low SDI countries in 2017. Non-communicable diseases were the leading causes of DALYs in the remaining SDI quintiles in the same year. Sex differences in HALE exist across SDI quintiles, with the smallest difference seen in low SDI countries. Although societal factors, biological factors, and men’s risk behaviours probably contribute to their shorter HALE, 22 , 23 the smaller male–female gap in HALE in low SDI countries might be explained by the increased risk of mortality in both sexes due to the high occurrence of infectious diseases. Additionally, women are affected by pregnancy-related conditions, and maternal disorders were a leading cause of DALYs in low SDI countries in 2017. In several high SDI countries, the difference between male and female HALE has decreased over time, which could partly be attributable to the decreasing gap between the sexes in the prevalence of specific risk factors—eg, smoking and alcohol use. 24 , 25 Our results show increased life expectancy and more years lived in poor health for women than men in most countries. This finding could be attributable to sex differences in the patterns of disease burden. For example, women are more likely than men to have a higher burden from disabling conditions (eg, most musculoskeletal disorders except for gout, iron-deficiency anaemia, and major depressive disorder), whereas men are more likely than women to be affected by fatal conditions including different types of cancer (eg, liver cancer, lung cancer, leukaemia, colorectal cancer, and pancreatic cancer), injuries, and ischaemic heart disease. Various explanations have been suggested for the sex difference in disease risk, including social norms (eg, heavy drinking is socially acceptable for men in Russian tradition), health-related beliefs and behaviours, and biological factors (eg, sex hormones). 22 , 26 In 2017, the widest difference between male and female HALE at birth was observed in the eastern European region (7·7 years [95% UI 6·8–8·4]), where the age-standardised DALY rate from all causes among men was also 1·7 times higher than among women. A large proportion of DALYs in men in these countries was attributable to alcohol use and smoking, which led to high incidence and prevalence of cardiovascular disease and other leading causes of DALYs. 27 Interventions aiming to reduce these risk factors could help decrease the difference. Ukraine and Lithuania had the biggest sex difference in HALE at birth globally (8·7 years [7·5–9·8] for Ukraine and 7·8 years [6·6–8·9] for Lithuania) in 2017. Age-standardised DALY rates due to cardiovascular diseases and alcohol use disorders were two times higher in both these countries for men than for women, whereas self-harm and drug use disorders were four or more times higher. With increasing life expectancy in almost all locations worldwide, the question of whether the years of life gained are spent in good health or poor health is increasingly relevant because of the associated policy implications, ranging from health-care provisions to extending retirement ages. In the context of improving longevity, the age of retirement versus extending working life has been much debated. Some key questions need to be considered in this debate, including whether the increase in life expectancy is accompanied by an equivalent increase in years in good health, and whether the common causes of early retirement, 28 , 29 such as mental disorders and musculoskeletal disorders, have improved over time. Our results showed that the additional years of life were accompanied by poor health to some extent, but varied widely across countries, and the burden from musculoskeletal disorders and mental disorders has not improved over time. A lack of progress in reducing the burden of these causes and other conditions associated with ageing, such as sense organ disorders and Alzheimer’s disease, might restrict the ability of older workers to contribute to the workforce. 30 Globally, 12 Level 2 causes have had significant increases in DALY counts since 2007. Causes with the largest increases in DALY counts include musculoskeletal disorders, sense organ diseases, neurological disorders, diabetes and kidney diseases, and neoplasms. The large and increasing number of people living with these diseases necessitates careful planning by governments and health-care providers to ensure adequate funding and staff for treatment and rehabilitation services. Some of these diseases are expensive to manage. 31 Despite the large amount of spending on health in some countries, such as India and Nigeria, the spending is inefficient because most of it is through out-of-pocket expenditures (ie, direct payments by individuals to health-care providers); additional health services could be provided if the money was consolidated. 32 Our results showed that substantial variations in health gains exist even between countries within the same SDI quintile. For example, between 1990 and 2017, the increase in HALE at birth among females in Singapore was more than six times greater than the increase among females in the USA (7·3 years [95% UI 6·6–7·9] vs 1·1 years [0·8–1·4]), and the females in Singapore spent a smaller proportion of the extra years in poor health (17·7% [12·7–23·1] vs 33·6% [25·1–42·2]). The comparatively small increase in HALE in the USA is probably due to inequalities in access to health care. 33 The burden attributable to drug use disorders has also been increasing in the USA, with the age-standardised DALY rates due to drug use disorders almost three times higher in 2017 compared with 1990. This trend coincides with a sharp increase between 2015 and 2016 in the use of and deaths from synthetic opioids such as fentanyl, suggesting the need to improve harm-reduction efforts in the USA. 34 , 35 Over the past 28 years, several countries in sub-Saharan Africa have had substantial increases in HALE. But at the same time, some countries in this region, especially those with major HIV epidemics, had a stagnation or reduction in HALE. Although HIV control efforts have resulted in a decrease in HIV-related mortality and an increase in life expectancy and HALE in the past decade, countries such as Lesotho and Swaziland (eSwatini) still have yet to catch up with the level of HALE in 1990. In addition