Menu
Menu
Menu
Menu
Background Healthy life expectancy (HALE) and disability-adjusted life-years (DALYs) provide summary measures of health across geographies and time that can inform assessments of epidemiological patterns and health system performance, help to prioritise investments in research and development, and monitor progress toward the Sustainable Development Goals (SDGs). We aimed to provide updated HALE and DALYs for geographies worldwide and evaluate how disease burden changes with development. Methods We used results from the Global Burden of Diseases, Injuries, and Risk Factors Study 2015 (GBD 2015) for all-cause mortality, cause-specific mortality, and non-fatal disease burden to derive HALE and DALYs by sex for 195 countries and territories from 1990 to 2015. We calculated DALYs by summing years of life lost (YLLs) and years of life lived with disability (YLDs) for each geography, age group, sex, and year. We estimated HALE using the Sullivan method, which draws from age-specific death rates and YLDs per capita. We then assessed how observed levels of DALYs and HALE differed from expected trends calculated with the Socio-demographic Index (SDI), a composite indicator constructed from measures of income per capita, average years of schooling, and total fertility rate. Findings Total global DALYs remained largely unchanged from 1990 to 2015, with decreases in communicable, neonatal, maternal, and nutritional (Group 1) disease DALYs offset by increased DALYs due to non-communicable diseases (NCDs). Much of this epidemiological transition was caused by changes in population growth and ageing, but it was accelerated by widespread improvements in SDI that also correlated strongly with the increasing importance of NCDs. Both total DALYs and age-standardised DALY rates due to most Group 1 causes significantly decreased by 2015, and although total burden climbed for the majority of NCDs, age-standardised DALY rates due to NCDs declined. Nonetheless, age-standardised DALY rates due to several high-burden NCDs (including osteoarthritis, drug use disorders, depression, diabetes, congenital birth defects, and skin, oral, and sense organ diseases) either increased or remained unchanged, leading to increases in their relative ranking in many geographies. From 2005 to 2015, HALE at birth increased by an average of 2·9 years (95% uncertainty interval 2·9–3·0) for men and 3·5 years (3·4–3·7) for women, while HALE at age 65 years improved by 0·85 years (0·78–0·92) and 1·2 years (1·1–1·3), respectively. Rising SDI was associated with consistently higher HALE and a somewhat smaller proportion of life spent with functional health loss; however, rising SDI was related to increases in total disability. Many countries and territories in central America and eastern sub-Saharan Africa had increasingly lower rates of disease burden than expected given their SDI. At the same time, a subset of geographies recorded a growing gap between observed and expected levels of DALYs, a trend driven mainly by rising burden due to war, interpersonal violence, and various NCDs. Interpretation Health is improving globally, but this means more populations are spending more time with functional health loss, an absolute expansion of morbidity. The proportion of life spent in ill health decreases somewhat with increasing SDI, a relative compression of morbidity, which supports continued efforts to elevate personal income, improve education, and limit fertility. Our analysis of DALYs and HALE and their relationship to SDI represents a robust framework on which to benchmark geography-specific health performance and SDG progress. Country-specific drivers of disease burden, particularly for causes with higher-than-expected DALYs, should inform financial and research investments, prevention efforts, health policies, and health system improvement initiatives for all countries along the development continuum. Funding Bill & Melinda Gates Foundation.
