Burden of disease in francophone Africa, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017

Background: Peer-reviewed literature on health is almost exclusively published in English, limiting the uptake of research for decision making in francophone African countries. We used results from the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2017 to assess the burden of disease in francophone Africa and inform health professionals and their partners in the region. Methods: We assessed the burden of disease in the 21 francophone African countries and compared the results with those for their non-francophone counterparts in three economic communities: the Economic Community of West African States, the Economic Community of Central African States, and the Southern African Development Community. GBD 2017 employed a variety of statistical models to determine the number of deaths from each cause, through the Cause of Death Ensemble model algorithm, using CoDCorrect to ensure that the number of deaths per cause did not exceed the total number of estimated deaths. After producing estimates for the number of deaths from each of the 282 fatal outcomes included in the GBD 2017 list of causes, the years of life lost (YLLs) due to premature death were calculated. Years lived with disability (YLDs) were estimated as the product of prevalence and a disability weight for all mutually exclusive sequelae. Disability-adjusted life-years (DALYs) were calculated as the sum of YLLs and YLDs. All calculations are presented with 95% uncertainty intervals (UIs). A sample of 1000 draws was taken from the posterior distribution of each estimation step; aggregation of uncertainty across age, sex, and location was done on each draw, assuming independence of uncertainty. The lower and upper UIs represent the ordinal 25th and 975th draws of each quantity and attempt to describe modelling as well as sampling error. Findings: In 2017, 779 deaths (95% UI 750–809) per 100 000 population occurred in francophone Africa, a decrease of 45·3% since 1990. Malaria, lower respiratory infections, neonatal disorders, diarrhoeal diseases, and tuberculosis were the top five Level 3 causes of death. These five causes were found among the six leading causes of death in most francophone countries. In 2017, francophone Africa experienced 53 570 DALYs (50 164–57 361) per 100 000 population, distributed between 43 708 YLLs (41 673–45 742) and 9862 YLDs (7331–12 749) per 100 000 population. In 2017, YLLs constituted the majority of DALYs in the 21 countries of francophone Africa. Age-specific and cause-specific mortality and population ageing were responsible for most of the reductions in disease burden, whereas population growth was responsible for most of the increases. Interpretation: Francophone Africa still carries a high burden of communicable and neonatal diseases, probably due to the weakness of health-care systems and services, as evidenced by the almost complete attribution of DALYs to YLLs. To cope with this burden of disease, francophone Africa should define its priorities and invest more resources in health-system strengthening and in the quality and quantity of health-care services, especially in rural and remote areas. The region could also be prioritised in terms of technical and financial assistance focused on achieving these goals, as much as on demographic investments including education and family planning. Funding: Bill & Melinda Gates Foundation. Before the GBD project was initiated in 1991, no comprehensive assessments of human health at a global level had been done. GBD is a global comparative risk assessment exercise, with the first preliminary results (for base year 1990) published in the World Development Report 1993.13 This international collaborative effort is currently led by the Institute for Health Metrics and Evaluation in Seattle, WA, USA. GBD uses methodologies for correcting the under-reporting of deaths and those assigned garbage codes. These codes are misclassifications of deaths present in the data. Some of these codes represent cases where the indicated cause cannot logically have caused the death, such as abdominal stiffness, senility, and yellow nail syndrome. Correction of the codes uses evidence from the medical literature, expert opinion, and statistical techniques to reassign each item to the most probable cause of death.13 In this study, we consider the 21 francophone African countries and how they differ from their non-francophone counterparts in three economic communities (appendix 2 p 3). We assess the evolution in the main causes of death in francophone Africa, compared with their evolution in their non-francophone counterparts, and rank the main causes of death in each of the francophone African countries. We assess burden of disease with years of life lost (YLLs) due to premature death, years lived with disability (YLDs), and disability-adjusted life-years (DALYs). We present the evolution of the main causes of YLLs, the total DALYs by sex and country, the expected burden based on SDI, the main risk factors contributing to DALYs, the drivers of change in burden of disease in francophone Africa, and a comparison of the disease burden between francophone and non-francophone countries within the three economic communities. Information about the data sources, estimation methods, computational tools, and statistical analysis used in the derivation of GBD estimates are available elsewhere.14 Data sources used for the GBD analysis in francophone Africa are listed in appendix 3. All data sources used in GBD are evaluated before being included in the analysis. A detailed description of our data sources, their limitations, and their use is published elsewhere.15 All GBD research is done on a public-domain secondary database, without nominal identification, in accordance with US Decree number 7724 of May 16, 2012, and Resolution number 510 of April 7, 2016; thus, there was no need to submit this study to a research ethics committee as no ethics approval was required. This analysis complies with the Guidelines for Accurate and Transparent Health Estimates Reporting.16 After