to examining improvements in health, we compared the observed HALE and disease burden with those expected based on development status to discern which countries are lagging behind and identify priority areas that need to be addressed to further improve HALE. In 2017, several countries, most of them from southern sub-Saharan Africa and eastern Europe, had higher disease burden and lower HALE than expected on the basis of their sociodemographic development. Our study also identified countries that outperformed expectations based on their development status (eg, in Niger and the Maldives, males had more than 9 years higher HALE at birth than expected based on the countries’ SDI). The approaches used in these countries to accelerate improvements in health (eg, the child survival programme in Niger 36 ) could inform successful programmatic strategies for countries with poor performance with similar levels of development. Although child and maternal undernutrition is the main risk factor for the leading causes of disease burden in low SDI countries (eg, lower respiratory infections, diarrhoea, and neonatal disorders), obesity remains a key driver of the leading causes of burden in high-middle and high SDI countries (eg, ischaemic heart disease, ischaemic stroke, diabetes, low back and neck pain). 27 With changing diets and lifestyle over time, obesity-related diseases are emerging, especially in low-middle and middle SDI countries where diseases related to both undernutrition and obesity contribute substantially to the total burden of disease. Evidence suggests that the disease burden associated with undernutrition and obesity can potentially be reduced. 37 For example, nutritional counselling, food supplementation, and conditional cash-transfer programmes have been shown to substantially reduce stunting and associated burdens of disease in settings with food insecurity. 37 Randomised controlled trials have shown that a lifestyle intervention with a small amount of weight loss (5–7%) decreases the incidence of type 2 diabetes by 58% during a 3-year follow-up and by 34% during a 10-year follow-up. 38 – 40 However, weight maintenance after weight loss is challenging and requires continuing support to help maintain diet, physical activity, and behavioural changes. 41 For countries with a dual burden of malnutrition, an integrated approach addressing both undernutrition and obesity to simultaneously reduce the burden associated with both risk factors has been recommended. 42 , 43 Dietary iron deficiency was the fifth leading cause of DALYs among women of reproductive age and the top leading cause of DALYs among children of both sexes at ages 5–14 years in 2017. Age-standardised DALY rates due to dietary iron deficiency were similar between boys and girls aged 5–14 years, but for adult women the rates were almost double those for men. Despite its high burden, iron deficiency remains commonly under-diagnosed and undertreated. 44 Iron deficiency is not only a problem in low SDI countries, but also in high SDI countries, and reduction of iron deficiency among children and women of reproductive age, for example, is one of the objectives of the US Public Health Service’s Healthy People 2020 initiative. 45 Various options are available to treat iron deficiency, ranging from iron supplementation to biofortification of crops. 46 , 47 Although iron supplementation has been shown to be efficacious, challenges exist in terms of distribution, cost and compliance, and inability to tolerate it because of potential adverse gastrointestinal effects. 46 Food fortification is highly cost-effective but can be a challenge in rural populations with little access to marketed fortified food. 47 Biofortification of crops is considered to be a sustainable strategy to prevent iron deficiency, especially in settings with few resources; this strategy has been shown to improve iron status in iron-deficient individuals, but the longer-term effects on functional outcomes has yet to be determined. 48 A global shift has occurred from communicable to non-communicable causes of disease burden with sociodemo-graphic development. However, a few key exceptions exist among communicable diseases. In most tropical and subtropical countries, the burden of dengue has been increasing over time, in terms of both DALY counts and age-standardised DALY rates, with the highest number of DALYs occurring in south Asia. A dengue vaccine has been licensed for use in people aged 9–45 years living in several endemic settings; 49 however, the vaccine efficacy varies from 76% in seropositive individuals to only 39% in seronegative individuals. 50 In the absence of specific treatment for dengue, prevention and control measures are crucial. Top-down vector-control efforts have restricted effectiveness and sustainability, and innovative community-based vector-control strategies are being developed. 51 All-age DALY counts and age-standardised DALY rates due to HIV/AIDS have been decreasing since 2007, but for adolescents (aged 15–19 years) age-specific DALY rates continued to increase after 2007. The HIV/AIDS YLL and YLD trends among adolescents also differ from the all-age trends. The all-age YLL rate increased after 1990, peaked around 2005, and then declined steadily afterwards 11 as antiretroviral treatment became more widely available. Among adolescents, the YLL rate continued to increase after 2005, reaching a plateau since 2013. The all-age YLD rate due to HIV/AIDS has decreased slightly since 2007, but the adolescent YLDs rate has continued to increase. 12 DALYs from HIV/AIDS are concentrated in sub-Saharan Africa. Adolescents have low rates of HIV testing and poor access to antiretroviral treatment, which might partly explain some of the increases in HIV/AIDS DALYs in this age group. 52 Since 1990, we have seen exceptional progress in many countries in reducing the burden from communicable diseases, especially vaccine-preventable diseases such as tetanus and measles. Much of this burden came from premature mortality, which has sharply decreased in the past 28 years. However, despite these interventions, DALYs due to these causes remain unnecessarily high in several low, low-middle, and middle SDI countries. Immunisation efforts have been helpful, but progress in immunisation coverage has slowed in the past decade; about 20 million children younger than 1 year, most of them in sub-Saharan Africa and south Asia, did not receive the measles vaccine and the three recommended doses of the diphtheria, tetanus, and pertussis vaccine in 2016. 