Detailed methods for estimating DALYs and HALE, including analytic approaches for mortality and non-fatal health loss estimation, are provided in related publications.12, 20 Additional detail on GBD metrics and definitions are found elsewhere.21 Interactive tools are also available to explore GBD 2015 results and data sources. This analysis follows the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER), which includes recommendations on documentation of data sources, estimation methods, and statistical analysis.22, 23 In brief, the GBD geographic hierarchy involves 519 total geographies within 195 countries and territories, 21 regions, and seven super-regions. This study reports results for all countries and territories. The GBD cause hierarchy has four levels of classification and causes reported within each level that are mutually exclusive and collectively exhaustive. The full GBD cause list with corresponding International Classification of Diseases (ICD)-9 and ICD-10 codes are available in our publications on cause-specific mortality12 and non-fatal health outcomes.20 We estimated all-cause and cause-specific mortality with a multistep computation process, which included systematically addressing known data challenges, such as different coding schemes, different age-group reporting, variation in certification, misclassification of HIV/AIDS deaths in some countries, misclassification of maternal HIV/AIDS deaths, and incorporation of population-based cancer registry data, before computation of cause-specific mortality with analytic tools such as Cause of Death Ensemble Modelling (CODEm). Each death could have only one underlying cause. Additional detail, including model specifications and data availability for each cause-specific model, can be found in the supplementary material of the GBD 2015 mortality and causes of death publication.12 We calculated normative life tables based on the lowest death rates for each age group among geographies with total populations greater than 5 million. We computed cause-specific YLLs by multiplying cause-specific deaths by the life expectancy at the age of death (ie, 86·59 years at age 0 years; 23·79 years at age 65 years) from this normative life table, and then used the GBD world population age standard to calculate age-standardised mortality rates and YLL rates.12 Our most commonly used analytic approach to estimate non-fatal health loss was DisMod-MR 2.1, a Bayesian meta-regression tool that synthesises diverse data sources to produce internally consistent estimates of incidence, prevalence, remission, and excess mortality. The use of other methods to estimate non-fatal health loss was determined by cause-specific data availability and epidemiological characteristics.24 Additional detail, including model specifications and data availability for each cause-specific model, can be found in the supplementary material of the GBD 2015 non-fatal publication.20 Each non-fatal sequela was estimated separately. We then applied a microsimulation framework to assess the occurrence of comorbidity in each age group, sex, geography, and year separately. Disability from comorbid conditions was apportioned to each of the contributing causes. GBD disability weights were based on population surveys with more than 60 000 respondents, and previous studies show that disability weights do not significantly vary across geographies, income, or educational attainment.25, 26 In this study, disability weights are invariant over geography and time, although the distribution of sequelae, and therefore the severity and cumulative disability per case of a condition, can differ by age, sex, geography, and year. DALYs are the sum of YLLs and YLDs as estimated in GBD 2015 for each cause, geography, age group, sex, and year.12, 20 Using methods developed by Sullivan,7 we calculated HALE by age group within abridged multiple-decrement life tables and estimates of YLDs per capita for each geography–age–sex–year from 1990 to 2015.8, 10, 27 For all results, we report 95% uncertainty intervals (UIs), which were derived from 1000 draws from the posterior distribution of each step in the estimation process. UIs are distinct from confidence intervals, because confidence intervals only capture the uncertainty associated with sampling error, whereas uncertainty intervals provide a method for propagation of uncertainty from multiple sources including sampling, model estimation, and model specification. 95% UIs represent the ordinal 25th and 975th draw of the quantity of interest. For mortality and YLLs, UIs reflect uncertainty that arises from sample sizes of studies used as data sources, adjustments to sources of all-cause mortality, parameter uncertainty in model estimation, and specification uncertainty for all-cause and cause-specific models. For prevalence, incidence, and YLDs, UIs reflect the uncertainty that arises from sample sizes of studies used as data sources, data adjustments from non-reference definitions, parameter uncertainty in model estimation, and uncertainty in the disability weights. In the absence of any direction information about the correlation between uncertainty in YLLs and YLDs, we assumed uncertainty in age-specific YLDs is independent of age-specific YLLs in DALYs and death rates in HALE. We examined the relationship between DALYs, HALE, and the Socio-demographic Index (SDI).28 SDI was constructed based on the geometric mean of three indicators: income per capita, average years of schooling among people aged 15 years or older, and the total fertility rate. SDI values were scaled to a range of 0 to 1, with 0 equalling the lowest income, lowest schooling, and highest fertility rate observed from 1980 to 2015, and 1 equalling the highest income, highest schooling, and lowest fertility rate assessed during that time. The average relationships between each summary health measure and SDI were estimated using spline regressions. These regressions were used to estimate expected values at each level of SDI. Additional detail on SDI computation and geography-specific SDI values are available in the appendix (pp 4–5 and pp 74–80). The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