addressing data-quality issues, GBD 2017 employed a variety of statistical models to determine the number of deaths from each cause, through the Cause of Death Ensemble model algorithm.17 To ensure that the number of deaths per cause did not exceed the total number of estimated deaths, a correction technique called CoDCorrect was used. This technique ensures that estimates of the number of deaths from each cause do not total more than 100% of deaths in a given year, age group, and sex strata.18 After producing estimates for the number of deaths from each of the 282 fatal outcomes included in the list of causes in the GBD 2017 study, YLLs were calculated. For every death due to a particular cause, the number of years lost was estimated on the basis of the highest life expectancy in the deceased individual’s age group.19, 20 YLDs were estimated as the product of prevalence and a disability weight for all mutually exclusive sequelae, corrected for comorbidity and aggregated to cause level.15 DALYs were calculated as the sum of YLLs and YLDs. GBD used a list of causes that placed 282 causes of death within a four-level hierarchy (appendix 2 pp 4–13). Level 1 divided causes into three groups: communicable, maternal, neonatal, and nutritional diseases; non-communicable diseases; and injuries. Level 2 consisted of 22 major causes of diseases such as neonatal disorders, cardiovascular diseases, and traffic injuries. Level 3 subdivided Level 2 causes into types such as neonatal preterm birth complications, cerebrovascular disease, and traffic injuries. Level 4 further subdivided those types in some cases—eg, ischaemic stroke and haemorrhagic stroke; and pedestrian road injuries, cyclist road injuries, motorcyclist road injuries, motor vehicle road injuries, and other road injuries. The leading causes of death were analysed using the Level 3 aggregation of causes of death from the GBD 2017 study (appendix 2 pp 4–13). Information on risk factors and their attributable DALYs have been described in detail previously.21 Briefly, GBD uses the comparative risk assessment framework developed for previous iterations to estimate levels and trends in exposure, attributable deaths, and attributable DALYs, by age group, sex, year, and location for 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks from 1990 to 2017. The GBD 2017 study included 476 risk–outcome pairs that met the GBD study criteria for convincing or probable evidence of causation. Relative risk and exposure estimates were extracted from 46 749 randomised controlled trials, cohorts, pooled cohorts, household surveys, census data, satellite data, and other sources, according to the GBD 2017 source counting methods. Using the counterfactual scenario of theoretical minimum risk exposure level, the portion of deaths and DALYs that could be attributed to a given risk were estimated. Since 2015, GBD has estimated the expected burden for each of the three principal measures—deaths, YLLs, and YLDs, with DALYs being the sum of the last two—as a function of each country’s SDI.22 SDI was first developed for GBD 2015 to provide an interpretable synthesis of overall development, as measured by lag-dependent income per capita, average educational attainment in the population over 15 years of age, and total fertility rates. In GBD 2017, SDI was computed by rescaling each component to a scale of zero to one, with zero being the fewest years of schooling, lowest income per capita, and highest fertility, and one being the most years of schooling, highest income per capita, and lowest fertility, and then taking the geometric mean of these values for each location-year.17 Starting with GBD 2016, some modifications have been made to better use each of the scales. The minimum and maximum have been set by examining the relationships each of the inputs had with life expectancy at birth and under-5 mortality and identifying points of limiting returns at both high and low values, if they occurred before theoretical limits.14 Furthermore, for GBD 2017, total fertility rate was replaced with fertility rate under 25 years of age, which provides a better measure of women’s status in society as it focuses on ages where childbearing disrupts the pursuit of education and entrance into the workforce, and income was replaced with the lag-distributed income per capita.23 The average relationship between SDI and disease burden was evaluated for each age-sex-cause group using a smoothing regression spline on SDI for each cause in the GBD cause hierarchy. The estimates were scaled from more detailed causes up to the most aggregated to ensure the total predicted burden at the highest level equalled the sum of the lower levels.22 To analyse the drivers of change in disease burden, GBD decomposes trends in diseases and attributable burden into contributions from population growth, changes in population age structures, changes in exposure to environmental and occupational risks, changes in exposure to behavioural risks, changes in exposure to metabolic risks, and changes due to all other factors, approximated as the risk-deleted death and DALY rates. These methods are detailed elsewhere.21 Uncertainty levels were propagated at multiple stages throughout the GBD modelling process. Uncertainty for mortality and YLLs reflected uncertainty in the levels of all-cause mortality and in the estimation of each mortality cause in each age group, sex, and year. Uncertainty in the disability weight for each sequela was propagated into the estimates of YLDs for each disease and injury. A sample of 1000 draws was taken from the posterior distribution of each estimation step; aggregation of uncertainty across age, sex, and location was done on each draw, assuming independence of uncertainty. The lower and upper uncertainty intervals (UIs) represent the ordinal 25th and 975th draws of each quantity and attempt to describe modelling as well as sampling error.24 95% UIs take into account the uncertainty in the epidemiological parameters used to estimate YLLs, YLDs, and DALYs. 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 responsibility for the decision to submit for publication.