53 Conflict, inadequate investment in national immunisation programmes, and vaccine stock outages were among the reasons for the stalled progress in immunisation coverage. 53 We have made substantial improvements in the estimation of mortality and life expectancy in GBD 2017 across our publications, including an independent estimation of population, a comprehensive update on fertility, adding a substantial amount of new data (from censuses, Demographic Surveillance Sites, and other sources), improvements to the GBD model life-table system, and enhancements to the modelling framework. 13 We made changes to the GBD cause hierarchy, which can restrict comparisons to estimates from previous GBD iterations. For example, we have combined maternal and neonatal conditions at Level 2 of the cause hierarchy in GBD 2017 and reported these causes separately at Level 3 (rather than Level 2). Because of these changes, neonatal disorders now appear for the first time as the first leading cause of DALYs in both 1990 and 2017. Estimates of DALYs due to congenital birth defects were higher in GBD 2017 than in GBD 2016 for two reasons. 9 First, all data sources that captured only a small subset of congenital causes (eg, only congenital heart disease and neural tube defects) were excluded for GBD 2017, leading to increased mortality estimates. Second, for GBD 2017 we implemented an algorithm to empirically identify data sources for registration of birth defects and for administrations (ie, hospital and claims) that had systematic under-reporting by age, sex, and defect. This change led to exclusion of several sources and, in many cases, higher prevalence estimates. The estimated DALYs due to malaria were lower in GBD 2017 than in GBD 2016 because of newly included Plasmodium parasite rate survey data and updates to insecticide-treated bednet coverage, which decreased incidence estimates, particularly in Nigeria. The decrease in incidence was in agreement with additional verbal autopsy data for Nigeria, which resulted in a decrease in estimated YLLs and DALYs. The global age-standardised DALY rate for drug use disorders in GBD 2017 was higher for the most recent decade than in GBD 2016, because of a better fit to the most recent years of cause of death data in the USA and the use of more appropriate covariates, including the prevalence of injecting drug use and an estimate of consumption (measured by import and domestic manufacture of scheduled substances, including opioids, expressed as daily doses per 1000 people per day). Estimates of DALYs due to acute hepatitis overall are very similar to GBD 2016, but our assessments of the relative contribution of different viruses changed. We used clinical administrative data sources to calculate location-specific and age-specific case-fatality ratios for acute hepatitis infection, pairing these ratios with incidence estimates of each virus. Previously all locations were assumed to have identical case-fatality ratios. This resulted in higher estimates of DALYs for acute hepatitis A, and lower DALY estimates for acute hepatitis B, C, and E. Finally, estimated DALYs due to HIV/AIDS among children younger than 5 years have increased in the GBD 2017 study because of methodological improvements. In countries with high-quality vital registration data, incidence among children was adjusted to produce mortality estimates that better align with recorded HIV/AIDS deaths. Additionally, we produced the paediatric estimates of HIV/AIDS mortality using the CD4 progression and CD4-specific mortality rates developed by UNAIDS. 54 This study has several limitations. First, the GBD 2017 YLL and YLD analyses had limitations as described in the GBD 2017 cause of death, 11 disease and injury, 12 and mortality papers. 13 Second, we assumed that uncertainty is independent between YLLs and YLDs because little empirical evidence exists to establish this correlation; however, this assumption could result in an under-estimation of the total uncertainty for DALYs. Third, HALE and DALY estimates are influenced by the availability of data for YLL and YLD estimations. Because of time lags in the reporting of health data by countries and their subsequent incorporation into the GBD estimation, recent changes in health states might not have been captured in our estimates. The scarcity of data for a particular location is reflected by wider uncertainty intervals. Fourth, although we have included several sources of uncertainty in our estimations, we have not been able to incorporate uncertainty into the covariates used by cause of death and non-fatal models. Fifth, SDI utility might be restricted in countries with high income inequality. The applicability of SDI could be enhanced in the future by taking into account social heterogeneity within countries. Finally, time trends for specific causes, such as cancer, might be influenced by changes in diagnostic technology, whereas in previous years under-reporting might have occurred when diagnostic tests were done infrequently. Understanding trends in the health status of the global population and changes in the leading causes of disease burden over time is needed to accurately inform policies and set priorities for action. Updating the GBD study on a regular basis provides an opportunity to assess the latest evidence and monitor these trends with time to understand where interventions are having an effect and how much they have affected the disease burden. Our results showed that, globally, enormous improvements in health have occurred over the past 28 years. Nevertheless, large inequalities in HALE, years lived in poor health, and disease burden exist across SDI quintiles and countries, and between sexes. Despite the progress made in reducing the total burden of disease, hundreds of millions of DALYs could still be averted and disparities could be minimised through targeted interventions ranging from the prevention of risk factors and extended vaccine coverage, to universal access to essential health services.